Chemical Transport by Methane Ebullition in a Freshwater Lake
by
Kyle Brook Delwiche
B.S. Civil and Environmental Engineering
University of California, Berkeley, 2007
M.S. Biological and Environmental Engineering
Cornell University, 2012
Submitted to the Department of Civil and Environmental Engineering in partial
fulfillment of the requirements for the degree of
Doctor of Philosophy
in Environmental Engineering
at the
Massachusetts Institute of Technology
June 2018
C Massachusetts Institute of Technology 2018. All rights reserved.
Signature redacted
Signature of Author:
Department of Civil and Environmental Engineering
May 18, 2018
Certifiedby: _Signature redacted
Harold F. Hemond
William E. Leonha d Professor of ivil and Environmental Engineering
Signature redacted Thesis Supervisor
Accepted by:
MASSACHUSETTS INSTITUTE Jesse H. KrollOF TECHNOLOGY Assodiate Professor of Civil and Environmental Engineering
Associate Professor of Chemical Engineering
JUL 2 6 2018 Chair, Departmental Committee for Graduate Students
LIBRARIES 1
ARCHIVES
2
Chemical Transport by Methane Ebullition in a Freshwater Lake
by
Kyle Brook Delwiche
Submitted to the Department of Civil and Environmental Engineering
on May 18, 2018 in partial fulfillment of the
requirements for the degree of Doctor of Philosophy in
Civil and Environmental Engineering
ABSTRACT
Methane bubbling from lakes contributes significantly to atmospheric methane levels,
and methane is second only to carbon dioxide in global warming potential. Microorganisms in
aquatic sediments produce methane while consuming organic matter, and the majority of this
methane is released via bubbling. Bubbles dissolve as they rise, and the fraction of original
methane that dissolves versus escapes to the atmosphere is strongly influenced by bubble size.
While bubble sizes are critical to methane fate, traditional methods of measuring bubbles sizes in
situ are resource intensive (i.e. sonar or video cameras).
In this work we design, build, and deploy a fleet of novel optical bubble size sensors
capable of measuring methane bubbles in situ for long periods of time. Data from our field
campaign on Upper Mystic Lake, MA illuminate spatial differences in bubble size distributions
and provide an estimate of the contribution from methane bubble dissolution to dissolved
methane accumulation. These results improve our understanding of processes governing the
emission of this important greenhouse gas.
In addition to transporting gas, bubbles effectively transport particles in water columns.
This process has been used extensively in industry since the 1900s to separate chemicals of
interest from bulk solutions. While bubbles also transport particulate matter in marine systems,
to date very little work has focused on the possibility that methane bubbles transport particles in
freshwater systems. We use laboratory and field experiments on Upper Mystic Lake to show that
bubbles can transport arsenic-containing sediment particles to the surface of the lake from depths
exceeding 15 m. While we estimate that arsenic transport is insignificant at the relatively modest
methane bubbling levels in Upper Mystic Lake, other water bodies experience an order of
magnitude more ebullition and bubbling may therefore constitute a significant contaminant flux
in these systems. Furthermore, bubbles may also transport organisms (or pathogens) from the
sediment to the water surface.
Thesis Supervisor: Harold F. Hemond
Title: William E. Leonhard Professor of Civil and Environmental Engineering
4
Acknowledgments
I am very grateful for the expertise, guidance, and intellectual freedom Harry Hemond has given
me during my six years at MIT. He has been a truly excellent advisor, both by offering me spot-
on suggestions and advice when I need it, and also by giving me the time and space to explore the
research ideas that excite me. It has been a true pleasure to work with him. I would also like to
thank the rest of my committee, Phil Gschwend and Heidi Nepf, for their suggestions that helped
guide and improve my research.
I owe a debt of gratitude to the scientific and social community of Parsons. I have benefitted
enormously from the diversity of scientific expertise within the building, and the collegial
atmosphere that encourages people to share their time and knowledge. Much of my field work
depended on recruiting volunteers, and it is a testament to the giving nature of Parsons that I never
struggled to recruit a willing helper. I will miss the lively lunches, Friday breakfasts, Halloween
parties, and innumerable daily conversations that have made working here so much fun.
Scientifically, the Supergroup meetings have been an invaluable learning experience. Supergroup
has sharpened my critical thinking skills and gotten me used to a barrage of pointed and helpful
questions when I present. Presenting is much less intimidating when you are used to having Phil
Gschwend in your audience (thank you, Phil)!
Ben Scandella, my fellow bubble chaser, taught me everything I know about boat-based field work,
was an excellent sounding board for thinking through challenges, kept me company on countless
visits to Upper Mystic Lake, and his sunny and goofy personality always made me smile. Irene
Hu has been an excellent labmate, friend, and willing field work helper. I would like to thank
Carolyn Ruppel and Amy Mueller for sharing their career advice and life experiences.
My favorite place on campus is the Edgerton machine shop, and this is thanks in large part to Mark
Belanger's genius. He is a superb teacher, and I am constantly amazed by his technical skill,
patience, and crystal-clear explanations. My research would not have been possible without the
skills he taught me. John MacFarlane has been my go-to person for all lab and safety questions,
and I am very grateful for his help and instruction over the years. Vicky Murphy has cheerfully
and expertly guided me through the tangle of University operating policies, making my work easier
in the process. I also thank my excellent summer UROPs William Popov, Tim Manganello, and
Elise Bickford who each spent many days out on the lake with me.
I thank my Mom for being my emotional rock, particularly in the last few years as I have navigated
my own transition to motherhood. I thank my Dad for his career wisdom and advice over the years
as I have worked to carve my own career path (and for doing the hard and dirty work of helping
me haul bubble traps out of the lake every November).
Above all, I would like to thank my husband, Joe, and my two precious children, Eli and Noa. Joe,
thank you for always being excited to hear my bubble stories, for being a pillar of support, love,
and patience in my life, and for being an outstanding partner in this crazy juggling game of dual
careers and parenting. Eli and Noa, you two are the light of my life. You motivate me to work
efficientLy so I can come home and play, you give me a sense of perspective when work is
challenging, and the earnest joy in your little faces makes my heart sing.
5
6
TABLE OF CONTENTS
C hapter 1: Introduction ............................................................................................................. 11
Chapter 2: A novel optical sensor designed to measure methane bubble sizes in situ.........20
A BSTRA CT .............................................................................................................................. 20
IN TRO D UCTIO N ..................................................................................................................... 21
M A TERIA LS AN D PROCED URES ................................................................................... 25
Sensor electronics design ................................................................................................... 25
Sensor body design................................................................................................................ 28
Sensor calibration .................................................................................................................. 1
Field deploym ent and m onitoring ..................................................................................... 33
D ata post-processing ............................................................................................................. .34
A SSESSM ENT ......................................................................................................................... 34
D ISCU SSIO N. ........................................................................................................................... 39
CO M M EN TS AN D RECOM M EN DA TION S..................................................................... 40
REFEREN C ES.......................................................................................................................... 41
SU PPLEM EN TA L IN FORM ATION ................................................................................... 44
Chapter 3: An enhanced bubble size sensor for long-term ebullition studies.......... 58
A BSTRA C .............................................................................................................................. 58
INTRODU CTIO N ..................................................................................................................... 58
M A TERIA LS A N D PROCEDU RES ..................................................................................... 61
Introduction to sensor hardw are .......................................................................................... 61
Introduction to bubble detection algorithm ........................................................................ 62
H ardw are upgrades ................................................................................................................ 63
Hardware and software modifications to improve sensor accuracy and precision ............ 67
Laboratory oscilloscope m onitoring procedure................................................................. 68
Field raw data collection procedure ...................................................................................... 69
Sensor hardware and software improvements based on oscilloscope and field data ...... 70
Calibration curve ................................................................................................................... 73
Field data post-processing steps .......................................................................................... 75
A SSESSM EN T ......................................................................................................................... 80
Sensor field perform ance................................................................................................... 80
Potential m echanism s for m issing or rejecting bubble data .................................................. 83
D ISCU SSION ........................................................................................................................... 96
CO M M EN TS AN D RECOM M EN DATION S..................................................................... 97
REFEREN CES.......................................................................................................................... 98
7
SU PPLEM EN TA L IN FO RM A TION ..................................................................................... 101
Chapter 4: Methane bubble size distributions, flux, and dissolution in a freshwater lake 124
IN TROD UCTION ................................................................................................................... 125
M A TERIA LS AN D M ETH OD S ............................................................................................ 127
Sensor deploym ent .............................................................................................................. 127
D ata collection and sam pling .............................................................................................. 129
Bubble dissolution m odel.................................................................................................... 129
RESU LTS AN D D ISCU SSION ............................................................................................. 130
Bubble size and flux characteristics .................................................................................... 130
Bubble dissolution and contribution to dissolved m ethane in lake ..................................... 137
REFEREN CES........................................................................................................................ 141
Chapter 5: Methane bubbles transport sediment and arsenic to a lake surface ................ 151
A BSTRA CT ............................................................................................................................ 151
IN TRO D U CTION ................................................................................................................... 151
Brief introduction to flotation theory .................................................................................. 153
M ETH OD S .............................................................................................................................. 157
U pper M ystic Lake field site history ................................................................................... 157
M aterials and supplies ......................................................................................................... 157
Field sam pling and analysis................................................................................................. 157
Laboratory colum n design and operation ............................................................................ 158
Bubble volum e norm alization ............................................................................................. 160
RESU LTS AN D D ISCU SSION ............................................................................................. 161
Particle transport.................................................................................................................. 161
Im plications for arsenic cycling .......................................................................................... 169
Im plications for organism cycling ....................................................................................... 172
CON CLU SION S..................................................................................................................... 173
REFEREN CES........................................................................................................................ 174
SU PPLEM EN TA L IN FO RM A TION :.................................................................................... 179
Contaminant ratios in bubble column particulate matter and sediment .............................. 179
A rsenic transport in bubble colum ns................................................................................... 180
pH effect on arsenic transport.............................................................................................. 181
C hapter 6: Sum m ary and future w ork ................................................................................. 183
SU M M A RY ............................................................................................................................ 183
8
B ub b le size sen so r ............................................................................................................... 183
Ebullition patterns and bubble dissolution .......................................................................... 184
Bubble particle transport ..................................................................................................... 185
F U T U R E W O R K .................................................................................................................... 186
BROADER IMPACTS ........................................................................................................... 188
APPENDIX................................................................................................................................ 191
Al: Additional information for sensor functioning and calibration........................................ 191
A2: Fine scale temporal spacing for bubbling events ............................................................. 196
A3: Evidence for phantom midge interference in bubble size sensor................. 198
A4: Spatial heterogeneity in bubble flux................................................................................. 200
A5: Turkey Hill weather station details .................................................................................. 201
A6: 2016 field data (including bulk methane and nitrogen content)....................................... 202
A7. Isotope data from methane bubble samples...................................................................... 204
A8. Bubble sensor supply list.................................................................................................. 205
A9. Bubble sizer assembly instructions .................................................................................. 207
A10: Bubble size sensor data processing code........................................................................ 210
A ll. Arduino code for bubble size sensor .............................................................................. 217
A 12. Arduino code for data collection buoy ........................................................................... 234
A13. Arduino library files for bubble size sensor: BDStorage.cpp ........................................ 240
A14: Arduino library files for bubble size sensor: BDStorage.h ............................................ 247
A15: Arduino library files for bubble size sensor: BubbleDetector.h..................................... 249
9
10
Chapter 1: Introduction
Methane is a potent greenhouse gas with critical implications for current and future
climate change. Methane is second only to carbon dioxide in contribution to atmospheric
radiative forcing, and represents around 30% of total radiative forcing from greenhouse gases
(Forster et al. 2007). Atmospheric methane levels have been rising for at least the last 100 years,
and are far above levels from pre-industrial times. Methane levels are expected to continue
rising in a warming climate, and studies of methane sources and fates are considered critical to
improving climate change projections (Bastviken et al. 2011, Kirschke et al. 2013).
Methane gas comes from a variety of natural sources (including wetlands, termites,
freshwaters, and hydrates) and anthropogenic sources (including ruminants, rice paddies,
biomass burning, natural gas production, and landfills) (Wuebbles and Hayhoe 2002). Methane
is produced either biotically or abiotically. Biotic methane production proceeds via
methanogenesis, a form of anaerobic respiration whereby microorganisms living in anoxic
environments consume organic matter and produce methane (Borrel et al. 2011). This
methanogenesis usually proceeds via C02 reduction or acetate fermentation (Walter et al. 2008).
The remainder of the methane is thermogenically produced when organic molecules are broken
down under the high temperatures and pressures at depths exceeding 1 km beneath the Earth's
surface (Judd et al. 2002). Isotopic signatures can distinguish between methane sources, with
lighter methane being indicative of a biogenic source (Flores et al. 2008).
Aquatic systems are large methane producers because sediments are typically anaerobic
and rich in organic matter. Methane accumulates in the sediment until being released by
diffusion, ebullition (bubbling), or transport through plant vascular tissue, although ebullition is
often considered the dominant emission pathway (Walter et al. 2006, Bastviken et al. 2011).
11
Ebullition rates presented in the literature vary widely based on factors such as organic matter
input, nutrient loading, and temperature (DelSontro et al. 2010, Deemer et al. 2016). Studies
have shown that increased temperature and nutrient loading cause a synergistic effect to greatly
increase methane ebullition (Davidson et al. 2018). The link between temperature increases and
more methane ebullition raise concerns that climate change will create a positive feedback loop
for methane generation (Aben et al. 2017).
Methane ebullition is often triggered by drops in hydrostatic pressure caused either by
changes in air pressure due to meteorological events, or changes in water level from tidal action
or reservoir draw-down (Chanton and Martens 1988, Joyce and Jewell 2003, Varadharajan and
Hemond 2012, Maeck et al. 2014). Ebullition can also occur spontaneously when gas generation
exceeds sediment storage capacity (Liu et al. 2016). Bubbles rising through the water column
are subject to dissolution, with methane evading the bubble and dissolved gases present in the
water column (such as nitrogen) entering the bubble. This dissolution reduces the amount of
methane emitted to the atmosphere, and contributes to dissolved methane which can be
biologically incorporated into the lake food web (Kankaala et al. 2006, Bastviken et al. 2013,
Sanseverino et al. 2012). Existing dissolution models are of the form:
dMi -K H i- 4r (1)
dz P Vb
where i is the gas, M is the number of moles of gas i in the bubble, z is vertical distance
(m), KL is the gas transfer coefficient (m/s), H is the Henry's constant (mol/m 3/bar), P is the
partial gas pressure within the bubble (bar), C is the dissolved gas content in the water (mol/m 3),
r is bubble radius (m), and Ub is the bubble rise velocity (m/s). The model adaptation presented in
McGinnis et al. 2006 is widely used in freshwater methane literature, and provides empirical
estimates for the gas transfer coefficient based on bubble diameter as a function of gas
12
diffusivity, rise velocity, and/or diameter. The bubble rise velocity is also determined for a range
of bubble sizes, and is based on factors such as drag coefficient, Reynolds number, gas density,
bubble diameter, and surface tension. We note that numerous parameters in this model have
temperature dependencies, including Henry's constant, gas diffusivity, and water viscosity. We
also note that the HiP product in Equation 1 is very high, and therefore dissolved methane levels
would have to be several orders of magnitude higher than those typically seen in natural systems
for bubble dissolution to be affected.
As seen in Equation 1, bubble dissolution is heavily dependent on bubble size. Small
bubbles can dissolve completely as they rise, whereas slower dissolution rates and faster rise
velocities mean that larger bubbles contribute a larger fraction of their initial methane to the
atmosphere. Despite the importance of bubble size to the fate of methane in rising bubbles,
existing methods to estimate bubble size are typically resource-intensive and therefore only used
for limited deployments (Greinert and Nfltzel 2004, Ostrovsky et al. 2008, DelSontro et al.
2014). These methods include hydroacoustic studies with calibrated sonar devices that can be
mounted on boats to generate bubble size measurements over multiple lake transects. Video
equipment can also gather bubble size information, though this technique is primarily used in
deep-water marine seeps (Sauter et al. 2006, Romer et al. 2012, Wang and Socolofsky 2015).
Given the episodic and heterogeneous natural of methane ebullition, single-day measurement
campaigns may not accurately capture typical bubble size distributions at a given location within
a water body. The ability to measure bubble sizes over longer time periods may therefore
improve estimates of atmospheric methane emissions from ebullition.
Given the importance of methane bubble sizes to dissolution rates and the partitioning of
methane between dissolved and atmospheric pools, Chapters 2 and 3 of this dissertation present
13
the design, construction, and testing of a new optical bubble size sensor. This sensor will be a
valuable addition to current bubble-size measurement techniques, and for the first time will allow
for the long-term study of bubble size distributions in situ. Data from this sensor will also allow
us to quantify the spatial and temporal heterogeneity in bubble sizes, as well as the importance of
methane bubble dissolution to methane cycling in Upper Mystic Lake near Boston, MA
(presented in Chapter 4). This work will provide valuable insights into the role bubbles play in
transporting methane from aquatic sediments to the atmosphere.
While bubbles are an effective methane conduit in natural systems, bubbles are also used
as particle transporters in industrial systems. During bubble particle flotation, surface-active
particles form a stable three-phase contact line with a bubble's gas/water interface and are
transported upward during bubble rise. The flotation process is used extensively in industries
such as mining, where flotation separates valuable minerals from gangue (Rodrigues and Rubio
2007, Min et al. 2008). Bubble particle flotation also occurs in the ocean, where wave-injected
bubbles scavenge a range of organic and inorganic particles and deposit them in a concentrated
surface microlayer (Wallace 1972, Blanchard 1975, and Aller et al. 2005).
Despite the abundant evidence that bubbles are effective particle transporters in industrial
and open-ocean conditions, very few studies have looked at the importance of bubble particle
transport by methane bubbles in freshwater systems. The few existing studies have found
evidence that methane bubbles can transport polycyclic aromatic hydrocarbons (Viana et al.
2012) and manufactured gas plant tar (McLinn and Stolzenburg 2009) from the sediment, but the
full extent of methane bubble particle flotation in aquatic systems remains unknown. Since
aquatic sediment contamination by heavy metals and organic pollutants from industry and
sewage discharge is a widespread problem (Nriagu et al. 1996, Taylor and Owens 2009, Pan and
14
Wang 2012), methane bubble particle transport has the potential to increase contaminant flux
beyond rates predicted from advection and diffusion alone.
Upper Mystic Lake, the field site we used to study methane transport by bubbles, also has
a long history of sediment contamination due to upstream industrial activities. Sediments
contain high levels of arsenic, chromium, and lead in a distinct layered structure that reflects
variable input rates over time (Spliethoff and Hemond 1996). We hypothesize that bubble-
particle transport may mobilize sediment contaminants and deposit toxins such as arsenic on the
lake surface. We therefore conduct field sampling and laboratory experiments to quantify
particle transport rates, identify particle sources, and to estimate the contribution of this transport
to overall arsenic cycling within the lake (presented in Chapter 5).
In summary, this work improves estimates of the fate and transport of methane and
associated particles in bubbles emitted from the sediments of Upper Mystic Lake, MA. This
work as three main goals:
1. Develop a rugged, economical, and low power sensor capable of measuring bubble
sizes in situ for long deployment periods (Chapters 2 and 3).
2. Use the new sensor to characterize spatial and temporal variability in methane bubble
size distributions, and to estimate the importance of methane dissolution to the dissolved
methane budget within the lake (Chapter 4).
3. Quantify bubble-particle transport rates in Upper Mystic Lake, and estimate the
importance of this transport to arsenic cycling (Chapter 5).
The findings from this work will improve our understanding of the unique role that
ebullition plays in methane and contaminant cycling in freshwater systems.
15
REFERENCES
Aben, R. C. H., Barros, N., Donk, E., Frenken, T., Hilt, S., Kazanjian, G., Lamers, L., Peeters,
E., Roelofs, J., Domis, L., Stephan, S., Velthuis, M., Waal, D., Wik, M. Thornton, B.,
Wilkinson, J., DelSontro, T., Kosten, S. (2017). Cross continental increase in methane
ebullition under climate change Cross continental increase in methane ebullition under
climate change. Nature Communications, (November).
Aller, J. Y., Kuznetsova, M. R., Jahns, C. J., Kemp, P. F. (2005). The sea surface microlayer as a
source of viral and bacterial enrichment in marine aerosols. Journal ofAerosol Science, 36,
801-812.
Bastviken, D., L. J. Tranvik, J. A. Downing, P. M. Crill, and A. Enrich-Prast. 2011. Freshwater
methane emissions offset the continental carbon sink. Science 331: 50.
Bastviken, D., Ejlertsson, J., Sundh, I., & Tranvik, L. (2013). Methane as a Source of Carbon and
Energy for Lake Pelagic Food Webs. Ecology, 84(4), 969-981.
Blanchard, D. C. (1975). Bubble Scavenging and the Water-to-Air Transfer of Organic Material
in the Sea. In R. E. Baier (Ed.), Applied Chemistry at Protein Interfaces (pp. 360-387).
American Chemical Society.
Borrel, G., J6zdquel, D., Biderre-Petit, C., Morel-Desrosiers, N., Morel, J.-P., Peyret, P., Fonty,
G. Lehours, A. (2011). Production and consumption of methane in freshwater lake
ecosystems. Research in Microbiology, 162(9), 832-47.
Chanton, J. P., & Martens, C. S. (1988). Seasonal variations in ebullitive flux and carbon
isotopic composition of methane in a tidal freshwater estuary. Global Biogeochemical
Cycles, 2(3), 289-298.
Davidson, T. A., Audet, J., Jeppesen, E., Landkildehus, F., Lauridsen, T. L., Sondergaard, M., &
Syvdranta, J. (2018). Synergy between nutrients and warming enhances methane ebullition
from experimental lakes. Nature Climate Change, 8 (February).
DelSontro, T., McGinnis, D. F., Sobek, S., Ostrovsky, I., & Wehrli, B. (2010). Extreme methane
emissions from a Swiss hydropower reservoir: contribution from bubbling sediments.
Environmental Science & Technology, 44(7), 2419-25.
DelSontro, T., McGinnis, D. F., Wehrli, B., & Ostrovsky, I. (2014). Size does matter: importance
of large bubbles and small-scale hot spots for methane transport. Environmental Science &
Technology, 49(3), 1268-76.
Flores, R., Rice, C., Stricker, G., Warden, A., Ellis, M. (2008). Methanogenic pathways of coal-
bed gas in the Powder River Basin, United States: The geologic factor. International
Journal of Coal Geology, 76(1-2), 52-75.
16
Forster, P., and others. 2007. Changes in atmospheric constituents and in radiative forcing, p.
131-217. In S. Solomon and others [eds.], Climate Change 2007: The Physical Science
Basis. Contribution of Working Group I to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge Univ. Press.
Greinert, J., & Nfitzel, B. (2004). Hydroacoustic experiments to establish a method for the
determination of methane bubble fluxes at cold seeps. Geo-Marine Letters, 24(2), 75-85.
http://doi.org/10.1007/s00367-003-0165-7Joyce, J., & Jewell, P. W. (2003). Physical
Controls on Methane Ebullition from Reservoirs and Lakes. Environmental and
Engineering Geoscience, IX(2), 167-178.
Judd, A. G., Hovland, M., Dimitrov, L. I., Garci, S., & Jukes, V. (2002). The geological methane
budget at Continental Margins and its influence on climate change. Geofluids, 2, 109-126.
Kankaala, P., Huotari, J., Peltomaa, E., Saloranta, T., & Ojala, A. (2006). Methanotrophic
activity in relation to methane efflux and total heterotrophic bacterial production in a
stratified, humic, boreal lake. Limnology and Oceanography, 51(2), 1195-1204.
Kirschke, S., and others. (2013). Three decades of global methane sources and sinks. Nature
Geoscience, 6(September), 813-823.
Liu, L., Wilkinson, J., Koca, K., Buchmann, C., & Lorke, A. (2016). The role of sediment
structure in gas bubble storage and release. Journal of Geophysical Research:
Biogeosciences, 121, 1-14.
McLinn, E. L., Stolzenburg, T. R. (2009). Ebullition-facilitated transport of manufactured gas
plant tar from contaminated sediment. Environmental Toxicology and Chemistry, 28(11),
2298-2306.
McGinnis, D. F., J. Greinert, Y. Artemov, S. E. Beaubien, and A. WOest. 2006. Fate of rising
methane bubbles in stratified waters: How much methane reaches the atmosphere? J.
Geophys. Res. 111: C09007, doi: 10.1029/2005JC003183.
Maeck, A., Hofmann, H., & Lorke, A. (2014). Pumping methane out of aquatic sediments -
ebullition forcing mechanisms in an impounded river. Biogeosciences, 11, 2925-2938.
Min, Q., Duan, Y. Y., Peng, X. F., Mujumdar, A., Hsu, C., & Lee, D. J. (2008). Froth flotation of
mineral particles: Mechanism. Drying Technology, 26, 985-995.
Nriagu, J. 0., Lawson, G., Wong, H. K. T., & Cheam, V. (1996). Dissolved Trace Metals in
Lakes Superior, Erie, and Ontario. Environmental Science & Technology, 30(1), 178-187.
Ostrovsky, I., McGinnis, D. F., Lapidus, L., & Eckert, W. (2008). Quantifying gas ebullition
with echosounder: the role of methane transport by bubbles in a medium-sized lake.
Limnology and Oceanography: Methods, 6, 105-118. Pan, K., & Wang, W.-X. (2012).
Trace metal contamination in estuarine and coastal environments in China. Science of the
Total Environment, 421-422, 3-16.
17
Rodrigues, R. T., & Rubio, J. (2007). DAF-dissolved air flotation: Potential applications in the
mining and mineral processing industry. International Journal of Mineral Processing, 82,
1-13.
R6mer, M., Sahling, H., Pape, T., Bohrmann, G., & Spie3, V. (2012). Quantification of gas
bubble emissions from submarine hydrocarbon seeps at the Makran continental margin
(offshore Pakistan). Journal of Geophysical Research, 117, C10015. Sanseverino, A. M.,
Bastviken, D., Sundh, I., Pickova, J., & Enrich-prast, A. (2012). Methane Carbon Supports
Aquatic Food Webs to the Fish Level, 7(8).
Sauter, E. J., Muyakshin, S. I., Charlou, J. L., Schltiter, M., Boetius, A., Jerosch, K., ... Klages,
M. (2006). Methane discharge from a deep-sea submarine mud volcano into the upper water
column by gas hydrate-coated methane bubbles. Earth and Planetary Science Letters,
243(3-4), 354-365.
Spliethoff, H. M., & Hemond, H. F. (1996). History of Toxic Metal Discharge to Surface Waters
of the Aberjona Watershed. Environmental Science & Technology, 30(1), 121-128. Taylor,
K. G., & Owens, P. N. (2009). Sediments in urban river basins: a review of sediment-
contaminant dynamics in an environmental system conditioned by human activities. Journal
ofSoils and Sediments, 9, 281-303.
Walter, K. M., S. A. Zimov, J. P. Chanton, D. Verbyla, and F. S. Chapin. 2006. Methane
bubbling from Siberian thaw lakes as a positive feedback to climate warming. Nature
443: 71-75.
Walter, K. M., Chanton, J. P., Chapin, F. S., Schuur, E. A. G., & Zimov, S. A. (2008). Methane
production and bubble emissions from arctic lakes: Isotopic implications for source
pathways and ages. Journal of Geophysical Research: Biogeosciences, 113(3).
Wuebbles, D. J., & Hayhoe, K. (2002). Atmospheric methane and global change. Earth-Science
Reviews, 57, 177-210.
Varadharaj an, C., & Hemond, H. F. (2012). Time-series analysis of high-resolution ebullition
fluxes from a stratified, freshwater lake. Journal of Geophysical Research, 117, G02004.
Wallace, G. T., Loeb, G. I., & Wilson, D. F. (1972). On the Flotation of Particulates in Sea
Water by Rising Bubbles. Journal of Geophysical Research, 77(27), 5293-5301.
Wang, B., & Socolofsky, S. A. (2015). A deep-sea, high-speed, stereoscopic imaging system for
in situ measurement of natural seep bubble and droplet characteristics. Deep Sea Research
Part I: Oceanographic Research Papers, 104, 134-148.
Viana, P. Z., Yin, K., & Rockne, K. J. (2012). Field measurements and modeling of ebullition-
facilitated flux of heavy metals and polycyclic aromatic hydrocarbons from sediments to the
water column. Environmental Science & Technology, 46, 12046-12054.
18
A novel optical sensor designed to measure methane bubble sizes in situ
Kyle Delwichel*, Schuyler Senft-Grupp1, Harold Hemond'
'Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, MA,
United States
*Corresponding Author. Phone: 510-928-6583, Email: kyled@mit.edu
Running Head: Sensor to measure CH 4 bubble sizes
Keywords: methane, ebullition, bubble volume distribution, lake, optical sensor
Publication citation: Delwiche, K., Senft-Grupp, S. and Hemond, H. (2015), A novel optical
sensor designed to measure methane bubble sizes in situ. Limnol. Oceanogr. Methods, 13: 712-
721. doi:10.1002/lom3.10060.
(Publication reformatted to fit this dissertation)
19
Chapter 2: A novel optical sensor designed to measure methane bubble
sizes in situ
ABSTRACT
This work presents a novel design for an optical bubble size sensor that is rugged, economical to
build, and capable of accurately measuring methane bubble sizes in aquatic environments over
long deployment periods. The sensor intercepts rising gas bubbles, elongates them in a thin glass
tube, and routes elongated bubbles past an optical detector. The optical detector records
information on bubble rise velocity and travel time, which can be combined with the flow path
geometry to calculate bubble volume at flow rates up to 3 bubbles/second. The sensor circuitry is
powered by 6-V C alkaline battery packs and is cased in a waterproof housing built from
commercially available PVC pipe fittings. Laboratory testing indicates the sensor can accurately
measure bubble volumes up to 1 mL in volume. We deployed the sensor in a lake with a history
of methane ebullition and gathered data on bubble size distributions (most bubbles were between
0.025 mL and 0.2 mL) as well as the precise timing of bubbling events. The sensor also includes
an optional gas collection system to allow for bulk gas sampling.
20
INTRODUCTION
Methane gas is a key climate change driver due to its high radiative efficiency (Forster et
al. 2007). Recent estimates of methane gas emission vary considerably, from 145-260 Tg
CH4/yr for natural methane emissions, and from 264-428 Tg CH4/yr for anthropogenic methane
emissions (Denman et al. 2007). This uncertainty stems in part from the difficulty in measuring
heterogeneous and time-varying natural methane emissions. In particular, lake and reservoir
sediments have recently been shown to produce methane via microbially-mediated anaerobic
decomposition of organic matter in significant volumes and at higher rates than previously
thought (St. Louis et al. 2000; Bastviken et al. 2004).
Methane produced in lake sediments is emitted from sediments by ebullition, diffusion,
or transport through plant vascular tissue. Ebullition is thought to often account for the majority
of methane lake emissions and in some cases 95% of methane emissions have been attributed to
ebullition (Casper et al. 2000; Walter et al. 2006; Bastviken et al. 2011). Although ebullition
may contribute the majority of lake-wide methane emissions, not all methane released from the
sediment as bubbles reaches the atmosphere. As bubbles rise, some methane dissolves into the
surrounding water and is subject to microbial oxidation and thereby can support the lake food
web (Kankaala et al. 2006). The rate of bubble dissolution depends on several factors, including
initial bubble size, temperature, dissolved gas concentrations, and bubble rise velocity (Leifer
and Patro 2002; McGinnis et al. 2006). McGinnis et al. 2006 and Leifer and Patro 2002 have
built bubble dissolution models to explain dissolution rates in marine environments, and these
models are widely used among ebullition researchers.
These models are particularly sensitive to the initial bubble size. The McGinnis et al.
model shows, for example, that bubbles less than 2 mm in diameter may dissolve completely
21
when rising through a 20 m water column. Large bubbles lose much less methane, and because
of the decreasing pressure gradient can actually increase in dimension as they rise. Clearly,
accurate estimates of bubble sizes are essential to estimating the fraction of bubble methane
emitted to the atmosphere and the fraction dissolved into the water column.
Current methods to quantify bubble size distributions typically require divers, underwater
video equipment, or hydroacoustic instruments (Greinert and NUtzel 2004; Ostrovsky et al. 2008;
Vagle et al. 2010; Salmi et al. 2011; Bussmann et al. 2013; Leblond et al. 2014). These methods
usually restrict data-collection opportunities to brief periods in time and/or space. Currently, no
feasible method exists to study how bubble size distributions may vary over long periods of time.
While numerous studies have looked at spatial heterogeneity in methane emissions (DelSontro et
al. 2011; Varadharajan and Hemond 2012; Maeck et al. 2013, Wik et al. 2013), few studies have
addressed spatial variability in bubble size distributions (Ostrovsky 2003, DelSontro et al. 2014).
This gap in data-collection abilities points to the need for a simple, autonomous device
that can measure bubble size distributions and be deployed unattended over long periods of time.
Such a device would allow for a thorough characterization of bubble size distributions over a
range of time and space. Detailed information about typical bubble sizes at different water
column elevations would make it possible to determine if existing bubble dissolution models
accurately predict bubble behavior in lake water columns, and ultimately to better understand the
fate of methane bubbles emitted from lake sediments and the role of methane in the lake carbon
balance.
We present a novel optical bubble size sensor that is rugged, uses relatively little power,
is economical to build, and is capable of accurately measuring bubble sizes over long
deployment periods. The bubble-size sensor measures bubble volumes by optically detecting
22
bubbles passing through a glass tube and recording information needed to calculate bubble rise
time and velocity. The optical sensor uses LEDs and phototransistors attached to a custom
printed circuit board (PCB), and the PCB is cased in a waterproof PVC housing constructed from
standard plumbing parts (Figure 1).
Internal o-ring for funnel
-- U-bolt
Waterproof LED - I Phototransistor
PVC
housing -Printed circuit board
<- Glass funnel stem
OE-Pyrex glassL
funnel
Figure 1: Schematic of bubble sizer housing and internal electronics.
The sensor is deployed underwater where rising methane bubbles are intercepted by an
inverted glass funnel. Larger bubbles elongate as they pass through the glass tube (the funnel's
stem), becoming approximately cylindrical in shape, and their volume is calculated from the
elongated bubble length and the cross-sectional area of the glass tube. In order to determine
elongated bubble length, we use a series of 3 optical detectors and record the time when the top
and bottom of each bubble passes each detector. Figure 2 describes the four data points recorded
for each rising bubble, and how this data is combined to yield bubble volume estimates. Bubbles
with diameters smaller than the tube diameter are not substantially elongated, but appropriate
signal processing still produces accurate estimates of their volumes (Section: Sensor calibration).
23
Detector Identification Number
LED Photocell
2
Detector 1 sees Detector 2 sees Detector 3 sees
bubble, tells bubble, records bubble, records
Detectors 2 and 3 time (Det2Start) time (Det3Start)
to turn on
Area = Flow cross-
sectional area
L = space between
* H Hdetectors
Detector 2 stops Detector 3 stops
seeing bubble, seeing bubble,
records time records time
(Det2End) (Det3End)
Bubble Volume Calculations
LMinimumVelocity = Det3Start - Det2Start 1St Lower Volume Estimate = Minimum Velocity x Det2Time x Area
MaximumVelocity = 2"d Lower Volume Estimate = MinimumVelocity x Det3Time x AreaDet3End - Det2End
Det2Time = Det2End - Det2Start 1 s Upper Volume Estimate = MaximumVelocity x Det2Time x Area
Det3Time = Det3End - Det3Start 2nd Upper Volume Estimate = Maximum Velocity x Det3Time x Area
Figure 2: Schematic of the five steps for detecting a bubble and recording the data
necessary to calculate estimated bubble volume.
As shown in Figure 2, the sensor derives 2 velocity estimates - the velocity of the leading
edge of the bubble (MinimumVelocity), and the velocity of the trailing edge of the bubble
(MaximumVelocity). In the case where bubble acceleration is zero, these estimates would be
equal. However, bubbles initially slow down upon reaching the constriction at the base of the
funnel stem. Once they elongate to fit inside the funnel stem they begin to accelerate. If the
glass tube is not long enough for bubbles to reach a terminal rise velocity before passing the
detectors, data from trailing edge of the bubble will produce a higher velocity estimate than the
leading edge. In order to account for this acceleration, we average the high and low velocity
readings and assume a constant acceleration. We also average the travel time measurements
produced by the detectors 2 and 3, though in practice these measurements are very similar.
24
A -- 1
MATERIALS AND PROCEDURES
Sensor electronics design
During initial electronics design, both optical and electrical conductivity sensors were
tested. Sensing bubble size based on conductivity signals produced inconsistent signals, whereas
optical sensors placed along the bubble flow path produced reproducible and robust signals.
Extensive tests with an oscilloscope (Tektronix DPO 7104) showed that optical sensors had a
high signal to noise ratio, allowing for reliable detection of the leading edge and trailing edge of
rising bubbles for bubbles of varying sizes (Figure 3).
a
b
Figure 3: Oscilloscope traces for bubble optical sensing for three separate elongated
bubbles (a), and 10 bubbles smaller than the tube diameter (b). Elongated bubbles
experience a partial increase in light transmittance when the tube is mostly filled with gas,
in between the leading and trailing edge of the bubble (a).
25
Erg I I I A - N I I
Key electric components used in the sensor design are described Table 1. The bubble size
sensor uses 3 pairs of infrared LEDs and phototransistors, facing each other along the bubble
flow path. The electrical signals from the phototransistors are wired to the analog input pins of
an Arduino Pro Mini circuit board. The Arduino's microcontroller (ATMEGA328) is
programmed to record baseline phototransistor output, identify bubbles when the phototransistor
output significantly deviates from this baseline, trigger the data gathering sequence shown in
Figure 2, and store the resulting data on an EEPROM chip. The Arduino program files are
available from the author upon request.
Table 1: Sensor electronic components.
Component Manufacturer Part Description Price in
Number 2015
Data chip Microchip 24LC1025- EEPROM IMbit $3.50
Technology I/P 400kHz 8DIP
Microcontroller Arduino DEV-1 1113 Arduino Pro Mini 328 - $10
5V/16MHz
Breakout board Future Technology DEV-09716 FTDI basic breakout - $15
Devices 5V
Infrared photo- Everlight Electronics PT928-6B-F Phototrans 1.5mm side $0.50
transistor Co. Ltd. facing
Infrared LED Everlight Electronics IR928-6C-F LED IR 1.5mm side $0.50
Co. Ltd. facing
PC board Advanced circuits
26
Arduino Pro-Mini
000000
he 0 O Z
_j z 0
OTXO RAWO-
ORXI GNDO-
- ORST RS1O
-0 GND Vcc0-
-02 A5 A30
,-3 A4 A20-
A LED -04 Al -
A LED 05 AOO-
15kG r LED 06 130
15kQ 07 120
15kO 08 110
09 100
A V_
VCC A2 SCL
VSS SD
EEPROM data chip
MicroFit 3.0 2 position
connector
-_-L.10
SpF
6V battery pack
10 10 10
vPhoto- "-Photo- v Photo-
'transisto "transistor transistor
o C.
Figure 4: Schematic diagram of bubble sizer circuitry showing power, LEDs,
phototransistors, Arduino Pro-Mini, and EEPROM data chip connections.
The complete circuit (Figure 4) is fabricated on a custom PCB as shown in Figure 5.
Design files for PCB manufacturing are available from the author upon request. PVC shims are
used to properly align the LEDs and phototransistors with the bubble flow path at a height of
0.43 cm above the PCB (See Figures S7 and S8 in the Supplemental Information for PVC shim
dimensions). The electronic circuitry is powered using 6-volt battery packs, each composed of 4
C-size alkaline batteries placed in series, with three 6-V packs in parallel for each sensor (Figure
5A).
27
I
a b
> Battery Connection
mm
LED
Micro-
computer Custom
piece for
LED
U-Bolt alignment
Phototransistor
Data Storage Chip
Figure 5: 6-V battery packs inside the waterproof PVC sensor housing (a). Custom PCB
(front and back view) (b). The PCB back holds the microcomputer and data chip, and the
front holds the LEDs and phototransistors.
The glass tube routing bubbles past the sensor electronics is the stem of a 6140 Pyrex
funnel with a 75mm diameter cone. While more fragile than plastic, preliminary testing showed
that the hydrophilic properties of the Pyrex material reduced instances of bubbles sticking either
at the constriction point or along the thin travel tube. The PCB was secured to the glass funnel
stem using U-bolts for 0.635 cm (0.25 inch) piping. Between the U-bolts and glass tube we
wrapped a 1 cm length of slitted 0.635 cm (0.25 inch) plastic tubing for cushioning and proper
alignment between the glass tube and optical sensors.
Sensor body design
A detailed materials list and construction guide with step-by-step instructions for the
construction of the sensor body can be found in the Supplemental Information. Briefly, the
sensor body is made from a nominal 4 inch PVC pipe coupling and 2 pipe plugs, as shown in
Figure 6. O-ring grooves and EPDM dash size 244 O-rings in the side of each pipe plug allow
28
for a water-tight seal between the coupling and the pipe plugs. The sensor electronics and
batteries are housed inside this PVC coupling. The Pyrex funnel that routes bubbles past the
optical sensor is passed through two holes drilled into the center of the top and bottom pipe
plugs. To create a water-tight seal between the funnel stem and the pipe plugs, 2 small o-ring
grooves are machined in to the holes in the top and bottom pipe plugs. The PCB containing the
sensor electronics is mounted on the funnel stem inside the housing.
In order to increase the number of intercepted bubbles and to collect the bubbles for gas
sampling, we also designed a detachable extension funnel and a gas collection chamber.
Detailed construction guidelines for each of these can also be found in the Supplemental
Information. The detachable extension funnel is made from a thin sheet of high-density
polyethylene (HDPE), rolled into a conical shape and secured with stainless steel nuts and bolts.
The extension funnel is attached to the sensor housing with three lengths of nylon rope running
from holes along the base of the cone to D-rings attached to the sensor housing with a hose
clamp (as shown in Figure 6). To protect the glass funnel from breaking, the extension funnel is
offset from the glass by approximately I cm with a short length of 4 inch diameter PVC pipe.
The size of this extension funnel can be changed to increase or decrease the number of
intercepted rising bubbles. We used a funnel diameter of 0.5 m for the field experiments
reported herein.
29
> Stainless steel s-hook
B d g Gas collection tubeBarbed tube fitting
Eyebolts
Detachable gas collection -> Nylon rope
chamber
Custom PVC fitting with
male-male PVC coupling > PVC pipe plug
Stainless steel hose clamp
> Stainless steel d-ring
PVC coupling
Eyebolts
Zip-tie Pyrex 6140 glass funnel
PVC extension piece to
protect glass funnel 
- Nylon rope
Custom detachable
HDPE extension
funnel
Gravel bag weights attached
to funnel base
Figure 6: Bubble size sensor complete assembly.
The gas collection chamber is mounted to the top of the sensor housing body, and collects
all gas as it exits the funnel. Flexible PVC tubing runs from the gas collection chamber to a
surface buoy (as shown in Figure 7), allowing the user to retrieve the contents of the gas
collection chamber at the water's surface. Gas can then be taken back to the lab for analysis.
Additionally, the sampled gas volume measurement serves as an independent check on the
cumulative volume recorded by the bubble size sensor.
30
I
Polyform AO series buoy - 3-way stopcock with Luer Lock
fitting
~2m Masterkleer PVC _yPolypropelene-
-2m tubing (0.125" I.D.) C covered elastic
cord (0.25" O.D.)
Zip-tie connections
every ~2 m n
Nylon rope
+- Bubble size sensor -
600 <- Cinderblock anchor
Figure 7: Bubble sizer deployment and sampling set-up.
Sensor calibration
Since the bubble size sensor directly measures bubble geometry, minimal calibration
should be required to achieve a 1:1 fit between actual bubble volume and measured bubble
volume. To test this, sensor calibrations were carried out in a bench-top water tank. To create
bubbles of known sizes, we used calibrated Eppendorf pipettes to release known volumes of air
underwater into an overturned beaker. Each air bubble was then manually released below the
bubble size sensor. This method allowed for the creation of bubbles with specific, repeatable
volumes. We used volumes from 0.01 mL to 1.5 mL. Gas volumes lower than 0.01 mL could
not consistently be ejected from the pipette tips, and bubbles larger than 1.5 mL tended to break
apart during passage through the inverted glass funnel.
As shown in Figure 2, the data collection sequence produces two velocity estimates and
two time estimates for a total of four bubble volume estimates. Since the observed flow times
for detectors 2 and 3 are very similar, the largest difference in estimates is due to the velocity
measurements. The velocity measurements differ because bubbles are still accelerating when
31
they pass each optical detector. Assuming a linear acceleration and averaging the two velocities
and flow times yields a nearly 1:1 fit between actual and measured bubble volumes between 0.01
mL and I mL (Figure 8). Beyond 1 mL the linear fit is less significant, potentially because
bubble acceleration becomes non-linear and the average velocity is less accurate.
1.8
*1st Lower Velocity
1.6 Estimate
o 2nd Lower Velocity
1.4 Estimate
4 ~ A 1st Upper Velocity
E .2Estimate
~1.2A
o 2nd Upper Velocity
E Estimate
1 x Average
S0.8 -/
a / * - 1:1 Line
as0.6 -
__ Small Bubble Linear Fit
Y = 1.05x + 0.013
0.4
.4------ Large Bubble Linear Fit
Y = 0.536x + 0.490
0.2
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
Actual volume (mL)
Figure 8: Bubble size sensor measurements for bubble volume plotted against actual
bubble volume (n = 3 for each data point). For clarity, only standard deviation error bars
for the 2nd Upper Velocity Estimate are shown. Standard deviations for the 1st Upper
Velocity Estimate are similar in magnitude, and error bars for the lower velocity estimates
are within the range of symbol size. Plot shows linear range between 0.1 mL and 1.0 mL
used for calibration equation, and approximate linear fit between 1.0 mL and 1.5 mL. 1 to
1 line drawn for comparison.
While simply averaging the velocities and flow times yields a nearly 1:1 fit between
actual and measured bubble volumes up to 1 mL, this fit can be improved by applying a linear
calibration equation for bubbles between 0.01 mL and I mL (equation given Figure 8, R2 =
32
0.9985). The fact that a calibration equation improves the volume estimates is not surprising
given that the precise moment a rising bubble triggers each detector is determined by the
complicated optical interaction between the cone of light emitted by the LEDs and the curved
meniscus of a rising bubble.
To estimate the flux contribution from large bubbles we have included a second linear
calibration line for bubbles between 1 mL and 1.5 mL (R 2 = 0.9261, Figure 8). While this curve
allows us to approximate large bubble volumes, the lower certainty of the calibration curve
between I mL and 1.5 mL adds additional uncertainty to the total flux measurement. Future
work should focus on improving the measurements for large bubbles.
Field deployment and monitoring
We installed 3 bubble size sensors in Upper Mystic Lake near Boston, MA during the late
summer and fall of 2014. Upper Mystic Lake is a dimictic kettle lake with an observed history
of methane ebullition (Varadharajan and Hemond 2012). The sensors were installed 1 m above
the hypolimnetic sediments at a depth of approximately 16 meters. The sensor deployment set-
up consists of 0.635 cm (0.25 inch) nylon rope, cinderblock anchors, four 20.32 cm (8 inch)
Polyform buoys, a length of elastic cord, and 0.318 cm (0.125 inch) plastic tubing running from
each sensor gas collection chamber to the surface buoy (Figure 7). The elastic cord allows the
center surface buoy to rise and fall in response to changes in the lake water level without
becoming submerged.
The bubble size sensors were left in place for 3 week intervals. After 3 weeks, a 50 mL
Luer-Lock syringe was used to sample all gas from the gas collection chambers and record the
collected volume. We then pulled the sensors to the surface for data retrieval and battery
33
replacement. Data retrieved from each sensor was processed to remove any abnormal data
points as described in the Data Post-Processing section below.
Data post-processing
There are several physical scenarios that in principle can lead to erroneous bubble data,
and thus a post-processing step is implemented to remove those events that clearly lead to error.
The first scenario addressed in the post-processing step is when detector 2 senses a bubble that
the detector 3 does not, or vice versa. This could occur when multiple bubbles coalesce within
the tube while traveling between the detectors, or when larger bubbles break apart between the
detectors. The second scenario is when detector 1 is triggered by something other than a bubble
(e.g. other opaque objects in the water column), and detectors 2 and 3 are not triggered. A
custom Matlab program identifies these scenarios based on data points with missing timing
information (e.g. Det3StartTime and Det3EndTime are 0) or when the calculated velocity is
negative.
For the data presented in the Assessment section, approximately 12% of the initial raw
data points were removed during post-processing. Of these removed data points, 82% were
removed due to data consistent with multiple bubbles joining together or larger bubbles breaking
apart within the tube.
ASSESSMENT
Results from our calibration tests demonstrate that the sensor can accurately measure
bubble sizes for individual bubbles up to 1 mL in volume. In these controlled experiments, only
one bubble at a time passed through the glass tube. However, real-world conditions are likely to
34
contain a wide range of ebullition flow rates, likely increasing the probability of errors due to
bubble coalescence.
In order to determine the sensor accuracy under higher ebullition rates, we ran a range of
bubble flow rates through the sensor using a syringe pump. The syringe pump was set up to
create a column of rising bubbles by ejecting air through an outlet submerged below the size
sensor. Two different outlet sizes were used to test flow rates with bubble volumes of 0.04
0.004 mL and 0.08 0.009 mL. The bubble volume passing through the sensor was collected for
flow rate verification. The resulting data were processed (see data post-processing section) and
corrected using the calibration curve discussed in the Sensor calibration section. The actual flow
rate in mL/min was then compared with the cumulative volume measurements recorded by the
size sensor.
30
.E 25
E
20 -
_ O0 K 0.04 0.004mL (4.2
15 -2mm diameter) bubbles
o0
S 0.08 0.009ml (5.3
a 10 3mm diameter) bubbles
-- 1:1 Line
5 -
0
0 10 20 30
Actual flow rate (mL/min)
Figure 9: Flow rate tests conducted for small bubbles (0.04 0.004 mL) and larger bubbles
(0.08 0.009 mL). Tests show that sensor can resolve approximately 3 bubbles per second
for both bubble sizes.
35
For smaller bubbles, the sensor had a nearly linear response between actual flow rate and
recorded flow rate for rates below 8 mL/min (approximately 200 bubbles/min, Figure 9). Above
8 mL/min the sensor could not reliably record flow rates. For the larger bubbles, the response
rate was nearly linear until 14 mL/min (approximately 175 bubbles/min), after which the sensor
began under-predicting the measured flow rate. The onset of under-prediction corresponds with
an increase in rejected data points which could be caused by bubbles coalescing inside the tube.
It is important to note that while we are testing flow rates on a mL/min basis, the temporal
spacing of the bubbles over short time scales is particularly critical (e.g., 5 mL/min of evenly
spaced bubbles may be easier to measure than 5 mL/min of bubbles that are clustered in several
groups). Our results indicate that the sensor can resolve approximately 3 bubbles per second for
both 0.04 mL and 0.08 mL bubbles. In comparison, a recent study looking at bubbling hot-spots
used an echosounder calibrated at a bubbling rate of approximately 1 bubble per second
(DelSontro et al. 2014). For field conditions with high bubble fluxes, the detachable extension
funnel described above could be made smaller to accommodate the sensor's flow rate limitations.
The 2014 field campaign yielded several bubble size distribution data sets, the largest of
which is presented in Figure 10. Of the 837 raw data points recorded by the sensor, 745 data
points were retained after post-processing. While Figure 10 only includes bubbles up to 1 mL in
size (the range over which our calibration curve is most robust), approximately 2% of the
measured bubbles were larger than 1 mL. While the absolute number of these large bubbles is
small (16 out of 735 bubbles), their large size could disproportionately affect the error between
measured bubble volume and collected volume. Indeed, applying the secondary calibration
curve for bubbles between 1 mL and 1.5 mL shows that these 16 larger bubbles comprise
approximately 20% of the total flux, highlighting the importance of large bubbles on atmospheric
36
methane emissions. More work should be done to appropriately extend the calibration curve
beyond 1 mL to allow for more accurate measurement of large but infrequent bubbles.
a
180
160
140
120
a) 100
C- 80 -
C 60
LL40
20
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
b Bubble Volume (mL)
120 -
100 -
O 80 -CC)
: 60 -
0r
2 40 -
20
0
1 2 3 4 5 6 7 8 9 10 11 12 13
Bubble Diameter (mm)
Figure 10: Bubble size distribution curve from Upper Mystic Lake, 3-11 October 2014 with
volume measured in mL (a) and diameter measured in mm (b).
Bubbling in Upper Mystic Lake was so vigorous during the deployment window of
October 3rd to the 1 7t* that the sensor data storage chip reached maximum capacity on October
1 1h. Because of this, the total volume of gas retrieved from the gas collection chamber cannot
be compared to the cumulative volume recorded by the sensor.
As shown in Figure 10, most measured bubbles volumes are between 0.025 mL and 0.2
mL (or 1.8 mm to 3.6 mm equivalent radii). This size distribution is remarkably similar to that
measured in Lake Kinneret, Israel where 90% of bubble radii were between 1.3 mm and 4.5 mm
(Ostrovsky et al. 2008). Our size distribution has 89% of bubble radii between 1.3 mm and 4.5
37
-E -
mm. These radii are three times larger than those reported for bubbles emitted at 42 m in
Sakinaw Lake, British Columbia (Vagle et al. 2010). While determining the factors that govern
bubble volume is outside the scope of this work, recent modeling work suggests that bubble size
is correlated with sediment characteristics (Katsman 2015).
In addition to the valuable bubble volume data obtained by the sensor, we also record
data on the timing of bubble events with millisecond accuracy (Figure 11). Examination of the
data shown in Figure 11 indicates that clusters of bubbles tend to be emitted over short time
intervals (e.g., 78% of bubbles emitted within 5 minutes of a previous bubble), and that larger
bubbles are usually accompanied by a cluster of smaller bubbles. The high level of temporal
detail in this data set has the potential to provide important insights into the mechanisms
controlling bubble release from sediments.
14
122
E E 10
1 0%E10
Taso ubl ees
6 03S
=3 all a 6 S 0
24:00 :00 120 18:0 240 6:0 120 18 0 2:0
Figure 11: Timing of bubbles emitted from Upper Mystic Lake, 8-10 October 2014, a
subset of the data shown in Fig. 10.
38
Our efforts in field deployment and data gathering demonstrated that the bubble size
sensors are adequately durable, with no noticeable degradation occurring over several months of
deployment. Sampling the gas collected in the gas collection chamber from a boat on the surface
was straightforward, allowing for a simple method to validate cumulative volume measurements
while simultaneously producing gas samples for laboratory analysis. Future sensor
improvements will include longer battery life and increased data storage capacity. Additionally, a
data and power cable could be run from the sensor to a surface buoy so that battery replacement
and data retrieval can happen on the lake surface. This would allow for faster data recovery and
long-term deployment above the same patch of sediment with minimal sediment disturbance.
DISCUSSION
Knowing the precise size distribution of methane bubbles emitted from lake sediments is
critical to predicting methane bubble fate within the water column. As explained earlier, bubble
dissolution models that predict the fraction of methane dissolved into the water column and the
fraction emitted to the atmosphere depend on knowing initial bubble size. Since current methods
to acquire accurate bubble size distributions are usually restricted to brief periods in time and/or
space, this relatively simple, low-power bubble size sensor is an important advance to bubble
volume measurements in situ. By analyzing gas collected from the gas collection chambers we
will also be able to determine average bubble methane concentrations, another important
parameter in bubble dissolution models.
In addition to making data collection easier, the sensor enables acquisition of bubble size
distributions over a wider temporal and spatial range than current methods allow. The sensor is
capable of gathering bubble size information over time periods up to approximately one month,
39
and future design improvements in battery life and data retrieval techniques will extend this to an
entire sampling season. In addition, the sensor is relatively easy and inexpensive to build (see
supplemental information for a construction guide), further facilitating the deployment of
multiple sensors to study the spatial variability in bubble size distributions throughout a lake.
The sensor also provides detailed information on the precise timing of bubble release.
Previous work on improving the temporal resolution of ebullition events has yielded important
insights into the mechanisms affecting bubble release (Varadharajan et al. 2010; Scandella et al.
2011), and this bubble size sensor produces time-series much more highly resolved than existing
data series.
COMMENTS AND RECOMMENDATIONS
The detachable design of both the gas collection chamber and the extension funnel allow
for easy customization of the bubble size sensor. The extension funnel could be made larger or
smaller to change the number of intercepted bubbles. If the size sensor were deployed in a
shallow water environment the gas collection chamber could either be shortened or removed
entirely, particularly if independent volume verification is not a priority. In this paper we have
presented a simple, efficient field deployment scheme that works well in the calm, low-traffic
environment of Upper Mystic Lake. However, the size sensor could be attached to a more robust
deployment system suitable for difficult field sites such as those with deeper waters, strong
currents, or heavy boat traffic.
40
ACKNOWLEDGMENTS
This material is based upon work supported by the National Science Foundation under
grant number EAR- 1045193 and Graduate Research Fellowship Program grant number DGE-
0707428, the MIT Martin Family Fellowship to K. Delwiche, the W.E. Leonhard 1941
professorship to H. Hemond, and the Singapore-MIT Alliance for Research and Technology
program. Timothy Manganello was funded by the MIT UROP program and assisted with
coding, equipment trouble shooting, and field work. The authors thank the personnel at the MIT
Edgerton machine shop for their invaluable knowledge and help with sensor housing fabrication.
The authors also thank co-PIs on the NSF grant Carolyn Ruppel and Ruben Juanes for their
guidance on methane ebullition processes, as well as numerous students in the MIT Parsons
laboratory for their help with field work (particularly Benjamin Scandella).
REFERENCES
Bastviken, J. Cole, M. Pace, and L. Tranvik. 2004. Methane emissions from lakes: Dependence
of lake characteristics, two regional assessments, and a global estimate. Global
Biogeochem. Cycles 18, doi:0886-6236/04/2004GB002238.
Bastviken, D., L. J. Tranvik, J. A. Downing, P. M. Crill, and A. Enrich-prast. 2011. Freshwater
Methane Emissions Offset the the Continental Carbon Sink. Science. 331: 6, doi:
10.1 126/science. 1196808.
Bussmann, I., E. Damm, M. Schltiter, and M. Wessels. 2013. Fate of methane bubbles released
by pockmarks in Lake Constance. Biogeochemistry 112: 613-623, doi: 10.1007/s10533-
012-9752-x.
Casper, P., S. C. Maberly, G. H. Hall, and B. J. Finlay. 2000. Fluxes of methane and carbon
dioxide from a small productive lake to the atmosphere. Biogeochemistry 49: 1-19.
DelSontro, T., M. J. Kunz, T. Kempter, A. Wiiest, B. Wehrli, and D. B. Senn. 2011. Spatial
heterogeneity of methane ebullition in a large tropical reservoir. Environ. Sci. Technol. 45:
9866-73, doi: 10.1021/es2005545.
41
DelSontro, T., D. F. McGinnis, B. Wehrli, and 1. Ostrovsky. 2014. Size Does Matter: Importance
of Large Bubbles and Small-scale Hot Spots for Methane Transport. Environ. Sci. Technol.,
doi:10.1021/es5054286
Denman, K., G. Brasseur, A. Chidthaisong, and others. 2007. Couplings Between Changes in the
Climate System and Biogeochemistry, p. 501-568. In M.T. and H.L.M. Solomon, S., D.
Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt [ed.], In: Climate Change 2007: The
Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report
of the Intergovernmental Panel on Climate Change. Cambridge University Press.
Forster, P., V. Ramaswamy, P. Artaxo, and others. 2007. Changes in Atmospheric Constituents
and in Radiative Forcing, p. 131-217. In M.T. and H.L.M. Solomon, S., D. Qin, M.
Manning, Z. Chen, M. Marquis, K.B. Averyt [ed.], Climate Change 2007: The Physical
Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the
Intergovernmental Panel on Climate Change. Cambridge University Press.
Greinert, J., and B. Ndtzel. 2004. Hydroacoustic experiments to establish a method for the
determination of methane bubble fluxes at cold seeps. Geo-Marine Lett. 24: 75-85, doi:
10. 1007/s00367-003-0165-7.
Kankaala, P., J. Huotari, E. Peltomaa, T. Saloranta, and A. Ojala. 2006. Methanotrophic activity
in relation to methane efflux and total heterotrophic bacterial production in a stratified ,
humic , boreal lake. Limnol. Oceanogr. 51: 1195-1204.
Katsman, R. 2015. Correlation of shape and size of methane bubbles in fine-grained muddy
aquatic sediments with sediment fracture toughness. J. Struct. Geol. 70: 56-64, doi:
10.1016/j.jsg.2014.11.002.
Leblond, I., C. Scalabrin, and L. Berger. 2014. Acoustic monitoring of gas emissions from the
seafloor. Part I: quantifying the volumetric flow of bubbles. Mar. Geophys. Res. 35: 191-
210, doi: 10.1007/si 1001-014-9223-y.
Leifer, I., and R. K. Patro. 2002. The bubble mechanism for methane transport from the shallow
sea bed to the surface: A review and sensitivity study. Cont. Shelf Res. 22: 2409-2428.
Maeck, A., T. Delsontro, D. F. McGinnis, H. Fischer, S. Flury, M. Schmidt, P. Fietzek, and A.
Lorke. 2013. Sediment trapping by dams creates methane emission hot spots. Environ. Sci.
Technol. 47: 8130-8137, doi: 10.1021/es4003907.
St. Louis, V. L., C. a. Kelly, t. Duchemin, J. W. M. Rudd, and D. M. Rosenberg. 2000.
Reservoir Surfaces as Sources of Greenhouse Gases to the Atmosphere: A Global Estimate.
Bioscience 50: 766-775, doi: 10.1641/0006-3568(2000)050[0766:RSASOG]2.0.C.
McGinnis, D. F.. J. Greinert, Y. Artemov, S. E. Beaubien, and A. Wtest. 2006. Fate of rising
methane bubbles in stratified waters: How much methane reaches the atmosphere? J.
Geophys. Res. 111: C09007, doi: 10.1029/2005JC003183.
42
Ostrovsky, I. 2003. Methane bubbles in Lake Kinneret: Quantification and temporal and spatial
heterogeneity. Lim 48: 1030-1036.
Ostrovsky, I., D. F. McGinnis, L. Lapidus, and W. Eckert. 2008. Quantifying gas ebullition with
echosounder: the role of methane transport by bubbles in a medium-sized lake. Limnol.
Oceanogr. Methods 6: 105-118.
Salmi, M. S., H. P. Johnson, I. Leifer, and J. E. Keister. 2011. Behavior of methane seep bubbles
over a pockmark on the Cascadia continental margin. Geosphere 7: 1273-1283, doi:
10.1 130/GES00648.1.
Scandella, B. P., C. Varadharajan, H. F. Hemond, C. Ruppel, and R. Juanes. 2011. A conduit
dilation model of methane venting from lake sediments. Geophys. Res. Lett. 38: L06408,
doi: 10.1029/2011GL046768.
Vagle, S., J. Hume, F. McLaughlin, E. MacIsaac, and K. Shortreed. 2010. A methane bubble
curtain in meromictic Sakinaw Lake, British Columbia. Limnol. Oceanogr. 55: 1313-1326,
doi: 10.4319/lo.2010.55.3.1313.
Varadharajan, C., and H. F. Hemond. 2012. Time-series analysis of high-resolution ebullition
fluxes from a stratified, freshwater lake. J. Geophys. Res. 117: G02004, doi:
10.1029/2011JG001866.
Varadharaj an, C., R. Hermosillo, and H. F. Hemond. 2010. A low-cost automated trap to
measure bubbling gas fluxes. Limnol. Oceanogr. Methods 8: 363-375, doi:
10:431 9/lom.2010.8.363.
Walter, K. M., S. a Zimov, J. P. Chanton, D. Verbyla, and F. S. Chapin. 2006. Methane bubbling
from Siberian thaw lakes as a positive feedback to climate warming. Nature 443: 71-5, doi:
10.1038/nature05040.
Wik, M., P. M. Crill, R. K. Varner, and D. Bastviken. 2013. Multiyear measurements of
ebullitive methane flux from three subarctic lakes. J. Geophys. Res. Biogeosciences 118:
1307-1321, doi: 10.1002/jgrg.20103.
43
SUPPLEMENTAL INFORMATION
A novel optical sensor designed to measure methane bubble sizes in situ.
Kyle Delwiche, Schuyler Senft-Grupp, Harold Hemond
Bubble Size Sensor Construction Guide
Sensor Body Materials
- 1 nominal 4 inch PVC coupling*
- 2 nominal 4 inch PVC pipe plugs
- 2 EPDM 0-rings, dash size 244
- 2 Buna-N 0-rings, 7mm inner diameter, 2.5mm wide
- 4 stainless steel eye bolts, 6-32 thread size and 1.9 cm (0.75 inch) length
- 2 stainless steel hex nuts, 6-32 thread size
- 10 cm long piece of nominal 4 inch PVC pipe
- 2 small zip ties, 10.16 cm (4 inch) in length
- Silicone vacuum grease
- 3 cm thick disc of solid PVC, 7.62 cm (3 inch) diameter
- I nominal 1 inch male adapter to male NPT fitting (PVC)
- PVC cement
- Stainless steel hose clamp for pipe sizes 92-165 mm
- 3 stainless steel D-rings for 3.8 cm (1.5 inch) webbing width
- Three 1.5 m lengths of 0.635cm (0.25 inch) diameter twisted nylon rope
*All PVC parts are schedule 40
44
Sensor Body Construction Directions (Refer to Figures S1, S2, and S3)
Repeat for top and bottom pipe plugs:
1. Using a lathe, cut PVC pipe plug to 3.38 cm in length. SAVE the trimmed ring-shaped
portion.
2. Using the lathe, reduce the outer diameter of the trimmed ring from the pipe plug until
ring fits inside of PVC coupling. This ring will keep battery packs in place.
3. Cut O-ring groove into side of the pipe plug. Groove should be 0.282 cm deep and 0.475
cm wide, and positioned approximately 0.8 cm from end of pipe plug (Fig Si-B). Make
sure that groove is smooth. It may help to reduce the lathe speed.
4. Drill a 0.87 cm (11/32 inch) diameter hole into the center of the pipe plug (Fig SI-A).
5. Use a small, hook-shaped lathe tool to cut an 0-ring groove into hole drilled in step 3.
Groove should be 0.33 cm wide and 0.154 cm deep. Refer to Figure S4 for example of
hook-shaped tool.
6. Insert small O-ring in to the small O-ring groove. Do not grease O-ring.
7. Grease the large 0-ring with silicone vacuum grease and insert into the large 0-ring
groove.
Instructions for top pipe plu2 only, Fig S1:
8. On the lathe, machine the flat top of top pipe plug to a smooth finish (Fig SI-C).
9. Drill a 3.33 cm diameter hole in the solid PVC disc. On the lathe, this can be achieved by
drilling a 3.175 cm (1.25) inch diameter hole and widening the hole to 3.33 cm diameter
using a boring bar.
45
10. Using PVC cement, glue the nominal 1 inch male adaptor to male NPT in to the hole in
the PVC disc (Fig SI-C)
11. Using PVC cement, glue the PVC disc and adaptor onto the flat top of the pipe plug,
approximately centered on the small O-ring hole (Fig SI-D).
Instructions for bottom pipe plug only, Fig S2:
12. Drill 2 small pilot holes horizontally into two opposite sides of the pipe plug top, being
careful not to puncture the pipe plug (for upper pair of eyebolts shown in Fig S2-B).
13. Screw two of the 6-32 stainless steel eyebolts into the pilot holes.
14. Drill 2 small pilot holes into 10 cm long piece of nominal 4 inch PVC, approximately I
cm from the top edge and equidistant from each other.
15. Screw the remaining 2 eyebolts horizontally into the side of the 10 cm long piece of
PVC. Secure with the hex nuts.
16. Use the 2 small zip ties to securely fasten the 10 cm piece of PVC to the bottom pipe
plug. This removable piece will protect the glass funnel during deployment (Fig S1-B).
Instructions on Sensor Body Assembly (Fig! S3):
17. Slide the 3 D-rings on to the hose clamp.
18. Slip the hose clamp around the mid-point of the nominal 4 inch coupling, and begin to
tighten the clamp.
19. Slide the D-rings around until they are equidistant from each other. Tighten the hose
clamp until rings are firmly in place and hose clamp is very snug (Fig S3). To facilitate
the placement of the D-rings around the hose clamp, 2 small zip ties can be used to secure
each D-ring to the appropriate location on the hose-clamp.
46
20. Tie the 3 lengths of nylon rope to the 3 D-rings.
Gas Collection Chamber Materials
- 90 cm of nominal 1 inch clear PVC pipe
- 1 nominal 1 inch PVC coupling
- 1 nominal 1 inch socket female to NPT female PVC adapter
- I nominal 1 inch male to 0.25 inch NPT female hex bushing (PVC)
- One nominal 0.25 inch NPT male to 1/8 inch barbed nylon tube fitting
- Teflon tape
- 3 stainless steel eyebolts, 0.25 inch-20 thread size with eye diameter of 0.375 inches.
- PVC cement
Gas Collection Chamber Construction Directions (Refer to Figure S5)
21. Using PVC cement, glue the nominal 1 inch unthreaded female to NPT female adapter on
to one end of the 90 cm clear PVC pipe (Fig. S5-B).
22. Using PVC cement, glue the nominal 1 inch coupling to the other end of the clear PVC
pipe, and then glue the nominal 1 inch male to 0.25 inch NPT female hex bushing in to
the other end of the coupling (Fig. S5-A).
23. Wrap Teflon tape around the barbed tube fitting and screw it in to the nominal 0.25 inch
hex bushing.
24. Cut eyebolt threads to 0.63 cm in length and polish the cut ends on a grinder.
25. Carefully drill three shallow pilot holes horizontally into the pipe coupling, equidistant
from each other and near the top edge of the coupling (Fig S5-A). Be careful not to break
through the wall of the gas collection chamber. Holes should be approximately 0.7 cm
47
deep, or until drill bit first starts cutting in to hex bushing underneath coupling. Tap the
holes.
26. Screw the trimmed eyebolts into the pilot holes, being careful not to puncture the gas
collection chamber. The eyebolts will serve as guides for 3 nylon ropes connecting the
sensor housing to the deployment structure.
27. Drill four 0.637 cm (0.25 inch) holes through the base of the gas collection chamber.
Holes should penetrate the top of the female pipe adapter and the clear PVC tube. These
holes will allow water to exit the chamber as bubbles enter (Fig S5-B).
Detachable Funnel Materials
- 1 sheet of HDPE, 0.16 cm (1/16 inch) thick and 1 m by 0.55 m
- 3 stainless steel bolts, thread size 0.25 inch-20
- 3 stainless steel nuts, thread size 0.25 inch-20
- Three stainless steel washers, 0.25 inch
- Three 1 m lengths of 0.635 cm (0.25 inch) nylon rope
- 4 zipper-top plastic sandwich bags
- 1200 g of small gravel
- 8 long zip ties
Detachable Funnel Construction Directions (Refer to Figure S6)
27. Cut a semi-circular pattern from the sheet of HDPE using Figure S6-A as a guide.
28. Drill three 0.66 cm (0.261 inch) holes equidistant along connection flap (Fig S6-A).
29. Drill holes for nylon rope and rock weight attachment as shown in Figure S6-A.
Ropes and weights should be equidistant from each other when cone is assembled.
48
30. Pull HDPE into a conical shape and secure with the bolts, nuts, and washers.
31. Add 300g of small gravel to each of the 4 sandwich bags. Roll each bag into a
cylindrical shape.
32. Attach rock weights to the cone using the zip ties (Fig S6-B).
33. Tie the 3 nylon ropes on to the 3 rope holes (Fig S6-B).
Electronics Materials
- Custom printed circuit board (PCB) as described in the main text, section titled Sensor
electronics design. Eagle files available from author upon request.
- Four 2 mm machine screws, 12 mm in length
- Four 2 mm nuts
- 2 PVC bars, minimum dimensions 0.9 cm by 1 cm by 3.1 cm
- Electronic parts as described in Figure 5 and Table 1 of main text, and PCB Eagle files
- 2 lengths of straight female headers, 12 pins long
- 1 length of straight female headers, 2 pins long
- 2 lengths of straight male headers, 12 pins long
- 1 length of straight male headers, 2 pins long
- I length of right-angle male headers, 6 pins long
- 3 6-V alkaline battery packs, size C. Packs should be assembled front to back in 2 rows of 2
cells (such as Digikey item P644-L022-ND).
- 2 position Molex Micro-Fit 3.0 I'm connector with female terminals
Electronics Construction Directions
34. Using a mill, cut the PVC bars to the specifications shown in Figures S7 and S8.
These shims will properly align the LED and photocells on the PCB.
49
35. Attach PVC shims to PCB using the 2 mm machine screws and nuts.
36. Solder all electronics and female headers for Arduino attachment to PCB. The 2-pin
length of female headers should be soldered to analog pins A4 and A5.
37. Solder straight male headers to Arduino Pro Mini with pins protruding from back
side. The 2-pin length of male headers should be soldered to analog pins A4 and A5.
38. Solder right-angle male headers to Arduino Pro Mini with pins protruding from the
front side.
39. Slide Arduino Pro Mini into female headers on PCB.
40. Slide EEPROM data chip into IC socket.
41. Wire battery packs so they are connected in parallel and all 3 are connected to the
Molex fitting. Use sufficient wire lengths to allow for battery pack placement inside
the sensor housing.
Complete Sensor Assembly Materials
- Two 0.635 cm (0.25 inch) U-bolts
- Four 0.635 cm (0.25 inch) nuts
- Two 1 cm lengths of 0.635 cm (0.25 inch) flexible plastic tubing, slit down the length of the
tube
- 1 6140 Pyrex glass funnel
- Trimmed ring-shaped portion of the pipe plugs made in step 2
- 10 cm length of 0.635 cm (0.25 inch) inner diameter plastic tubing
50
Complete Sensor Assembly Instructions (See Figure 6 in main text for final product)
42. Push the stem of the Pyrex glass funnel through o-ring hole in the bottom pipe plug.
43. Wrap the two 1 cm lengths of 0.25 inch flexible plastic tubing around the funnel
stem, 5 cm apart and approximately centered along the length of the stem.
44. Using the u-bolts, attach the PCB to the funnel stem at the locations of the flexible
tubing.
45. Push the 4 inch coupling on to the pipe plug (make sure the large o-ring is greased).
46. Slide the trimmed portion of the pipe plug in to the coupling. This will keep the
battery packs from moving around too much.
47. Arrange the 3 battery packs around the PCB, and connect batteries to PCB.
48. Slide the top pipe plug on to the coupling (make sure large o-ring is greased). To
accomplish this, temporarily pull the glass funnel downwards to release pressurized
air from the housing. Once the top pipe plug is snug on the coupling, push the funnel
stem back through the top pipe plug.
49. Push the 10 cm length of plastic tubing on to the top of the funnel stem. This tube
will route bubbles past the 4 holes drilled in to the bottom of the gas collection
chamber, ensuring that all gas is trapped within the gas collection chamber.
50. Tie the nylon ropes from the extension cone to the D-rings on the sensor housing.
51. Screw the gas collection chamber on to the top pipe plug.
52. Take the 3 nylon ropes tied to the D-rings and pass them through the 3 eye-bolts at
the top of the gas collection chamber.
51
These 3 nylon ropes can then be tied to a stainless steel s-hook for attachment to the
deployment structure shown in Figure 8 of the main text.
SUPPLEMENTAL INFORMATION FIGURES
a - Bottom view
0.87 cm (11/32
inch) diameter
hole with internal
o-ring groove
0.272 cm wide
and 0.154 cm
deep
c - Top side view
Nominal 1 inch
male unthreaded
to 1 inch male
NPT adapter
b - Side view
3 cm
33 c{
EPDM o-ring size 244
d - Top view
0-ring groove 0.475 cm
wide, 0.282 cm deep
0.8 cm
3.32 cm diameter hole in
PVC disc
7.62 cm (3 inch) diameter
PVC disk
Pipe plug surface
machined to smooth finish
Figure S1 -Top pipe plug
a - Bottom View b - Side View
Stainless -
steel eye
bolts, 6-32
thread size
Figure S2 - Bottom pipe plug
52
Side View
Stainless steel
hose clamp
D-ring for 3.8 -
cm (1.5 inch)
webbing
Optional zip ties
to position D-
rings
Figure S3 -Assembled sensor housing
Figure S4 - Hook-shaped lathe tool
53
a - Top view c - Side view
Nominal 0.25 -
inch NPT
male to
0.125 inch
barbed fitting
Stainless
steel eye-
bolts, 0.25
inch-20
thread size b - Bottom view
0.637 cm
(0.25 inch)
hole
T, -1
I
Nominal 1 inch
coupling
Nominal 1 inch
clear PVC pipe
Nominal 1 inch
unthreaded female to
female NPT adapter
Figure S5 - Gas collection chamber
a - Cone template
0.64 cm (0.25
inch) hole for nuts
1im
5 cm wide
overlap flap - ---- . . .
0.66 cm (0.261 -,
inch) hole for
nylon rope
0.52 cm (0.203
inch) holes for
attaching rock
weights with zip
ties
0.16 cm (1/16
inch) think
HDPE
b - Cone top side
it
0.635 cm (1/4 Rock bag
inch) nylon rope weights attached
with zip ties
Figure S6 - Detachable extension funnel
54
Stainless steel
bolts
-1
a - Right PVC shim
Figure not drawn to
scale
b - Top view
0.206 cm diameter (0.08
0. 102 cm diam
P-C
ML KJI
D
E
F
G
H
c - Side View
-uQ
--R
V U T
Figure S7 - Right PVC shim
1 inches)
eter (0.04 inches)
Section Length Length
(inch) (cm)
A-B 0.100 0.2540
A-C 0.200 0.5080
A-D 0.780 1.9812
A-E 0.880 2.2352
A-F 1.000 2.5400
A-G 1.100 2.7940
A-H 1.200 3.0480
I-J 0.085 0.2159
I-K 0.105 0.2667
I-L 0.235 0.5969
l-M 0.385 0.9779
N-0 0.150 0.3810
N-P 1.050 2.6670
Q-R 0.182 0.4623
Q-S 0.350 0.8890
T-U 0.105 0.2667
T-V 0.385 0.9779
55
a - Left PVC shim
Figure not drawn to
scale
b - Top view
0.206 cm dia
0.10
(0.0
R- 0
N M LKJI
c Side view
meter (0.081 inchi
2 cm diameter
4 inches)
A
B
D
E
F
G
H
T
Y X W
Figure S8 - Left PVC Shim
Section Length Length
(inch) (cm)
A-B 0.100 0.2540
A-C 0.200 0.5080
A-D 0.320 0.8128
A-E 0.420 1.0668
A-F 1.000 2.5400
A-G 1.100 2.7940
A-H 1.200 3.0480
1-J 0.085 0.2159
i-K 0.105 0.2667
I-L 0.135 0.3429
I-M 0.235 0.5969
i-N 0.385 0.9779
O-P 0.150 0.3810
O-Q 0.300 0.7620
O-R 0.900 2.2860
O-S 1.050 2.6670
T-U 0.182 0.4623
T-V 0.350 0.8890
W-X 0.105 0.1050
W-Y 0.385 0.9779
56
An enhanced bubble size sensor for long-term ebullition studies
Kyle Delwichel*, Harold F. Hemond'
'Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, MA,
United States
*Corresponding Author. Phone: 510-928-6583, Email: kyled@mit.edu
Running Head: Bubble size sensor for long-term studies
Keywords: methane, ebullition, bubble volume distribution, lake, optical sensor
Publication citation: Delwiche, K., and Hemond, H. (2017), An enhanced bubble size sensor
for long-term ebullition studies. Limnol. Oceanogr. Methods, 15: 821-835.
doi: 10.1002/lom3.10201
(Publication reformatted to fit this dissertation)
57
Chapter 3: An enhanced bubble size sensor for long-term ebullition
studies
ABSTRACT
Methane ebullition from freshwater sediments is a significant source of greenhouse gas to
the atmosphere. Methane bubble size determines the fraction of methane that dissolves into the
water column instead of entering the atmosphere, but current methods to measure bubble sizes in
situ are limited. Previously we reported the design for a novel optical bubble size sensor capable
of measuring methane bubble sizes in situ. Here we report sensor enhancements to allow for
continuous long-term sensor deployment, uninterrupted data collection under high ebullition
flux, and improved bubble volume measurement accuracy. The enhanced design includes a data
and power cable running from the sensor to a custom data buoy housing batteries and an SD
memory card. Batteries and SD cards are replaced every 2-4 weeks through the data buoy,
allowing for continuous, undisturbed sensor deployment. In addition, sensor parameters have
been tuned to reduce error from particle intrusion and fluctuations in raw signals that were
observed during transit of some bubbles through the sensor. Field data show that on average the
sensor measures cumulative bubble volumes equaling approximately 86% of collected gas
volumes. We discuss potential mechanisms of remaining error, some of which may be due to
zooplankton intrusion, and suggest directions for future sensor enhancements.
INTRODUCTION
Atmospheric methane currently contributes about 30% as much radiative forcing as
carbon dioxide to the globe (Forster et al. 2007). Recent bottom-up estimates of global methane
58
emissions suggest that approximately 50% of methane comes from non-anthropogenic sources,
11% of which comes from fresh water lakes and rivers (Kirschke et al. 2013). Man-made
reservoirs may be even more prolific methane generators than natural lakes (Deemer et al. 2016).
Methane is generated in lake and river sediments by anaerobic decomposition of organic matter
and enters the atmosphere by diffusion, ebullition, or transport through plant tissue, although
ebullition is often thought to be the dominant emission pathway (Casper et al. 2000, Walter et al.
2006, Bastviken et al. 2011). Significant uncertainty remains with respect to both quantifying
current ebullition flux and predicting future ebullition flux in a warming climate (Walter et al.
2007, Bastviken et al. 2011, Wik et al. 2016a). This uncertainty arises largely from the difficulty
in accurately measuring highly heterogeneous spatial and temporal ebullition patterns (Wik et al.
2016b).
Methane ebullition is often triggered by drops in hydrostatic pressure, either from
changes in air pressure, tide, or water level, or from wind-induced shear stress in sediments,
although such forcing is not required for bubble release (Joyce and Jewell et al. 2003, Chanton
and Martens 1988, Varadharajan and Hemond 2012, Maeck et al. 2014). Importantly, bubbles
rising through the water column are subject to dissolution and invasion by dissolved gases from
the surrounding water, and these processes are highly dependent on the initial size of the bubble
(Leifer and Patro 2002, McGinnis et al. 2006). Small bubbles can dissolve completely into the
water column, thereby contributing via a microbial loop to the lake food web (Kankaala et al.
2006). Larger bubbles have both a lower relative dissolution rate and faster rise velocities, and
therefore contribute a larger fraction of their initial methane content to the atmosphere.
Despite the importance of bubble size to bubble fate in the water, relatively few
techniques are available to measure individual methane bubble volume in situ. Some
59
hydroacoustic studies use calibrated sonar devices to estimate methane bubble size distributions,
and these techniques can provide an estimation of bubble sizes over multiple transects in a lake
(Greinert and NUtzel. 2004, Ostrovsky et al. 2008, DelSontro et al. 2015). However, most
hydroacoustic studies are limited to relatively short deployment windows in part because long-
term hydroacoustic equipment deployment requires additional infrastructure for power delivery
and data storage (Scandella et al. 2015). Video equipment can also aid in bubble size
determination, although this technique is primarily used in deep water marine environments
(Sauter et al. 2006, R6mer et al. 2012, Wang and Socolofsky. 2015). -
Recently our research group developed an optical bubble sizing sensor capable of
measuring bubble sizes in situ (Delwiche et al. 2015). This device is inexpensive to build and
rugged, and the first generation sensor design was capable of continuously measuring bubbles in
one location for deployment periods of up to 2 weeks. The bubble sensor routes bubbles upwards
through a glass tube (the tube is the stem of a glass funnel that intercepts rising bubbles) and past
a series of three infrared beams that are attenuated during bubble passage, providing optical
signals from which bubble numbers and volumes are determined. A microcontroller records
signals from the detectors, and bubble volumes are calculated from these data. Optional sensor
attachments include an external funnel to increase the sampled area, and a gas collection
chamber to provide gas samples as well as an independent verification of cumulative gas volume
passing through the sensor.
This initial sensor design is described in Delwiche et al. 2015. Based on subsequent
design modifications, software improvements, and field deployments in 2015 and 2016, we
describe here a number of hardware and software enhancements. These enhancements greatly
increase the sensor's maximum deployment period and provide increased sensor accuracy,
60
particularly in high bubble flux environments. We also present procedural steps for processing
anomalies in field data, as well as a discussion of potential causes for anomalous field data. In
summary, we present:
1. Design and testing of a companion power supply and data storage buoy to greatly simplify
battery replacement and data downloading while avoiding sediment disturbance, and to increase
data storage capacity for uninterrupted data collection in high flux areas. We also describe an
improved external funnel to increase sampled area (see Materials and Procedures).
2. Hardware and software modifications to improve sensor accuracy and precision (see Materials
and Procedures).
3. Detailed data post-processing steps to properly handle anomalies in field data (see Materials
and Procedures).
4. Field data demonstrating that the sensor works well and measures, on average, 86% of the
cumulative gas volume collected from the gas collection chamber. We include a detailed
analysis of potential sources of missing data, and an assessment of the relative importance of
each (see Assessment).
MATERIALS AND PROCEDURES
Introduction to sensor hardware
The bubble sensor routes bubbles upwards through a glass tube and past a series of
infrared beams that are interrupted by bubble passage, providing optical signals from which
bubble numbers and volumes are determined. The glass tube is the stem of an inverted glass
funnel which is cased in a waterproof PVC housing. A circuit board contains three pairs of
61
infrared LEDs and phototransistors facing each other across the glass funnel stem. We will
henceforth refer to each pair of infrared LEDs and phototransistors as individual
detectors, with detector 1 being on the bottom, detector 2 in the middle, and detector 3
being at the top. The glass funnel protrudes from the bottom of the waterproof PVC housing to
intercept rising bubbles. An Arduino Pro-Mini microcontroller records and stores signals from
the detectors, and the data are used to calculate bubble volumes (detailed calculations provided
in Delwiche et al. 2015). Additional sensor attachments include an expansion funnel to increase
the sampled area if desired, and a gas collection chamber to allow for independent volume
verification of the amount of gas passing through the sensor between sampling periods. Full
sensor schematic, detailed description of bubble volume calculations, and a step-by-step design
guide for the first generation sensor are provided in Delwiche et al. 2015.
Introduction to bubble detection algorithm
The sensor detects the top of a rising bubble by comparing real-time phototransistor
analog to digital converter (ADC) readings with background values. Background ADC
measurements are taken from the phototransistors approximately every second, and the rolling
average of eight consecutive background measurements is used to calculate the threshold to
which the sensor compares future ADC readings. Each sensor records separate background
readings for each detector to accommodate differences in individual LEDs and phototransistors,
or their respective alignment. Bubbles substantially decrease the amount of light reaching each
phototransistor, causing a precipitous drop in ADC readings from background levels. The
sensor will assume a detector is seeing a bubble while ADC readings are 0.1 5V below the
average background reading. The microcontroller records the times detectors 2 and 3 each see
62
the beginning and end of a bubble, and combining this information with the space between
detectors and the flow path cross sectional area yields volume estimates. If an incoming bubble
reaches detector 1 before the previous bubble clears detector 3, the microcontroller will record a
multi-bubble event with the corresponding timing information for each bubble. The
microcontroller can keep track of up to 12 bubbles per bubbling event, and writes the bubble data
to memory in 4-bubble increments.
Hardware upgrades
Custom service buoy and data/power cable for long-term deployment
Sensor battery replacement and data downloading for the first generation sensor
described in Delwiche et al. 2015 required the entire sensor to be pulled to the water surface,
thus limiting the undisturbed deployment interval to the battery life of 1-2 weeks, or significantly
less in high bubble flux locations (due to data storage restrictions). In order to continuously
operate the bubble size sensor without disturbing sediments, we designed a custom service buoy
so all routine service functions can be carried out easily from the water surface. The service
buoy holds 7.4 V lithium ion batteries to power the sensor, and a SD card data storage system.
The PVC service buoy is built with a polyurethane float cast around the buoy body using a
modified angel food cake pan (as was done in Gardner et al. 2009). The float is coated with
epoxy to increase durability. A four-stranded waterproof neoprene-insulated cable carries power
to, and data from, the bubble sensor. A tube for gas sample retrieval is also provided. The buoy
contains a waterproof bulkhead fitting that the data cable connects to, an air release valve, and a
63
cap sealed with an O-ring. Buoy is shown in Figure 1 and detailed instructions on buoy
construction can be found in the Supplementary Information.
a Custom Data Buoy 4-- 3-way stopcock with Luer Lock fitting b
Elastic cord
-2m PVC Tubing - (0.25 O.D.)
(0.125" I.D.) P lfr
- - AO series
Neoprene buoy
data and --
power Nylon
cable rope
+- Bubble size
sensor
S600\ Cinderblock anchor
Figure 1: (a) Bubble sizer deployment and sampling set-up with the new data cable. (b)
Custom service buoy containing batteries and data storage circuit board for long-term
sensor deployment.
The data system on the buoy contains an Arduino Pro-Mini microcontroller which
periodically communicates with the bubble size sensor, typically every 12 hours. During
communication the bubble sizing sensor transmits all data stored on its EEPROM and then clears
the chip. These data are transmitted serially using USB protocol to the buoy's microcontroller
which stores the data on a micro SD card (using an Adafruit MicroSD card breakout board
(Figure 2)). In instances of high bubble flux the microcontroller could be programmed to
transmit data more frequently to ensure that the EEPROM chip on the bubble size sensor does
not reach its maximum capacity of 1023 events. To minimize power consumption, the service
buoy's microcontroller can be programmed to sleep when it is not retrieving data. The service
buoy circuit board contains a real-time clock to record the precise time of data transmission.
Design files for the data buoy PCB and the sensor PCB are available from the author upon
request.
64
PCBs connected through neoprene data cable with bulkhead fittings
MicroFit 3.0 4 position connector Battery clip MicroFit 3.0 4 position connector
7.4V Li-ion
ooo6 - +1rechargeable 00 0 UUV Battery clipbatr Arduino J p 5.Z 7.4V Li-ion C B attery ci e agal
Pro-Mini T .. 10 pF rechargeable Arduin T 0 F
0 G D RST0 Pro-Mini 0 S RST0
02 -OAs A30 02-OA4 A3003 A4 2 0 03 A5O A2 0V
A4 A10 0 3V MicroSDs AoO0- - 05 AO 0 11& k- GND crd
12 06 130 -0 ---O 130- -O CLK c
k() 12 07 120 0 120- -0 O breakout
kb 12 08 110 08 110-o Dr board
-~kQ 0 9 10 O 10 10 10 0 9 100 Cs
-- 
-L 0.1 pF
Photo- -
AU VCCgr transistortenson
AI WP F<0V A2 scLo > >n >n 0:o9 >
extnsin fDS3231 Real Ti iee Clock
EEPROM data
chip
Bubble sizing sensor PCB Data buoy PCB
Figure 2: Circuit diagram for ubnwi stinlss circuit board and data buoy circuit
board.
The service buoy circuit board also contains two surface mounted battery clips for 7.4V,
8800mAh rechargeable lithium-ion battery packs (Tenergy product #31010). One battery pack
powers the bubble sizing sensor, and one pack powers the data buoy circuit board. One battery
can power the bubble size sensor for 2-3 weeks, but this time can be doubled by using two
battery packs in parallel, for a total of 3 battery packs in the buoy.
Upgraded Extension Funnel
In addition to the data cable modifications, we also updated the first generation detachable
extension funnel described in Delwiche et al. 2015 to improve performance in the field (Figure
3). The new extension funnel is constructed from a 0.25 m2 commercial plastic funnel and
secured to the bubble sensor housing with stainless steel rope and carabiners (commercial plastic
funnel originally used by the Global Change and Watershed Biogeochemnistry research group at
65
---j
Washington State University). The steel rope connection to the bubble sensor body is more
robust and easier to attach than the ropes included in the original design. More detailed
instructions on funnel construction can be found in the construction guide in the Supplementary
Information. While this funnel is easier to deploy in the field, it has a fixed collection area. The
original funnel design presented in Delwiche et al. 2015 has a variable collection area and works
well as long as the ropes connecting the funnel to the sensor housing are tightly secured with
appropriate knots.
- Stainless steel s-hook
- Gas collection tube
Barbed tube
fitting Eyebolts
Detachable gas - Nylon rope
collection chamber
Custom PVC fitting ,Bulkhead fitting for data cable
PVC pipe plug
Stainless steel hose clamp
Carabiner < > Stainless steel d-ring
Eyebolts CPVC coupling
Zip-tie Pyrex 6140 glass funnel
Stainless steel wire rope <- Extension piece to protect
Standoff to align sensor glass funnel
on funnel > Eyebolts
12 oz lead <- Commercial ABS
fishing weight plastic funnel
Figure 3: Bubble size sensor complete assembly.
Field Deployment Procedure
We deployed eight sensors in Upper Mystic Lake, MA during the summers of 2015 and
2016. The sensor deployment structure is shown in Figure 1 and consists of a "W" shaped rope
structure suspended between ropes and buoys. The structure contains the data cable as well as
the PVC tubing necessary for gas sampling. The minimum sensor depth with this deployment
66
structure is around 2 meters which is beneficial in a lake with significant public recreation, but
sensors could also be placed closer to the water surface with a different deployment structure.
Specific instructions for deploying sensors on the structure shown in Figure 1 are provided in the
Supplemental Information. We visited deployed sensors every 2 weeks to replace batteries,
download data, and remove collected gas. In sufficiently high bubble flux environments the gas
collection chamber will need to be emptied more often. If using two batteries in parallel to power
the bubble size sensor, the sensor could operate for at least 4 weeks between maintenance visits.
Specific instructions for servicing the sensor and buoy are also provided in the Supplemental
Information. Field data allowed us to analyze bubble sensor performance in the more
challenging environment of a natural lake where the sensor is subject to physical stressors,
contact with micro and macro organisms, and other environmental factors. We also compared
gas volumes collected from the gas collection chambers to the cumulative bubble volume as
measured by the sensor for a further assessment of sensor performance.
Hardware and software modifications to improve sensor accuracy and precision
Data gathered during 2015 and early 2016 field campaigns pointed to the need for a
deeper understanding of sensor functioning under challenging heterogeneous field conditions
such as particle or zooplankton intrusion into the bubble flow path, or during high-flux bubble
events. To understand and improve sensor performance we developed two methods of
measuring detector waveforms and logic levels within the sensor under a range of bubble
conditions in the laboratory and field, as described below.
67
Laboratory oscilloscope monitoring procedure
To directly observe sensor functioning under high-flux bubble conditions, we retrofitted a
bubble sensor cap to include a bundle of wires running directly from individual output pins on
the sensor circuit board to an oscilloscope. The sensor was deployed underwater in a benchtop
fish tank in a stream of bubbles, and the oscilloscope was connected to directly observe raw
signal output from two of the three phototransistors at a time. Figure 4 shows the oscilloscope
traces for the phototransistors on detectors 2 and 3, respectively the yellow and pink traces. The
signal is high when no bubbles are present, and drops when a bubble passes. The signals are
plotted against time, so bubbles are detected by the 2nd detector (yellow trace) slightly before
detection by the 3rd detector (pink trace).
Bubble top passes Detector Two
Bubble bottom passes Detector Two
0
Raw signal from Detector
Two phototransistor0
Microcontroller response
to raw signal
W Raw signal from Detector
Three phototransistor
0
.~-. Microcontroller response
to raw signal
200 ms/division
Figure 4: An example of oscilloscope traces with a description of each trace. The blue and
green traces in Figure 4 represent the microcontroller's response to the passing bubble.
68
Incoming ADC readings from each detector are compared to background ADC values as
described in Materials and Procedures. When an ADC reading drops below the
background threshold the microcontroller records the start of the bubble, and the blue and
green traces jump high. The blue and green traces remain high while the ADC readings
remain low, and once ADC readings recover towards background at the end of the bubble
the blue and green traces drop low again.
Field raw data collection procedure
After using the oscilloscope to record raw sensor waveforms in the laboratory, we
acquired similar data in the field by programming one bubble size sensor to record full time
series of all of the raw phototransistor readings taken during a bubble event from detectors one,
two, and three. The field data resolution was lower than the laboratory oscilloscope raw data due
to timing limitations of the microcontroller. However, even with more coarse data resolution,
field raw data show clear bubble traces similar to those observed in the laboratory with the
oscilloscope (Figure 5). These raw field data allow us to confirm whether sensor behavior seen
in laboratory oscilloscope traces are also observed in the field.
69
Detector Detector
One records One records
start of end of
( bubble bubble
S1000 -
0j 0 i t
0-
0 10 20 30 40 50
ADC Reading (sequential number)
Bubble rises to
0-~ 1000 Detector Two
D0000
4)< 0 E
0 10 20 30 40 50
ADC Reading (sequential number)
()
1000 -
0 500 -
.
.
< 0
o 0 10 20 30 40 50
60
60
60
ADC Reading (sequential number)
Figure 5: Raw field data showing consecutive ADC readings for each detector. Bubble
beginning and end is highlighted for detector 1.
Sensor hardware and software improvements based on oscilloscope and field data
The laboratory oscilloscope work and field raw data collection allowed us to determine
that, in general, the phototransistor signal drops sharply as a bubble passes, and recovers rapidly
once the bubble is past. This clear delineation between background signal and bubble events
makes for robust bubble detection (refer to Materials and Procedures for algorithm description).
However, oscilloscope work revealed that while the body of the bubble passes each
phototransistor, the ADC signal can fluctuate as shown in Figure 6. The frequency and
amplitude of the fluctuations vary based on bubble size and bubble spacing, which indicates that
this is a physical effect as opposed to an electrical phenomenon. Signal fluctuations were also
observed in raw field data (Figure 7).
70
- I
a)
a) -c
Signal fluctuation amplitude Fluctuations appear to
can approach background intensify with decreased
signal levels bubble spacing
Figure 6: Fluctuations apparent in laboratory oscilloscope data (see Fig. 4 for trace
descriptions). Fluctuation frequency and amplitude appears to intensify as bubbles become
closer together and have correspondingly higher rise velocities.
0
o
a
4-0
0
1000
( 500-
0 -
0
1000 -
0
0
.eSOSee0006600060000066 6.606.0 0
gO 0.
I I I I
10 20 30 40 50
ADC Reading (sequential number)
60 70
00
*..0000000000000600e0 0 0 *
0 0
0 00
10 20 30 40 50
ADC Reading (sequential number)
60 70
Figure 7: An example of raw field data showing a signal fluctuation at the trailing edge of
the bubble (right hand side).
71
Interestingly, signal fluctuations seem to increase with decreased bubble spacing and this
phenomenon can also be seen in Figure 6. Bubbles that are closer together have higher rise
velocities, and comparing bubble rise velocity in the tube to a manual classification of the
number of fluctuations observed in the bubble signal yields a significant, positive correlation.
We hypothesize that these signal fluctuations are the product of a complex optical interaction
between the upper and lower menisci of the rising bubble, as well as changes in the thickness of
the water film coating the glass funnel stem during bubble passage. More work is needed to
elucidate the precise physical mechanisms underlying these fluctuations.
These fluctuations present a potential problem if the amplitude becomes high enough for
the sensor to mistake a fluctuation for the termination of one bubble and beginning of a new
bubble. This could lead the sensor to falsely report one bubble event as two bubble events, or for
detectors 1, 2, and 3 to record a different number of bubbles for the same bubble event. Indeed,
field data from 2015 and early 2016 show significant occurrence of events where detectors 2 and
3 saw a different number of bubbles. While we note that there are multiple potential causes for
this data mismatch (as discussed in Assessment), we designed a data smoothing algorithm to
mitigate the contribution of signal fluctuations to anomalous field data.
The smoothing algorithm is designed to minimize the possibility that a short-term
fluctuation could be mistaken for the termination of a bubble. Instead of comparing each
individual phototransistor reading to the bubble detection threshold, the microcontroller uses the
average of 8 successive phototransistor readings. This smoothing algorithm decreases the
likelihood that a short-term fluctuation in the bubble signal will be interpreted as the end of a
bubble. We also increased the sensor sensitivity by increasing LED output by changing the
resistors associated with the LEDs from 15k ohms to 12k ohms, a change that increased the
72
difference between background and bubble readings. Data gathered after the implementation of
the smoothing code and enhanced sensitivity show a 45% decrease in the times detectors 2 and 3
recorded different bubble quantities, indicating that the smoothing code improved sensor
measurement accuracy. We note that even with the smoothing code, 10% of unprocessed field
data contain a mismatch between the number of bubbles seen by detectors 2 and 3, and the
possible causes of this mismatch are more numerous than signal fluctuation alone (see Table 2).
Calibration curve
The sensor improvements described above necessitated an update of the original
calibration curve presented in Delwiche et al. 2015. The sensor was calibrated in a benchtop fish
tank with a range of gas volumes released by an Eppendorf pipette. Each bubble volume
generated four distinct volume estimates as described in Delwiche et al. 2015, and all four
estimates are plotted in Figure 8. The calibrations for bubbles from 0.01 mL to 1.0 mL (0.26 cm
to 1.24 cm equivalent diameter) and 1.0 mL to 1.8 mL (1.24 cm to 1.51 cm equivalent diameter)
were linear with low variability between repetitions. Bubbles below 0.01 mL (0.26 cm
equivalent diameter) were difficult to accurately produce with the pipettes, and sensor
measurement accuracy decreased substantially for bubbles over 1.8 mL (1.51 cm equivalent
diameter). The positive intercept on the small bubble linear fit means tiny bubbles cannot be
calibrated for, as discussed in the Assessment.
73
2.2 +T st lower volume
2 estimate
0 2" lower volume
estimate
E 1.6 0 18 upper volume
' 14 estimate
0 2n upper volume
5 1.2 estimate
-- ---- 1:1 line
0.8  x Average
CD
.
-.- Large bubble linear fit
0.4 y =0.45 x + 0.732
0.2 Small bubble linear fit
0 y = 1.174x + 0.021
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
(0.73) (0.91) (1.05) (1.15) (1.24) (1.32) (1.39) (1.45) (1.51)
Actual volume (mL)
(Equivalent diameter (cm))
Figure 8: Bubble size sensor measurements for bubble volume plotted against actual
bubble volume (n = 4 for each data point). Plot includes the four volume estimates
generated during each bubble event, as well as the average of these estimates. Plot shows
linear ranges from 0.01 mL to 1.0 mL (0.26 cm to 1.24 cm equivalent diameter) and 1.0 mL
to 1.8 mL (1.24 cm to 1.51 cm equivalent diameter) used for calibration equations. The 1 : 1
line is drawn for comparison.
We note that in Delwiche et al. 2015 we attributed the sensor's inability to accurately
measure large bubble sizes to the tendency for large bubbles to break apart within the glass
funnel stem. However, further analysis of the data showed that typically at least 97% of the
original gas volume is contained in a single large bubble, followed by 1-3 small bubbles, and
therefore bubble decomposition is likely an insignificant contributor to sensor inaccuracy. We
surmise that sensor inaccuracy for large bubbles stems from the complicated effect on bubble
velocity of the bubble exiting the top of the funnel stem, and the fact that large bubbles will exit
the funnel stem before the trailing edge velocity is measured. This aspect of the sensor
functioning warrants further investigation.
74
MMI
Field data post-processing steps
Unprocessed field data description
After bubble sensor deployment and data collection, field data must be processed to
remove incomplete data, or data that do not accurately measure bubble volumes. The potential
causes of incomplete or inaccurate data are numerous and will be described in detail in the
Assessment, but could include issues such as zooplankton intrusion, sensor confusion during
high flux bubble events, or noisy background signals. While there is no way to tell from the
field data which physical or electronic mechanism may be contributing to erroneous data,
specific data post-processing guidelines can distill reliable bubble data from the un-processed
data stream.
Table 2: Possible error codes for each bubble event (summed if more than one occurs for a
given bubble event)
Error Code Error Description
I Data collection time exceeds maximum timeout threshold of 5 seconds
10 Detector 2 saw more bubbles than detector 3
100 Detector 3 saw more bubbles than detector 2
1000 Detector 1 still looking for the end of a bubble
10000 Detector 2 still looking for the end of a bubble
100000 Detector 3 still looking for the end of a bubble
1000000 Detectors 2 and 3 do not see a bubble
The microcontroller is programmed to record the precise timing of each bubble event, as
well as specific raw analog signals from each detector. Figure 9 presents an example of the data
75
stream for one bubble with no errors, along with a description of each number stored during a
bubble event. Unprocessed data also contain bubble data with errors, and error codes are
described in Table 1. Depending on the error code the bubble event may still contain enough
information to calculate bubble volume, but often these data must be discarded because some of
the values shown in Figure 9 will be zero.
Typical raw data for a single bubble event:
6530759, 1, 0, 525, 460, 641, 2236042908, 398, 2236125676, 463, 2236123268, 580, 2236208916,648
A B C D E F G H I J K L M N
A Time of bubble event (in milliseconds since Arduino power was connected)
B Number of bubbles in this event (0 to 4)
C Error code (see Table I for error code descriptions)
D,E,F Background ADC values for detectors 1, 2, and 3, respectively
G, I Time detector 2 sees beginning and end of bubble, respectively (microseconds)
H, J Detector 2 ADC value at start and end of bubble, respectively
K, M Time detector 3 sees beginning and end of bubble, respectively (microseconds)
L, N Detector 3 ADC value at start and end of bubble, respectively
Figure 9: Example of raw bubble sizer data and description of each entry in the data
stream.
Description of data types removed during post-processing
There are four categories of bubble data that must be rejected during post-processing.
Table 2 presents each category, includes references to explain potential causes for each data type,
and provides an estimate of data category prevalence in field data. The first two categories are
composed of incomplete bubble event data where some or all of the data points associated with
detectors 2 and 3 are zero. The third category of data to be rejected includes large bubbles
outside of the calibration range described in the Assessment.
76
The fourth data type to be rejected includes data where the four volume estimates
acquired for each bubble event do not follow the pattern observed in a controlled laboratory
setting, and therefore bubble volume estimates are not reliable. As shown in Figure 8, laboratory
work demonstrates that the two lower bubble volume estimates should be similar, and we can
quantify this similarity using the coefficient of variation (the ratio of the standard deviation to the
mean) We have calculated the coefficient of variation for a range of bubble sizes measured in a
controlled laboratory setting, yielding a clear threshold for accepting or rejecting field bubble
data (Figure 10). Coefficients of variation increase as bubble volumes decrease, and bubbles
smaller than the internal diameter of the funnel stem have much higher coefficients, likely
because these bubbles are subject to lateral movement within the flow path (see Assessment for
details). Bubble volume estimates with coefficients of variation outside of the acceptable range
have unknown accuracy, and are therefore discarded during post-processing. While this
approach will result in under-counting cumulative volume measurements, it will improve
accuracy of individual measured bubble sizes. Possible bias to the total bubble size distribution
from under-counting bubbles is discussed below in the Assessment.
77
0.8
CO
10- *CO
E
0.6
E
75 0.4
a) S
. 0.2a) 1
0 0.1 0.2 0.3 0.4 0.5
(0.58) (0.73) (0.83) (0.91) (0.98)
Bubble Volume (mL)
(Equivalent Diameter (cm))
Figure 10: CV between the two lower volume estimates plotted against bubble volume for a
series of bubbles created in a laboratory tank. The shaded area delineates acceptable CV
values based on bubble volume, as used by the data processing algorithm.
78
Table 3: Description of different types of data removed during post-processing.
Data Type Description Potential Causes Frequency of
occurrencein
field data
Type 1 - Non- Detector 1 sees something, but See Potential 64% of summer
bubble events detectors 2 and 3 see nothing. Mechanism #5 in 2016 field data
Assessment.
Type 2 - Bubble Detector 2 sees a different number See Potential 10% of summer
number mismatch of events than detector 3. Mechanisms #2, 3, 5 2016 field data
and 6 in Assessment
Type 3 - Uncalibrated volume measurement See Potential 0.26% of
Large bubbles falls outside calibration range Mechanism #8 in summer 2016
outside of described in Materials and Assessment. field data after
calibration range Procedures. removing Types
1, 2, and 4 data
Type 4 - High CV between the two lower volume See Potential 8.5% of summer
coefficient of estimates* falls above the Mechanism #6 in 2016 field data
variation (CV) acceptable range shown in Figure Assessment. after removing
between volume 10. Type 1 and
estimates Type 2 data.
*As described in Delwiche et al. 2015, each bubble event produces 4 volume estimates. See
Figure 8 for typical relationship between volume estimates.
79
ASSESSMENT
Sensor field performance
Sensor ease of installation and maintenance, and afield data example
Sensor deployment at Upper Mystic Lake in 2015 and 2016 showed that the deployment
rope structure shown in Figure 1 was straightforward to install in low wind conditions.
Replacing SD cards, changing batteries, and sampling the gas was all done from the custom
service buoy with minimal effort. Battery life with one battery pack powering the bubble sizing
sensor and one battery pack powering the data buoy circuit board was approximately 2 weeks,
and this could be doubled by adding an additional battery for the bubble size sensor. Figure 11
shows an example of the data acquired from one of the bubble size sensors. These data were
taken in 2016 after sensor hardware and electronic upgrades described in Materials and
Procedures were implemented. The figure demonstrates the sensor's unique ability to measure
bubble volume and event timing over long deployment periods. Bubble volume data can also be
used to calculate the Sauter Mean Diameter, the diameter of a sphere having the same volume-to-
surface area ratio as the complete bubble size distribution (Orsat et al. 1993). The Sauter Mean
Diameter can be used in discrete bubble dissolution models to simulate dissolution of the
complete bubble size distribution (McGinnis et al. 2002).
80
15 - 15
a * . b
* Mean diameter = 4.6 mm
*5 * . 0 . Sauter mean diameter =
* 6.1mm
-10- . 10
4P *0 ** * Sc g rl JL
E . . 0 . .
0 0
so 0
eu ep 0 2040698 10 12
Aug Sep* 0** 68101
Bubble event date Frequency (%)
Figure 11: Time series of bubble events from 12 July 2016 to 15 September 2016 and
respective bubble diameters, as measured by one bubble size sensor. Each dot represents
one bubble.
Assessing sensor performance using collected gas volumes
In order to assess the sensor's ability to accurately measure total bubble volumes, we
downloaded bubble data and collected the bulk gas samples every 1 to 2 weeks from mid-July to
mid-November 2016 and compared measured versus sampled gas volumes. On average, our
sensor measured 86% of the total gas collected in the trap (Figure 12). While a small portion of
the data points are significantly below the 1:1 line, the majority of points show only a small
under-measurement of collected gas. Some under-measurement is expected, and potential
reasons for this under-measurement will be discussed later in the Assessment section. While the
sensor appears to miss a portion of the ebullition flux (particularly during a smaller number of
sampling windows), the relative consistency between measured and collected volumes and the
81
fact that measured volumes do not exceed collected volumes suggests that collected bubble size
distributions are reasonable. We also note that due to dissolution inside the gas collection
chamber, collected gas volumes are likely lower than total gas volumes. In situ incubation trials
suggest that total gas volumes may be up to 5-10% higher than the reported collected gas
volumes for this reason alone.
180
160
7140al
W120E '4
00OD
-a 80
20
0
0 20 40 60 80 100 120 140 160 180
Collected volume (ml)
Figure 12: Cumulative measured gas volumes as a percentage of the corresponding
collected gas volume. The 1:1 line is drawn for comparison.
Correlation between low measured/collected ratio and Types 2, 3, and 4 data
While many of the data points shown in Figure 12 show relatively close agreement
between measured and collected gas volumes, several points fall far from the 1: 1 line. Time
periods with large discrepancies are likely caused by a combination of factors. The occurrence
of Types 2, 3, and 4 data are all correlated with a decreased measured to collected ratio (refer to
Table 2 for data types). Discarding large bubbles outside of the calibration range therefore has a
82
negative impact on the sensor's ability to measure total cumulative volume. However, most data
sets only had two or fewer discarded large bubbles and the highest number of discarded bubbles
was six, indicating that the extent of this effect is likely limited. While we do not know what
causes higher occurrences of Types 2 and 4 data, one possible explanation would be the
temporary presence of a large particle or zooplankter in the flow path obstructing smooth bubble
flow.
To understand if bubble under-measurement biases the bubble size distribution, we
looked at the correlation between the bubble size and the ratio of the cumulative measured gas
volume to the collected gas volume for each individual sensor. While excluding large bubbles
will undoubtedly affect the bubble size distribution, we expect this effect to be minimal because
only 0.26% of the data are from these large bubbles (after removing Types 1, 2, and 4 data). By
contrast, Types 2 and 4 data represent 10% and 8.5% of the data set, respectively, and it is
therefore important to understand if removing these two types of data bias the overall
distribution. We found no correlation for 7 of the 8 sensors, and for the 8th sensor the
correlation effect was a 0.03 mm decrease in bubble size per 1 percentage point drop in the
measured to collected ratio. This correlation could serve as grounds for rejecting bubble size
distribution data from this particular sensor during periods of low agreement between measured
and collected cumulative gas volumes.
Potential mechanisms for missing or rejecting bubble data
To understand the potential causes behind the bubble volume under-measurement
reported above, we assess here several physical and electronic effects that could lead the sensor
to miss a bubble completely, record insufficient data for bubble volume calculations, or record
83
data that are discarded in post-processing. Where possible we assessed the likely importance of
each effect by analyzing field data or by examining oscilloscope or raw field data traces. Some
of the outstanding issues that could lead to missing bubble data might be mitigated by additional
sensor modifications beyond the scope of this work, while others are a product of the
heterogeneous nature of aquatic ecosystems and would be difficult to mitigate.
Potential mechanism #1: Bubbles not accelerating smoothly through tube, leading to Type 4 data
Issue Summary: Accurate bubble volume estimates depend on a smooth bubble acceleration
through the glass funnel stem, so any obstruction to smooth bubble flow will result in inaccurate
bubble volume estimates.
Implication: Bubble event data sets may be complete, but the coefficient of variation between the
two lower volume estimates will be outside of the acceptable range described in Materials and
Procedures, and data will be rejected in post-processing.
Relevance: Likely significant because 8.6% of summer 2016 field data were discarded due to
high coefficients of variation, and prevalence of high CV data correlates with lower
measured/collected ratios. However, we cannot separate the contribution of inconsistent bubble
flow to high CVs from other potential sources of high CVs, and future work could include
adding a camera to distinguish between physical flow restrictions and sensor measurement
artifacts.
Detailed discussion: Accurate bubble volume estimates depend on smooth bubble acceleration
through the glass funnel stem because the sensor indirectly measures the velocity of the leading
and trailing edge of each bubble. To accurately estimate volume, the true average bubble
velocity must be a predictable function of the average velocity of the leading and trailing bubble
84
edge. In a controlled laboratory setting, bubbles are observed to move smoothly up the funnel
stem. However, in a field setting bubble flow through the funnel stem could be temporarily
restricted by the presence of particulate matter or zooplankton.
Bubbles that do not accelerate smoothly will lead to data with high coefficients of
variation between lower volume estimates (described in Materials and Procedures). If the
coefficient of variation is higher than the acceptable threshold for that particular bubble volume
(as shown in Figure 10), the bubble event data will be rejected during post-processing. These
rejected data points are a significant source of missing bubble data because 8.6% of field data
were discarded due to high coefficients of variation. Additionally, we found a positive
correlation between increased frequency of high CV data and a decreased measured to collected
volume ratio.
It is important to note that circumstances other than unsmooth bubble flow can contribute
to high coefficients of variation among volume estimates. High coefficients of variation could
also be caused by the sensor not properly recognizing the leading and/or trailing edge of each
bubble. This could occur if the signal fluctuations discussed in Materials and Procedures cause
one detector to prematurely record the trailing edge of a bubble, resulting in higher variability
between volume estimates. Different volume estimates would also occur if detector 2 only
records part of a passing bubble because of the 'dead-time' issue discussed under Potential
Mechanism #3. Additionally, high CVs could also occur if detectors 2 and 3 record a different
number of bubbles in a stream, which causes bubble data between the two detectors to be mis-
paired. At present there is no way to determine the cause behind individual high CV data points,
but redesigning the sensor to include a small camera may help distinguish between physical flow
restrictions and sensor measurement artifacts. Camera data could also provide a means of
85
corroborating sensor volume estimates, thereby improving sensor accuracy and reducing the
number of data points rejected in post-processing. We note that we do not think sensor
measurement frequency is an important contributor to high coefficients of variation because the
measurement frequency is approximately an order of magnitude higher than bubble rise velocity
within the funnel stem.
Potential mechanism #2: Background readings too noisy for bubble detection
Issue Summary: If background sensor readings are noisy the detectors will be unable to detect
incoming bubbles.
Implication: If detector 1 is put offline due to noisy background readings, the sensor will be
unable to record any new bubble events until readings stabilize. If detectors 2 and/or 3 are
offline, the sensor will only be able to record incomplete bubble event data.
Relevance: Likely insignificant because field data showed that detectors 2 and 3 were only
offline due to background noise 0.05% of the time, though the instances of all three detectors
being offline is likely somewhat higher.
Detailed discussion: As is typical of environmental sensors, the bubble size sensor relies on
relatively stable background readings to record accurate data (a description of the bubble
detection algorithm can be found in Materials and Procedures). If background reading variability
exceeds a predefined threshold (two times the difference between the maximum and minimum
background values) the sensor will be unable to detect bubbles until subsequent background
readings establish a less noisy baseline. If detector I is offline due to noise, no field data will be
gathered even if detectors 2 and 3 are not offline. We cannot assess the frequency of this
scenario because it does not result in any recorded data. As in-lab measurements invariably show
86
stable backgrounds, it seems likely that noise in the background during field deployment arises
from particulate material in the water, although we have not confirmed this. However, a subset
of the field data from 2016 includes detection thresholds for detectors 2 and 3, allowing us to
determine that 0.05% of unprocessed field data are gathered during times when either detectors 2
or 3 are offline but detector 1 is online. While the occurrence of all three detectors being offline
due to noisy background is likely somewhat higher than 0.05%, we still assume that noisy
background levels are likely not a large source of missing bubble data. If the sensor were
deployed in a more heterogeneous environment the noise requirements could be made less
stringent to reduce the number of times the sensor is offline to bubble detection.
Potential mechanism #3: Sensor 'dead-time' during data writing
Issue Summary: The sensor requires a finite amount of time to record bubble event data, during
which time the sensor will be unable to detect the arrival of new bubbles.
Implication: Bubbles rising through the flow path during sensor 6dead-time' (while the sensor is
recording data from previous bubble events) will be missed entirely. If the sensor comes back
online while the bubble is already partially in front of the detectors, incomplete bubble event data
will be recorded.
Relevance: Likely unimportant in Upper Mystic Lake because dead-time comprises only
approximately 0.2% of total time during a high-flux bubble events.
Detailed discussion: The Arduino Pro-Mini microcontroller used in this sensor requires a
minimum data storage time of approximately 7ms + 8ms x (# of bubbles per event). For
87
example, if three bubbles pass through the sensor during a bubble event, the sensor will be
offline to new bubbles for 31 ms during data storage, plus a few additional milliseconds for code
completion. If another bubble passes the first detector during this 'dead-time' window it will be
completely missed by the sensor, as shown in Figure 13. This 'dead-time' limitation means that
during high bubble flux events the sensor will under-count total cumulative bubble volumes.
To assess the significance of this dead-time, we used field data to estimate that only 0.2%
of time during high-flux bubble periods is comprised of 'dead-time'. This indicates that data loss
to dead-time is likely minor. If reducing dead-time were still a concern (for example if the
sensor were to be used in a much higher bubble-flux environment), the sensor could be
redesigned to use a faster microcontroller.
Sensor is busy storing 2 previous
bubbles and misses this bubble event.
C
0
C-)
a)
a)
a)
0 0 M
Figure 13: Dead-time demonstration for closely spaced bubbles.
Potential mechanism #4. Bubbles coalescing or breaking apart within flow path
Issue Summary: Proper sensor functioning relies on individual bubbles passing each detector
intact. If bubbles coalesce or break apart within the flow path the sensor will be unable to record
sufficient data for volume estimation.
88
-1
Implication: This could lead to a mis-match in the number of bubbles seen by detectors 1, 2, and
3, resulting in incomplete bubble event data sets.
Relevance: Appears minimally important within the glass funnel stem, but bubble coalescence
could be occurring at funnel stem constriction which would impact the measured bubble size
distribution. More work would be needed to assess the importance of this effect.
Detailed Discussion: Two bubbles entering the glass funnel stem in close succession and then
merging in to one bubble as they pass the detectors would make it impossible for the sensor to
estimate their volume using the existing algorithm. Fortunately, our laboratory oscilloscope
work provided no evidence that bubble coalescence within the flow path was a significant issue.
Smaller bubbles separated by even a thin lens of water overwhelmingly remained separated as
they rose past detectors two and three. However, a rapid burst of bubbles emitted underneath the
glass funnel did tend to coalesce into a larger bubble at the funnel constriction point. One way to
reduce this coalescence would be to decrease the total funnel area and thus decrease the
possibility of numerous bubbles approaching the funnel at once. However, the merit of this
approach must be weighed against the resulting decrease in the amount of data gathered during a
given field campaign.
Similar to bubble coalescence, a large bubble breaking into smaller bubbles inside the glass
funnel stem would lead to incomplete bubble event data. Oscilloscope data demonstrated that
large bubbles passing detector 1 remained intact past detectors 2 and 3. Very large bubbles (over
2 mL or 1.6 cm equivalent diameter) often shed smaller bubbles (typically below 0.05 mL or
0.46 cm equivalent diameter), but this cleavage occurs prior to the bubble passing detector 1 and
therefore should not affect cumulative volume measurements.
89
Potential mechanism #5. False triggering of sensor by particles or organisms
Issue Summary: Zooplankton or other opaque particles passing detector 1 may trigger a bubble
detection sequence, but detectors 2 and 3 will not record any data (a "false-start"). If 3 false-
starts are recorded within 10 seconds the microcontroller resets the background values, during
which time the sensor will be unable to detect incoming bubbles.
Implication: This results in incomplete bubble event data, as well as potential sensor dead-time if
background levels are reset after 3 consecutive false-starts.
Relevance: 64% of unprocessed field data consist of events where detector 1 sees something but
detectors 2 and 3 see nothing. The frequent occurrence of these data points increase the required
data storage capacity, but do not impact cumulative volume measurements because sensors are
not offline to bubble detection during these events. While 3 consecutive false-starts that trigger
background resets do occur, they only put the sensors offline for approximately 0.03% of the
sampling time.
Detailed Discussion: One type of data inconsistency frequently observed in field data is when
detector 1 sees a bubble and initiates a data collection sequence, but detectors 2 and 3 fail to see
a bubble. This false triggering could be due to opaque particles such as organic matter or
zooplankton obstructing light transmittance at detector 1. Particles or zooplankton settling
downwards through the glass tube will trigger a data collection sequence, but if the particle is
moving downwards detectors 2 and 3 will not record a bubble start and stop. Likewise, if
zooplankton travel slowly upwards past detector 1 but fail to reach detectors 2 and 3 before the
data collection sequence times out (5 seconds), the bubble event data will be missing values.
These "false-start" data points are quite common in field data, accounting for 64% of all raw data
recorded during the second half of the 2016 sampling season. However, these data points can
90
easily be removed during post-processing and the SD card data storage system in the service
buoy means that even large numbers of false-start data points will not exceed the sensor data
storage capacity.
False-start data will only affect cumulative volume measurements if they trigger a
background reset within the sensor. This reset happens if 3 consecutive false-starts occur, and
the sensor is offline to new bubble detection during the approximately 8 seconds it takes to re-
establish the background detection thresholds. Data gathered in the first half of 2016 included
enough information to calculate that background reset put sensors offline only 0.03% of the time.
While the sensors could miss passing bubbles during this time, it is unlikely to significantly
affect cumulative volume measurements.
While the sensor data stream does not provide sufficient information to determine the
cause of these false-start data events, one potential contributor to this type of data is the phantom
midge (the larvae of a small fly, Chaoborus). Phantom midges migrate vertically through the
water column at night to feed and down during the day to escape predation (Stratton 2011). We
routinely find phantom midges in our gas collection chambers, indicating that midges are likely
traveling through the sensor. In addition, false-start data more commonly occur during the
nighttime hours, consistent with the prime vertical travel time for midges. There is also a
positive correlation between the date and the occurrence of "false-start" data, which has
numerous possible explanations including zooplankton discovery of sensors over time, seasonal
increases in zooplankton abundance, or biofilm build-up on the glass funnel stem.
91
Potential mechanism #6: Small bubbles wobbling during rise through tube
Issue Summary: Bubbles smaller than 4 mm diameter are not elongated as they rise past the
detectors, and may therefore be subject to lateral migration within the flow path.
Implication: This lateral migration could lead to the detectors missing small bubbles entirely, or
to different volume estimates from detectors 2 and 3. Additionally, bubbles with average volume
below 0.002 mL (0.16 cm equivalent diameter) are outside the calibration range and are therefore
discarded as described under Potential Mechanism #7.
Relevance: Twenty-seven percent of bubbles from Upper Mystic Lake are smaller than 4 mm in
diameter. The possibility of detectors 2 and 3 recording different volume estimates has been
mitigated by the lenient coefficient of variation threshold for smaller bubbles shown in Figure
10. While some smaller bubbles could be missing the sensor entirely or generating incomplete
data, this is likely to be a small percentage of the total bubble volume.
Detailed discussion: Bubbles smaller than the glass funnel stem (4 mm) will not be elongated as
they rise past the detectors. Rather, they will remain spherical and be subject to lateral migration
within the tube during rise. This lateral migration could result in data from detectors 2 and 3
yielding slightly different bubble lengths, resulting in higher coefficients of variation between
volume estimates. Indeed, laboratory results show that small bubbles have substantially higher
coefficients of variation (Figure 10), and we mitigate this effect by using a higher coefficient of
variation threshold for small bubbles. This issue could be addressed directly by using a more
narrow glass funnel stem that would elongate a greater percentage of rising bubbles, but care
would be necessary to ensure the funnel stem was still wide enough to prevent bubbles from
becoming stuck. Small bubbles may also be missed entirely by one or more detectors, as shown
92
in Figure 14. Field data show that 27% of processed bubbles in Upper Mystic Lake are below
the 4 mm threshold.
Signal size varies Detector 2 cannot see bubbles
0
0
0
Detector 3 cannot see bubble
Figure 14: Small bubble traces showing significant variability in signal between detectors 2
and 3, likely because bubbles are smaller than flow path diameter and their position
relative to the detectors is not constant (bubbles are approximately 3-4 mm diameter, tube
diameter is 4 mm). In some cases,. bubbles are not seen at all.
Potential mechanism #7: Tiny bubbles (smaller than 0.002 mL or 0.15 cm equivalent diameter)
are outside of calibrated sensor range
Issue Summary: Due to the positive intercept on the small bubble calibration curve presented in
Figure 8, bubbles below a certain lower threshold do not lie in the physical Iy-meaningful portion
of the best-fit calibration curve.
Implication: Data points for tiny bubbles are not reliable, so data must either be rejected or
included with only approximate volume estimates.
Relevance: In Mystic Lake, tiny bubbles represent 3 % of total unprocessed data (10% of data
after removing Type I non-bubble event data). Although this represents a substantial occurrence
93
of tiny bubbles in terms of bubble event count, the impact on cumulative measured volumes is
negligible since these bubbles are below 0.002 mL.
Detailed discussion: We have not been able to calibrate the sensor for tiny bubbles due to the
difficulty in forming accurate tiny bubbles in a laboratory tank. The existing calibration curve
presented in Figure 8 was generated for bubbles as small as 0.01 mL (0.27 cm equivalent
diameter). Additional calibration attempts for bubbles smaller than 0.01 mL showed that sensor
performance was still roughly linear from 0.002 mL to 0.01 mL with a similar calibration
equation, but volume estimates lacked precision. This variability could be due to sensor
measurement error or variability in actual bubble size. Since 0.002 mL was the smallest gas
volume we could dispense, we were unable to verify the sensor's lower detection limit or the
shape of the calibration curve below 0.002 mL. However, even though 3% of total unprocessed
field data consist of tiny bubbles, the impact on cumulative volume estimates by discarding these
data will be negligible.
It is also possible that modification of the bubble flow path could optimize measurements
for tiny bubbles. The present geometry provides both robustness and simplicity, but a
noncircular bubble channel geometry and additional optical elements, such as a cylindrical lens
to create a light source more nearly approximating a line source, could be worth exploring.
Potential mechanism #8: Large bubble volumes
Issue Summary: Bubbles larger than 1.8 mL (1.51 cm equivalent diameter) cannot accurately be
measured by the sensor, so any data exceeding the 1.8 mL threshold must be discarded.
Implication: Discarding data from larger bubbles may disproportionately affect the measured to
collected ratios.
94
Relevance: Only 0.26% of field data were from bubbles over 1.8 mL (after removing Types 1, 2,
and 4 data as described in Table 2). If we add 0.26% more bubbles to the data set shown in
Figure 11 of volumes 2mL, 3mL, or 4mL (1.56 cm, 1.79 cm or 1.97 cm equivalent diameters),
the Sauter Mean Diameter increases by 4%, 7%, or 10%.
Detailed Discussion: The calibration curve presented in Materials and Procedures shows that
the sensor can accurately measure bubbles up to 1.8 mL (1.51 cm diameter). Beyond this
volume the sensor measurement accuracy breaks down. In practice, this means we reject any
field data for bubbles larger than 1.8 mL and in 2016 this comprised only 0.26% of field data
(after removing Types 1, 2, and 4 data as described in Table 2). If we artificially add 0.26%
more bubbles to the size distribution presented in Figure 11 and recalculate the Sauter Mean
Diameter, we find bubble volumes of 2mL, 3mL, or 4mL (1.56 cm, 1.79 cm or 1.97 cm
equivalent diameters) raise the Sauter Mean Diameter by 4%, 7%, or 10%. Since the Sauter
Mean Diameter has been shown to be useful for estimating bubble dissolution rates for an entire
bubble size distribution, the potential increase in Sauter Mean Diameter compared to measured
data could be taken into consideration in dissolution calculations. For sites such as the one
described in DelSontro 2015 where bubbles over 2 mL (1.56 cm equivalent diameter) are found
to contribute significantly to total ebullition flux, the sensor could be altered to contain a wider
glass funnel stem which may improve large volume estimates. However, a larger funnel stem
means fewer rising bubbles would be elongated, potentially reducing sensor accuracy as
described under Potential Mechanism #6.
95
DISCUSSION
The long-term data from the bubble size sensor, with the upgrades described here, give
the unique ability to study temporal trends in ebullition flux and bubble size over long time
intervals with minimal sensor servicing and sediment disturbance. These trends are otherwise
impossible to detect in short-duration campaigns. For example, the field data show that on
average 70% of the total cumulative flux occurred on only 32% of the sampling days, indicating
that bubble events tend to be temporally clustered. Temporal clustering means that limited
duration sampling campaigns could artificially bias data used to estimate long-term ebullition
flux. Recent work to quantify this artificial bias has found that short sampling campaigns are
likely to underestimate ebullition (Wik et al. 2016b).
While some existing ebullition measurement techniques also have the ability to estimate
daily heterogeneity in flux (Varadharajan and Hemond 2012, Maeck et al. 2014), our sensor has
the unique ability to measure variations in bubble flux on the time scale of seconds, as well as
individual bubble volumes. These data could be important to estimating bubble dissolution rates
because previous studies have found that bubbles emitted less than 3 seconds apart have
significantly faster rise times, though the increase in rise velocity is variable and dependent on
bubble size (Garner and Hammerton 1954). Current single-bubble dissolution models do not
account for potential inter-bubble interactions (McGinnis et al. 2006, Leifer and Patro 2002), and
prior to the availability of data from this sensor there was no way to measure the distribution of
temporal spacing between bubble events.
This sensor will also allow us to compare size distributions from different locations
within a water body. While bubble-sizing sonar devices can be moved to multiple locations
within a water body, the relatively short time duration of these sampling campaigns means that
96
observed differences in bubble size distributions could be an artifact of limited sampling
windows. Our sensor is cost-effective enough for multiple sensors to be deployed for long
periods of time, thus discerning the spatial variability in bubble size distributions.
In addition to studying spatial heterogeneity in bubble size, the ability to measure bubble
sizes for long time periods also allows us for the first time to track changes in bubble size with
time and ebullition flux. Previous studies have found seasonal variability in ebullition flux
(DelSontro et al. 2010, Maeck et al. 2013). However, since traditional bubble sizing techniques
are limited to short campaigns we do not currently know if these changes in ebullition flux are
correlated with a change in average bubble size. Since a change in mean bubble size would lead
to a change in bubble dissolution rate, it is important to understand if bubble size distributions
are constant or variable throughout the year.
COMMENTS AND RECOMMENDATIONS
We have presented critical upgrades to our first generation bubble sizing sensor. These
upgrades allow for long-term uninterrupted bubble size and ebullition flux measurements. If
sensors were to be deployed in a remote field location where bi-monthly or monthly field
campaigns to replace batteries were infeasible, the sensor could be modified to run on solar
power. The data buoy could also be enhanced with Bluetooth capability, as was done in the
original buoy design described in Gardner et al. 2009. Additionally, the sensor could be
equipped with a small camera to image bubbles rising through the funnel stem. Camera images
could be used to further investigate potential causes of missing or rejected bubble data.
97
ACKNOWLEDGMENTS
This material is based on work supported by the National Science Foundation under grant
number EAR-1045193 and Graduate Research Fellowship Program grant number DGE-
0707428, the MIT Martin Family Fellowship to K. Delwiche, the W.E. Leonhard 1941
professorship to H. Hemond, and the Singapore-MIT Alliance for Research and Technology
program. Authors thank many people in the MIT Parsons laboratory for their help with field
work, particularly Benjamin Scandella and Irene Hu.
REFERENCES
Bastviken, D., L. J. Tranvik, J. A. Downing, P. M. Crill, and A. Enrich-prast. 2011. Freshwater
methane emissions offset the the continental carbon sink. Science. 331: 6. doi:
10.1 126/science. 1196808
Casper, P., S. C. Maberly, G. H. Hall, and B. J. Finlay. 2000. Fluxes of methane and carbon
dioxide from a small productive lake to the atmosphere. Biogeochemistry 49: 1-19
Chanton, J. P., C. S. Martens. 1988. Seasonal variatiosn in ebullitive flux and carbon isotopic
composition of methane in a tidal freshwater estuary. Global Biogeochem. Cy. 2:3: 289-
298
Deemer, B., J. Harrison, S. Li, J. Beaulieu, and others. Greenhouse gas emissions from reservoir
water surfaces: A new global synthesis. BioScience 66 (11): 949-964.
DelSontro, T., D. F. McGinnis, B. Wehrli, and I. Ostrovsky. 2014. Size does matter: Importance
of large bubbles and small-scale hot spots for methane transport. Environ. Sci. Technol., 49
(3): 1268-1276. doi:10.1021/es5054286
DelSontro, T., Mcginnis, D., Sobek, S., Ostrovky, I.,Wehrli, B. 2010. Extreme methane
emissions from a Swiss hydropower reservoir: Contribution from bubbling sediments.
Environ. Sci. Technol., 44: 2419-2425. doi: 10.1021/es9031369
Delwiche, K., S. Senft-Grupp., H. Hemond. 2015. A novel optical sensor designed to measure
methane bubble sizes in situ. Limnol. Oceanogr.: Methods, 13: 712-721
Forster, P., V. Ramaswamy, P. Artaxo, and others. 2007. Changes in atmospheric constituents
and in radiative forcing, p. 131-217. In M.T. and H.L.M. Solomon, S., D. Qin, M. Manning,
98
Z. Chen, M. Marquis, K.B. Averyt [ed.], Climate Change 2007: The Physical Science Basis.
Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental
Panel on Climate Change. Cambridge University Press
Garner, F., D. Hammerton. 1954. Circulation inside gas bubbles. Chem. Eng. Sci. 3 (1)
Gardner, A.T., H. Karam, A. Mulligan, C. Harvey, T. Hammar, H. Hemond. 2009. A differential
pressure instrument with wireless telemetry for in-situ measurement of fluid flow across
sediment-water boundaries. Sensors 9: 404-429. doi: 10.3390/s90100404
Greinert, J., B. Nltzel. 2004. Hydroacoustic experiments to establish a method for the
determination of methane bubble fluxes at cold seeps. Geo-Mar. Lett. 24: 75-85. doi:
10.1007/s00367-003-0165-7
Joyce, J. P. Jewell. 2003. Physical controls on methane ebullition from reservoirs and lakes.
Environ. Eng. Geosci. Vol IX, No. 2: 167-178
Kankaala, P., J. Huotari, E. Peltomaa, T. Saloranta, and A. Ojala. 2006. Methanotrophic activity
in relation to methane efflux and total heterotrophic bacterial production in a stratified ,
humic , boreal lake. Limnol. Oceanogr. 51: 1195-1204
Kirschke, S., P. Bousquet, P. Ciais, and others. 2013. Three decades of global methane sources
and sinks. Nat. Geosci. 6: 813-823. doi: 10.1038/NGEO 1955
Leifer, I., and R. K. Patro. 2002. The bubble mechanism for methane transport from the shallow
sea bed to the surface: A review and sensitivity study. Cont. Shelf Res. 22: 2409-2428
Maeck, A., DelSontro, T., McGinnis, D., Fischer, H., Flury, S., Schmidt, M., Fietzek, P., Lorke,
A. 2013. Sediment trapping by dams creates methane emission hot spots. Environ. Sci.
Technol. 47: 8130-8137. doi: 10.1021/es4003907
Maeck, A., H. Hofmann, A. Lorke. 2014. Pumping methane out of aquatic sediments - ebullition
forcing mechanisms in an impounded river. Biogeosciences. 11: 2925-2938. doi:
10.5194/bg-1 1-2925-2014
McGinnis, D. F., J. Little. 2002. Predicting diffused-bubble oxygen transfer rate using the
discrete-bubble model.Water Research. 36: 4627-4635.
McGinnis, D. F., J. Greinert, Y. Artemov, S. E. Beaubien, and A. Wtiest. 2006. Fate of rising
methane bubbles in stratified waters: How much methane reaches the atmosphere? J.
Geophys. Res. 111: C09007. doi: 10.1029/2005JC003183
Orsat V., C. Vigneault, G. Raghavan. 1993. Air diffusers characterization using a digitized image
analysis system. Appl. Eng. Agric. 9:115-21.
99
Ostrovsky, I., D. F. McGinnis, L. Lapidus, and W. Eckert. 2008. Quantifying gas ebullition with
echosounder: the role of methane transpot by bubbles in a medium-sized lake. Limnol.
Oceanogr.: Methods 6: 105-118
ROmer, M., H. Sahling, T. Pape, G. Bohrmann, V. SpieB. 2012. Quantification of gas bubble
emissions from submarine hydrocarbon seeps at the Makran continental margin (offshore
Pakistan). J. Geophys. Res. 117: C10015, doi:10.1029/2011JC007424
Sauter, E., S. I. Muyakshin, J. Charlou, and others. 2006. Methane discharge from a deep-sea
submarine mud volcano into the upper water column by gas hydrate-coated methane
bubbles. Earth Planet. Sci. Lett. 243: 354-365. doi: 10.1016/j.epsl.2006.01.041
Scandella, B. P., L. Pillsbury, T. Weber, C. Ruppel. H. Hemond, R. Juanes. 2016. Ephemerality
of discrete methane vents in lake sediments. Geophys. Res. Lett. 43: 4374-4381. doi:
10.1002/2016GL068668
Stratton, M., D, Kesler. 2011. The role of light and oxygen in Chaoborus punctipennis (Insecta:
Diptera) diel vertical migration. J. Freshwater Ecol. 22:1: 101-106
Varadharajan, C., and H. F. Hemond. 2012. Time-series analysis of high-resolution ebullition
fluxes from a stratified, freshwater lake. J. Geophys. Res. 117: G02004. doi:
10.1029/2011JG001866
Walter, K. M., S. A. Zimov, J. P. Chanton, D. Verbyla, and F. S. Chapin. 2006. Methane
bubbling from Siberian thaw lakes as a positive feedback to climate warming. Nature 443:
71-5. doi: 10.1038/nature05040
Walter, K.M., L. C. Smith, F. S. Chapin III. 2007. Methane bubbling from nothern lakes: present
and future contributions to the global methane budget. Philos. Trans. R. Soc., A. 365: 1657-
1676. doi: 10.1098/rsta.2007.2036
Wang, B., S. Socolofsky. 2015. A deep-sea, high-speed, stereoscopic imaging system for in situ
measurement of natural seep bubble and droplet characteristics. Deep Sea Res. Part 1. 104:
134-148. doi: 10.1016/j.dsr.2015.08.001
Wik, M. R. K. Vamer, K. W. Anthony, S. Maclntyre, D. Bastviken. 2016. Climate-sensitive
northern lakes and ponds are critical components of methane release. Nat. Geosci., 9: 99-
106. doi: 10.1038/NGE02578
Wik, M., B. F. Thornton, D. Bastviken, J. Uhlback, P. M. Crill. 2016. Biased sampling of
methane release from northern lakes: A problem for extrapolation. Geophys. Res. Lett. 43:
1256-1262 doi: 10.1002/2015GL066501
100
An enhanced bubble size sensor for long-term ebullition studies
Kyle Delwiche, Harold Hemond
SUPPLEMENTAL INFORMATION
SENSOR DEPLOYMENT INSTRUCTIONS
1. Measure rope, cable, and tubing lengths according to the water depth and desired
sensor deployment depth. Excess data cable length can be coiled and secured with zip
ties.
2. Tie a stainless steel hook on the rope where the sensor will be attached.
3. Tie the rope, PCV tubing, and data cable structure together every 1-2 meters using zip
ties. To ensure the rope bears the cable weight during deployment, include a small
amount of slack in the cable and tubing between each zip-tie.
4. To counteract the weight of the data cable, attach multiple small floats to the data
cable.
5. Once assembled, the cable, tubing, and rope structure can be coiled on to a large
wooden spool to facilitate field deployment.
6. In the field, lower the anchor farthest from the bubble sensor while uncoiling one
diagonal leg of the rope structure. Once the anchor is on the bottom, move the boat in
the desired deployment direction while uncoiling the other diagonal leg with the
attached data cable and PVC tubing.
101
7. At the pre-determined sensor deployment location, stop uncoiling the rope. Attach
the detachable expansion funnel and gas collection chamber to the properly
assembled and sealed sensor housing. Connect the data cable and PVC tubing to the
sensor, clip the sensor to the hook on the rope, and put the sensor in the water.
8. Uncoil the remainder of the rope, and lower the second anchor. If the underwater
center buoy is not submerged, move the anchors farther apart.
9. Attach the data buoy to the data cable, empty the PVC tubing of any gas, make sure
data buoy circuit board has an SD card and a desiccant pack, connect the batteries,
and record time of battery connection.
SERVICE BUOY MAINTENANCE INSTRUCTIONS
1. Pull buoy in to boat, open air-release valve at buoy base, and remove buoy cap.
2. Unplug batteries from printed circuit board, noting which battery powered which
Arduino microcontroller.
3. Remove SD card and replace with blank SD card (do not remove SD card while
batteries are still connected).
4. Plug fresh batteries in to printed circuit board. If using two batteries in parallel,
batteries should be of equal voltage rating, chemistry, and charge state.
5. Record time batteries were connected. Replace cap on buoy, shut air-release valve,
and return buoy to the water.
102
BUBBLE SIZE SENSOR CONSTRUCTION GUIDE
Sensor Body Materials
- I nominal 4 inch PVC coupling*
- 2 nominal 4 inch PVC pipe plugs
- 2 EPDM 0-rings, dash size 244
- 2 Buna-N 0-rings, 7 mm inner diameter, 2.5 mm wide
- 4 stainless steel eye bolts, 6-32 thread size and 1.9 cm (0.75 inch) length
- 2 stainless steel hex nuts, 6-32 thread size
- 11.5 cm long piece of 5 inch diameter acrylic pipe
- 2 small zip ties, 10.16 cm (4 inch) in length
- Silicone vacuum grease
- 3 cm thick disc of solid PVC, 8.9 cm (3.5 inch) diameter
- 1 nominal 1 inch male adapter to male NPT fitting (PVC)
- PVC cement
- Stainless steel hose clamp for pipe sizes 92-165 mm
- 3 stainless steel D-rings for 3.8 cm (1.5 inch) webbing width
- Three 1.5 m lengths of 0.635 cm (0.25 inch) diameter twisted nylon rope
- I four pin bulkhead fitting, MacArtney Underwater Technology part MCBH4F
- 1 7/16"-20 hex nut
- 1 7/16" washer
- 4 Molex Micro-Fit 3 0TM female sockets for 24 gauge wire (Digikey WM7082CT-ND)
- I Molex 4 pin housing connector receptacle (Digikey WM 1847-ND)
103
*All PVC parts are schedule 40
Sensor Body Construction Directions (Refer to Figures St, S2, and S3)
Repeat for sensor housing top and base:
28. Using a lathe, cut PVC pipe plug to 3.38 cm in length.
29. Cut O-ring groove into side of the pipe plug. Groove should be 0.282 cm deep and 0.475
cm wide, and positioned approximately 0.8 cm from end of pipe plug (Fig Sl-B). Make
sure that groove is smooth. It may help to reduce the lathe speed.
30. Drill a 0.87 cm (11/32 inch) diameter hole into the center of the pipe plug (Fig S 1-A).
31. Use a small, hook-shaped lathe tool to cut an O-ring groove into hole drilled in step 3.
Groove should be 0.33 cm wide and 0.154 cm deep. Refer to Figure S4 for example of
hook-shaped tool.
32. Insert small O-ring in to the small O-ring groove. Do not grease O-ring.
Instructions for sensor housing top only, Fig Si:
33. On the lathe, machine the flat top of top pipe plug to a smooth finish (Fig SI-C).
34. Drill a 3.33 cm diameter hole in the solid PVC disc. On the lathe, this can be achieved by
drilling a 3.175 cm (1.25) inch diameter hole and widening the hole to 3.33 cm diameter
using a boring bar.
35. Drill a 7/16" hole (0.4375 cm) in the top pipe plug 2.5 cm from one edge.
36. Using an end mill, cut a slight indentation centered on the 7/16" hole to the outer
diameter of the 7/16" washer. Indentation must be cut on the under-side of the pipe plug.
Washer needs to sit in this indentation, flush against the pipe plug (Fig SI -A).
104
37. Measure 2.5 cm from one edge of the solid PVC disc towards the center of the disc.
Draw a chord through this point, perpendicular to the diameter of the disc. Remove this
crescent-shaped segment of PVC material, either by cutting with a band saw or milling.
38. Using PVC cement, glue the nominal 1 inch male adaptor to male NPT in to the hole in
the PVC disc (Fig SI-C).
39. Using PVC cement, glue the PVC disc and adaptor onto the flat top of the pipe plug,
centered on the small O-ring hole (Fig Si -D) and positioned to maximize clearance
around the 7/16" hole.
40. Pass the bulkhead fitting through the 7/16" hole in the top of the pipe plug, and secure
with the washer and nut.
41. Crimp the female Molex sockets on to the ends of the bulkhead wires. Insert sockets into
the 4-prong male Molex clip. Hold Molex clip such that you are looking in to holes
where you will insert the sockets, and the plastic latch is facing upwards. From left to
right, insert black, red, green, and white wires.
42. Grease the large O-ring with silicone vacuum grease and insert into the large O-ring
groove.
Instructions for sensor base only, Fit S2:
43. Drill 2 small pilot holes horizontally into two opposite sides of the pipe plug top, being
careful not to puncture the pipe plug (for upper pair of eyebolts shown in Fig S2-B).
44. Screw two of the 6-32 stainless steel eyebolts into the pilot holes.
45. Use a 0.635 cm (0.25 inch) end mill to cut 3 notches into the 11.5 cm long piece of
acrylic. Notches should be equidistant to each other, and 4.42 cm (1.75 inches) deep.
Check that this piece fits onto funnel with the hex standoffs (Fig S2-B).
105
46. Drill and tap two 6-32 holes into the piece of acrylic, approximately 1 cm from the top
edge and equidistant from each other. Holes should be on the opposite end of the pipe
from the notches.
47. Screw the remaining 2 eyebolts horizontally into the side of the 10 cm long piece of
acrylic, using Teflon tape as necessary to keep eyebolts snug. Secure with the hex nuts.
48. Use the 2 small zip ties to securely fasten the acrylic piece to the bottom pipe plug. This
removable piece will protect the glass funnel during deployment (Fig S 1-B).
49. Grease the large O-ring with silicone vacuum grease and insert into the large O-ring
groove.
Instructions on Sensor Body Assembly (Fis! S3):
50. Slide the 3 D-rings on to the hose clamp.
51. Slip the hose clamp around the mid-point of the nominal 4 inch coupling, and begin to
tighten the clamp.
52. Slide the D-rings around until they are equidistant from each other. Tighten the hose
clamp until rings are firmly in place and hose clamp is very snug. To facilitate the
placement of the D-rings around the hose clamp, 2 small zip ties can be used to secure
each D-ring to the appropriate location on the hose-clamp.
53. Tie the 3 nylon ropes to the 3 D-rings
Gas Collection Chamber Materials
- 90 cm of nominal I inch clear PVC pipe
- 1 nominal 1 inch PVC coupling
- 1 nominal 1 inch socket female to NPT female PVC adapter
106
- I nominal 1 inch male to 0.25 inch NPT female hex bushing (PVC)
- One nominal 0.25 inch NPT male to 1/8 inch barbed nylon tube fitting
- Teflon tape
- 3 stainless steel eyebolts, 0.25 inch-20 thread size with eye diameter of 0.375 inches.
- PVC cement
Gas Collection Chamber Construction Directions (Refer to Figure S5)
54. Using PVC cement, glue the nominal 1 inch unthreaded female to NPT female adapter on
to one end of the 90 cm clear PVC pipe (Fig. S5-B).
55. Using PVC cement, glue the nominal 1 inch coupling to the other end of the clear PVC
pipe, and then glue the nominal 1 inch male to 0.25 inch NPT female hex bushing in to
the other end of the coupling (Fig. S5-A).
56. Wrap Teflon tape around the barbed tube fitting and screw it in to the nominal 0.25 inch
hex bushing.
57. Cut eyebolt threads to 0.63 cm in length and polish the cut ends on a grinder.
58. Carefully drill three shallow pilot holes horizontally into the pipe coupling, equidistant
from each other and near the top edge of the coupling (Fig S5-A). Be careful not to break
through the wall of the gas collection chamber. Holes should be approximately 0.7 cm
deep, or until drill bit first starts cutting in to hex bushing underneath coupling. Tap the
holes.
59. Screw the trimmed eyebolts into the pilot holes, being careful not to puncture the gas
collection chamber. The eyebolts will serve as guides for 3 nylon ropes connecting the
sensor housing to the deployment structure.
107
60. Drill four 0.637 cm (0.25 inch) holes through the base of the gas collection chamber.
Holes should penetrate the top of the female pipe adapter and the clear PVC tube. These
holes will allow water to exit the chamber as bubbles enter (Fig S5-B).
Detachable Funnel Materials
- 1 commercial funnel designed for 55 gallon barrel deer feeders (Texas Hunter, LFBF)
- 3 stainless steel female hex standoffs, /2" length and 6-32 thread size
- 3 stainless steel screws, /2" length and 6-32 thread size
- 3 stainless steel washers, #6 screw size
- 3 stainless steel '%"-20 eyebolts with a 1 inch shank
- 3 stainless steel %"-20 locknuts
- 3 stainless steel /4"-20 nuts
- 6 stainless steel washers for the '/4"-20 eyebolts
- 3 stainless steel split rings, approximately 1 inch diameter
- 3 12oz lead cannonball sinkers (Cabelas IK- 118098)
- 3 lengths of stainless steel wire rope, 1/16" diameter and 24 cm long
- 6 stainless steel rope compression sleeves for 1/16" diameter wire rope
- 3 stainless steel carabiners (McMaster 3716T51)
Detachable Funnel Construction Directions (Refer to Figure S6)
61. Drill a 0.635 cm (0.25 inch) hole 4 cm up from the base of each funnel ridge.
62. Thread the regular '/4"-20 nut on one eyebolt, add a washer, and pass the eyebolt through
the top of the hole. Add another washer, and secure with the /4"-20 lock nut. Make sure
to align eyebolt such that it is parallel with the funnel ridge. Repeat 3 times.
63. Drill a 0.35 cm (0.138 inch) hole 18.5 cm up from the base of each funnel ridge.
108
64. Place a washer on the 6-32 screw and pass the screw upwards through the hole. Screw
on the standoff.
65. Attach one 12 oz cannonball sinker to each eyebolt using the split rings.
66. Pass the wire rope through the eyebolt, and use the compression sleeve to form a small
loop around the eyebolt. Crimp sleeve with appropriate crimping tool.
67. Pass the remaining end of the wire rope through the carabiner and use the compression
sleeve to form another small loop. All three wire ropes should be the same length after
crimping. Crimp sleeve with appropriate crimping tool.
Sensor Electronics Materials
- Custom printed circuit board (PCB) for bubble size sensor (Eagle files available from author
upon request).
- Four 2M machine screws, 12 mm in length
- Four 2M nuts
- 2 PVC bars, minimum dimensions 0.9 cm by 1 cm by 3.1 cm
- Electronic parts described in Table I in Delwiche et al., 2015.
- 2 lengths of straight female headers, 12 pins long
- 1 length of straight female headers, 6 pins long
- 1 length of straight female headers, 2 pins long
- 2 lengths of straight male headers, 12 pins long
- 1 length of straight male headers, 6 pins long
- I length of straight male headers, 2 pins long
- I four position Molex Micro-Fit 3 .0"m connector (Molex product #43650-0413)
- 2 2kf2 surface mount resistors, size 0603
109
- 3 1 0kQ surface mount resistors, size 0603
- 3 12kM surface mount resistors, size 0603
- 1 eight pin (4 x 2) IC socket with 0.1 inch pitch and 0.3 inch row spacing
- 1 0.1 tF surface mount capacitor, size 0603
- 1 10p F surface mount capacitor, size 1210
Sensor Electronics Construction Directions
68. Using a mill, cut the PVC bars to the specifications shown in Figures S7 and S8. These
shims will properly align the LED and photocells on the PCB.
69. Attach PVC shims to PCB using the 2 mm machine screws and nuts. Shims are on
opposite side of board from surface mount components.
70. Solder all electronics and female headers for Arduino attachment to PCB. Female
Arduino headers are on same side of board as surface mount components.
71. Solder the two 2k resistors to the back of the Arduino near the A4 and A5 pins. Solder
straight male headers to Arduino Pro Mini with pins protruding from back side. The 2-
pin length of male headers should be soldered to analog pins A4 and A5.
72. Slide Arduino Pro Mini into female headers on PCB.
73. Slide EEPROM data chip into IC socket.
Complete Sensor Assembly Materials
- Two 0.635 cm (0.25 inch) U-bolts
- Four 0.635 cm (0.25 inch) nuts
- Two 1 cm lengths of 0.635 cm (0.25 inch) flexible plastic tubing, slit down the length of the
tube
- 1 6140 Pyrex glass funnel
110
- 10 cm length of 0.635 cm (0.25 inch) inner diameter plastic tubing
- 1 Silica Gel desiccant bag
Complete Sensor Assembly Instructions (See Figure 3 in main text for final product)
74. Push the stem of the Pyrex glass funnel through o-ring hole in the bottom pipe plug.
75. Wrap the two 1 cm lengths of 0.635 cm (0.25 inch) flexible plastic tubing around the
funnel stem, 5 cm apart and approximately centered along the length of the stem.
76. Using the u-bolts, attach the PCB to the funnel stem at the locations of the flexible
tubing.
77. Push the 4 inch coupling on to the pipe plug (make sure the large o-ring is greased).
78. Attach bulkhead fitting wire clip to sensor PCB.
79. Add desiccant bag to housing.
80. Slide the top pipe plug on to the coupling (make sure large o-ring is greased). To
accomplish this, temporarily pull the glass funnel downwards to release pressurized air
from the housing. Once the top pipe plug is snug on the coupling, push the funnel stem
back through the top pipe plug.
81. Push the 10 cm length of plastic tubing on to the top of the funnel stem. This tube will
route bubbles past the 4 holes drilled in to the bottom of the gas collection chamber,
ensuring that all gas is trapped within the gas collection chamber.
82. Place the assembled sensor housing on top of the extension funnel, making sure to line up
the standoffs with the grooves in the sensor housing.
83. Loosen the hose clamp holding the D-rings to the sensor housing. Lower the hose clamp
until the carabiners on the funnel can be clipped on to the D-rings. Once the carabiners
are attached, raise the hose clamp until the metal wires are taught and tighten hose clamp.
111
Make sure the metal wires are tight enough that the extension funnel is securely fastened
to the sensor housing.
84. Screw the gas collection chamber on to the top pipe plug.
85. Take the 3 nylon ropes tied to the D-rings and pass them through the 3 eye-bolts at the
top of the gas collection chamber.
86. Tie the 3 nylon ropes to a stainless steel s-hook for attachment to the deployment
structure shown in Figure 1 of the main text.
Data Buoy Materials
- 30 cm length of nominal 3 inch PVC pipe
- 3 nominal 3 inch PVC couplings
- 2 nominal 3 inch PVC pipe plugs
- 1 EPDM O-ring, dash size 236
- 1 Angel food cake pan
- Partall #2 mold release paste
- Dow Corning high vacuum grease
- US Composites 2-part expanding polyurethane foam, 81b density
- West Marine 105 epoxy resin
- West Marine 207 epoxy hardener
- 1 7/16"-20 thread size nut
- 1 7/16" washer
- 1 1/8" stainless steel NPTF male straight connector
- 1 PVC ball valve, 1/8" NPT female (such as McMaster 4757K11)
- Stainless steel hose clamp for pipe sizes 80-150 mm
112
- 2 stainless steel D-rings for 3.8 cm (1.5 inch) webbing width
- 3 Li-ion battery packs, 18650 7.4V 8800mAh (Tenergy #3 1010)
- 1 30 cm length each of black, red, green, and white 24 gauge stranded wire
- 2 four pin male Molex wire receptacle housings (Molex #43645-0400)
- 8 female Molex connecting sockets (Molex #0462350001)
- 1 four pin female Molex housing connector plug (Molex #0436400401)
- 4 male Molex connecting sockets (Molex #0430310002)
- Sandpaper
- Disposable paintbrush
Data Buoy Assembly Instructions
Instructions for buoy end plugs (refer to Fit S9):
87. On the lathe, machine the top of the first pipe plug to a flat, smooth finish.
88. Drill a 7/16" diameter hole in the center of the first plug.
89. Drill and tap a hole for the 1/8" straight connector. Hole should be 1.5 cm from one flat
edge of the PVC pipe plug. Screw the connector into the hole using Teflon tape to form a
tight seal.
90. Screw the PVC ball valve onto the 1/8" straight connector such that valve handle points
downwards when in open position.
91. Insert the bulkhead fitting in to the 7/16" hole and secure with the nut and washer.
92. Crimp the female Molex sockets on to the ends of the bulkhead wires. Insert sockets into
the 4-prong male Molex clip. Hold Molex clip such that you are looking in to holes
where you will insert the sockets, and the plastic latch is facing upwards. From left to
right, insert black, red, green, and white wires.
113
93. Make a 4-stranded 30 cm extension cable using the black, red, green, and white wires.
Place a female Molex clip on one end and a male Molex clip on the other, making sure to
keep the wire colors consistent. Connect this extension cable to the bulkhead fitting
Molex clip.
94. On the lathe, trim the second pipe plug to 3 cm in length.
95. Cut O-ring groove into side of the second pipe plug. Groove should be 0.282 cm deep
and 0.475 cm wide, and positioned approximately 0.6 cm from end of pipe plug (Fig S9 -
C). Make sure that groove is smooth. It may help to reduce the lathe speed.
96. Grease the O-ring with silicone vacuum grease and insert into the O-ring groove.
Instructions for buoy body (refer to Fig S9):
97. Cut a 1 cm ring from one of the pipe couplings. Using the PVC cement, glue this ring 7.5
cm from one edge of the PVC pipe.
98. Cut an 8.9 cm (3.5 inch) diameter hole in the center of the angel food cake pan.
99. Coat the inside of the angel food cake pan with the mold release paste.
100. Spread a thin ring of vacuum grease right around the hole in the angel food cake
pan, on the inside of the pan.
101. Pass the short end of the PVC pipe through the hole on the inside of the angel
food cake pan. Let the wider ring formed by the pipe coupling rest on the base of the
angel food cake pan, on top of the ring of vacuum grease (grease will keep the
polyurethane foam from seeping through this crack). Support the angel food cake pan
from below with the pipe resting in the center.
114
102. Following the instructions on the polyurethane foam container, mix 300-350 total
milliliters of foam mix. To achieve consistent mixing, we mounted a disposable paddle
on to a hand-held drill.
103. Quickly pour the foam mixture into the angel food cake pan and let it rise. After
foam is fully risen and cooled, remove the angel food cake pan.
104. Use the sandpaper to roughen the surface of the foam and remove the mold
release paste.
105. Following the instructions on the epoxy container, mix a small batch of epoxy.
Paint the foam with the epoxy, using multiple batches and layers until foam is fully
coated in a thick shell of epoxy.
106. Using the PVC cement, glue the first pipe plug in to one of the couplings (make
sure it already has the extension cable built in step 66). Glue this coupling on to the long
end of the buoy.
107. Use the PCV cement to glue the remaining coupling to the short end of the buoy.
108. Use the hose clamp to secure the D-rings towards the base of the buoy body. D-
rings should be on opposite sides of the buoy (Fig S9 - A).
Buoy Electronics Materials
- Arduino Pro Mini 328 - 5V/16MHz
- MicroSD card breakout board (Adafruit product #254)
- DS3231 precision RTC breakout (Adafruit product #3013)
- 1 four position Molex Micro-Fit 3 .0 TM connector (Molex product #43650-0413)
- 2 two position through-hole Molex connector (Molex product #0432550059)
- 2 two position connector plug (Molex product #0050841025)
115
- 2 0.1 pF surface mount capacitor, size 0603
- 1 1OptF surface mount capacitor, size 1210
- 2 lengths of straight male headers, 8 pins long
- 2 lengths of straight female headers, 8 pins long
- 2 length of straight female headers, 12 pins long
- 2 length of straight female headers, 6 pins long
- 1 length of straight female headers, 2 pins long
- Two 0.437" nylon spacers, unthreaded #2
- Two 2M machine screws, 12 mm in length
- Two 2M nuts
- Two 2-56 nylon machine screws, 7/16" length
- Two 2-56 female hex standoffs, 7/16" in length
Buoy Electronics Construction Directions
109. Repeat step 44 for the buoy PCB Arduino.
110. Solder the 8 pin male headers to the MicroSD breakout board and the DS3231
clock.
111. Solder all female headers and remaining electronic components to PCB.
112. Pass the 2-56 nylon screws through the holes in the DS3231 clock, and screw in
to the 2-56 nylon standoffs. DS3231 can now be installed on the PCB.
113. Place the MicroSD breakout board on the PCB. Pass the 2 mm machine screws
through the holes on the breakout board, through the 0.437" nylon spacers, and through
the holes on the buoy PCB. Secure the screws with the 2 mm nuts.
Optional Battery Sled Materials
116
- 1 37 cm long piece of schedule 80 PVC pipe, nominal 2 /2" diameter
Optional Battery Sled Instructions (refer to Fit S10)
114. Cut the PVC pipe in half length-wise.
115. Use CNC mill programming software such as Mastercam to create a program for
the cuts shown in Figure S 10.
116. Mount the half-pipe on the mill and run the program.
117. Drill a 1.27 cm (0.5 inch) diameter hole 1.9 cm from the top edge of the half-pipe.
117
SUPPLEMENTARY INFORMATION FIGURES
a - Bottom view
7/16" washer
and nut
0.87 cm (11/32
inch) diameter
hole with
internal o-ring
groove 0.272
cm wide and
0.154 cm deep
c - Top side view
Nominal 1 inch
male
unthreaded to 1
inch male NPT7
adapter
b - Side view
3 cm
3.3 cm {
EPDM o-ring size 2
d - Top view
0-ring groove 0.475 cm
wide, 0.282 cm deep
0.8 cm
3.32 cm diameter hole
in PVC disc
Bulkhead fitting through a
7/16" diameter hole
7.62 cm (3 inch)
diameter PVC disk
Pipe plug surface
machined to smooth
finish
Figure S1 - Top pipe plug
a - Top view b - Side view
Zip tie
> 11.5 cm of
nominal 5
inch acrylic or
PVC pipe
4.42 cm tall
and 0.635 cm
k wide notch
Figure S2 - Bottom pipe plug
118
Stainless
steel eye
bolts, 6-32
thread
size
Side view
Stainless
steel hose
clamp
D-ring for 3.8
cm (1.5 inch)
webbing
Pyrex glass
funnel passing
through sensor
housing
Figure S3 - Assembled sensor housing
Figure S4 - Hook-shaped lathe tool
119
Nominal 0.25
inch NPT male
to 0.125 inch
barbed fitting
Stainless steel
eye-bolts, 0.25
inch-20 thread
size
0.637 cm
(0.25 inch)
hole
a - Top view
b - Bottom view
c - Side view
r
Figure S5 - Gas collection chamber
Nominal 1
inch coupling
Nominal 1
inch clear
PVC pipe
Nominal 1 inch
unthreaded female
to female NPT
adapter
a - Top view b - Top side view
Stainless steel standoff,
18-8 and 1%" length
P, [
Commercial funnel from
Texas Hunter
12 oz. lead
fishing weight
Stainless steel carabiner
crimped to 1/16" diameter
stainless steep rope
Stainless steel eyebolt,
'"-20 with 1" shank
Figure S6 - Detachable extension funnel
120
-A
AO 
-
b - Top view
0 206 cm diameter (
.102 cm dia
7 A
0BP--~~~ 0 0.-
a - Right PVC shim
* Figure not drawn
to scale
D
-E
-F
c - Side View
--Q
--R
V U T
Figure S7 - Right PVC shim
a - Left PVC shim b
0
0
0 0
* Figure not drawn
to scale
- Top view
0.206 cm diameter (0.081 inches:
0.102 cm diameter
0.04 inches)
-A
sO
R 0 D
E
H
N M L K J I
c - Side view
--U
Y X W
Figure S8 - Left PVC Shim
0
0
00 0
N
M L K JIl
121
).081 inches)
meter (0.04 inches)
Section Length Length
(inch) (cm)
A-B 0.100 0.2540
A-C 0.200 0.5080
A-D 0.780 1.9812
A-E 0.880 2.2352
A-F 1.000 2.5400
A-G 1.100 2.7940
A-H 1.200 3.0480
1-J 0.085 0.2159
i-K 0.105 0.2667
I-L 0.235 0.5969
i-M 0.385 0.9779
N-O 0.150 0.3810
N-P 1.050 2.6670
Q-R 0.182 0.4623
Q-s 0.350 0.8890
T-U 0.105 0.2667
T-V 0.385 0.9779
Section Length Length
(inch) (cm)
A-B 0.100 0.2540
A-C 0.200 0.5080
A-D 0.320 0.8128
A-E 0.420 1.0668
A-F 1.000 2.5400
A-G 1.100 2.7940
A-H 1.200 3.0480
1-J 0.085 0.2159
I-K 0.105 0.2667
I-L 0.135 0.3429
I-M 0.235 0.5969
I-N 0.385 0.9779
O-P 0.150 0.3810
0-Q 0.300 0.7620
O-R 0.900 2.2860
O-S 1.050 2.6670
T-U 0.182 0.4623
T-v 0.350 0.8890
W-X 0.105 0.1050
W-Y 0.385 0.9779
b - View of bottom
Bulkhead fitting
through a 7/16"
diameter hole
Polyurethane
foam donut
cast in angel
food cake pan
> D-ring for 3.8
cm (1.5 inch)
webbing
PVC ball
valve for
pressure
release
c - Buoy cap
0-ring groove
0.475 cm wide,
0.282 cm deep
0.6 cm
Figure S9 - Custom data buoy
a - Side view B - Top view
A B C D
Section Length Length
(inch) (cm)
A-B 1.00 2.54
A-C 4.00 10.16
A-D 4.50 11.43
A-E 7.50 19.05
A-F 9.25 23.50
A-G 12.25 31.12
A-H 14.50 36.83
A-1 15.25 38.74
E F G H I
1.27 cm (1/2")
diameter hole
1.9 cm from
edge
Figure S10 - Optional battery sled
122
30 cm of
nominal 3
inch PVC
pipe
a - Side view
3 cm
ib
Methane bubble size distributions, flux, and dissolution in a freshwater
lake
Kyle B. Delwichel*, Harold F. Hemond'
'Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, MA,
United States
*Corresponding Author. Email: kyled(mit.edu
Publication citation: Delwiche, K.; Hemond, H. Methane bubble size distributions, flux, and
dissolution in a freshwater lake. Environ. Sci. Technol. 2017, 51, 13733-13739.
(Publication reformatted to fit this dissertation)
123
Chapter 4: Methane bubble size distributions, flux, and dissolution in a
freshwater lake
ABSTRACT
The majority of methane produced in many anoxic sediments is released via ebullition.
These bubbles are subject to dissolution as they rise, and dissolution rates are strongly influenced
by bubble size. Current understanding of natural methane bubble size distributions are limited
by the difficulty in measuring bubble sizes over wide spatial or temporal scales. Our custom
optical bubble size sensors recorded bubble sizes and release timing at 8 locations in Upper
Mystic Lake, MA continuously for 3 months. Bubble size distributions were spatially
heterogeneous even over relatively small areas experiencing similar flux, suggesting that
localized sediment conditions are important to controlling bubble size. There was no change in
bubble size distributions over the 3 month sampling period, but mean bubble size was positively
correlated with daily ebullition flux. Bubble data was used to verify the performance of a
widely-used bubble dissolution model, and the model was then used to estimate that bubble
dissolution accounts for approximately 10% of methane accumulated in the hypolimnion during
summer stratification, and at most 15% of the air-water methane flux from the epilimnion.
124
Bubble Diameters at 3 m
10
0 1i b
Bubble size Bubble Diameter (mm)
sensors
Bubble Diametem at 16 m
af 1 10
12
Bubble Diameter ( m)
INTRODUCTION
Freshwater lakes and reservoirs are a significant source of methane to the atmosphere,
14
and methane is a potent greenhouse gas with large implications for future climate change'-4.
Ebullition is often the dominant methane transport pathway compared with dissolution from
sediments and transport through plant vascular tissue, and can represent up to 95% of lake
methane emissions 4. Since rising bubbles are subject to dissolution, only a portion of the
methane released from the sediment as bubbles is emitted to the atmosphere. The remainder
dissolves into the water column where it is available for biological incorporation into the lake
food web5 7 .
Bubble dissolution rates depend strongly on bubble size and to a lesser extent on water
chemistry, temperature, and dissolved gas concentrations8-1 0. Despite the importance of methane
bubble size to dissolution and atmospheric emissions, relatively few estimates of lacustrine
methane bubble size distributions exist'"-13. Existing bubble size distribution measurements are
typically limited to short periods in time or space, and therefore little is known about the spatial
and temporal changes in methane bubble size distributions within a water body.
125
Given the limited knowledge about bubble size distributions, it is difficult to estimate the
importance of bubble dissolution to the dissolved carbon budget in a lake. Some data suggest
that dissolution may contribute significantly to methane in the hypolimnion' 3. and insignificantly
to methane in the epilimnion 14. Recent modeling work suggests that bubble dissolution can be
important to both hypolimnetic and epilimnetic methane accumulations at sufficiently high
ebullition flux'. In all cases the bubble dissolution equations used to calculate contributions to
dissolved methane were developed for deep marine environments and have not been
9independently verified for shallow, freshwater conditions
The purpose of this work is to characterize the temporal and spatial variability of methane
bubble size distributions in a dimictic kettle lake, to test the performance of a widely-used bubble
dissolution model under relatively shallow freshwater conditions, and to quantify the importance
of dissolving bubbles to dissolved methane levels within both the hypolimnion and epilimnion.
Detailed bubble size distribution data were gathered using our recently-developed bubble size
sensors which can measure in situ bubble volumes over long time periods' 6-7 . By deploying
multiple bubble size sensors at 2 depths within the lake for a multi-month period we are able to
address the following questions:
1. What is the distribution of bubble sizes in this lake and how spatially and temporally variable
are these sizes?
2. Do existing methane dissolution models accurately predict the evolution in bubble size
distribution as bubbles rise through a water column?
3. How important is bubble dissolution to accumulated methane in the hypolimnion, or to the air-
water flux of dissolved methane from the epilimnion?
126
MATERIALS AND METHODS
Bubble size distributions were measured using a bubble size sensor and data processing
algorithms described in Delwiche et al. " and Delwiche and Hemond1 7 . Briefly, the sensor uses
infrared LEDS, phototransistors, and a microcontroller to optically detect and measure bubbles
rising through an inverted glass funnel stem. The sensor electronics are encased in a custom
water-proof housing, and a custom surface data buoy holds batteries and electronics capable of
powering the sensor via a data cable for up to 4 weeks between battery replacements. Raw sensor
data is downloaded at the data buoy and post-processing algorithms convert the data to estimated
bubble volumes. The sensor also includes optional attachments to increase the number of
intercepted bubbles and to sample their bulk methane content.
Sensor deployment
Eight bubble size sensors were deployed at two different depths in Upper Mystic Lake,
MA from July-November of 2016. Upper Mystic Lake is a dimictic, eutrophic kettle lake near
Boston, MA with a history of methane ebullition'8-1 9. The sensor deployment site shown in
Figure 1 was chosen because of its mild bathymetry gradients and observed history of ebullition
in preliminary field work. Four sensors were deployed at 16m depth (sensors TI, T2, T3, and
T4), and four sensors were deployed at 3m depth (sensors T6, T7, T8, and T9) using the
deployment system described in Delwiche and Hemond' . Initial deployments prior to July had
more sensors in the vicinity of T3, but field results showed that collected bubble volumes were
consistently lower around T3 and much higher around T7, so in July sensors TI, T6, and T4
were moved southward to capture more bubble data. Sensor T3 remained in place and continued
to see negligible ebullition for the remainder of the deployment window. We believe the lack of
127
measured ebullition represents a real phenomenon because the sensor continued to record data
consistent with particulate matter floating past the detectors (indicating the sensor was working),
and we did not collect gas in the gas collection chamber. The lack of ebullition in a location with
similar bathymetry and expected organic matter deposition to a bubbling area has intriguing
implications for controls on ebullition flux, although this topic was outside the scope of the
present study.
Figure 1: Upper Mystic Lake field site showing the sensor deployment locations (sensor IDs
visible in inlay). Bathymetric lines drawn at 3.05 m (10 foot) increments.
128
Data collection and sampling
We visited sensors every 1-2 weeks to download data and collect gas accumulated in the
gas collection chambers. Collected gas was stored in wetted glass syringes and analyzed within
24 hours on a thermal conductivity gas chromatograph. Methane results were calibrated using
commercial standards of 10%, 50%, and 100% methane, and nitrogen results were calibrated
against 10%, 50% and atmospheric nitrogen concentrations. We periodically measured
temperature and dissolved oxygen using a Hydrolab MiniSonde probe, and gathered water
samples for dissolved methane analysis using a sampling device described in Peterson.
Dissolved gases were extracted into a helium-filled headspace by shaking samples for 20
minutes and then analyzing the headspace gas on a Shimadzu 2014 gas chromatograph using a
flame ionization detector.
To quantify potential changes in bulk methane content within the gas collection chambers
in between sampling intervals, we incubated a known quantity of methane in a trap modified to
block any incoming bubbles and deployed at 3 m depth. Multiple measurements of the change in
gas volume and methane content over 1-2 week periods allowed us to modify measured methane
contents in the other traps to account for this modest in-trap dissolution. In addition to the
modified trap at 3 m, we also ran incubation trials in T3 (at 16 m) when it became apparent that
T3 was deployed in a location with negligible ebullition flux.
Bubble dissolution model
Bubble dissolution rates are calculated using the following mass transfer equation:
129
-- = -K(H P - C,) -
dz Vb (1)
where i is the gas, M is the number of moles of gas i in the bubble, z is vertical distance, KL is
the gas transfer coefficient, H is the Henry's constant, P is the partial gas pressure within the
bubble. C is the dissolved gas content in the water, r is bubble radius, and Ub is the bubble rise
velocity. Many approximations for the bubble gas transfer coefficient and the bubble rise
velocity exist, and the equations used in this study are those included in the model from
McGinnis et al." The bubble dissolution model was written in Matlab, integrated using the Euler
method, and run with Upper Mystic Lake dissolved oxygen, dissolved methane, and temperature
profiles (profiles in S.I.). Nitrogen was assumed to be in equilibrium with the atmosphere.
RESULTS AND DISCUSSION
Bubble size and flux characteristics
Spatial heterogeneity in bubble size distributions
During the 2016 field season sensors were fully operational from August through mid-
November, excluding times during power loss or sensor malfunction. Of the 8 sensors deployed,
7 recorded a substantial number of ebullition events throughout the season. We believe that
sensor T3, which only recorded 4 bubbles all season, was deployed in a low flux area as
described in Materials and Methods.
Previous studies of methane bubble size distributions have either included distributions at
a single location in a water body, or distributions measured over a short period of time. These
previous studies have therefore left open the question of spatial heterogeneity in bubble size
distributions within a water body. The 2016 field data shows that there are statistical differences
130
between some distributions even over the relatively small and bathymetrically homogeneous
sampling area chosen for this study (Figure 2), implying that localized sediment structure may
play an important role in determining bubble size distributions. We also note that several size
distributions had a double-peak structure with one peak between 3-4 mm diameter, and a peak
around the mean bubble diameter. Further work is necessary to determine prevalence and cause
of this double peak distribution structure.
Dissolution modeling using data from sensors TI, T2, and T4 to predict distributions at
T6, T7, and T9 yields size distributions that are consistent with distributions measured at T6, T7,
and T9 (modeled results of 5.77 mm, 5.50 mm, and 5.46 mm, respectively, compared to
measured results of 5.6 mm, 5.7 mm, and 5.7 mm, respectively). Sensor T8 appears to have a
significantly larger bubble size distribution, and it therefore may not be appropriate to use one
bubble size distribution to calculate dissolution rates over an entire water body.
131
15
10
S
0
5
Sensor 1 - 16m depth
is- a Mean damter (mm)=4.8 -
SMD (mm) =6.2
10
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1
Sensor 2 -6m depth
LS b Mean dameter (mm) =4.5 -
SMD (mm) 6A
LO -
5 -
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1
Sensor 3 - im depth
Mean dameter (mm) =8.1
SMD (mm) 8A
10 
I
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1!
Sensor 4- 16m depth
Is - CMean dameter (mmn) =4.6
10. SMD (mm) 46
5
15
10
5S
5
15
10
5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Sensor 6 -3m depth
d Mean dameter (mm) =5.6 -
SMD (mm) =7.5
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1
Sensor 7-3m depth
d Mean dameter (mm) =5.7
SMD (mm)=7.4
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 1
Sensor 8-3m! depth
- e Mean dameter (mm) =7.1 -
SMD (mm) 9.6
AF- lTFTf~fFF=
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Sensor 9-3m depth
- dMean damneter (mm) =6.7
SMD (mm) 7.9
-ffT~Fh1
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
Bubble Diameter (mm)
Figure 2: Bubble size distributions from each sensor deployed during 2016, including
mean bubble diameters and Sauter Mean Diameters (SMD). Letter in the top left corner
shows statistically different distributions (comparisons are only made between sensors at
the same deployment depth, 16m or 3m). Sensor 3 only saw 4 bubbles during the 2016
season.
132
LL2
LL
I
15
10
5
0
5
5
0
Temporal heterogeneity in bubble size
As shown in Figure 2, bubble size distributions vary spatially even over relatively short
distances. The question remains as to whether size distributions vary temporally. Although our
bubble size data only span 3 months, within this limited context we saw no evidence of
systematic changes to the bubble size distributions. However, we did find that average daily
bubble size was positively (though non-linearly) correlated with daily ebullition flux (Figure 3).
This indicates that higher ebullition flux is attributed both to more bubbles and to increased
bubble size. The natural logarithm of daily ebullition flux was a significant predictor for average
daily bubble size (p<0.05, excluding days with fewer than 10 bubbles), and the regression
coefficient of 0.46 indicates that, for example, a 50% increase in daily ebullition flux yields a
0.08 mm increase in daily average bubble size. To determine whether this would have an
appreciable impact on net methane dissolution, we modeled the amount of methane dissolution
per mL gas emitted from all bubbles on days when flux was 40 - 60 mg CH4/m 2/day. We then
modeled the effect of the predicted increase in bubble size distribution due to a 100% and 200%
increase in flux and found that this resulted in a 3% and 5% decrease, respectively, in the amount
of methane dissolved per mL gas emitted to atmosphere. Therefore, bubble dissolution volumes
calculated with size distributions gathered on relatively low flux days may mildly over-represent
in situ dissolution rates.
133
8 8 8
E66 6
0e~~~ I. * ks--
4 4 4
E 2 2 2
a,0 0 0
100 10 102 100 10 1 102 100  101 102
) Sensor6 Sensor7 Sensor8 Sensor9
CM
8 8 * 8 8
6 6 -L 6 6
L,41 4 4
4 0
2 2 2 2
0 - 0 0 0
100 101 102 100 10 1 102 10 0  101 102 100 101 102
Daily ebullition flux (mg CH 4/m /day))
Figure 3: Relationship between average daily bubble diameter (mm) and daily ebullition
flux (mg CH4/m 2/day). Best fit lines drawn for data sets with significant correlation
between ebullition flux and daily average bubble diameter. To improve accuracy of the
daily bubble diameter, days with fewer than 10 bubbles are not included.
Temporal heterogeneity in ebullition flux
In addition to measuring bubble size, the sensors provide detailed data on the timing of
bubble release. Figure 4 shows the complete time series for bubbles measured by each sensor.
As has been found with previous studies'19,22, ebullition events often coincided with drops in
hydrostatic pressure (see S.I.), and high flux events often occurred simultaneously at multiple
sensor locations.
134
Sensori Sensor2 Sensor4
Sensor T43
15
10so T1 
-
Sensor T8 2
(at 31 m)
Aenso T9 Ot
F(gar 4: Tiesre fidvdanbbl msin srcrddb)ahsnowt
corresponding~0 di ee mesrmn (m ) Gre bascrepn0otmssnoswr
ofln. lecice crepodt dt reiulypbise n ewch1n0Hmn"
Penosre r4 e ae ndpomn oainfomNrht ot ssoni iue1
135
In contrast with earlier studies, temporal resolution for data shown in Figure 4 is
exceptionally high (on the sub-second scale) which provides a powerful tool for studying fine-
scale temporal dynamics of methane bubble release. Bubbles in Upper Mystic Lake are often
tightly clustered in time. For example, 25% of bubbles are found to enter the sensor within 1
second of a previous bubble (Figure 5). While it cannot be shown from these data that bubble
clusters are emitted from the exact same bubble vent, visual observations of bubble events within
Upper Mystic Lake show that several bubbles in quick succession often break the surface in the
same location. The degree of temporal clustering observed in this data set is similar to that
found using Upper Mystic Lake ebullition data gathered using a sonar device 18. Temporal
clustering on the daily scale (see S.I.) corroborated previous findings that to accurately estimate
ebullition rates, bubble flux must be measured for many days23
100
R0
80-
60-
40-
25% of bubbles are I sec apart
E 20-
0
10'1 101 101 105
Time between bubbles (sec)
Figure 5: Cumulative frequency percentage of the time between bubble arrivals (in
seconds). The horizontal and vertical lines show that 25% of the bubble data is recorded
within 1 second of the previous bubble.
The close temporal spacing of rising bubbles could be significant to bubble dissolution.
Previous studies suggest that bubbles rising within a few seconds of each other may experience
136
higher velocities 2426 While an increase in velocity would by itself reduce the bubble travel
time and therefore reduce total dissolution, this effect is partially compensated for by the fact that
increased velocity will keep bubbles larger, and larger bubbles have higher rates of dissolution
(see equation 1). For example, a 5mm diameter bubble released at 16m and subjected to a 30%
velocity increase (30% being an upper bound based on existing literature 24-26) will experience
5% less dissolution (by volume) than a 5 mm bubble with no velocity increase. The effect is
larger for smaller bubbles, and a 1 mm diameter bubble will experience 25% less dissolution by
volume.
Close bubble spacing could also contribute to bubble coalescence. Because dissolution
rates are highly dependent on bubble size, coalescence has the potential to significantly impact
dissolution. The physical and chemical mechanisms of bubble coalescence are complex, and
changes with water chemistry, bubble size, and bubble rise trajectory27-29 , so more work would
be needed to quantify the significance of methane bubble coalescence in freshwater
environments.
Bubble dissolution and contribution to dissolved methane in lake
Comparing bubble dissolution model to field data
The bubble dissolution model introduced by McGinnis et al. 9 is often used to estimate
bubble dissolution rates, but bubble dissolution rates vary based on water chemistry and the
model has not previously been tested for performance in shallow, freshwater environments. To
test the model under these conditions, we modeled the dissolution of bubbles measured at 16m
and compared the results to bubbles measured at 3m (using data from August to minimize
temperature changes, and bubble data from all sensors to limit the impact of variability in bubble
sizes between deployment locations), as shown in Figure 6. Bubbles were assumed to be 80%
137
methane and 20% at 16 m depth based on the bulk gas samples, taking into consideration in situ
dissolution as discussed in Materials and Procedures. We found the results were statistically
equivalent when the diffusion exponent n presented in Equations S2-S3 in the Supporting
Information was 0.55, which is between the values used for clean and dirty water (0.5 vs 0.67,
respectively). For a sensitivity analysis of the impact of clean versus dirty water and the variable
exponent n, please see the Supporting Information. The weighted methane fraction for the bubble
size distribution was 74% which is within the range expected based on field bulk gas samples
(corrected for in situ dissolution). We therefore conclude that the model adequately represents
bubble dissolution for the eutrophic freshwater conditions present in Upper Mystic Lake.
10
-Model Results
8 .--- Field Data
2 -
0
0 5 10 15 20
Bubble Diameter at 3m (mm)
Figure 6: Using field data to verify the bubble dissolution model. The dashed line
represents the combined bubble size distribution measured at 3 m. The solid line is the
predicted size distribution at 3 m based on bubble sizes measured at 16 m, as calculated by
the bubble dissolution model. Distributions are statistically similar, indicating that the
model performs well.
138
Contribution from dissolving bubbles to hypolimnetic methane accumulation
Dissolved methane profiles taken in August 2016 and November 2016 show an
accumulation of approximately 300 mmol CH 4/m 2 in the bottom 3 meters of the sampling area
(Figure S 1 in S.I.). We use the aforementioned bubble dissolution model with bubble data
gathered at 16 m to estimate that approximately 32 mmol CH4/m 2 of this accumulation was due
to bubble dissolution (similar estimation achieved using the Sauter Mean Diameter
approximation discussed in McGinnis et al.30 , equation included in SI). Methane bubble
dissolution therefore represents a significant but modest contribution to methane input to the
hypolimnion in this lake at the observed ebullition flux rates. Diffusion from porewater likely
contributes the majority of the remainder of the dissolved methane in the hypolimnion, though
sources such as lateral inflow, in situ production, and sediment disturbance (e.g.: by internal
waves or biota) may play a role given their importance elsewhere -34.
Contribution from dissolving bubbles to epilimnetic methane flux across air/water interface
Methane in epilimnetic waters is subject to biological oxidation and therefore does not
accumulative significantly. Therefore, we estimated the importance of bubble dissolution to the
methane budget in the epilimnion by comparing it to dissolved methane flux across the air/water
interface:
Fal, = k X (Ca, - pa Kq) (2)
where k is the gas transfer velocity (calculation method presented in the Supporting
Information), C is the dissolved gas concentration, p is the gas partial pressure, and K is Henry's
constant. The estimated diffusive flux of methane to the atmosphere during August was 0.8
mmol/m 2/day, although this may be an underestimate given the recently reported importance of
139
micro-bubbles in enhancing the air/water diffusive flux of methane 35-36. The estimated bubble
dissolution into the epilimnion calculated using the model was approximately 0.1 mmol/m2 /day.
Since this is at most 13% of the total estimated diffusion flux from the epilimnion, there must be
other sources of methane to the epilimnion to account for dissolved methane levels. This
conclusion is supported by previous studies showing that littoral sediments and in situ methane
production are of primary importance to epilimnetic methane' 4,3 ' 37-38. These findings are also
in line with recent modeling work suggesting that at the ebullition levels observed in Upper
Mystic Lake, the contribution from dissolving bubbles to the lacustrine carbon balance is likely
to be minor'.
Supporting Information
Calculation for the Sauter Mean Diameter (eq SI), the gas transfer rate equations used in the
bubble dissolution model (eqs S2-S4), the method for approximating epilimnetic air/water
methane flux (eqs S5-S9), profiles for temperature, dissolved oxygen, and dissolved methane
(Fig SI), the relationship between flux and hydrostatic pressure (Fig S2), a discussion of the
daily temporal heterogeneity in ebullition flux (Fig S3), and a discussion of the diffusion
exponent n in the bubble dissolution model (Fig S4).
Acknowledgements
We would like to thank students of the Parsons lab for their fieldwork assistance, and Mark
Belanger for his help in the machine shop. This material is based on work supported by the
National Science Foundation under grant number EAR- 1045193 and Graduate Research
Fellowship Program grant number DGE-0707428, the MIT Martin Family Fellowship to K.
Delwiche, the W.E. Leonhard 1941 professorship to H. Hemond, and the Singapore-MIT
Alliance for Research and Technology program.
140
REFERENCES
(1) Kirschke, S.; Bousquet, P.; Ciais, P.; Saunois, M.; Canadell, J. G.; Dlugokencky, E. J.;
Bergamaschi, P.; Bergmann, D.; Blake, D. R.; Bruhwiler, L.; et al. Three decades of global
methane sources and sinks. Nat. Geosci. 2013, 6, 813-823.
(2) Deemer, B.; Harrison, J.; Li, S.; Beaulieu, J.; DelSontro, T.; Barros, N.; Bezerra-Neto, J.;
Powers, S.; Dos Santos, M.; Vonk, J. Greenhouse gas emissions from reservoir water surfaces:
A new global synthesis. BioSci. 2016, 66 (11): 949-964.
(3) Bastviken, D.; Tranvik, L. J.; Downing, J. A.; Crill, P. M.; Enrich- prast, A. Freshwater methane
emissions offset the continental carbon sink. Science 2011, 331,50-50.
(4) Walter, K. M.; Zimov, S.A.; Chanton, J. P.; Verbyla, D.; Chapin, F. S.. Methane bubbling from
Siberian thaw lakes as a positive feedback to climate warming. Nature 2006, 443, 71-75.
(5) Kankaala, P.; Huotari, J.; Peltomaa, E.; Saloranta, T.; Ojala, A. Methanotrophic activity in
relation to methane efflux and total heterotrophic bacterial production in a stratified, humic,
boreal lake. Limnol. Oceanogr. 2006, 51, 1195-1204.
(6) Bastviken, D.; Ejlertsson, J.; Sundh, I.; Tranvik, L.; Methane as a source of carbon and energy
for lake pelagic food webs. Ecology 2003, 84(4), 969-981.
(7) Sanseverino, A.; Bastviken, D.; Sundh, I.; Pickova, J.; Enrich-Prast, A.; Methane carbon supports
aquatic food webs to the fish level. PloS ONE 2012, 7(8), e42723.
(8) Leifer, I.; Patro, R. K. The bubble mechanism for methane transport from the shallow sea bed to
the surface: A review and sensitivity study. Cont. Shelf Res. 2002, 22, 2409-2428.
(9) McGinnis, D. F.; Greinert, J.; Artemov, Y.; Beaubien, S. E.; Wilest, A. Fate of rising methane
bubbles in stratified waters: How much methane reaches the atmosphere? J Geophys. Res.
2006, 111, C09007.
(10) Vasconcelos, J.; Orvalho, S.; Alves, S.; Gas-liquid mass transfer to single bubbles: Effect of
surface contamination. AiChE J. 2002, 48(6), 1145-1154.
(11) Greinert, J.; Nuetzel, B. Hydroacoustic experiments to establish a method for the determination
of methane bubble fluxes at cold seeps. Geo-Marine Lett. 2004, 24, 75-85.
(12) DelSontro, T.; McGinnis, D.F.; Wehrli, B.; Ostrovsky, I. Size does matter: Importance of large
bubbles and small-scale hot spots for methane transport. Environ. Sci. Technol., 2014, 49(3):
1268-1276.
141
(13) Ostrovsky, I.; McGinnis, D. F.; Lapidus, L.; Eckert, W. Quantifying gas ebullition with
echosounder: the role of methane transport by bubbles in a medium-sized lake. Limnol.
Oceanogr. Methods 2008, 6, 105-118.
(14) Tang, K. W.; McGinnis, D.F.; Ionescu, D.; Grossart, H. P.; Methane production in oxic lake
waters potentially increases aquatic methane flux to air. Environ. Sci. Technol., 2016, 3(6), 227-
233.
(15) Schmid, M.; Ostrovsky, I.; McGinnis, D. Role of gas ebullition in the methane budget of a deep
subtropical lake: What can we learn from process-based modeling? Limnol. Oceanogr. 2017;
DOI 10.1002/lno.10598.
(16) Delwiche, K.; Senft-Grupp, S.; Hemond, H. A novel optical sensor designed to measure
methane bubble sizes in situ. Limnol. Oceanogr.: Methods 2015, 13, 712-721.
(17) Delwiche, K.; Hemond, H. An enhanced bubble size sensor for long-term ebullition studies.
Limnol. Oceanogr.: Methods 2017; DOI 10.1002/lom3.10201.
(18) Scandella, B. P.; Pillsbury, L.; Weber, T.; Ruppel, C.; Hemond, H.; Juanes, R. Ephemerality of
discrete methane vents in lake sediments. Geophys. Res. Lett. 2016, 43, 4374-4381.
(19) Varadharajan, C.; Hemond, H. F. Time-series analysis of high- resolution ebullition fluxes from
a stratified, freshwater lake. J. Geophys. Res. Biogeosci. 2012, 117, G02004.
(20) Peterson, E. Carbon and electron flow via methanogenesis, S04'. N03, and Fe 3 reduction in
the anoxic hypolimnia of Upper Mystic Lake. 2005, M.S. Thesis, M.I.T., Cambridge, MA.
(21) Clift, R.; Grace, J. R.; Weber, M. E. Bubbles, Drops, and Particles. Academic Press: New
York, 1978.
(22) Chanton, J. P.; Martens, C. S. Seasonal variations in ebullitive flux and carbon isotopic
composition of methane in a tidal freshwater estuary. Global Biogeochem. Cy. 1988, 2(3), 289-
298.
(23) Wik, M.; Thornton, B.; Bastviken, D.; Uhlback, J.; Crill, P. Biased sampling of methane
release from northern lakes: A problem for extrapolation. Geophys. Res. Lett. 2016, 43, 1256-
1262.
(24) Garner, F.; Hammerton, D. Circulation inside gas bubbles. Chem. Eng. Sci. 1954, 3(1).
(25) Marks, C. H. Measurements of the terminal velocity of bubbles rising in a chain. J. Fluid.
Siruct. 1973, March 17-22.
(26) Coppock, P.; Meiklejohn, G. The behavior of gas bubbles in relation to mass transfer. T. I.
Chem. Eng.-Lond. 1951, 75-86.
142
(27) Otake, T.; Tone, S.; Nakao, K.; Mitsuhashi, Y.; Coalescence and breakup of bubbles in liquids.
Chem. Eng. Sci. 1977, 32, 377-383.
(28) Nevers, N.; Jen-Liang, W.; Bubble coalescence in viscous fluids. AIChE Journal 1971, 17(1),
182-186.
(29) Kok, J. Dynamics of a pair of gas bubbles moving through liquid: Part 1. Theory. Eur. J. Mech.,
B/Fluids 1993, 12(4), 515-540.
(30) McGinnis, D. F.; Little, J. C. Predicting diffused-bubble oxygen transfer rate using the discrete-
bubble model. Water Res. 2002, 36, 4627-463 5.
(31) Murase, J.; Sakai, Y.; Kametani, A.; Dynamics of methane in mesotrophic Lake Biwa, Japan.
Ecol. Res 2005, 20, 377-385.
(32) Rimmer, A.; Eckert, W.; Nishri, A.; Agnon, Y.; Evaluating hypolimnetic diffusion parameters
in thermally stratified lakes. Limnol. Oceanogr. 2006, 51(4), 1906-1914.
(33) Imboden, D.; Lemmin, U.; Joller, T.; Schurter, M.; Mixing processes in lakes: mechanisms and
ecological relevance. Schw'eiz. Z Hydrol. 1983, 45(1), 11-44.
(34) Biderre-Petit, C.; Jezequel, D.; Dugat-Bony, E.; Lopes, F.; Kuever, J.; Borrel, G.; Viollier, E.;
Fonty, G.; Peyret, P. Identification of microbial communities involved in the methane cycle of a
freshwater meromictic lake. FEMS Microbiol. Ecol. 2011, 77, 533-545.
(35) Prairie, Y.; Giorgio, P. A new pathway of freshwater methane emissiosn and the putative
importance of microbubbles. Inland Waters, 2013, 3, 311-320.
(36) McGinnis, D.; Kirillin, G.; Tang, K.; Flury, S.; Bodmer, P.; Engelhard, C.; Casper, P.; Grossart,
H.; Enhancing surface methane fluxes from an oligotrophic lake: Exploring the microbubble
hypothesis. Environ. Sci. Technol. 2015, 49, 873-880.
(37) Hofmann, H.; Federwisch, L.; Peeters, F. Wave-induced release of methane: Littoral zones as a
source of methane in lakes. Limnol. Oceanogr. 2010, 55(5), 1990-2000.
(38) Fernandez, J.; Peeters, F.; Hofmann, H. On the methane paradox: Transport from shallow water
zones rather than in situ methanogenesis is the major source of CH 4 in the open surface water of
lakes. J. Geophys. Res. -Biogeo. 2016, 2717-2726.
143
SUPPLEMENTAL INFORMATION
Methane bubble size distributions, flux, and dissolution in a freshwater lake
Kyle Delwichel*, Harold F. Hemond'
'Civil and Environmental Engineering, Massachusetts Institute of Technology, Cambridge, MA,
United States
*Corresponding Author. Email: kyled@mit.edu
Supporting Information Contents
Contents
A. M ethods................................................................................................................................. 145
1. Calculating sauter mean diameter ....................................................................................... 145
3. Calculating epilimnetic methane flux across air/water interface ........................................ 145
B. Results and Discussion......................................................................................................... 146
1. Temperature, dissolved oxygen, and dissolved CH4 profiles.............................................. 146
2. Hydrostatic pressure drops trigger ebullition ...................................................................... 147
3. Daily temporal heterogeneity .............................................................................................. 147
C. Figures................................................................................................................................... 148
F ig u re S I ................................................................................................................................. 14 8
F ig u re S 2 ................................................................................................................................. 14 8
F ig u re S 3 ................................................................................................................................. 14 9
D. References............................................................................................................................. 149
144
A. Methods
1. Calculating sauter mean diameter
Previous research has demonstrated that the Sauter Mean Diameter (SMD) can be used to
simulate dissolution of an entire bubble size distribution'. The Sauter Mean Diameter is the
equivalent diameter of a bubble with the same volume to surface area ratio as the entire bubble
size distribution:
SMD = d(S)
where d is the diameter of each bubble in the bubble size distribution.
2. Calculating epilimnetic methane flux across air/water interface
The total dissolved methane flux across the air/water interface is:
Falw = k x (C - PaKH) (S2)
where C is the dissolved gas concentration, p, is the gas partial pressure (1803 ppb methane 2),
and Ki1 is Henry's constant 3. The k, represents the piston velocity, calculated as:
ki = (6)a x k60   (S3)
Where a = -0.67 when itio < 5m/s (from Liss et al., 19864, uio is the wind speed 10 m above the
earth's surface), Sc(T) is the Schmidt number for methane at the surface water temperature 5 and
k600 = 2.07 + 0.215u, 0 1 7 (cm/hr) (S4)
From Cole and Caraco 6. The u10 is estimated using Mackay and Yuen':
( 10.4 \
U 10 = In(z)+8.1)xuz (S5)
145
The Cole and Caraco relationship for k6 oo was developed based on data from a lake with
relatively low wind speeds. While other studies have highlighted the superiority of site-specific
k600 measurements8 , their results are in good agreement with the Cole and Caraco relationship
above 1.0 m/s. Local wind data is available through the Turkey Hill weather station in nearby
Arlington, MA, and the average wind speed in August 2016 was 1.3 m/s. Though the
anemometer height at Turkey Hill is unknown, previous work at Upper Mystic Lake comparing
Turkey Hill wind data to anemometer readings from a buoy on the lake estimated that the
relationship between these wind speeds is approximately 9:
UUML,1m( m/s) = Ustation( m /S) + 0.5 (S6)
The methane concentration in the epilimnion was approximately 1.2 uM at 1 m depth, yielding an
air/water methane flux of approximately 0.8 mmol/m 2/day. This flux calculation method was
used previously on Upper Mystic Lake by Varadharajan 9. We note that the average ebullition
flux was also 0.8 mmol/m 2/day, and the similar magnitude between diffusion and ebullition
reflects the fact that ebullition in the deep part of Upper Mystic Lake is relatively low compared
to other lakes where ebullition flux dominates101
B. Results and Discussion
1. Temperature, dissolved oxygen, and dissolved CH4 profiles
The temperature, dissolved oxygen, and dissolved methane profiles presented in Figure
SI were used in the bubble dissolution model. The profiles presented here are in line with data
from previous sampling campaigns at Upper Mystic Lake measuring dissolved oxygen and
temperature more frequently (e.g.: Varadharajan 20099 and Peterson 2005 12). The dissolved
methane profile was also used to calculate that approximately 0.3 mol CH4/m 2 of dissolved
146
methane accumulated in the bottom 3 meters of the water column between August 5* and
November 13th, 2016.
2. Hydrostatic pressure drops trigger ebullition
As has been widely demonstrated before' 3 -14, drops in hydrostatic pressure at Upper
Mystic Lake tend to trigger ebullition events. We monitored hydrostatic pressure using a
commercial sensor (the Solinst Model 3001 Levelogger Gold) that is permanently installed at
Upper Mystic Lake. As can be seen in Figure S2, large ebullition events are typically preceded
by large drops in pressure. Flux peaks also tend to occur simultaneously in multiple traps.
3. Daily temporal heterogeneity
Methane ebullition from natural waters is widely known to be a temporally
heterogeneous process, with emissions concentrated over short time intervals 14-15. Ebullition at
Upper Mystic Lake is also temporally clustered, as is visible in Figure 4 of the main text. To
quantify the temporal variability in methane ebullition, we calculate daily ebullition flux and
compared this to the cumulative ebullition flux and the number of days contributing to this
cumulative flux (Figure S3). The separation between the two lines in Figure S3 demonstrates
that the majority of the ebullition flux occurs on a small number of days (for reference, note that
in the hypothetical case of constant flux the two lines would be the same). For example, the
dotted line indicates that 60% of the sampled days only account for about 22% of the total flux.
This corroborates previous findings' 6 that short-duration sampling campaigns may under-
estimate ebullition flux unless they happen to coincide with higher-flux time periods.
147
C. Figures
0 0 0
2 - 2 2 -
4 - 4 4
6 6 6
EE E
4- 8 88
a -L C- 5Aug 2016
010 0 10 10 3 Oct 2016
............. 13Nov2016
12 - 12:e' 12
14 14 14 .
16 - 16 - 16 IB
0 20 40 0 10 20 0 500
Temperature (*C) Dissolved Oxygen (mg/I) Dissolved Methane (uM)
Figure Si: Temperature profile in degrees C (left), dissolved oxygen profile in mg/L (middle),
and dissolved methane in pM (right) taken in August. October, and November of 2016.
0.4 | C140
10.3 -
C.2 120
'60.1
E 100
E
60
40
20
0
Aug Sep Oct Nov
Figure S2: Hydrostatic pressure in m of water relative to the low point on Sept. 1st, 2016 (left
axis), and daily ebullition flux in mg CH4/M 2/day (right axis). Vertical dashed grey lines show
that ebullition peaks often coincide with dropping hydrostatic pressure. Daily ebullition is
plotted for sensor 1 and other sensors show similar event timing.
148
100
0.
60
U-
40
2- - Flux Volume
E- Number of Days
0 0 50 100 150
Daily Flux (mg CH4Im2/day)
Figure S3: Cumulative sum of daily ebullition flux (dashed line) and the number of days
contributing to the cumulative flux (solid line), plotted against the daily ebullition flux in mg
CH4/m 2/day. The separation between the Flux Volume and Number of Days demonstrates that
ebullition flux is not homogeneous, but rather occurs in bursts over a smaller subset of days. For
example, dotted vertical line shows that 60% of the sampled days only account for 22% of the
total observed flux.
D. References
1. McGinnis, D.; Little, J. Predicting diffused-bubble oxygen transfer rate using the
discrete-bubble model. Water Res. 2002, 36, 4627-4635.
2. International Panel on Climate Change (IPCC). The physical science basis: Summary for
policymakers. Fifth Assessment Report. 2013, Cambridge University Press.
3. Sander, R. Compilation of Henry's law constants (version 4.0) for water as a solvent.
Atmos. Chem. Phys. 2015, 15, 4399-4981.
4. Liss, P.; Merlivat, L., Eds. Air-Sea Gas Exchange Rates: Introduction and Synthesis, The
Role ofAir-sea Exchange in Geochemical Cycling. 1986, D. Reidel Publishing,
Dordrecht, The Netherlands.
149
5. Schwarzenbach, R.; Gschwend, P.; Imboden, D., Eds. Environmental Organic Chemistry,
Second Edition. John Wiley & Sons, Hoboken, New Jersey. 2003. p 911.
6. Cole, J. J.; Caraco, N. F. Atmospheric exchange of carbon dioxide in a low-wind
oligotrophic lake measured by the addition of SF6. Limnol. Oceanogr. 1998, 43(4), 647-
656.
7. Mackay, D.; Yeun, A. Mass transfer coefficient correlations for volatilization of organic
solutes from water. Environ. Sci. Technol. 1983, 17(4), 211-217.
8. Schilder, J.; Bastviken, D.; Hardenbroek, M. van; Heiri, 0. Spatiotemporal patterns in
methane flux and gas transfer velocity at low wind speeds: Implications for upscaling
studies on small lakes. Journal of Geophysical Research: Biogeosciences, 2016, 1457-
1467.
9. Varadharajan, C. Magnitude and spatio-temporal variability of methane emissions from a
eutrophic freshwater lake. Ph.D. Dissertation, Massachusetts Institute of Technology,
Cambridge, MA, 2009.
10. Casper, P.; Maberly, S. C.; Hall, G. H.; Finlay, M.J. Fluxes of methane and carbon
dioxide from a small productive lake to the atmosphere. Biogeochemistry 2000, 49, 1-
19.
11. Keller, M.; Stallard, R. F. Methane emission by bubbling from Gatun Lake, Panama. J.
Geophys. Res. 1994, 99, 8307-8319.
12. Peterson, E. Carbon and electron flow via methanogenesis, S042-, NO3-. and Fe 3
reduction in the anoxic hypolimnia of Upper Mystic Lake. M.S. Thesis, Massachusetts
Institute of Technology: Cambridge, MA, 2005.
13. Chanton, J. P.; Martens, C. S. Seasonal variations in ebullitive flux and carbon isotopic
composition of methane in a tidal freshwater estuary. Global Biogeochem. Cy. 1988,
2(3), 289-298.
14. Varadharajan, C.; Hemond, H. F. Time-series analysis of high- resolution ebullition
fluxes from a stratified, freshwater lake. J Geophys. Res. Biogeosci. 2012, 117, G02004.
15. Maeck, A.; Hofmann, H.; Lorke, A. Pumping methane out of aquatic sediments -
ebullition forcing mechanisms in an impounded river. Biogeosciences 2014, 11, 2925-
2938.
16. Wik, M.; Thornton, B.; Bastviken, D.; Uhlback, J.; Crill, P. Biased sampling of methane
release from northern lakes: A problem for extrapolation. Geophys. Res. Lett. 2016, 43,
1256-1262.
150
Chapter 5: Methane bubbles transport sediment and arsenic to a lake
surface
ABSTRACT
Bubbles effectively adsorb and transport particulate matter both in industrial and marine
systems. While methane bubbles emitted from anoxic sediments are found extensively in aquatic
ecosystems, relatively little attention has been paid to the possibility of methane bubble particle
transport. We quantified methane bubble-facilitated particle transport on Upper Mystic Lake,
MA, and microscope-based evidence demonstrated that transported particles were likely from the
sediment. Concentrations of arsenic, chromium, and lead in the particulate matter were similar
to bulk sediment measurements, further indicating that bubbles are transporting sediment
particles from depths exceeding 15 m. Testing in a 15 m tall laboratory bubble column showed
that over 90% of bubble particulate matter originated in the sediment instead of being scavenged
from the water column. We conclude that bubbles transport particles upwards through a
stratified water column through which significant movement of sedimentary material would
otherwise not be expected. This could have important implications for chemical or biological
cycling in natural ecosystems, although we found that bubble-facilitated arsenic transport on
Upper Mystic Lake is likely a minor component of total arsenic cycling.
INTRODUCTION
Bubble particle flotation is a process by which amphiphilic particles attach to a bubble's
gas-water interface and are transported upwards during bubble rise. This phenomenon is used
extensively in industry for applications such as separating valuable minerals from gangue
151
(Rodrigues et al. 2007, Min et al. 2008), removing ink during paper recycling (Vashisth et al.
2011), recovering desirable proteins and microorganisms from industrial bioreactors (SchUgerl
2000), and treating wastewaters (Lin and Lo 1996, Aldrich and Feng 2000, Rubio et al. 2002).
Bubble-mediated particle transport also occurs in the open ocean and contributes to an
accumulation of surface-active particles at the surface of the ocean (Wallace et al. 1972,
Blanchard 1975, Aller et al. 2005). The ocean surface microlayer (operationally defined based
on sampling protocol, with sampling thickness ranging from 50-400 um (Cunliffe et al. 2013))
has been found to contain microorganisms, pesticides, heavy metals, organic contaminants, and
other chemicals far in excess of concentrations in underlying waters (Wurl and Obbard 2004).
The bubbles responsible for transporting particulate matter are injected into the ocean surface by
breaking waves and scavenge particles as they rise (Liss 1975).
Despite the abundant evidence that bubbles are effective particle transporters in industrial
and open-ocean conditions, very little is known about the importance of bubble particle transport
in freshwater systems. Although smaller freshwater systems may lack the white-cap waves
necessary to inject bubbles into the water column, methane bubbles produced in anoxic
sediments are prevalent in these systems. These bubbles originate in the sediment as a product
of microbial organic matter decomposition, and are released to the surface during drops in
hydrostatic pressure, sediment disturbance, or upon sufficient gas accumulation (Chanton et al.
1988, Joyce and Jewell 2003, Varadharajan and Hemond 2012, Maeck et al. 2014, Liu et al.
2016).
Since methane bubbles can rise from great depths through density-stratified waters,
bubble flotation could provide a chemical link from deep water to surface waters that would
otherwise not occur through diffusion alone. This potential transport could have important
152
implications for aquatic chemical and biological cycles. Since aquatic sediments are often
contaminated with heavy metals and organic pollutants (Nriagu et al. 1996, Taylor and Owens
2009, Pan and Wang 2012), methane bubbles have the potential to transport contaminated
particles to the water surface. To date, very few studies have addressed the particle-transport
capability of methane bubbles generated in aquatic sediments. While the few existing studies
have found evidence that methane bubbles can transport polycyclic aromatic hydrocarbons
(Viana et al. 2012) and manufactured gas plant tar (McLinn and Stolzenburg 2009), the full
extent of methane bubble particle flotation in aquatic systems remains unknown.
To begin addressing the many questions related to methane bubble particle flotation in
freshwater systems, we have quantified bubble-particle transport under field conditions in Upper
Mystic Lake, MA, an urban lake with a history of sediment contamination. We also used
laboratory bubble columns to further study the potential for bubble-particle transport. Given the
importance of bubble size to particle flotation, we used a previously-designed bubble size sensor
(Delwiche et al. 2015, Delwiche and Hemond 2017) to measure bubble diameter distributions
both in the lake and in the laboratory. We address the following questions:
1. Can bubbles emitted from lake sediment transport particles to the lake surface?
2. Where do particles originate? Do bubbles shed or scavenge particles as they rise?
3. How does bubble volume affect particle transport?
4. What implications does this transport pathway have for arsenic cycling in Upper Mystic Lake?
Brief introduction to flotation theory
Rising bubbles collide with particles and under the right conditions the particle and
bubble will form a stable three-phase contact line. Bubble-particle flotation efficiency is highly
153
dependent on bubble-particle interactions, and particle capture efficiency (Ecap) is determined by
the effectiveness of three sub-steps:
Ecap Ec Ea Es
where E,, Ea, and Es are the particle collision, attachment, and stability efficiencies, respectively
(Yoon and Lutrell 1989, Dai et al. 2000).
Particle-bubble collision efficiency, Ec
Bubbles collide with particles as they rise through the water column. Particles move
along streamlines around the bubble, and collision occurs if the particle radius is greater than the
smallest distance between the fluid streamline and the bubble interface. Many models exist to
predict Ec, the collision efficiency (Dai et al. 2000). Earlier models focused only on
hydrodynamic effects, and found that the probability of collision was proportional to particle
diameter and inversely proportional to bubble diameter (Yoon and Lutrell 1989). However,
recent efforts have demonstrated that particle inertia also affects collision efficiency: initially,
inertia propels a particle toward the bubble interface, but later the centrifugal force generated as
the particle is swept outwards along the bubble interface decreases collision efficiency (Dai et al.
1998). A thorough review of collision models has been presented in Dai et al. 2000, and the
Generalized Sutherland Equation has been shown to predict bubble particle collision well for
particles between 1 - 1 00um in size and bubble diameters between 0.08 and 0.15 cm
(Hassanzadeh et al. 2017).
154
Particle-bubble adhesion efficiency, Ea
After a particle collides with a bubble, a stable three-phase contact line between particle,
gas, and water must develop for stable particle adhesion. For this to occur, the liquid film
between the particle and bubble must thin and rupture in less time than it takes the particle to be
swept past the bubble interface (Tao 2005). Film thinning and rupturing depend on particle
factors such as roughness, hydrophobicity, and electric charge. The more hydrophobic a particle
is, the less stable the thin film separating bubble and particle is likely to be. Rougher particle
surfaces are able to pierce the bubble interface more readily, and to hold micro gas bubbles that
can increase attachment efficiency (Krasowska and Malysa 2007, Hassas et al. 2016). Evidence
also suggests that adhesion increases for smaller particles and smaller bubbles, up to a lower size
limit (Yoon and Lutrell, 1989).
Particle-bubble detachment efficiency, Ed
Particles detach from the interface if a collision between the bubble and another particle
imparts enough kinetic energy to overcome the particle attachment forces (Ralston et al. 1999).
The forces acting on the particle when it is attached to the bubble include the capillary force
acting along the three-phase contact line, the particle weight in the liquid, and the hydrodynamic
drag force (Tao 2005). Additionally, when a rising bubble collides with a particle in solution it
induces a vibration in the bubble. This vibration causes the adhered particles to move in a
circulatory pattern, thus generating centrifugal force which, if strong enough, can detach the
particles (Cheng and Holtham 1995).
155
Role of bubble and paricle size inflotation
In general, flotation rates per unit of gas volume increase as particle size increases and
bubble diameter decreases. Smaller bubbles increase the particle collision probability and will
therefore increase flotation as long as bubbles maintain sufficient buoyancy to rise (Tao 2005).
Larger particles are more likely to collide with a rising bubble which increases collision
efficiency. Due to the complexity of the physics and dependence on particle characteristics,
different studies offer different optimum bubble and particle sizes (Ahmed and Jameson 1985,
Yoon and Lutrell 1989, Feng and Aldrich 1999). Although the range of optimum bubble
volumes reported in the literature varies, industrial flotation systems typically use bubble
diameters much smaller than the 2mm and larger bubble diameters often seen in natural aquatic
systems (DelSontro et al. 2005, Ostrovsky et al. 2008, Bussman et al. 2013). Since very large
bubbles have mobile interfaces (Clift et al. 1978), and bubble interface vibrations are known to
increase the particle detachment probability, we expect particles on very large bubbles to have a
higher probability of detachment. However, since we could find no literature on flotation with
large bubbles, further work is needed to understand the physics of particle detachment for
bubbles of large diameters. Another difference between flotation systems and natural aquatic
systems is that aquatic systems can be quite deep, and therefore bubble volumes change due to
hydrostatic pressure changes across the water column. We could not find literature discussing
the effect of bubble volume change on flotation, but we surmise that as bubbles grow and their
interfaces become more mobile, the detachment probability may increase.
156
METHODS
Upper Mystic Lake field site history
Upper Mystic Lake in Arlington, MA is an urban, dimictic kettle lake with an average
depth of 15 m, a maximum depth of 24 m, and a surface area of 0.5 km2 . The lake is used
extensively for recreational and scientific purposes, and previous studies have focused on
methane ebullition on the lake (Varadharajan and Hemond 2012, Scandella et al. 2016, Delwiche
and Hemond 2017). Chemical manufacturing and leather tanning industries during the late 1800s
and 1900s produced toxins such as arsenic, chromium, and lead and deposited them in waste
piles and settling ponds along the Aberjona River which feeds Upper Mystic Lake. These toxins
flowed into the lake during manufacturing and later during earth-moving activities. Arsenic,
lead, and chromium were deposited in layers in the Upper Mystic Lake sediments, and sediment
cores reveal a distinct layered pattern with peak metal/metalloid concentrations traceable to years
of peak manufacturing or earth-moving (Spliethoff and Hemond 1996).
Materials and supplies
Sample digestions were done with nitric acid from Fisher Scientific (Optima grade for
ultra-trace elemental analysis). All plastic ware was soaked in 5-10% HN03 for 24 hours and
rinsed with Milli-Q water prior to use.
Field sampling and analysis
We ran field tests in an area of the lake previously found to have relatively high ebullition
rates (Delwiche and Hemond 2017). To efficiently collect multiple samples in a single visit and
minimize algae build-up in our sampling containers, we triggered bubble events by dropping an
anchor into the sediment (calculations showed that the resulting bubble diameter distributions
157
were statistically similar to those of un-triggered bubbles). We then manually positioned a
custom bubble size sensor above the rising bubble plume (sensor described previously in
Delwiche et al. 2015 and Delwiche and Hemond 2017). Bubbles exiting the sensor, and any
particles adhered to the bubble/water interface, were collected in a removable sample cup
attached to the top of the funnel stem. Several anchor drops within an area of roughly 10 m by
10 m were required to intercept a sufficient number of bubbles, and we intentionally collected
samples with different total gas volumes. We also collected a blank water sample to correct for
background contributions of particulate matter and arsenic concentrations in bubble-transported
particle data.
We filtered samples within 24 hours with pre-weighed Whatman Grade 41 quantitative
cotton filters (nominal pore size 20 pm, 25 mm diameter). Due to filter clogging, we typically
used multiple filters for each sample. After filtering we air-dried filters, weighed them,
transferred each to microwave digestion vessels, and added 10 mL of nitric acid. Samples were
digested in a MARS6 microwave oven, diluted with 30 mLs of Milli-Q water, and then filtered
with a 0.2 um polyethersulfone membrane syringe filter. For analysis, we diluted samples to 2 %
nitric acid, added a rhodium internal standard, and analyzed on an Agilent 7900 ICP-MS with a 5
point calibration curve from 0.05 - 10 ppb. Blank analysis to determine background arsenic
concentrations in the Whatman cotton filter paper found levels at least two orders of magnitude
below sample concentrations.
Laboratory column design and operation
Large Column
To study bubble particle shedding and scavenging, we built a 15 m tall bubble column in
the laboratory stairwell. The column is comprised of four sections of 6 inch (15.3 cm) nominal
158
diameter transparent polyvinyl chloride (PVC) pipe joined by threaded unions with o-ring seals.
The base of the column is a reducing tee fitting with a removable spigot for drainage, and the
column was filled at the top. We built a sediment container connected to 1/8 inch copper tubing
that could be lowered into the column and secured at any depth. We used a syringe pump to
push air into the sediment through the tubing at a controlled rate. Sediment was originally
collected with an Ekman dredge from the same place in Upper Mystic Lake used for field
sampling.
For experimental runs, we injected 50 mL of air at 0.7 mL/min. Prior to the start of each
run we took background water samples to correct for background contributions to particulate
matter and arsenic concentrations in bubble-transported particle data. Bubbles passed through a
customized bubble size sensor fitted with the same sample collection cup design used in the field
setting. We filtered bubble column samples using pre-weighed 5.0 Pm and 0.2 ptm Whatman
Nuclepore membrane filters with 47mm diameter which allowed us to use only one of each filter
per sample. Filters were dried, weighted, digested, diluted, and analyzed as described above.
Blank analysis on Nuclepore membranes and lab filtering procedures found arsenic
contamination levels approximately two orders of magnitude below total (total being the sum of
results from the 5 pmol filter and 0.2 pmol filter) sample concentrations.
Small column
To study the effect of water pH we constructed a 1.5 m tall column from 7.4 cm diameter
transparent PVC pipe. The column base was fitted with an o-ring sealed port for gas injection in
to the sediment via a syringe pump. For the first two trials comparing pH, we placed
approximately 500 mL of thoroughly mixed Upper Mystic Lake sediment into the column and
159
filled with tap water (with pH = 9 prior to equilibration with carbon dioxide in atmosphere). We
injected 50 mL of air at 0.7 mL/min, and measured bubble volumes using a custom-built bubble
size sensor built with a smaller-diameter waterproof housing to fit the smaller width of the
bubble column. We conducted three replicate experiments with the same water and sediment
and tracked water pH. We then removed the water from the column (leaving the original
sediment in place) and replaced it with tap water that had been acidified to pH = 6.5 with trace
metal grade HNO 3 and ran three replicate experiments. The second set of trials comparing pH
were conducted similarly, except we replaced the sediment along with the water before
beginning the pH = 6.5 experiments. Sampling and analysis details were the same as those for
the large bubble column.
Bubble volume normalization
For both field and laboratory data, we measured bubble volumes at the top of the water
column. However, bubble volumes at the surface differ from those at the sediment because
rising bubbles undergo both gas dissolution and expansion due to pressure reduction (Leifer and
Patro 2002). In order to effectively compare data gathered from different systems, we used the
bubble dissolution model of McGinnis et al. 2006 to estimate initial bubble volumes based on
surface measurements. This model requires water temperature and dissolved gas concentration
profiles, which we estimated from profiles collected in October 2016 for field bubble volumes
(Delwiche and Hemond 2017). The laboratory column was assumed to be 21 degrees C, in
equilibrium with the atmosphere with respect to oxygen and nitrogen, with negligible dissolved
methane.
160
RESULTS AND DISCUSSION
Particle transport
Figure 1: Picture of the lake surface after a triggered bubble event showing an
accumulation of particulate matter (visible as light specs on the water surface, though in
reality particles are black).
Field daci shows bubbles transport particles
During field tests we observed particles accumulating on the water surface during bubble
triggering events (Figure 1), and individual bursting bubbles often left particles distributed in a
ring pattern. This visual evidence of bubble-particle transport was further supported by the
sample data showing a clear upwards trend between total particle mass and gas volume (Figure
2). Higher gas volumes led to significantly more particle mass transport (r 2 = 0.79, p<0.05),
with some minor spread in the data indicating that other factors in addition to total gas volume
affect particle transport rates. Some of this variability may have arisen because samples were
collected in an area of tens of square meters and localized sediment characteristics may impact
161
particle transport rates. Variability could also come from a difference in bubble volumes since
bubble volumes affect flotation rates (Yoon and Luttrell 1989), although we found no statistical
trend between particle mass transport and average bubble diameter (see discussion below). We
also note that triggering bubbles by dropping an anchor undoubtedly disturbed the sediments,
and this artificial perturbation may have affected particle transport rates.
15
E
E
cc 10
CD,
0
06
0 50 100 150 200
Total bubble volume (mL)
Figure 2: Total particle mass (in mg) associated with the bubbles captured during each
bubble triggering event. Triggering events yielded different bubble volumes (given in mL).
Particles may be from the sediment
Field mass transport data demonstrate that bubbles are capable of transporting particles to
a lake surface, but do not determine the particle origins. Microscope images of the particles
show biological material as well as material that may be inorganic (Figure 3). Many of the
biological structures in the particles appear to be the carapaces and head shields of Bosmina sp-p,
which have been found extensively in other freshwater lake sediments (Kerfoot 1995). To
further characterize the particulate matter, we compared metal concentrations to concentration
162
ratios from sediment samples and found that the ratios of arsenic, chromium, and lead to the bulk
sediment mass were within the range (but on average only 70%) of the ratios measured for
bubble-transported particulate matter (Figure 4). The fact that the average metals content per
particle mass for the particulate matter are lower than the sediment ratios could suggest that
bubbles preferentially transport fractions of the sediment that have lower metals concentrations
(e.g. organic matter such as the Bosmina spp remains).
400 urnm0u<~
Figure 3: Light microscope images of bubble-transported particulate matter showing many
apparent Bosmina spp. carapaces and head shields, along with material that may be
primarily inorganic in nature.
Our data strongly suggest that bubbles transport sediment directly from the bottom of the
lake to the water surface. This direct transport occurs over a 15 m deep water column, (and does
not preclude the possibility that bubbles can transport particles over significantly larger depths.)
This transport provides a direct chemical link between the sediment and surface waters, a link
that otherwise would not occur during months of stratification (sediment transport to the lake
surface could theoretically occur during unstratified periods, but transport would likely not be as
rapid as bubble transport). Furthermore, the rapid rise of bubbles limits the time available for
163
oxidation reactions, and suggests that sediment particulate matter may reach the lake surface in a
reduced state, with possible consequences for both toxicity and reactivity.
600
500
CU
E
*> 400 -
CU
0 A Sediment
r_ 300
Cu 0 Bubble Particles
200
E 100
0'
Arsenic Chromium Lead
Figure 4: Comparing mass of arsenic, chromium, and lead per kg of sediment (open
triangles) and bubble-transported particulate matter (solid circles). Standard deviation
scale similar to point size and therefore omitted for figure clarity.
Bubble column data indicates minimal particle shedding and scavenging
While field sampling data indicate that bubbles transport particles for distances in excess
of 15 m from sediments to the lake surface, field data do not indicate whether bubbles shed
particles as they rise. Particle shedding would contribute to more sediment mass mobilization
per bubble than would be predicted based on measurements from surface samples alone. To
quantify the importance of particle shedding, we conducted experiments in a 15 m tall bubble
column in our laboratory stairwell (as described in Methods). Bubble tests showed significant
particle transport even in the absence of the disturbance caused by dropping the anchor in field
164
experiments (though fluxes were approximately 4 times lower than field data, see discussion
below). Bubbling tests conducted at 5 m, 10 m, and 15 m showed no significant difference in
total particle transport rates (Figure 5). Data also showed that on average, greater than 98% of
particle mass transport was composed of particles larger than 5 pm, and less than 2% was
between 0.2 [tm and 5 ptm.
The lack of a trend between mass transport and depth indicates that particle shedding is
insignificant, though the large degree of variability between replicates could have masked
underlying physical processes. In particular, we riote that samples taken after raising or lowering
the sediment bed had markedly higher particle mass transport rates (noted by the extra black
circles around data points in Figure 5). This could be because transportable particles get
depleted during replicate bubbling events, and moving the sediment bed mixes the sediment.
This surmised influence of localized sediment conditions also supports our hypothesis based on
field data that heterogeneous sediment conditions contribute to variability in particle mass
transport.
165
0.14 r
0 0.12 0
a -0.1
0.08
CU 0.06 0E cue
0.04
0.02 -
0
0 5 10 15
Bubble release depth (m)
Figure 5: Transported particle mass per L of gas bubbled in the large bubble column, as a
function of bubble release depth. Solid circles represent samples where bubbles were
emitted from the sediment bed, diamonds represent samples where gas was bubbled
directly above the sediment bed. Hollow circles around solid circles denote samples with
recently-disturbed sediments.
Another possible explanation for the relative similarity of particulate mass transported
from different depths is that bubbles also scavenge particles as they rise, and shedding and
scavenging rates are similar. To quantify the bubble scavenging capacity, we compared the data
from 5 m and 10 m to samples gathered when gas was bubbled several centimeters above the
sediment. Particle mass scavenging represented approximately 10% of the mean particle
transport for 5 m and 10 m (Figure 5), indicating that while scavenging rates are non-zero, the
large majority of the particulate matter in these experiments was from the sediment.
Furthermore, as with the field samples, bubble column particulate matter element concentrations
166
Normal samples
Normal samples (recently
disturbed sediment)
Bubbles emitted
above sediment
are similar to the bulk sediment (see Figure SI in S.I.). Interestingly, contaminant concentrations
in the scavenged particulate matter follow a different trend (discussed in S.I.) and suggest that
lead, chromium, and arsenic are not evenly distributed among sediment particulate matter.
Contaminant levels in regular bubble column samples support our conclusion that bubbles are
primarily transporting sediment matter to the lake surface, despite the relatively deep water
column, and particle shedding appears to be minimal. Since particle shedding rates appear to be
minimal, particle transport measured at the surface of the lake are likely representative of total
bubble-facilitated particle transport rates.
Explaining differences between field and bubble column data
While both field data and bubble column data clearly demonstrate that bubbles are
capable of transporting particulate matter, average transport rates in the field were approximately
4 times higher than average transport rates in the large bubble column (0.07 0.03 mg/L for the
columns vs 0.26 0.2 mg/L for the field). While bubble diameters were different between the
two sample sets (see Figure 6), as we discuss below there did not appear to be a trend between
bubble diameter and particle transport over our data sets. Given the observed relationship
between recent sediment disturbance and higher particle transport rates, we think the higher field
transport rates can be explained (at least in part) by physical sediment disturbance from the
dropped anchor. Another possibility is that the sediment sample we used for the bubble column
happened to contain particulate matter less conducive to bubble transport. Yet another
possibility is that the dissolved and particulate matter present in Upper Mystic Lake water
enhances bubble particle flotation. Future work should focus on trapping naturally-emitted
167
methane bubbles from the lake to determine whether sediment disturbance caused the increased
transport rates in the field, or if observed rates are indeed typical of in situ lake conditions.
One further experimental factor to consider was that the city tap water used to fill the
bubble column is maintained around pH = 9 to minimize pipe corrosion. Typical pH levels in
the lake are between pH 6-7 (Senn 2001), and previous work on bubble particle flotation has
shown that increasing the pH can inhibit natural organic matter particle transport (Shi et al.
2017). To test the effect of pH on sediment flotation, we ran experiments in our small bubble
column with water pH starting at 9 (and dropping from 8.5 - 8.0 during the course of the
experiment due to atmospheric equilibration), and water pH starting at 6.5 (and remaining close
to 6.5 during the experiments). These experiments did not show a statistically significant
difference related to water pH (0.084 0.03 mg/mL for pH = 8-8.5 versus 0.062 1 0.02 mg/mL
for pH = 6.5). See Supplemental Information for figures.
Effect of bubble diameter
Despite the documented importance of bubble diameter in flotation systems (Yoon and
Luttrell 1989), we found no trend between average bubble diameter and particle mass transport
(Figure 6, note that we only have bubble diameter distributions for 5 of the field data points).
However, we do note that bubble diameter is not a static measurement and in deep water
columns bubbles will change volume significantly as they rise, which complicates the search for
a trend. Average bubble diameter varies between data sets, and we attribute the variation
between the small and large laboratory columns to differences between the sediment height
above the gas injection point (preliminary work suggested that sediment depth influenced bubble
volume). It is also important to note that while we are reporting average bubble diameters, the
168
sensor only measured on average 35% of the total flux due to a combination of flow anomalies
caused by the sample cup attached to the bubble sensor outflow and to previously-documented
limitations in sensor data storage and measurement capability at high bubble flux (Delwiche et
al. 2015a). Anecdotally, we did observe that large bubbles entering the sampling cup carried no
perceptible particles, whereas smaller bubbles had visible attached particles. Thus we conclude
that sediment particle transport rates are likely significantly lower for very large bubbles, but
more work is be needed to characterize transport rates for the range of bubble diameters
observed in natural aquatic systems.
0.8-
L
o 0.6 -
(U..
E cn * Small column
a)f 0.4 -a Large column
On 4 Field data
-E
C. 0.2
- 0'
2 3 4 5 6
Average bubble diameter at sediment (mm)
Figure 6: Bubble-transported particle mass (g/L) versus the average bubble diameter
(mm).
Implications for arsenic cycling
Our finding that methane bubbles transport particulate matter over depths of at least 15 m
could have important implications for chemical and biological cycles in natural systems.
Aquatic sediments are often contaminated with heavy metals and organic pollutants, so bubble
particle flotation could resuspend contaminants and thereby increase human exposure to legacy
contaminants. Upper Mystic Lake sediments contain high levels of arsenic, and ICP data from
169
our field samples shows that bubble transported particles contain arsenic at an average ratio of
100 tg/kg particle, and 8 ig arsenic per liter of gas bubbled (Figure 7). Bubble column arsenic
data showed similar ratios of particulate matter arsenic to sediment (Figure Si), and the pH
effect on arsenic transport appeared to be insignificant (though more information may be needed.
see Supplemental Information for more details). Given the average daily gas flux of 45
mL/m 2/day estimated in a 2016 field campaign for this general sampling area, this corresponds to
an estimated arsenic flux of 0.005 ptmol/m 2/day. This estimate likely represents the high end of
transport, because as discussed earlier the bubble-particle transport rates in the field may have
been artificially enhanced by the anchor disturbing the sediments.
1.2 o 1.5 2.5
03 AL~
EE CU 1 C0.8 -h A E
cu 1.5
A
C- 0.6 - -(n, A 0. 1 *
U, cc 0.5 1)
cc E A C,E0.4 - W ,.
E E2 - ' -0.5'
Cu 100 20 .00 2  10 20
Cu 0 *
3 0
H-0.2 0 -0.51 -0.51
0 100 200 0 100 200 0 100 200
Total bubble volume (mL) Total bubble volume (mL) Total bubble volume (mL)
Figure 7: Arsenic, chromium, and lead mass (in pg) transported versus the volume of each
sample (in mL, as measured at the lake surface). Error bars represent standard deviation
and are mostly smaller than marker size.
This estimate of expected daily arsenic flux can be compared with historical
measurements of arsenic cycling in Upper Mystic Lake. In 2000, Knauer et al. 2000 measured
170
approximate arsenic accumulations on the order of 0.5 Vmol/m 2/day in the epilimnion of Upper
Mystic Lake. This flux is two orders of magnitude larger than our estimate for bubble transported
arsenic of 0.005 [tmol/m2/day, indicating that bubble-arsenic transport may not be a major
contributor to total arsenic input to the epilimnion of Upper Mystic Lake. However, a
significant fraction of the As input to the epilimnetic waters of UML may be due to inflow from
the Aberjona River, estimated at 92 kg arsenic/year by Hemond 1995 (though this amount has
likely decreased due to ongoing remediation activities upstream from the lake). Similar flux of
bubble transported arsenic may therefore represent a larger fraction of epilimnetic input in other
lakes. In addition, arsenic-containing particles transported by bubbles are deposited directly at
the water surface, which makes it particularly susceptible to direct ingestion by swimmers. The
World Health Organization limits for arsenic in drinking water is 10 pg/L, and at the average
human water intake of 2.7-3.7 L/day (Dietary reference intakesfjbr water, potassium, sodium,
chloride, and sulfate, 2004) this corresponds to a daily arsenic intake of 27-37 fig. In this
context, the bubble-transported arsenic flux of 0.4 pg/m2/day is still quite small. This indicates
that for Upper Mystic Lake, bubble-facilitated arsenic transport likely does not contribute
significantly to total arsenic cycling or to concerning human exposure.
In addition, although bubble-facilitated transport does not appear to dominate arsenic
transport in Upper Mystic Lake, much higher ebullition rates have been reported elsewhere in the
world. Numerous hydropower reservoirs have been found to experience at least an order of
magnitude more methane ebullition than we measure at Upper Mystic Lake (Deemer et al. 2016).
Co-occurrence of high ebullition rates and contaminated sediment could lead to significant
bubble-facilitated contaminant cycling, and future efforts should be made to quantify
contaminant transport rates in contaminated, high bubble flux reservoirs.
171
Implications for organism cycling
In addition to transporting chemical contaminants, bubbles may also transport organisms
(or pathogens) directly from the sediment to the water surface, potentially introducing a novel
component to aquatic life cycles. For example, our field and laboratory particulate matter
samples contained many small black spheroids on the order of 0.5mm in diameter. Microscopic
images of these objects (Figure 8) indicate that they are likely ephippia, the protective cases on
diapausing eggs produced by zooplankton such as Daphnia. Ephippia can overwinter in lake
sediments or survive periods of desiccation. providing a seed bank to recolonize the water
column when favorable conditions return (Hairston 1996, Caceres and Tessier 1998). Diapause
termination cues often include physical stimuli such as temperature and photoperiod changes,
and more work needs to be done to determine if biological cues (ie: food abundance) also play a
role (Gyllstr6m and Hansson 2004).
Figure 8: Microscopic image of potential Daphnia ephippia found in bubble-
transported particulate matter. Scale bar is 400 pm.
172
A
While the cues that trigger egg hatching from ephippia vary between (and potentially
among) species and are not fully understood, it is reasonable to assume that the warmth and light
at the top of the lake could provide a favorable hatching environment for ephippia previously
stored in the dark, cold, anoxic sediments of the hypolimnia. Bubble transport may therefore
provide a previously-unstudied mechanism for diapausing eggs to hatch and contribute to
zooplankton growth in the water column. Additionally, we note that a variety of organisms
overwinter in lake sediments. For example, the potentially toxic cyanobacterium Microcystis can
be found in large amounts in the sediments of some lakes (Verspagen et al. 2004). Modeling
efforts have found that the absence of Microcystis recruitment from the benthos to the water
column could reduce the summer bloom by 50% (Verspagen et al. 2005). Recruitment is
thought to largely be a passive process caused by wind-induced resuspension, or bioturbation
(Verspagen et al. 2004). Bubble transport of Microcystis could represent a novel recruitment
pathway, though more work would be needed to determine if bubble transport could significantly
contribute to algal blooms. Future work could also address the possibility that sediment
resuspension triggered by bubble release (as opposed to direct transport by adsorption to the
bubble interface) may recruit organisms or pathogens to the water column.
CONCLUSIONS
Bubble-particle transport between the sediment and surface of Upper Mystic Lake is a
novel transport pathway capable of moving particulate matter upwards in a stratified water
column through which significant movement of sedimentary material would otherwise not be
expected. Bubbles appear capable of transporting arsenic-containing sediment particles over
depths exceeding 15 m, and do not appreciably shed particles as they rise. While bubble-
facilitated arsenic transport in Upper Mystic Lake appears minor from both the perspective of
173
total arsenic accumulation in the epilimnion and of human exposure, lakes with higher ebullition
flux may experience more significant chemical transport. Additionally, since bubbles also
appear to transport biological material, studies should focus on whether this represents a
meaningful transport pathway to aquatic species.
REFERENCES
Ahmed, N., & Jameson, G. J. (1985). The effect of bubble size on the rate of flotation of fine
particles. International Journal ofMineral Processing, 14, 195-215.
Aldrich, C., & Feng, D. (2000). Removal of Heavy Metals from wastewater effluents by
biosorptive flotation. Minerals Engineering, 13(10-11), 1129-1138.
Aller, J. Y., Kuznetsova, M. R., Jahns, C. J., Kemp, P. F. (2005). The sea surface microlayer as a
source of viral and bacterial enrichment in marine aerosols. Journal ofAerosol Science, 36,
801-812.
Blanchard, D. C. (1975). Bubble Scavenging and the Water-to-Air Transfer of Organic Material
in the Sea. In R. E. Baier (Ed.), Applied Chemistry at Protein Jnterfaces (pp. 360-387).
American Chemical Society.
Bussmann, I., Damm, E., Schliiter, M., & Wessels, M. (2013). Fate of methane bubbles released
by pockmarks in Lake Constance. Biogeochemistry, 112(1-3), 613-623.
Cdceres, C. E., & Tessier, A. J. (2004). To sink or swim: Variable diapause strategies among
Daphnia species. Limnology and Oceanography, 49(4 part 2), 1333-1340.
Chanton, J. P., Martens, C. S., & Kelley, C. a. (1989). Gas transport from methane-saturated,
tidal freshwater and wetland sediments. Limnology and Oceanography, 34(5), 807-819.
Cheng, T.-W., & Holtham, P. N. (1995). The Particle Detachment Process in Flotation. Minerals
Engineering, 8(8), 883-891.
Clift, R., Grace, J.R., Weber, M.E. (1978). Bubbles, Drops, and Particles, Academic Press: New
York.
Cunliffe, M., Engel, A., Frka, S., Gasparovid, B. Z., Guitart, C., Murrell, J. C., Salter, M., Stolle,
C., Upstill-Goddard, R. Wurl, 0. (2013). Sea surface microlayers: A unified
physicochemical and biological perspective of the air-ocean interface. Progress in
Oceanography, 109, 104-116.
174
Dai, Z., Fornasiero, D., & Ralston, J. (2000). Particle-bubble collision models--a review.
Advances in Colloid and Interfaice Science, 85, 231-56.
Dai, Z., Dukhin, S., Fornasiero, D., & Ralston, J. (1998). The inertial hydrodynamic interaction
of particles and rising bubbles with mobile surfaces. Journal of Colloid and Interface
Science, 197, 275-292.
Deemer, B., Harrison, J., Li, S., Beaulieu, J., DelSontro, T., Barros, N., Bezerra-Neto, J., Powers,
S., Dos Santos, M., Vonk, J. (2016). Greenhouse Gas Emissions from Reservoir Water
Surfaces: A New Global Synthesis Manuscript. BioScience, 66(11) 949-964.
Delsontro, T., McGinnis, D. F., Sobek, S., Ostrovsky, I., & Wehrli, B. (2010). Extreme methane
emissions from a Swiss hydropower reservoir: contribution from bubbling sediments.
Environmental Science & Technology, 44(7), 2419-2425.
Delwiche, K., Senft-Grupp, S., Hemond, H. (2015a). A novel optical sensor designed to measure
methane bubble sizes in situ. Limnology and Oceanography: Methods, 13, 712-721.
Delwiche, K., Hemond, H. (2015b). An enhanced bubble size sensor for long-term ebullition
studies. Limnology and Oceanography: Methods,. 15, 821.
Delwiche, K., Hemond, H. (2017). Methane bubble size distributions, flux, and dissolution in a
freshwater lake. Environmental Science and Technology, 51, 13733-13739.
Dietary Reference Intakes for Water, Potassium, Sodium, Chloride, and Sulfate. (2004). Panel on
Dietary Reference Intakes for Electrolytes and Water, Standing Committee on the Scientific
Evaluation of Dietary Reference Intakes. National Academies Press.
Feng, D., & Aldrich, C. (1999). Effect of particle size on flotation performance of complex
sulphide ores. Minerals Engineering, 12(7), 721-731.
Guidelines for Drinking-water Quality, Fourth Edition. (2011). World Health Organization.
Gyllstr6m, M., & Hansson, L.-A. (2004). Dormancy in freshwater zooplankton: Induction,
termination and the importance of benthic-pelagic coupling. Aquatic Sciences, 66(3), 274-
295.
Hairston, N. G. J. (1996). Zooplankton Eggs as Biotic Reservoirs in Changing Environments.
Limnology and Oceanography, 41(5), 1087-1092.
Hassanzadeh, A., Kouachi, S., Hasanzadeh, M., & Qelik, M. S. (2017). A new insight to the role
of bubble properties on inertial effect in particle-bubble interaction. Journal of Dispersion
Science and Technology, 38(7), 953-960.
175
Hassas, B. V., Caliskan, H., Guven, 0., Karakas, F., Cinar, M., & Celik, M. S. (2016). Effect of
roughness and shape factor on flotation characteristics of glass beads. Colloids and Surfaces
A.: Physicochemical and Engineering Aspects, 492, 88-99.
Hemond, H. (1995). Movement and distribution of arsenic in the Aberjona watershed.
Environmental Health Perspectives, 103, 35-40.
Joyce, J., & Jewell, P. W. (2003). Physical Controls on Methane Ebullition from Reservoirs and
Lakes. Environmental and Engineering Geoscience, IX(2), 167-178.
Kerfoot, W. C. (1995). Bosmina remains in Lake Washington sediments: Qualitative
heterogeneity of bay environments and quantitative correspondence to production.
Limnology and Oceanography, 40(2), 211-225.
Knauer, K., Nepf, H. M., & Hemond, H. F. (2000). The production of chemical heterogeneity in
Upper Mystic Lake. Limnology and Oceanography, 45(7), 1647-1654.
Krasowska, M., & Malysa, K. (2007). Kinetis of bubble collision and attachment to hydrophobic
solids: Effect of surface roughness. Int. J. Miner. Process, 81, 205-216.
Leifer, I., & Patro, R. K. (2002). The bubble mechanism for methane transport from the shallow
sea bed to the surface: A review and sensitivity study. Continental Shelf Research, 22,
2409-2428.
Lin, S. H., & Lo, C. C. (1996). Treatment of textile wastewater by foam flotation. Environmental
Technology, 17(8), 841-849.
Liss, P.S. (1975). Chemistry of the sea surface microlayer. In Chemical Oceanography, ed. J. P.
Riley, G. Skirrow, 2:193. London, New York, San Francisco: Academic 2 ",d ed.
Liu, L., Wilkinson, J., Koca, K., Buchmann, C., & Lorke, A. (2016). The role of sediment
structure in gas bubble storage and release. Journal of Geophysical Research:
Biogeosciences, 121, 1-14.
Maeck, A., Hofmann, H., & Lorke, A. (2014). Pumping methane out of aquatic sediments -
ebullition forcing mechanisms in an impounded river. Biogeosciences, 11, 2925-2938.
McGinnis, D. F., Greinert, J., Artemov, Y., Beaubien, S. E., & Wfiest, A. (2006). Fate of rising
methane bubbles in stratified waters: How much methane reaches the atmosphere? Journal
of Geophysical Research, 111, C09007.
McLinn, E. L., Stolzenburg, T. R. (2009). Ebullition-facilitated transport of manufactured gas
plant tar from contaminated sediment. Environmental Toxicology and Chemistry, 28(11),
2298-2306.
176
Min, Q., Duan, Y. Y., Peng, X. F., Mujumdar, A., Hsu, C., & Lee, D. J. (2008). Froth flotation of
mineral particles: Mechanism. Drying Technology, 26, 985-995.
Nriagu, J. 0., Lawson, G., Wong, H. K. T., & Cheam, V. (1996). Dissolved Trace Metals in
Lakes Superior, Erie, and Ontario. Environmental Science & Technology, 30(1), 178-187.
Ostrovsky, I., McGinnis, D. F., Lapidus, L., & Eckert, W. (2008). Quantifying gas ebullition
with echosounder: the role of methane transport by bubbles in a medium-sized lake.
Limnology and Oceanography: Methods, 6, 105-118.
Pan, K., & Wang, W.-X. (2012). Trace metal contamination in estuarine and coastal
environments in China. Science of the Total Environment, 421-422, 3-16.
Ralston, J., Fornasiero, D., & Hayes, R. (1999). Bubble-particle attachment and detachment in
flotation. Int. J. Miner. Process, 56, 133-164.
Rodrigues, R. T., & Rubio, J. (2007). DAF-dissolved air flotation: Potential applications in the
mining and mineral processing industry. International Journal of Mineral Processing, 82,
1-13.
Rubio, J., Souza, M. L., & Smith, R. W. (2002). Overview of flotation as a wastewater treatment
technique. Minerals Engineering, 15, 139-155.
Scandella, B. P., Pillsbury, L., Weber, T., Ruppel, C., Hemond, H. F., & Juanes, R. (2016).
Ephemerality of discrete methane vents in lake sediments. Geophysical Research Letters,
49(9), 4374-4381.
Schfigerl, K. (2000). Recovery of proteins and microorganisms from cultivation media by foam
flotation. In T. Scheper (Ed.), Advances in biochemical engineering/Biotechnology (Vol. 68,
pp. 191-233). Springer-Verlag.
Senn, D. B. (2001). Coupled Arsenic, Iron, and Nitrogen cycling in Arsenic-Contaminated
Upper Mystic Lake. PhD Dissertation, Massachusetts Institute of Technology.
Shi, Y., Yang, J., Ma, J., & Luo, C. (2017). Feasibility of bubble surface modification for natural
organic matter removal from river water using dissolved air flotation. Front. Environ. Sci.
Eng., 11(6).
Spliethoff, H. M., & Hemond, H. F. (1996). History of Toxic Metal Discharge to Surface Waters
of the Aberjona Watershed. Environmental Science & Technology, 30(1), 121-128.
Tao, D. (2005). Role of Bubble Size in Flotation of Coarse and Fine Particles-A Review.
Separation Science and Technology, 39(4), 741-760.
177
Taylor, K. G., & Owens, P. N. (2009). Sediments in urban river basins: a review of sediment-
contaminant dynamics in an environmental system conditioned by human activities. Journal
of Soils and Sediments, 9, 281-303.
Varadharajan, C., & Hemond, H. F. (2012). Time-series analysis of high-resolution ebullition
fluxes from a stratified, freshwater lake. Journal of Geophysical Research, 117, G02004.
Vashisth, S., Bennington, C. P. J., Grace, J. R., & Kerekes, R. J. (2011). Column Flotation
Deinking: State-of-the-art and opportunities. Resources, Conservation and Recycling, 55,
1154-1177.
Verspagen, J. M. H., Snelder, E. 0. F. M., Visser, P. M., Huisman, J., Mur, L. R., & Ibelings, B.
W. (2004). Recruitment of benthic Microcystis (Cyanophyceae) to the water column:
Internal buoyancy changes or resuspension? Journal of Phycology, 40(2), 260-270.
Verspagen, J. M. H., Snelder, E. 0. F. M., Visser, P. M., J6hnk, K. D., Ibelings, B. W., Mur, L.
R., & Huisman, J. (2005). Benthic-pelagic coupling in the population dynamics of the
harmful cyanobacterium Microcystis. Freshwater Biology, 50(5), 854-867.
Viana, P. Z., Yin, K., & Rockne, K. J. (2012). Field measurements and modeling of ebullition-
facilitated flux of heavy metals and polycyclic aromatic hydrocarbons from sediments to the
water column. Environmental Science & Technology, 46, 12046-12054.
Wallace, G. T., Loeb, G. I., & Wilson, D. F. (1972). On the Flotation of Particulates in Sea
Water by Rising Bubbles. Journal of Geophysical Research, 77(27), 5293-5301.
Wurl, 0., & Obbard, J. P. (2004). A review of pollutants in the sea-surface microlayer (SML): a
unique habitat for marine organisms. Marine Pollution Bulletin, 48, 1016-1030.
Yoon, R. H., & Luttrell, G. H. (1989). The Effect of Bubble Size on Fine Particle Flotation.
Mineral Processing and Extractive Metallurgy Review, 5(1-4), 101-122.
178
SUPPLEMENTAL INFORMATION:
Methane bubbles transport sediment and arsenic to a lake surface
Contaminant ratios in bubble column particulate matter and sediment
Ratios of arsenic and chromium to total mass of bubble-transported particulate matter are
in line with ratios found in the sediment, while lead levels are lower in the bulk sediment sample
(Figure SI). The differences for lead may be because sediment used in the bubble columns was
acquired during a different sampling campaign than the sediment used for the digestion.
Additionally, Figure SI shows results for the two scavenged particle samples. Neither sample
contained chromium, one sample had arsenic at levels consistent with regular column samples,
and both samples had high lead concentrations. The different scavenging rates among
contaminants are quite interesting, and suggest that lead, chromium, and arsenic may be
associated with particles having different affinities for bubble scavenging. More work should be
done to understand contaminant distribution among sediment particulate matter and how this
affects transport.
179
1500
CD
E
CD 1000 A Sediment sample
Y Large column
100 O 0 Small column pH =8CU
0 Small column pH=6.5
500 - 8 o Scavenged sample
E 13
0
Arsenic Chromium Lead
Figure Sl: Comparing mass of arsenic, chromium, and lead per kg of particulate matter
for laboratory experiments (large bubble column, small column with pH = 8, small column
with pH = 6.5) and bulk sediment. Data from trials with the same water pH are lumped
together. Open triangles are ratios measured in sediment sample. Open squares are from
samples where gas was emitted directly above the sediment and particulate matter was
therefore scavenged from the water column. Error bars representing standard deviations
are of same approximate size as symbols, and are therefore omitted for clarity.
Arsenic transport in bubble columns
The amount of transported arsenic in the bubble column was independent of bubble
release depth (Figure S2), and reflects trends observed in total particulate mass transported with
depth (Figure 5 in the main text). Arsenic transport by gas emitted above the sediment rather
than into the sediment represented 1% and 12% of arsenic transport from 5m and 1 Orn,
respectively. As with particle mass transport, this indicates that the large majority of arsenic in
180
bubble-transported particulate matter originates in the sediment even in water columns exceeding
15 m depth.
20-
- 15 -
.0
A Normal samples
M Denotes recently
0)10- disturbed sediment
Bubbles emitted
above sediment
O A
E -
0
0 5 10 15
Bubble release depth (m)
Figure S2: Transported arsenic per L of gas bubbled in the large column, as a function of
bubble release depth. Solid triangles represent samples were gas was emitted in to the
sediment bed, and diamonds are samples were gas was emitted directly above the sediment
bed. Large circles indicate recently mixed sediment conditions. Error bars representing
standard deviations are of same approximate size as symbols, and are therefore omitted for
clarity.
pH effect on arsenic transport
Figure S3 shows the relationship between water pH and bubble-particle transport. The
two data sets are statistically equivalent (p>0.05) when considering all values. However, if the
data points associated with recently-mixed sediment are excluded, than particle transport at pH
181
8-8.5 is significantly higher than transport at pH 6.5. Data associated with arsenic transport at
different show a similar trend (Figure S3, right panel), with pH having no significant difference
except when excluding the recently mixed data points. More work would be needed to
determine if these trends continued under repeated sampling, especially since existing literature
suggests that particle transport should be lower at higher pH (Shi et al. 2017). We note that after
correcting for background concentrations, arsenic transport on particles between 0.2 and 5 pm is
negligible, therefore we present arsenic quantities on particles larger than 5 ptm.
() 0.2
U16-
0.18 -)
-0 14 -0.16 -. 0cc0
0.14 - 12 -
CO
1 0.12 0)10 -
0.1 - 8 -E*-0.08 -
S-E 6 -
0.06 - )
76 C,,
Q 4 -0.04 -
0.02 . 0 2-
o 0 0}- Water pH-6.5 Water pH-8 Water pH-6.5 Water pH-8
Figure S3: Effect of water pH on total transported particle mass in g/L (left) and arsenic in
ptg/L (right). Larger black circle around data points indicates recently mixed sediment
conditions. Error bars represent standard deviation calculated from ICP data uncertainty.
Supplemental information references
Shi, Y., Yang, J., Ma, J., & Luo, C. (2017). Feasibility of bubble surface modification for natural
organic matter removal from river water using dissolved air flotation. Front. Environ. Sci.
Eng., 11(6).
182
Chapter 6: Summary and future work
SUMMARY
The work presented in these chapters introduces a unique tool to measure methane bubble
sizes in situ, increases our understanding of methane bubble emission patterns and dissolution in
a freshwater lake, and begins to explore the importance of methane bubble particle transport to
contaminant cycling in a lake. Given the importance of methane to global climate change, and
the increased recognition that aquatic systems produce significant amounts of methane bubbles,
this work represents a meaningful advancement in a critical environmental field.
Bubble size sensor
Methane bubble diameter has a large impact on dissolution rate, which in turn influences
the amount of methane entering the Earth's atmosphere from aquatic systems. Therefore,
understanding methane bubble diameter distributions is fundamental to understanding how
methane partitions between the water column and the atmosphere. Prior to this work, measuring
bubble volumes was resource intensive and only done for short sampling campaigns. The novel
optical bubble size sensor we developed makes it more cost effective to measure methane bubble
volumes, and for the first time allows us to track methane bubble volumes over long time
periods.
The sensor itself is built around an inverted glass funnel that is encased in a waterproof
housing. The funnel intercepts rising methane bubbles, and a custom circuit board mounted to
the funnel stem uses pairs of LEDS and light detectors and a microprocessor to record timing
information associated with each bubble. Timing data combined with channel geometry leads to
183
accurate bubble volume estimations, and laboratory calibrations further improve the data fit. We
also document through extensive laboratory testing the current sensor limitations, including
issues related to hardware processing speeds, data storage limitations, data irregularities caused
by flow obstructions in the funnel stem, and channel geometry limits that put a lower bound on
the detectable bubble size (0.15 cm diameter). Some of these issues could be addressed with
future hardware and software improvements, which we will discuss below.
We deployed this sensor in the field with a detachable gas collection chamber, a separate
data storage buoy filled with batteries, and an extension funnel to increase the bubble collection
area. Four months of data from 8 deployed sensors show that the sensors measure on average
86% of the total ebullition volume passing through the sensor (see Figure 12 in Chapter 2).
Given the complexity of field systems and the inherent difficulty with long-term underwater
deployments (ie: algae build-up), the sensors perform well and are a valuable addition to bubble
volume measurement techniques.
Ebullition patterns and bubble dissolution
We deployed eight bubble size sensors at two depths within Upper Mystic Lake near
Boston, MA for four months. Collecting bubble size data at two depths allowed us to track the
change in bubble diameter distribution between 16 m and the lake surface. We compared this
diameter evolution with model results from a popular existing bubble dissolution model, and
found that the model adequately predicts our recorded change in bubble diameter. We then used
the data, coupled with dissolved methane profiles taken over the course of the season, to estimate
that dissolving bubbles contribute approximately 10% of the total dissolved methane
184
accumulation in the hypolimnion of Upper Mystic Lake. Dissolving bubbles also represent at
most 15% of the diffusive air-water methane flux from the epilimnion of Upper Mystic Lake.
Bubble diameter distributions measured over the course of the season did not appear to
change from summer to fall, but there was spatial variability. Two sensors deployed at the
northern most extent of the sampling area experienced substantially less ebullition and much
larger bubble diameters, pointing to relatively large variability in the ebullition process on the
scale of tens of meters within the lake. The other six sensors experienced similar flux, but bubble
diameter distributions were still statistically different which indicates that localized sediment
conditions are also important in controlling bubble diameter.
Bubble particle transport
Bubbles effectively scavenge and transport particles in water columns, a process that has
been used extensively in industry since the 1900s to separate chemicals of interest from bulk
solutions. While bubbles have also been found to transport particulate matter in marine
systems, to date very little work has focused on the possibility of methane bubble particle
transport in freshwater systems. Anecdotally, we routinely see that bubbles bursting on the
surface of Upper Mystic Lake leave a visible ring of particles, and this observation led us to
study methane bubble particle transport. We used a modified bubble size sensor to sample
bubbles and their associated particles in Upper Mystic Lake, and found a linear relationship
between total bubble volume and mass of associated particles. Particulate matter concentrations
of arsenic, chromium, and lead, as well as microscope images strongly indicated that the
particulate matter originated in the sediment, even in water depths exceeding 15 m deep. We
185
therefore conclude that bubbles transport particles upwards through portions of a stratified water
column that would not otherwise exchange material.
This sediment particle transport could have important implications for chemical cycling
in natural systems, particularly in systems with contaminated sediments. Since Upper Mystic
Lake sediments are contaminated with arsenic due to historical upstream chemical
manufacturing facilities, we estimated the arsenic cycling due to methane bubbling. We found
that potential arsenic transport was two orders of magnitude lower than previously-reported
arsenic accumulations in the epilimnion of the lake. While bubble-facilitated arsenic cycling
may be minor at the ebullition levels typical for Upper Mystic Lake, many lakes and reservoirs
are known to experience an order of magnitude more ebullition. Therefore, there remains a
need to study bubble-particle transport in systems with high flux and contaminated sediments.
Additionally, bubbles may be transporting organisms or pathogens from the sediment to the
surface, and the importance of this potential transport to aquatic life cycles should be studied.
FUTURE WORK
Our field work showed that in general, the bubble size sensors accurately measure
methane bubble volumes and total cumulative fluxes passing through the sensor. However, on
several occasions the sensors measured flux much lower than the actual flux, and in one case a
sensor stopped transmitting data. Future work should focus on improvements to sensor
hardware and software that will make them more robust and reliable in the challenging
conditions of long-term field deployment. Sensors need to be able to handle lake flora growing
in the glass funnel stem, lake fauna crawling or swimming through the funnel, and other
heterogeneous conditions inherent to natural systems. These conditions can diminish the
186
sensor's ability to detect bubbles, or lead to false detection of non-bubble events. The
microprocessor code accommodates some of this variability by continually updating
background detection thresholds, but more work could be done to improve these algorithms.
These improvements could narrow the gap between sensor-measured gas flux and actual gas
flux.
We also envision numerous hardware improvements that would make sensor field
deployment less labor intensive. Currently the sensor is powered by rechargeable lithium ion
battery packs that must be replaced every 2-4 weeks. Future work could include redesigning the
buoy to wirelessly transmit data and adding solar panels to eliminate the need for routine field
trips. Sensor circuitry could also be modified to contain a small video camera to film bubble
passage through the sensor in field conditions. Such footage may contribute to improving bubble
sensing algorithms, and may also shed light on the lake fauna that like to traverse the funnel.
Our work on bubble-particle transport in a lake setting presents many interesting questions
for future work:
" How much do particle transport rates vary based on different sediment characteristics?
* Do bubbles emitted naturally from the sediment have similar transport rates to artificially
triggered bubbles?
" Over what depth can bubbles transport particles?
* What fraction of bubble-transported particles adhere to the air/water interface at the
surface of the lake, and how long do the particles remain here?
* In contaminated water bodies with high ebullition flux, does bubble-facilitated particle
transport contribute to significant contaminant accumulation at the water surface?
187
* Do bubbles transport organisms or pathogens, and could this transport represent a
meaningful component of life cycles?
Investigating these questions would undoubtedly lead to more interesting unknowns, and we see
this as an exciting area for future work.
BROADER IMPACTS
Interest in methane ebullition from aquatic systems continues to grow as improved
measurement techniques show that methane ebullition comprises a significant (and potentially
growing) component of atmospheric methane. Given the importance of methane to climate
change and the overwhelming thread that climate change poses to human civilizations, it is
critical that we understand the processes affecting methane ebullition flux to the atmosphere.
However, the spatial and temporal heterogeneity of methane ebullition makes it a difficult
process to study, and extensive field campaigns are necessary to adequately estimate ebullition
flux. Measuring bubble sizes in situ traditionally requires even more equipment than simply
measuring total flux, so any improvements to bubble measurement techniques will greatly assist
future field efforts. Our simple, economical bubble size sensor provides a valuable tool that will
allow researchers to expand the understanding of methane fate and transport in aquatic systems.
The sensors have already been deployed in projects spanning three continents, and continue to be
used by our colleagues in the methane community. We anticipate that data from ongoing
international projects will provide valuable insight into the sediment mechanics of bubble
release, and may help improve methods to estimate ebullition rates from aquatic systems. These
outputs should increase our ability to accurately predict methane emissions, though many
unanswered questions about the fate and transport of methane from aquatic systems will remain.
188
The particle transport work presented here draws on the established fields of industrial
foam flotation and marine particle transport, and applies these well-known bubble transport
principles to a new system, freshwater lakes. Given the pervasive extent of sediment
contamination and high ebullition flux, we believe there is a possibility that bubble particle
transport could contribute meaningfully to chemical or organismal transport. However, much
work remains to be done to understand the complexities of particle transport by methane bubbles
in natural ecosystems.
189
190
APPENDIX
At: Additional information for sensor functioning and calibration
Sensor tunctioning details
The LEDs on the sensor circuit board emit infrared light at 940 nm to a phototransistor
with peak sensitivity at 940 nm (see table A8-1 for specific part numbers). When the bubble
passes in front of the LED, the change in refractive index between the water and the gas cause
the light to bend away from the phototransistor. This results in a decrease of light, as can be seen
in Figure 3 in Chapter 2.
Calibration curve
The calibration curve presented in Figure 8 of Chapter 4 shows that the sensor can
estimate bubble volume very well from 0.02 mL to 1 mL. Assessment below 0.02 mL is less
accurate, as shown in Figure Al-I and discussed below. Sensor performance between I mL and
1.8 mL is still reasonable, though noise increases as volumes approach 1.8 mL. Anything
beyond 1.8 mL is deemed outside the range of the sensor. The larger the bubble, the higher the
buoyancy force, and therefore the faster the acceleration through the funnel stem. The trailing
meniscus for bubbles beyond 1.8 mL moves so quickly that the sensor is unable to accurately
measure the timing of the meniscus passage, and therefore the variability in volume estimates
increases. Bubbles accelerate through the glass funnel stem, presumably because their rise
velocity is reduced at the constriction where they first enter the funnel stem, and then buoyancy
force causes them to accelerate as they rise past the detectors. By averaging the four volume
191
estimates we are assuming that this acceleration is linear, but as can be seen in Figure 8 of
Chapter 4, this assumption appears to break down as bubble volume increases past 1 mL.
+ 1st Lower Volume
Estimate
0.05
2' 2nd Lower Volume
E Estimate
E / 1st Upper Volume
.3 Estimate
- o2nd Upper Volume
Estimate
C,,
-1:1 Line
x Average
-_Small Bubble Linear
0.005 Fit
0.001 0.01 0.1
Actual Volume (mL)
Figure Al-1: The lower end of the calibration curve presented in Chapter 4, Figure 8,
plotted on a log-log scale. Small volume estimates show the increased variability that arises
when bubbles are smaller than the funnel stem diameter.
The fact that the raw sensor estimates are slightly larger than the 1:1 line is likely a
function of the fact that the radius of the elongated bubble within the funnel stem is actually
smaller than the funnel stem itself due to a thin layer of water film between the bubble and glass
tube wall. This film of water helps the bubble move upwards within the funnel stem, and we can
plot the potential thickness of this film by calculating the internal bubble radius that would be
required to produce the volume over-estimates we see for bubbles from 0.05 to 1 mL (Figure Al-
192
2). This thickness appears to be around 0.2 mm. Above 1 mL we can no longer use this simple
method to estimate film thickness because the flow regime has changed. Below 0.05 mL the
water film will be thicker because bubbles do not fill the entire glass stem diameter. We also
note that an additional explanation for the deviation from the 1:1 line in the mid-volume range
could be that the bubbles do not form exact cylinders, and the top of the bubble in particular is
likely to have a curved shape. However, these issues are accounted for with the calibration
curve.
0.16
0.12
E 
0.08
00
0.04
~ 0.5 i1 1.5** 2.5
Bubble Volume (mL)
Figure A1-2: Estimated wetted perimeter thickness (cm) between bubble and glass
funnel stem, plotted against the bubble volume (mL).
Calibration curve error estimates
The calibration curve shown in Figure 8 of Chapter 4 has a small amount of associated
error. The curve for bubbles below 1.0 mL is y = (1.17 0.009)x + 0.02 0.0038 (r2 = 0.99,
mean square error = 1.9 * 1 0A-4). The curve for bubbles between 1.OmL and 1.8mL is y = (0.45
0.045)x + 0.73 0.064 (r2 = 0.93, mean square error = 0.0012). We can use this uncertainty to
193
estimate uncertainties in the bubble diameters and cumulative volume sums presented throughout
this thesis. We use the error propagation formula generically described as:
= dw) 2 dw )2 V dw )2dx V dy dz
where w = x + y + z and and x, y, and z are independent variables (Peters et al. 1974).
When we propagate errors for volume and diameter estimates, we find that standard deviation
error bars are smaller than the size of points on data plots.
Small bubble limitations
To make the sensor' performance for small bubbles easier to see, we have included the
lower end of the calibration curve shown in Figure 8 of Chapter 4 and plotted it on a log-log plot
(Figure Al-1). The multiple volume estimates per bubble are relatively similar down to a bubble
volume of 0.02 mL (3.4 mm diameter). For bubbles below 0.01 mL (2.6 mm diameter), bubble
volume estimates have much larger variability. We assume this is because these very small
bubbles can move horizontally within the funnel stem (which has an inner diameter of 4 mm),
and this horizontal movement will increase variability in the sensor measurements. While a
smaller diameter tube would likely improve these small-volume estimates, it would also increase
the likelihood that particulate matter or living organisms in the lake could obstruct the flow path.
We also note that while recent studies have highlighted the importance of microbubbles in
methane transport from freshwater lakes (McGinnis et al. 2015), these bubbles have a diameter
smaller than 1mm and are therefore below the lowest reliable detection limit of the sensor.
194
References
McGinnis, D. F., Kirillin, G., Tang, K. W., Flury, S., Bodmer, P., Engelhardt, C., Casper, P.,
Grossart, H. P. (2015). Enhancing surface methane fluxes from an oligotrophic lake:
Exploring the microbubble hypothesis. Environmental Science and Technology, 49(2), 873-
880.
Peters, D., Hayes, J., and Heiftje G. (1974). Chemical Separation and Measurements. Theory
and Practice ofAnalytical Chemistry. Saunders: Philadelphia, PA.
195
A2: Fine scale temporal spacing for bubbling events
One of the major advantages of the bubble size sensor over other existing techniques to
measure ebullition flux is the sensor's ability to capture fine-scale temporal dynamics for
bubbling events. For example, Figure A2-1 shows four separate bubble event clusters (chosen
approximately randomly from field data) spanning 5-10 seconds each. For these four clusters,
the event begins with the largest bubble, followed by numerous smaller bubbles. This pattern
has numerous possible explanations. It could a sign that larger bubbles are necessary to open a
conduit in the sediment before smaller bubbles can be released. The pattern could be caused
because larger bubbles typically rise faster, thus leading to the largest bubble being measured
first. Alternatively, several smaller bubbles could be coalescing to form the larger, leading
bubble. While more work would be needed to determine if this pattern persists throughout the
bubble diameter data and to explore potential causes, this pattern shows the level of fine-scale
information the bubble size sensor is capable of capturing.
196
10 August 30 2016 bubble stream
8
6
4
2
0
-0
E
E~
E
M-
5
S
S
4,
4 eptember 7 2016 bubble stream
0
8
6
4
2
0
0
0
Q0g
1 eptember 18 2016 bubble stream
8
6
4
8
0
0
..
.0
L.
EE-
E
0
20
01
li,.4?
"9.
October 19 2016 bubble stream10r
0
06-
4-
0
* 0.
0
eg
.4,t~ .4,
Figure A2-1: Close temporal spacing on 4 bubble release clusters. Event clusters span 5-10
seconds, and for these 4 events it appears that the largest bubble in a cluster comes first.
197
E
E
M
E
E.
E
4-
0s
2
A3: Evidence for phantom midge interference in bubble size sensor
As discussed in Chapter 3, we find evidence in our field data that phantom midges pass
through the bubble size sensors. These midges collect at the gas/water interface within the gas
collection chamber, and during sampling events we remove them from the traps (see Figure A3-1
for midge image). Sensors also record a lot of "falsestart" data that may be attributed (at least in
part) to phantom midge migration through the funnel stem. This migration is a nocturnal event,
and for many of the sensors we see an increase in "falsestart" events during night hours (Figure
A3-2). Future sensor adaptions could involve creating a zooplankton sensor by adding features
such as a camera.
Figure A3-1: Microscope image of phantom midge found in bubble size sensor bulk gas
trap (picture is a composite of two separate images).
198
Sensor2
1800 
.
1600
1400
1200
1000
8004
600
350
300
250
200
'150
100
-C
-c
4-J
a)
4-.J
4-J
U,
0
E
z
0..
000
1 1
S 5 10 15
.0 800 [
700 .
0
0
600
500
20 25
Sensor3
0
00.0 0
0
0
00
0 00
0 00
0.
0
0 5 10 15 20 2
Sensor6
00 0
0 5 10 15 20 2
SensorS
0 .0
gO 0
S . 0 .
900
800
700
600
500
400
5
5
0 5 10 15 20 25
350
300
250
200
150
100~
200
150
100
5C
00
00 0
0
0.
0
0
.. .
e * 0
0 5 10 15 20 2
Sensor4
. 0
0
o 0 1 202
Sensor7
0
0 5 10 15 20 25
Sensor9
0
*0
0
. 0
00-
0
0 0
0 0
0
0
.
0
0
0 5 10 15 20 25
Hour of Day
Figure A3-2: Number of "falsestart" events recorded by each bubble size sensor per hour
of day (with hour 0 being 12:00 am). Note the different scales for each y-axis.
199
1000
900
800
700
600
500
1000
800
600
Sensori
I
5
5
A4: Spatial heterogeneity in bubble flux
As discussed in Chapter 5, the bubble size sensor deployment location chosen for the
2016 field campaign was chosen in part to increase the amount of gathered data. Preliminary
work during the 2015 and 2016 field season demonstrated that a small portion of the lake tended
to have relatively homogeneous and high ebullition rates, and we chose this area for bubble
sensor deployment to maximize our data collection. However, this area is not reflective of the
broader ebullition heterogeneity across the lake. In 2008, Varadharajan deployed 8 bubble traps
across the entire lake, and found average daily ebullition fluxes from 1 to 80 ml/m 2/day, with an
overall average of 28 ml/m 2/day. In contrast, aside from sensor 3, the flux range measured in
this work was 15 to 49 ml/m 2/day, with an overall average of 38 ml/m 2/day. Therefore,
estimates of methane accumulation attributable to bubble dissolution are likely an overestimate
for the anoxic portions of Upper Mystic Lake. This further highlights the relatively minor
contribution of dissolving bubbles to methane accumulation that we discussed in Chapter 5.
We note that fluxes per meter squared were determined by dividing the flux measured by
each sensor by the collection funnel surface area of 0.2 in2 . Since bubbles could have
experienced lateral advection during rise, the flux measured by one sensor may reflect flux from
a sediment patch some distance away. Fricker and Nepf (2000) estimated that seiche-induced
bed velocities were on the order of 1 cm/sec. For a typical bubble rise velocity of 20 cm/sec, this
would correspond to a potential lateral bubble displacement of 75 cm over 15 m of bubble rise.
References:
Fricker, P.D., Nepf, H.M. Bathymetry, stratification, and internal seiche structure. J. Geophys.
Res. 2000, 105, C6, 14237-14251.
Varadharajan, C. Magnitude and spatio-temporal variability of methane emissions from a
eutrophic freshwater lake. Ph.D. Dissertation, Massachusetts Institute of Technology,
Cambridge, MA, 2009.
200
A5: Turkey Hill weather station details
The Turkey Hill weather station used for wind speed estimates in Chapter 5 is a personal
weather station reporting weather data through the website www.wunderground.com (weather
station identification KMAARLIN21). This site is to the west of Upper Mystic Lake, the
northwest of Turkey Hill (peak elevation around 107 m), and the self-reported weather station
elevation is 78 m. The weather station appears to be in a residential area, and may be affixed to a
residence (the exact station location is not provided). The station is approximately 400 m from
the Turkey Hill peak.
201
A6: 2016 field data (including bulk methane and nitrogen content)
Table A6-1: Sensor ID, gas collected per sampling event, approximate midge count per
sampling event, and methane/nitrogen contents in bulk gas.
Sensor
Sensor Depth
ID m)
1 16
1 16
1 16
1 16
1 16
1 16
1 16
1 16
1 16
2
2
2
2
2
2
2
2
2
16
16
16
16
16
16
16
16
16
4 16
4 16
4 16
4 16
4 16
4 16
4 16
4 16
4 16
4 16
6 3
Sample
Date
8/15/2016
8/24/2016
9/2/2016
9/13/2016
9/22/2016
10/3/2016
10/13/2016
10/27/2016
11/13/2016
8/15/2016
8/24/2016
9/2/2016
9/13/2016
9/22/2016
10/3/2016
10/13/2016
10/27/2016
11/13/2016
7/26/2016
8/15/2016
8/24/2016
9/2/2016
9/13/2016
9/22/2016
10/3/2016
10/13/2016
10/27/2016
11/13/2016
8/15/2016
Methane
content
(%)
Nitrogen
content
(%)
35
35
32
32
33
36
38
Gas
collected
from
collection
chamber
(mL)
108
102
126
115
70
38
29
119
166
142
46
140
114
54
66
28
86
158
58
133
83
127
105
45
58
20
116
168
85
Midges
collected
1
2
2
0
5
0
0
~10
0
~10
5
~10
~10
0
0
0
~10
~10
3
-15
5
0
-10
~10
14
-20
-20
20-30
midges
0
Days
since
previous
collection
10
9
9
11
9
11
10
14
17
20
9
9
11
9
11
10
14
17
14
20
9
9
11
9
11
10
14
17
10
67
65
69
68
67
65
62
76
73
79
79
76
77
70
73
76
74
76
75
70
71
64
70
24
27
22
20
24
23
32
32
25
26
23
25
29
29
37
202
32
8/24/2016
9/2/2016
9/13/2016
9/22/2016
10/3/2016
10/13/2016
10/27/2016
11/13/2016
7/26/2016
8/15/2016
8/24/2016
9/2/2016
9/13/2016
9/22/2016
10/3/2016
10/13/2016
10/27/2016
11/13/2016
7/26/2016
8/15/2016
8/24/2016
9/2/2016
9/13/2016
9/22/2016
10/3/2016
10/13/2016
10/27/2016
11/13/2016
8/15/2016
8/24/2016
9/2/2016
9/13/2016
9/22/2016
10/3/2016
10/13/2016
10/27/2016
11/13/2016
46
118
98
44
43
20
97
132
69
87
41
133
115
36
36
26
120
109
12
5
3
32
75
14
26
1.5
67
46
57
29
73
65
40
33
13
128
110
9
9
11
9
11
10
14
17
14
20
9
9
11
9
11
10
14
17
14
20
9
9
11
9
11
10
14
17
10
9
9
11
9
11
10
14
17
64
72
72
58
58
46
37
28
29
40
40
43
67
61
65
70
71
63
64
54
6
38
41
36
29
30
36
37
35
6 off chart
45
60
37
45
54
47
60
61
58
61
48
41
52
40
41
40
40
52
203
58
52
40
61
77
A7. Isotope data from methane bubble samples
During the summer 2016 sampling campaign we sent a subset of bubble sensor bulk gas
samples to the UC Davis Stable Isotope Facility. Results are presented in Table A7- 1:
Table A7-1: Carbon 13 isotope readings for bulk gas samples.
Sensor Methane Fraction
Sensor ID Depth (m) 613 CVPDB (volume %)
TI 16 -68.66 72
T2 16 -70.05 58
T4 16 -70.91 52
T7 3 -70.35 51
T9 3 -71.27 42
204
A8. Bubble sensor supply list
Table A8-1: Parts supplier and product numbers for many of the specific products
required to build a complete bubble size sensor and data buoy.
Item
Clear PVC Pipe
Stainless steel hook
Solid PVC rod
Socket male - threaded male NPT fitting
Bolts (stainless steel)
Hex bushing adapter
Barbed tube fitting
Socket female -threaded female NPT fitting
Eye bolt (stainless steel)
Large black funnel
Carabiner (stainless steel)
Standoff
Washer for standoff
Funnel eyebolt
Eyebolt locking nut
MCBH4F-SS Subconn connector
MCDLS-F locking sleeve plus snap ring
MCIL4F //I 00ft 18-4 //MCIL4M
MCBH4M-SS Subconn connector
Hex nut (7/16"-20)
Washer (7/16")
Expanding foan for buoys
West Marine epoxy hardener
PVC ball valve
Pipe fitting for ball valve (stainless steel)
O-ring (Dash no. 236, 1/8" width)
Arduino Pro Mini
Adafrlit MicroSD card breakout board
Standoff for SD breakout board
Micro SD card
Device
Gas collection trap
Gas collection trap
Gas collection trap
Gas collection trap
Gas collection trap
Gas collection trap
Gas collection trap
Gas collection trap
Gas collection trap
Funnel
Funnel
Funnel
Funnel
Funnel
Funnel
Data cable
Data cable
Data cable
Data cable
Data buoy/Bubble
sizer
Data buoy/Bubble
sizer
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Supplier
McMaster
McMaster
McMaster
McMaster
McMaster
McMaster
McMaster
McMaster
McMaster
Texas Hunter
McMaster
McMaster
McMaster
McMaster
McMaster
Macartney
Macartney
Macartney
Macartney
McMaster
McMaster
USComposites
West Marine
McMaster
McMaster
McMaster
Sparkfun
Adafrulit
Digikey
Adafruit
Product #
49035K85
6043T6
8745K65
4880K653
93190A536
4880K825
5463K439
4880K83
9489T51 1
LFBF
3716T51
91115A282
90107A007
9489T56
95878A300
customl item
custom item
custom item
custom item
92676A465
92916A206
#0804
#207-SA
4757K1 I
51205K471
9557K79
DEV-I 1113
254
492-1047-ND
1294
205
Adafruit DS3231 clock breakout board
Standoff for clock
Plastic screw for clock standoff
2pin surface mount wire housing for
batteries
4 pin surface mount wire housing
Battery wire housing connector plug
Battery connectors
4 pin wire housing, male
Connecting sockets for male clip
4 pin wire housing, female
Connecting sockets for female clip
Arduino Pro Mini
USB cable for Arduino
4 pin wire housing, male
FTDI breakout board
Eye bolt (stainless steel)
Female headers
Male headers
Data storage chip
Phototransistor
LED
IC socket
Pipe plug
D-rings
Sizer body (4" pipe coupling)
Hose clamp
U-bolts
M2 screws
M2 nuts
M2 washers
O-ring (Dash no. 244, 1/8" width)
O-rings, 7mm x 2.5 mm
Pyrex glass funnel
Data buoy Adafruit
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Data buoy
Bubble sizer/Data
buoy
Bubble sizer/Data
buoy
Bubble sizer/Data
buoy
Bubble sizer/ Data
buoy
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Bubble sizer
Mouser
Digikey
Digikey
Mouser
Digikey
Digikey
Mouser
Digikey
Digikey
Digikey
Sparkfun
Sparkfun
Mouser
Sparkfun
McMaster
Sparkfun
Sparkfun
Digikey
Digikey
Digikey
Digikey
McMaster
McMaster
McMaster
McMaster
McMaster
Mcmaster
Mcmaster
Mcmaster
McMaster
McMaster
Cole Parmer
3013
761-2056-256-N
H534-ND
WM4055-ND
538-43650-0413
WM2880-ND
WM1271-ND
43645-0400
WM7082CT-ND
WMI 1857-ND
WM I 127CT-ND
DEV-11113
CAB-i 1301
43645-0400
DEV-09716
9489T511
PRT-00l 15
P RT-00 116
24LC 1025-I/P-ND
1080-1 167-ND
1080-109 1-ND
A24807-ND
4880K851
34905T91
4880K78
501 1T37
2936T65
901 16A009
94150A305
90965Al 10
9557K243
9262K727
EW-34560-52
206
A9. Bubble sizer assembly instructions
The following instructions were provided to colleagues deploying bubble sensors that I had sent
them:
Assembling Sensor Housing for Deployment
1. Push a glass funnel up through the base of the bubble sensor housing. Remove the
housing top and side, leaving just the base. Make sure funnel is inserted as far as
possible.
2. Wrap the ~Icm pieces of tubing onto the glass funnel stem. Tubing pieces should be as
far apart as the holes in the sensor circuit board, and should be positioned around the
center of the funnel stem (several centimeters each from top and bottom).
3. Bolt the sensor circuit board onto the funnel stem using the U-bolts. Make sure the
plastic tubing is cushioning the funnel stem both where it touches the U-bolt and where it
touches the circuit board. Make sure circuit board is securely attached to funnel
stem.
4. Re-install the pipe coupling that forms the body of the sensor housing. If necessary,
spread a little vacuum grease on to the large O-ring before pushing the pipe coupling on.
5. Put a desiccant pack into the sensor housing.
6. Gently pull the glass funnel downwards until the circuit board is resting against the base
of the sensor housing. Take the housing top and attach the wire clip to the circuit
board. Then push the housing top on to the pipe coupling. Gently maneuver the glass
funnel until the funnel stem can be pushed up through the small O-ring in the center of
the housing top. (Be careful not to get any vacuum grease on the glass funnel while you
are handling it).
Assembling Sensor Attachments for Deployment
1. Use a screw driver to loosen the hose clamp on the outside of the sensor housing. Push
the hose clamp downwards until it is a few centimeters from the bottom of the pipe
coupling.
2. Place the sensor housing on to the black extension funnel, making sure to match the
grooves in the sensor housing with the posts on the funnel.
3. Attach the carabiners to the D-rings (adjusting D-rings to align with the eyehooks on the
extension funnel).
4. Loosen the hose clamp again, and now pull the hose clamp up as high as possible until
the metal rope is taut - tighten hose clamp.
5. Screw the gas collection chamber on to the top of the sensor housing.
6. Take the set of 3 ropes tied together, and pass one rope through each eye-hook at the top
of the gas collection chamber. Tie these ropes to the D-rings. Adjust knots so sensor
hangs vertically when suspended from hook.
7. When attaching the data cable to the bulkhead fitting on the sensor, make sure to grease
the female fitting ( spread enough grease on the female fitting to cover the holes, but no
need to force more grease into the holes)
207
Buoy Instructions
1. Plug the buoy circuit board in to the wire clip inside the buoy.
2. Make sure there is a blank SD card in the Micro-SD breakout board.
3. Place one battery pack in each of the 3 grooves on the battery sled. The 2 batteries
closest together at the base of the sled should be clipped in to the small circuit board
designed to put the batteries into parallel (these two batteries MUST have similar
voltages!). Wires from this board will then plug in to the outer-most clip on the main
buoy circuit board. The 3rd battery will plug in to the inner clip on the main buoy circuit
board.
Note #1- Only connect the batteries when the buoy is connected to the data cable!
Otherwise it would be easy to short the metal pins, potentially making the batteries
explode.
Note #2 - The two battery packs wired in parallel MUST have similar voltages to
prevent battery explosion.
4. Slide the battery sled to the bottom of the buoy. Make sure the buoy contains a desiccant
pack.
5. Open the air release valve, push the buoy top on, and close the air release valve.
6. When attaching the data cable to the bulkhead fitting on the buoy, make sure to grease
the female fitting ( spread enough grease on the female fitting to cover the holes, but no
need to force more grease into the holes)
Collecting Data
1. Open the air release valve on the buoy and remove the buoy top.
2. Unplug the batteries.
3. Remove the micro-SD card and replace it with a blank micro-SD card.
4. Replace batteries if needed.
5. Reconnect the batteries and record the time batteries were reconnected. (Note - Once
the batteries have been reconnected, do not remove the SD card! If card is accidently
removed while the batteries are connected, you must reinsert the SD card and push the
small reset button on the Arduino.)
6. Put the top back on the buoy and close the air-release valve.
Downloading Data Manually (if this becomes necessary)
1. Install the Arduino software. I have been using version 1.0.6. and had some issues when
I tried updating, so use version 1.0.6:
(https://www.arduino.cc/en/Main/OldSoftwareReleases#previous
2. Install the FTDI drivers necessary for communication between the computer and
Arduino https://learn.sparkfun.com/tutorials/how-to-install-ftdi-drivers/windows---quick-
and-easy.
3. Disconnect the buoy circuit board from the wire clip, and attach the small circuit board
for downloading data.
4. Make sure the double-sided 6-pin piece is in the female header on the circuit board. Push
the red FTDI breakout board on to the 6 pins with the top of the board pointing
towards the battery clip. Connect the red USB cable to the computer and the FTDI
board.
208
5. Open the Arduino program and make sure the proper serial port is selected under
Tools>Serial Port.
6. Open the serial monitor by clicking on the magnifying glass icon in the upper right
corner.
7. Make sure baud rate is set to 57600 using the pull down menu in the lower right corner of
the serial monitor window.
8. Type any letter in to the command line. This should cause Menu text to appear.
9. To download data, type 'A'. Cut and paste this data into a .txt file. Then delete data
from the sensor by typing 'X', then 'Y'.
209
A10: Bubble size sensor data processing code
Please contact Kyle Delwiche (k.delwiche@alum.mit.edu) for latest version.
%% MATLAB program to process, filter, and plot data from the bubble size
sensor described in Delwiche and Hemond, 2017
% Written by Kyle Delwiche
% Please contact Kyle Delwiche for future code improvements
% (k.delwiche@alum.mit.edu)
% Prior to running this program, upload .csv data file using Matlab's "Import
Data" tool. Make sure to:
% - upload .csv files only
% - select Numeric Matrix under "Output Type",
% - under column delimeters/more options choose "treat delimeters as one"
% - start selection at first row of numerical data and not above
% - save file after uploading
%% PROCESS DATA **********************
%%~ *********************************************************
clear all; close all; cdc;
****************** Import Data**********************************
filename = 'filename'; %CHANGE THIS, file to be processed
exportfilename = 'exportfilename'; %CHANGE THIS, place to save the
processed file
prompt = 'Enter time batteries were connected (m/d/yyyy hh:mm:ss AM/PM)'
%prompt user to input sensor start time and date
StartTime = input(prompt,'s'); %this date will be used to
determine calendar date/time of bubble events
StartTime = datenum(StartTime);
a = load(filename); %load data (will load as a structure)
b = struct2cell(a); %convert to cell
Fielddata = cell2mat(b); %convert to matrix
%***********Deal with "falsestart" data structure (future Arduino updates
could restructure Falsestart data so this bit is
unnecessary)************************
%********Code could likely be improved to make this faster***********
%%"I****** rearrange data set one row at a time************
if size(Fielddata,2)>14 %If there is "falsestart" data the matrix will have
more than 14 columns
jj=0;
210
x = 15; %number of data points for each line of data +1
while sum(isfinite(Fielddata(:,x))) > 0 %continue loop as long as there
is "falsestart" data remaining
spacer = 0; %starts off at 0, spacer will accomodate table
referencing as Fielddata grows
jj=jj+1;
while isnan(Fielddata(jj,x)) == 0 %for all rows that extend beyond
the usual #x-1 values, i.e. rows with falsestartdata
%split the data matrix into rows above and below the current row
topoftable = Fielddata(1:jj-1+spacer,:);
bottomoftable = Fielddata(jj+spacer+1:end,:);
%pull apart the current row one "falsestart" section at a time
middleoftablel =
[Fielddata(jj+spacer,1:6),zeros(1,8),NaN*ones(1,(size(Fielddata,2)-(x-1)))];
% pull out the first "falsestart" data point in the set
middleoftable2 = [Fielddata(jj+spacer,7:end),NaN*ones(1,6)];
% tack on NaN's to complete the line
Fielddata =
[topoftable;middleoftablel;middleoftable2;bottomoftable]; %put everything
together in to table with 1 more row
spacer = spacer+1;
end
end
Fielddata = Fielddata(:,l:x-1); %remove all columns beyond
these are full of NaNs)
column x,
end
Fielddata(isnan(Fielddata))
in case any NaN's remain
0; %replace remaining NaN's with zeros, just
%************Reformat data structure for multi-bubble
events****** ********************************** ******
%First pull out certain rows in data table
ClockTime = Fielddata(:,1);
since sensor was powered on)
Bubbles = Fielddata(:,2);
Errors = Fielddata(:,3);
with bubble event
Det2StartTime = Fielddata(:,7);
detector 2 saw start of bubble
Det2EndTime = Fielddata(:,9);
detector 2 saw end of bubble
Det3StartTime = Fielddata(:,11);
detector 3 saw start of bubble
Det3EndTime = Fielddata(:,13);
detector 3 saw end of bubble
%Time of event (in msec
%# of bubbles in event
%error code associated
%time (in usec) that
%time (in usec) that
%time (in usec) that
%time (in usec) that
%Add missing data for multi-bubble events (Time, bubble #, and error code
%are only stored for the first bubble in an "event")
for k=l:size(Fielddata,l) %Fielddata rows
211
if ClockTime(k) == 0
ClockTime(k) = ClockTime(k-1);
Bubbles(k) = Bubbles(k-1);
Errors(k) = Errors(k-1);
end
end
index = find(Errors(:)==0);
Errors(index,:) = 9;
with zeros later on
%Fine all bubble events with error code "0"
%convert these to "9" so I can delete rows
%Remove all bubble events with incomplete information
combinedtable =
horzcat(ClockTime,Bubbles,Errors,Det2StartTime,Det2EndTime,Det3StartTime,Det3
EndTime);
combinedtable(any(-combinedtable,2), ) []; %remove any row in table
containing a zero. These correspond to events with incomplete info (ie;,
det2endtime' wasn't detected)
-*************Calculate bubble
ClockTime = combinedtable(:,l);
Bubbles = combinedtable(:,2);
Errors = combinedtable(:,3);
Det2StartTime = combinedtable(:,4);
Det2EndTime = combinedtable(:,5);
Det3StartTime = combinedtable(:,6);
Det3EndTime = combinedtable(:,7);
StartStart = Det3StartTime - Det2StartTime;
3 each seeing start of bubble (usec)
EndEnd = Det3EndTime - Det2EndTime;
3 each seeing end of bubble (usec)
Det2Time = Det2EndTime - Det2StartTime;
front of detector 2 (usec)
Det3Time = Det3EndTime - Det3StartTime;
front of detector 3 (usec)
spacing = 0.5588;
%Time between detectors 2 and
%Time between detectors 2 and
%length of time bubble was in
%length of time bubble was in
%detector spacing in cm
VelStartStart = spacing./(StartStart./1000000);
in cm/sec
VelEndEnd = spacing./(EndEnd./1000000);
in cm/sec
%velocity (lower estimate)
%velocity (upper estimate)
%Bubble length calculations
Length2StartStart = VelStartStart .* Det2Time/1000000; %bubble length
calculated using det2 time and start-start velocity, cm
Length2EndEnd = VelEndEnd .* Det2Time/1000000;
Length3StartStart = VelStartStart .* Det3Time/1000000;
Length3EndEnd = VelEndEnd .* Det3Time/1000000;
AverageLength = (1/4)*(Length2StartStart + Length2EndEnd + Length3StartStart
+ Length3EndEnd);
212
%Bubble volume calculations
area = 0.1294; %cross-sectional area of
funnel stem, cmA2
Volume2StartStart = Length2StartStart * area; %volume in mL
Volume2EndEnd = Length2EndEnd * area;
Volume3StartStart = Length3StartStart * area;
Volume3EndEnd = Length3EndEnd * area;
AverageVolume =
(Volume2StartStart+Volume2EndEnd+Volume3StartStart+Volume3EndEnd)/4;
%Implement calibration curve
smallbub = find(AverageVolume < 1.1); %index of volumes below 1.1mL
threshold
bigbub = find(AverageVolume >= 1.1); %index of volumes above 1.1mL
threshold
%******Future code versions may use predict() function with regression model
%************ ******* ******
AverageVolume Calibrated(smallbub,1) = (AverageVolume(smallbub) -
0.0213)./1.174; %calibration parameters published in Delwiche and Hemond,
2017 (DOI: doi: 10.1002/lom3.10201)
AverageVolumeCalibrated(bigbub,1) = (AverageVolume(bigbub) - .4501)./0.7324;
%calibration parameters published in Delwiche and Hemond, 2017 (DOI: doi:
10.1002/lom3.10201)
%*************Calculate Coefficients of Variation for start-start
measurements and end-end measurements*************
%*************These will be used later to discard data with high CV
values**********
stdevSS = std([Volume2StartStart,Volume3StartStart],0,2);
meanSS = mean([Volume2StartStart,Volume3StartStart],2);
stdevEE = std( [Volume2EndEnd,Volume3EndEnd] ,0,2);
meanEE = mean( [Volume2EndEnd,Volume3EndEnd] ,2);
CV SS = stdevSS./meanSS;
CVEE = stdevEE./meanEE;
% Create date in string format
DateTime = StartTime + ClockTime./(1000*60*60*24); %ClockTime values are
stored in milliseconds and datenum puts things in terms of dates, so divide
ClockTime by # of milliseconds in a day
DateTimeStr = datestr(DateTime);
%Combine data, remove any
data = horzcat(DateTime,
AverageVolumeCalibrated,
data(data<O)=0;
will have negative volume
in = find(any(-data,2));
data(in, : ) = [];
DateTimeStr(in,:) = [];
remaining zeros
ClockTime, Bubbles,Errors,
CVSS, CVEE);
%convert all negative values to 0 (tiny bubbles
due to calibration equations)
%index of zero values
%remove all zero values
%remove zero values from date string as well
%Save data
213
column header = {'Date', 'Matlab DateTime','ClockTime (ms since Arduino
started)', 'Bubbles (# bubbles in event)', 'Errors (refer to Delwiche and
Hemond 2017 for error codes)', 'AverageVolumeCalibrated (mL)', 'CVSS',
'CVEE'};
xlswrite (exportfilename, cellstr(columnheader), 'Processed','Al');
xlswrite (exportfilename, cellstr (DateTimeStr), 'Processed', 'A2');
xlswrite(exportfilename, data, 'Processed','B2'); %export processed data to
table
clearvars -Except data exportfilename
%% *******************FILTER PROCESSED DATA*************************
%%9 *****************************************************************
%filter according to critera in Delwiche and Hemond, 2017 (DOI:
10.1002/lom3.10201)
ClockTime = data(:,2);
DateTime = data(:,l);
Volume = data(:,5);
Diameter = 20 * ((3 * Volume)./(4*3.1415)) .^ (1/3);
CV SS = data(:,6);
NumBubbles = data(:,3);
Data Compilation = [DateTime, ClockTime, Diameter, Volume, CVSS,
NumBubbles];
CV1 = 0.75; % CV cut off for bubbles smaller than 0.02 mL
CV2 = 0.15; % CV cut off for bubbles between 0.02 and 0.06 mL
CV3 = 0.075; % CV cut off for bubbles between 0.06 and 0.15 mL
CV4 = 0.05; % CV cut off fur bubbles larger than 0.15 mL
% CV cut offs presented in Delwiche and Hemond, 2017 (DOI: %
10.1002/lom3.10201)
% remove large CVs with variable ranges based
deleteCV1 = find(DataCompilation(:,5) > CV1
0.02);
DataCompilation(deleteCV1,:) = []; %remove
volume below 0.02mls
deleteCV2 = find(DataCompilation(:,5) > CV2
& DataCompilation(:,4) < 0.06);
DataCompilation(delete CV2,:) = []; %remove
volume between 0.02 and .06 mLs
deleteCV3 = find(DataCompilation(:,5) > CV3
& DataCompilation(:,4) < 0.15);
Data Compilation(delete CV3,:) = []; %remove
volume between 0.06 and .15 mLs
on bubble volume
& DataCompilation(:,4) <
all rows with CV over 0.75 and
& DataCompilation(:,4) >= 0.02
all rows with CV over 0.15 and
& DataCompilation(:,4) >= 0.06
all rows with CV over 0.075 and
214
deleteCV4 = find(DataCompilation(:,5) > CV4 & DataCompilation(:,4) >= 0.15
DataCompilation(deleteCV4,:) = []; %remove all rows with CV over 0.05 and
volume over 0.15mL
% Remove bubbles over 1.543mLs
VolumeCutoff = 1.543; %maximum measured bubble size allowed (corresponds
with actual value of 1.8mL, see calibration curve for upper limit)
deletevol = find(DataCompilation(:,4)>VolumeCutoff); %index for rows to
delete
DataCompilation(deletevol,:) = []; %remove all rows with volumes larger
than volume cutoff
volume-sum = sum(DataCompilation(:,4)) %total bubble volume after filters
are applied
columnheader = {'DateTime (Matlab format)', 'ClockTime (usec since Arduino
started)', 'Diameter (mm)', 'Volume (mL)', 'CVSS', 'NumBubbles (# bubbles
in event)'};
xlswrite(exportfilename, cellstr(column header), 'Processed and
trimmed', 'Al');
xlswrite(exportfilename, DataCompilation, 'Processed and trimmed','A2');
%% ******** PLOT DATA *********************
bins for displaying diameter data = [-1:0.25:121;
binsfor displayingvolumedata = [0:0.05:1.1];
fontsize = 12;
fontsize axis = 8;
fontsizeletter = 20;
fontname = 'Calibri';
%********************* Plot histogram**************************************
figure
subplot (1,2,1)
hold on
realDiameter = hist(DataCompilation(:,3),binsfordisplayingdiameterdata);
%create histogram of bubble sizes
realDiameter = realDiameter./length(DataCompilation).*100; %convert
to frequency percentage
bar(binsfordisplayingdiameterdata,realDiameter,l,'FaceColor', [0.8, 0.8,
0.8])
xlabel('Bubble Diameter(mm) ', 'FontSize',fontsize, 'FontName', fontname)
ylabel ('Normalized Frequency', 'FontSize', fontsize, 'FontName', fontname)
title('Bubble Size Distribution');
xlim([0,13])
ylim([0,15])
set(gca,'XTick', 0:1:13, 'YTick', 0:5:25);
***** Plot Time Series******************************************
215
subplot (1,2,2)
dates = DataCompilation(:,l);
diameters = DataCompilation(:,3);
%dates
%diameters in mm
scatter(dates, diameters, 15, 'filled', 'MarkerEdgeColor', [0 0 0],
'MarkerFaceColor', [0 0.4 0.9], 'LineWidth', 0.7)
ylabel('Bubble Diameter (mm)','FontSize',fontsize, 'FontName', fontname)
datetick('x',15)
title('Bubble Release Timing');
ylim([0,15])
xlim([min(dates) max(dates)]); %set x-axis limit to encompass the whole
data set
216
All. Arduino code for bubble size sensor
Please contact Kyle Delwiche (k.delwiche@alum.mit.edu) for latest version.
/*** **************** ************** ********************************************
Title: Bubble Sizing Detector Firmware
Version: 1.1
Date Started: 01 December 2015
Description: This code is the firmware for the bubble size detector
Authors: Schuyler Senft-Grupp (skysg@alum.mit.edu) and Kyle Delwiche
(k.delwiche@alum.mit.edu)
License: This work is licensed under the Creative Commons Attribution-NonCommercial-
ShareAlike
4.0 International License. To view a copy of this license, visit
http://creativecommons.org/licenses/by-nc-sa/4.0/.
* *** *********** ** **************** *** ***** ******************
* LIBRARIES - For instructions on how to install libraries, see
http://arduino.cc/en/Guide/Libraries
#include <Wire.h>
#include <digitalWriteFast.h>
#include <BubbleDetector.h>
#include <BDStorage.h>
// Arduino library for 12C communications
* PIN DECLARATIONS
**** ** ****** ** ********* ******** *************************** **********
217
uint8_t ledOne 4;
uint8_t ledTwo 3;
uint8_t ledThree = 2;
uint8_t detOne AO;
uint8_t detTwo A1;
uint8_t detThree = A2;
// Detector 1 LED
// Detector 2 LED
// Detector 3 LED
// Detector 1 ADC input
// Detector 2 ADC input
// Detector 3 ADC input
* GLOBAL CONSTANTS
const uint8_t LOOKFORSTART = 0;
const uint8_t LOOKFOREND = 1;
const unsigned long MAXTIMEOUT = 5000000; // Maximum time allowed for data-collection
sequence
const uint16_t MINTHRESHOLD = 30;
to trigger data collection sequence
const uint8_t NOISE MULTIPLILER = 2;
max/min backgound readings
// The minimum change in ADC value required
// The amount to multiply the difference betwen
// for estimate of background noise threshold
const uint16_t BKGDSTORAGE = 1000;
in background readings
* **V** ** *L* D*
* GLOBAL VARIABLE DECLARATIONS
// Store every BKGDSTORAGE ADC reading
unsigned long timeOfLastTimeout = 0;
uint8_t numberTimeouts = 0;
uint8 t numberFalsestarts = 0;
const uint8_t MAXNUMTIMEOUTS = 3;
const uint8_t MAXNUMFALSESTARTS = 3;
unsigned long timeoutClockStart = 0;
//Each of the detectors has a state variable. The detector is always either
//waiting to see the start of a bubble (LOOKFORSTART) or waiting to see
218
const
const
const
const
const
const
//the end of a bubble (LOOKFOREND).
uint8_ t detOneState = LOOKFORSTART;
uint8_t detTwoState = LOOKFORSTART;
uint8_t detThreeState = LOOKFORSTART;
//Structures defined in BubbleDetector.h file
//They store estimates of background readings and noise
eventinfo eventinfo; // A struct to store information about bubble event
bubbledata bData [MAX NUMBERBUBBLES]; // An array of structs to store bubble data
backgrounddata detOneBkgd; // A struct with information on detector 1 background
ADC value
backgrounddata detTwoBkgd; // A struct with information on detector 2 background
ADC value
backgrounddata detThreeBkgd; // A struct with information on detector 3 background
ADC value
ADCdata detOneADC; // A struct with information on detector I ADC values
during bubble event
ADCdata detTwoADC; // A struct with information on detector 1 ADC values
during bubble event
ADCdata detThreeADC; // A struct with information on detector 1 ADC values
during bubble event
uint8_t detOneNumBubbles; // Number of bubbles Detector 2 sees during a bubbling
event
uint8_t detTwoNumBubbles; // Number of bubbles Detector 2 sees during a bubbling
event
uint8_t detThreeNumBubbles; // Number of bubbles Detector 3 sees during a bubbling
event
uint8_t mostBubblesSeen; // The higher of either detTwoNumBubbles or
detThreeNumBubbles
uintI6_t bkgdCounter; II Variable to keep track of when to record background ADC
reading
* HELPER VARIABLES - these are generic variables but
* declaring them here gives better estimate of RAM usage
uint16_t adcReading = 0;
uint8_t i; // Used for counting in loops
uint8 t numBubblesInTube; // Used to keep track of number of bubbles in tube at once
uint8 t nextState; // Will either be LOOKFORSTART or LOOKFOREND
uint32-t tempTime;
219
/********************************************************************** *******
* PERFORM SETUP TASKS
void setupo{
// Initialize pins
pinMode(ledOne, OUTPUT);
pinMode(ledTwo, OUTPUT);
pinMode(ledThree, OUTPUT);
pinMode(detOne, INPUT);
pinMode(detTwo, INPUT);
pinMode(detThree, INPUT);
// uncomment if pin 3 on memory chip is not hard wired to 5V
// pinMode(13, OUTPUT);
// digitalWrite(13, HIGH);
//setup serial for debugging purposes
Serial.begin(57600); // Baud communication rate
analogReference(INTERNAL); // Use the built-in reference voltage
InitializeBkgdStructs(; /Initialize background data structs.
InitializeADCDatao; // Initialize ring buffers holding ADC values
bkgdCounter = 0; I/ This counter increments every time detector 1 does not sense a
bubble.
bdstorage.begino; //this may be necessary to begin storing data on chip if I don't have the
serial.print statement below
/******************* ** ********* ************** **** *** **** ***********
** ****** ** *
* BEGIN LOOKING FOR BUBBLES
void loopo{ // Create loop that checks if a bubble has entered glass tube
220
// Detector 1 is initially looking for start of
bubble
//clear start and end time arrays after each detection event
for(i = 0; i < MAXNUMBERBUBBLES; i++){
COMPLETE
//ABOUT 100 MICROSECONDS TO
bData[i].d2stime 0;
bData[i].d2sval 0;
bData[i].d2etime 0;
bData[i].d2eval 0;
bData[i].d3stime 0;
bData[i].d3sval 0;
bData[i].d3etime 0;
bData[i].d3eval 0;
while(detOneState == LOOKFORSTART)
if(Serial.availableO > 0){ / Allows
computer if connected
talkToComputero; // talkToC
}
//Take ADC reading
digitalWriteFast(ledOne, HIGH); // Tu
// Before the first bubble enters the tube
.or communication between Arduino and
)mputer defined below
rn on the LED for detector 1
//Check if it is time to take a background reading and if so turn on the other detector LEDs
if(bkgdCounter == BKGDSTORAGE) I
digitalWriteFast(ledTwo, HIGH); // Turn on the LED for detector 2
digitalWriteFast(ledThree, HIGH); // Turn on the LED for detector 3
}
//this is a throw away reading that takes ~100 microseconds
analogRead(detOne);
delayMicroseconds(200); // De
tempTime = micros(); // reco
adcReading = analogRead(detOne);
UpdateADC(&detOneADC, adcReading);
lay necessary to allow photocells to stabilize
rd time the ADC reading is going to be taken
// Read the ADC for detector 1
// Compare ADC reading to background readings. If sensor determines there is a bubble in the
tube, detOneState will switch to LOOKFOREND which will break while loop
221
detOneState = LOOKFORSTART;
detOneState = CheckForBubble(&detOneBkgd, detOneADC.totalADC,
LOOKFORSTART);
// If no bubble has been detected (detOneState is still LOOK_FOR_START), check if it is
time to store a background reading
if(bkgdCounter >= BKGDSTORAGE && detOneState == LOOKFORSTART) {
//Read each detector and update its background value
UpdateBkgd(&detOneBkgd, adcReading);
adcReading = analogRead(detTwo);
UpdateBkgd(&detTwoBkgd, adcReading);
UpdateADC(&detTwoADC, adcReading);
adcReading = analogRead(detThree);
UpdateBkgd(&detThreeBkgd, adcReading);
UpdateADC(&detThreeADC, adcReading);
bkgdCounter = 0;
analogRead(detOne);
// Reset background counter
// Throw away reading for detector one
bkgdCounter = bkgdCounter + 1; // Increment background counter
delay(1); //this used to be at beginning of while loop, I moved it 160708
to try to increase turn-around time between bubble detection events (ie, if deti is seeing a bubble
as soon as the arduino finishes storing data and enters this while loop
* A BUBBLE HAS NOW BEEN DETECTED ENTERING THE GLASS TUBE
eventlnfo.eventTime = milliso;
// Turn on all three LEDs
digitalWriteFast(ledOne, HIGH);
digitalWriteFast(ledTwo, HIGH);
digitalWriteFast(ledThree, HIGH);
delayMicroseconds(100);
// Begin by recording the event time
// Delay necessary to allow phototransistors to stabilize
222
}
numBubblesInTube = 1;
detected
// Initially set to I because the first bubble has been
/Initialize detTwo and detThree state and the number
// of bubbles they've seen
detTwoState = LOOKFOR START;
detThreeState = LOOKFORSTART;
detTwoNumBubbles =0;
detThreeNumBubbles = 0;
mostBubblesSeen = 0;
// Get the current time so that we know when
timeoutClockStart = microso;
tempTime = timeoutClockStart;
analogRead(detOne); // Throw
adds a small delay for everything to stabilize)
analogRead(detTwo);
analogRead(detThree);
to timeout the operation of looking for a bubble
out 1st ADC reading, it tends to be noisy. (also
while ((numBubblesInTube > 0) && (tempTime - timeoutClockStart < MAXTIMEOUT)) {
Take ADC reading from detector 1 to keep track of whether more bubbles enter the tube
tempTime = microso;
adcReading = analogRead(detOne); I/ Read the ADC for detector 1
UpdateADC(&detOneADC, adcReading); // Update the ADC ring buffer with the new
value (keeps the past 3 events)
nextState = CheckForBubble(&detOneBkgd, detOneADC.totalADC, detOneState);
the 3-point running average to check if a bubble beginning or end is seen
// If another bubble is detected, update the number of bubbles in the tube
if(detOneState == LOOKFORSTART && nextState == LOOKFOREND){
numBubblesInTube += 1;
//Use
detOneState = nextState; // Update detector 1 status
Take ADC reading from detector 2. At this detector we want to capture both the beginning and
end of the bubble
tempTime = microso; II Record current time
adcReading = analogRead(detTwo); I/ Read the ADC for detector 2
223
UpdateADC(&detTwoADC, adcReading); // Update the ADC ring buffer with the new
value (keeps the past 3 events)
nextState = CheckForBubble(&detTwoBkgd, detTwoADC.totalADC, detTwoState);
// If detector 2 sees the start of a bubble, record the current time
if(detTwoState == LOOKFORSTART && nextState == LOOKFOREND){
bData[detTwoNumBubbles].d2stime = tempTime;
bData[detTwoNumBubbles].d2sval = adcReading;
i}
// If detector 2 sees the end of a bubble, record the current time
if(detTwoState == LOOKFOREND && nextState == LOOKFORSTART){
bData[detTwoNumBubbles].d2etime tempTime;
bData[detTwoNumBubbles].d2eval adcReading;
detTwoNumBubbles += 1; /Increment total number of bubbles detector 2 has seen
i
detTwoState = nextState; // Update detector 2's status
Take a measurement for detector 3. This is almost the same as detector 2.
tempTime = microso; // Record current time
adcReading = analogRead(detThree); // Read the ADC for detector 3
UpdateADC(&detThreeADC, adcReading); // Update the ADC ring buffer with the new
value (keeps the past 3 events)
nextState = CheckForBubble(&detThreeBkgd, detThreeADC.totalADC, detThreeState);
// If detector 3 sees the start of a bubble, record the current time
if(detThreeState == LOOKFORSTART && nextState == LOOKFOREND) {
bData[detThreeNumBubbles].d3stime = tempTime;
bData[detThreeNumBubbles].d3sval = adcReading;
}
//If detector 3 sees the end of a bubble, record the current time
if(detThreeState == LOOKFOREND && nextState == LOOKFORSTART)
bData[detThreeNumBubbles].d3etime = tempTime;
bData[detThreeNumBubbles].d3eval = adcReading;
detThreeNumBubbles += 1;
numBubblesInTube -= 1; // Decrease total number of bubbles in the tube
I
detThreeState = nextState; // Update detector 3's status
224
/****************************** * **********************************
Error checking to make sure we don't overflow the maximum number of bubbles that can be
counted
if(detTwoNumBubbles == MAX NUMBERBUBBLES 11 detThreeNumBubbles
MAXNUMBERBUBBLES){
timeoutClockStart = 0;
}
// This will cause while loop to exit
//Code to limit # of timeouts //i think this needs to get moved up
if ((tempTime - timeoutClockStart > MAXTIMEOUT)) {
into while loop
if ((tempTime - timeOfLastTimeout) <(MAX_TIMEOUT*2)) {
numberTimeouts++;
}
else{
numberTimeouts = 1;
}
timeOfLastTimeout = tempTime;
if(numberTimeouts >= MAXNUMTIMEOUTS){
InitializeBkgdStructso;
numberTimeouts = 0;
S
* STORE DATA
//No more bubbles in tube, so save the data we have
/or transmit if in debug mode.
// First turn off all LEDs
// digitalWriteFast(ledOne, LOW);
//digitalWriteFast(ledTwo, LOW);
//digitalWriteFast(ledThree, LOW);
// Keep track of errors
eventlnfo.errorVal = 0;
// Error: Sensor timed out before finishing data collection sequence, 00000001
if((tempTime - timeoutClockStart > MAXTIMEOUT)){
225
eventlnfo.errorVal = bdeTIMEOUT;
}
// Error: Neither detector 2 or 3 saw a bubble, 01000000
if(detTwoNumBubbles == 0 && detThreeNumBubbles == 0){
eventlnfo.errorVal = eventlnfo.errorVal I bdeFALSESTART;
}
// Error: Detector 1 is still looking for the end of a bubble, 00001000
if(detOneState == LOOKFOR END) {
eventlnfo.errorVal = eventlnfo.errorVal I bdeDET1NOEND;
}
// Error: Detector 2 is still looking for the end of a bubble, 00010000
if(detTwoState == LOOKFOR END){
eventlnfo.errorVal = eventlnfo.errorVal I bdeDET2NOEND;
I
// Error: Detector 3 is still looking for the end of a bubble, 00100000
if(detThree State == LOOKFOR END){
eventlnfo.errorVal = eventlnfo.errorVal I bdeDET3NOEND;
}
// Error: Detector 2 saw more bubbles than detector 3, 00000010
if(detTwoNumBubbles > detThreeNumBubbles){
eventlnfo.errorVal = eventlnfo.errorVal I bdeDET2MOREBUBBLES;
}
// Error: Detector 3 saw more bubbles than detector 2, 00000100
if(detTwoNumBubbles < detThreeNumBubbles) {
eventlnfo.errorVal - eventlnfo.errorVal I bdeDET3MOREBUBBLES;
}
eventlnfo.numBubbles = max(detTwoNumBubbles, detThreeNumBubbles); // Maximum
number of bubbles seen by detectors 2 and 3
eventlnfo.detl avg = detOneBkgd.total/NUMBKGDPOINTS; // Average
background ADC reading for detector 2 during bubble event
eventlnfo.detlstartval = detOneBkgd.startdetvalue; // ADC value that triggered
beginning of detector 2 data collection sequence
eventlnfo.detlendval = detOneBkgd.enddetvalue; // ADC value that triggered
of detector 2 data collection sequence
eventlnfo.det2avg = detTwoBkgd.total/NUMBKGDPOINTS; // Average
background ADC reading for detector 2 during bubble event
eventlnfo.det2startval = detTwoBkgd.startdetvalue; // ADC value that triggered
beginning of detector 2 data collection sequence
eventlnfo.det2endval = detTwoBkgd.enddetvalue; // ADC value that triggere
of detector 2 data collection sequence
end
I end
226
eventlnfo.det3 avg = detThreeBkgd.total/NUMBKGDPOINTS;
background ADC reading for detector 3 during bubble event
eventlnfo.det3startval = detThreeBkgd.startdetvalue; /
beginning of detector 3 data collection sequence
eventlnfo.det3endval= detThreeBkgd.enddetvalue; //
of detector 3 data collection sequence
//For debugging purposes, data can be printed while the Arduino i
//For field deployment, keep this commented out.
// PrintData();
// Average
ADC value that triggered
ADC value that triggered end
s connected to a computer.
// PrintData() defined BELOW
//Save the data to the EEPROM chip
bdstorage.saveBubbleEvent(&eventlnfo, bData); //make sure you printdata first so you can see
bubbles 1-4 in groups of 4+ bubbles!
} // end of loop()
* CHECK TO SEE IF DETECTOR SEES THE BEGINNING OR END OF A BUBBLE
uint8_t CheckForBubble(backgrounddata* bkgd, uint16_t newValue, uint8_t state){
if(state == LOOKFOR START){ // While the detector is looking for the start of a bubble
if(newValue < bkgd->startdetvalue)
return LOOKFOREND; // A new bubble has been detected
}
else{
return LOOKFORSTART;
}-
// No bubble has been detected
else{
if(newValue > bkgd->enddetvalue){ // While the detector is looking for the end of a bubble
return LOOKFORSTART; // End of bubble has been detected
else{
return LOOKFOREND;
}
}
// Detector still looking for end of bubble
227
}
* UPDATE MAX, MIN, AND DETECTION VALUE OF THE BACKGROUNDDATA
STRUCT
void UpdateBkgd(backgrounddata * bkgd, uint 16_t newValue){
// Update the total value
bkgd->total = bkgd->total - bkgd->rb[bkgd->pos] + newValue;
// Store the new value
bkgd->rb[bkgd->pos]= newValue;
// Update the pos value
bkgd->pos= bkgd->pos + 1;
if(bkgd->pos == NUMBKGDPOINTS){
bkgd->pos =0;
}
// Update max/min and total values of ringbuffer
bkgd->maxvalue 0;
bkgd->minvalue 1024; // Maximum value of 10 bit ADC
for (uint8_t j = 0; j < NUMBKGDPOINTS; j++){
if(bkgd->rb[j] > bkgd->maxvalue){
bkgd->maxvalue = bkgd->rb[j];
if(bkgd->rbUj] < bkgd->minvalue)
bkgd->minvalue = bkgd->rbj];
}
// Pick the more conservative noise threshold
uint16_t noiseThreshold = max((bkgd->maxvalue - bkgd->minvalue) *
NOISE_ MULTIPLILER,
MIN_THRESHOLD);
// If the threshold is greater than the average value, then detection is impossible
if((bkgd->total/NUMBKGDPOINTS) <noiseThreshold){ /MAYBE THIS ISN'T EVEN
NECESSARY ANY MORE.......
bkgd->startdetvalue 0;
bkgd->enddetvalue 1023; // 2A10-1
I
else{
// Pick the more conservative detection value
228
bkgd->startdetvalue = (bkgd->total/NUMBKGDPOINTS) - 30;
bkgd->startdetvalue = bkgd->startdetvalue*8;
// The enddetvalue is arbitrarily larger than the startdetvalue by half the noise threshold.
// This should add significant hysteresis so that a bubble is not instantly detected and then
"undetected"
bkgd->enddetvalue = bkgd->total/NUMBKGDPOINTS - 20;
bkgd->enddetvalue = bkgd->enddetvalue*8;
f}
* UPDATE STRUCT CONTAINING ADC VALUES TAKEN DURING BUBBLE EVENTS
void UpdateADC(ADCdata * adc, uintI6_t newValue){
// Update the total value
adc->totalADC = adc->totalADC - adc->rbADC[adc->posADC] + newValue;
// Store the new value
adc->rbADC[adc->posADC] = newValue;
// Update the pos value
adc->posADC = adc->posADC + 1;
if(adc->posADC == NUMADCPOINTS) {
adc->posADC = 0;
f
}
* INITIALIZE THE BACKGROUNDDATA STRUCTS USED IN THE PROGRAM
//This program must be called in setupo
void InitializeBkgdStructs(){
for (i = 0; i < NUM BKGDPOINTS; i++){
detOneBkgd.rb[i] 0;
detTwoBkgd.rb[i] 0;
229
detThreeBkgd.rb[i] = 0;
}
detOneBkgd.pos 0;
detTwoBkgd.pos 0;
detThreeBkgd.pos = 0;
detOneBkgd.total 0; //new addition
detTwoBkgd.total 0; //new addition
detThreeBkgd. total = 0; //new addition
detOneBkgd.startdetvalue 0; /new addition
detTwoBkgd.startdetvalue 0; //new addition
detThreeBkgd.startdetvalue = 0; //new addition
/Why don't I initialize the enddetvalue??
UpdateBkgd(&detOneBkgd, 0);
UpdateBkgd(&detTwoBkgd, 0);
UpdateBkgd(&detThreeBkgd, 0);
*** ** ***
* INITIALIZE THE ADCDATA STRUCTS USED IN THE PROGRAM
************************ ***** ** ******* ************** ********** ** *******
//This program must be called in setupO
void InitializeADCData()
for (i = 0; i < NUMADC POINTS; i++){'
detOneADC.rbADC[i] 0;
detTwoADC.rbADC[i] 0;
detThreeADC.rbADC[i] = 0;
detOneADC.posADC 0;
detTwoADC.posADC 0;
detThreeADC.posADC 0;
detOneADC.totalADC 0; I/new addition
detTwoADC.totalADC 0; //new addition
detThreeADC.totalADC = 0; /new addition
UpdateADC(&detOneADC, 0);
UpdateADC(&detTwoADC, 0);
UpdateADC(&detThreeADC, 0);
*** * ** *
* COMMUNICATION BETWEEN ARDUINO AND COMPUTER
230
//This program must be called in setupo
void talkToComputero {
char c = Serial.reado;
uint32_t startTime = milliso; /initialize startTime which is used to return to logging after
a period of inactivity
boolean talkingToComputer = true;
Serial.println("Menu");
Serial.print(bdstorage.getNumRecords();
Serial.println(" Events Stored");
bdstorage.resetReado; // Reset the data read address when entering menu
Serial.println("A: All H: Header N: Next R: Reset\nX: Delete E: Exit");
while(talkingToComputer) {
if(Serial.availableO > 0)
c = Serial.read(;
startTime = milliso;
else if(millis()
c = E';
else{
c
- startTime > 60000) {// Go back to logging after minute of inactivity
// E to exit the talking to computer state
if(c =='A'){ // If user enters 'A', read the data and print to screen
bdstorage.resetReado;
uint16_ t numR = bdstorage.getNumRecordso;
for (uint16 _t t = 0; t < numR; t++){
bdstorage.readEvent(&eventlnfo, bData);
PrintDatao; // PrintData() defined below
}
}
else if(c =='E'){ / If user enters 'E', exit the talking to computer state and return to
logging
talkingToComputer = false;
i
else if(c == 'H'){ /If user enters 'H', display data column headers
Serial.println("ms,Bubbles,Error,D2 Back,D2 Det,D2 End,D3 Back,D3 Det,D3
End,D2s,D2val,D2e,d2val,D3s,D3val,D3e,d3val");
}
else if(c =='N') //If user enters 'N', display next line of data
int r = bdstorage.readEvent(&eventlnfo, bData);
if(r==0) {
231
}
else{
Serial.println("End of Data");
}
}
else if(c =='R'){ / If user enters 'R', reset N so it displays first line of data
bdstorage.resetReado;
}
else if (c =='X'){ // If user enters 'X', delete all data upon verification
Serial.println("Delete?(Y/N)"); // Verify user wants to delete all data, Y yes and N no
while(Serial.availableo == 0){
}
c = Serial.reado;
if(c == Y){
Serial.print("Deleting...");
bdstorage.deleteDatao;
Serial.println("Done");
}
}
else if(c == 'W'){
Serial.print("WrA: ");
Serial.println(bdstorage.getWrAddro, HEX);
Serial.print("StartAddr: ");
Serial.println(bdstorage.getDataStartAddr(, HEX);
Serial.print("RdA: ");
Serial.println(bdstorage.getRdAddro, HEX);
}//end else if
} /end while
Serial.println("Logging");
} /end talking to computer
* PRINTING DATA TO COMPUTER DURING SERIAL CONNECTION
* **** ** * ***
void PrintDataO{
Serial.print(eventlnfo.eventTime); // Time in microseconds of bubble event
Serial.print(",");
Serial.print(eventlnfo.numBubbles); // Number of bubbles recorded during bubble event
Serial.print(",");
Serial.print(eventlnfo.errorVal, BIN); // All error codes associated with bubbling event
Serial.print(",");
232
PrintData(); // PrintData() defined farther down
Serial.print(eventInfo.detl avg); // Bac
event
Serial.print(",");
Serial.print(eventlnfo.det2avg); // Bac
event
Serial.print(",");
Serial.print(eventlnfo.det3avg); // Bac
event
Serial.print(",");
for (i = 0; i < eventInfo.numBubbles; i ++){
if(i > 0)
Serial.print(",,,,,,");
Serial.print(bData[i] .d2stime); // Bubb
Serial.print(",");
Serial.print(bData[i] .d2sval); // ADC
collection sequence
Serial.print(",");
Serial.print(bData[i].d2etime); // Bubb
Serial.print(",");
Serial.print(bData[i].d2eval); // ADC
sequence
Serial.print(",");
Serial.print(bData[i].d3 stime); // Bubb
Serial.print(",");
Serial.print(bData[i] .d3sval); // ADC
collection sequence
Serial. print(",");
Serial.print(bData[i].d3etime); // Bubb
Serial.print(",");
Serial.println(bData[i].d3eval); // ADC
sequence
kground ADC value for detector 2 at time of bubble
kground ADC value for detector 2 at time of bubble
kground ADC value for detector 3 at time of bubble
le start time for detector 2
value that triggers beginning of detector 2 data
le end time for detector 2
value that triggers end of detector 2 data collection
le start time for detector 3
value that triggers beginning of detector 3 data
le end time for detector 3
value that triggers end of detector 3 data collection
}
}
233
A12. Arduino code for data collection buoy
Please contact Kyle Delwiche (k.delwiche@alum.mit.edu) for latest version.
Title: Bubble Sizing Data Buoy Firmware
Version: 1.1
Date Completed: 17 October 2016
Description: This code is the firmware for the data buoy Arduino that controls the bubble size
detector.
Authors: Kyle Delwiche
The portions of this code that put the Arduino to sleep are from Donal Morrissey - 2011.
http://donalmorrissey.blogspot.com/2010/04/sleeping-arduino-part-5-wake-up-via.html
#include <avr/sleep.h>
#include <avr/power.h>
#include <avr/wdt.h>
#include <SD.h>
#define LED-PIN (13)
volatile int f wdt=1;
int counter = 1; /initialize to 1
int X = 2700; //# of watchdog interrupt cycles between data downloading sequence. Each
cycle is ~8 seconds, so 2700 cycles = 6 hours
File myFile; //File we will be storing on SD card
int byteNumber = 1; //Incoming serial data
char a[9]; //Placeholder array for serial.readBytes to put data in to
int placeHolder = 1; //Dummy variable making while loop go continuously
int dummyVar = 1; //Dummy variable making void loopO go continuously
int tempTimeStart; /A timing variable that will help us break out of while loop when
serial.readbytes stops receiving data
int tempTimeEnd; /A timing variable that will help us break out of while loop when
serial.readbytes stops receiving data
int variable = 3;
234
/*************** ************ *** *** ***** ** *******
* Name: ISR(WDTvect)
* Returns: Nothing.
* Parameters: None.
* Description: Watchdog Interrupt Service. This
* is executed when watchdog timed out.
ISR(WDT-vect)
if(fwdt ==0)
f_wdt=1;
else
//Serial.println("WDT Overrun!!!");
}
}
* Name: enterSleep
* Returns: Nothing.
* Parameters: None.
* Description: Enters the arduino into sleep mode.
void enterSleep(void)
setsleepmode(SLEEP MODEPWRSAVE); /* EDIT: could also use
SLEEPMODEPWRDOWN for lowest power consumption. */
sleepenableo;
/* Now enter sleep mode. */
sleepmodeo;
/* The program will continue from here after the WDT timeout*/
sleepdisable(); /* First thing to do is disable sleep. */
/* Re-enable the peripherals. */
powerallenableO;
}
* Name: setup
* Returns: Nothing.
* Parameters: None.
* Description: Setup for the serial comms and the
* Watch dog timeout.
235
void setupO
{
pinMode( 13, OUTPUT);
/*** Setup the WDT ***/
/* Clear the reset flag. */
MCUSR &= ~(1<<WDRF);
/* In order to change WDE or the prescaler, we need to
* set WDCE (This will allow updates for 4 clock cycles).
WDTCSR 1= (1<<WDCE) I (1<<WDE);
/* set new watchdog timeout prescaler value */
WDTCSR = <<WDPO I l<<WDP3; /* 8.0 seconds */
/* Enable the WD interrupt (note no reset). */
WDTCSR 1= _BV(WDIE);
pinMode(1O, OUTPUT); //Set this up for serial communication
a[8] = 0;
Serial.setTimeout(5000); /make serial.readbytes wait 5 seconds before timing out
delay(7000); //delay 7 seconds every time battery is restarted so the bottom
arduino has time to initialize. It can't be more than 8 seconds or the watchdog timer runs over.
SD.begin(4); //Open communication lines with SD card
}
* Name: enterSleep
* Returns: Nothing.
* Parameters: None.
* Description: Main application loop.
void loop()
{
if(f wdt == 1) //Every 8 seconds the Arduino will wake. Every X wakeups it will toggle the
LED
{
if(counter == X) { //If the Arduino has woken up X number of times, run through data-
gathering sequence
downloadDataO;
counter = 1; /reset the counter variable to 1 for sleep cycle code
}
else{
236
counter = counter + 1; //update counter variable
}
/* Don't forget to clear the flag. */
f _wdt = 0;
/* Re-enter sleep mode. */
enterSleep(;
}
else
{
/* Do nothing. */
}
* Name: downloadData
* Returns: Nothing.
* Parameters: None.
* Description: Enters the arduino into sleep mode.
** ****** *** * **************** ** *** **** * *** *** ** ***/
void downloadData(void)
{while (dummyVar > 0){
myFile = SD.open("SensorData.txt", FILEWRITE); //Open a file called "Sensor Data"
on SD card
Serial.begin(57600); /set the Baud rate
delay(1000); //Delay necessary because it might take a while for the serial
connection to establish
Serial.print('k'); //Start the Menu function, this could be any letter
tempTimeStart = millis(; /record temporary time
while(! Serial.available() {
tempTimeEnd = millisO; /record temporary time
if ((tempTimeEnd - tempTimeStart)> 5000){ //5 seconds of no serial contact has
elapsed, re-start while loop by setting dummyVar to 0.
dummyVar = 0; /causes while loop to end when Serial.readBytes times
out
}
}; /Wait until there is data in the buffer
while(Serial.availableO> 0) { //As long as there is data in the buffer, keep reading data
Serial.readBytes(a,8); //Read 8 bits of data from the buffer
delay(1); //Add in small delay here in case it takes longer to send over
Menu text
237
//Send data to SD card
Serial.print('A') ; //Ask bottom Arduino to send first line of data from EEPROM
tempTimeStart = milliso; //record temporary time
while(! Serial.availableO) {
tempTimeEnd = millisO; /record temporary time
if ((tempTimeEnd - tempTimeStart)> 1000){ //1 seconds of no serial contact has
elapsed, re-start while loop by setting dummyVar to 0.
dummyVar = 0;
out
} //Wait until there is data in the buffer
i
/causes while loop to end when Serial.readBytes times
tempTimeStart = milliso;
while (placeHolder > 0) I
any more data in buffer
byteNumber = Serial.readBytes(a,8);
tempTimeEnd = milliso;
'/record temporary time
//Keep looping until Serial.readBytes doesn't see
/read 8 bytes at a time
/record temporary time
if ((tempTimeEnd - tempTimeStart) > 10000){ //once it takes Serial.readBytes longer
than XX seconds to get more data, exit while loop
placeHolder = 0;
times out
I
myFile.print(a);
/causes while loop to end when Serial.readBytes
//Send data to SD card
myFile.print("Bytes available on buffer right NOW: ");
myFile.println(Serial.avail
Serial.print(X');
Serial.print('Y');
Serial.print(E');
myFile.closeo;
Serial.end();
placeHolder = 1;
able(); //Make sure no data remains in the buffer
//delete data
/confirm delition
/exit from logging mode
//Close file (otherwise it doesn't save properly)
//Close serial connection
/reset placeholder to I
break; /end the downloadData function once data has been collected
}
238
chip
myFile.print(a);
/*****************If serial connection doesn't establish, restart loop **************/
myFile.close(; // Close file (otherwise it doesn't save properly)
Serial.end(; // Close serial connection
dummyVar = 1; // reset dummyVar to 1 to enter while loop again, this way Arduino
gets another shot at establishing serial connection
}
239
A13. Arduino library files for bubble size sensor: BDStorage.cpp
Please contact Kyle Delwiche (k.delwiche@alum.mit.edu) for latest version.
/* *
Authors: Schuyler Senft-Grupp and Kyle Delwiche
Version: 1.
License: This work is licensed under the Creative Commons Attribution-NonCommercial-
ShareAlike 4.0 International License. To view a copy of this license, visit
http://creativecommons.org/licenses/by-nc-sa/4.0/.
* */
#include "BDStorage.h"
BDStorage bdstorage;
// Public methods
uintl6 t BDStorage::begin({
//Start 12C communication
Wire.begin(;
//find data
rdAddr = 0;
uint8 t val;
uint8_t prevHasData;
rdAddr = 0x00020000 - 128;
Wire.beginTransmission(DEV ADDR HIGH);
Wire.write((uint8_t) (rdAddr >> 8));
Wire.write((uint8_t) (rdAddr & OxFF));
Wire.endTransmission();
val = Read8bit((uint8_t) true);
if(val == OxFF)
prevHasData = 0;
else
prevHasData = 1;
rdAddr = 0;
wrAddr = OxFFFFFFFF;
dataStartAddr = OxFFFFFFFF;
numRecords = 0;
//MSB = most significant bits
while(rdAddr < Ox00020000 &&
OxFFFFFFFF)){
(wrAddr == OxFFFFFFFF 11 dataStartAddr ==
if(rdAddr > OxFFFF){
Wire.beginTransmission(DEVADDRHIGH);
}else{
Wire.beginTransmission (DEVADDRLOW);
}
Wire.write((uint8_t) (rdAddr >> 8));
Wire.write((uint8_t) (rdAddr & OxFF));
//MSB = most significant
240
bits
Wire.endTransmission();
val = Read8bit((uint8 t) true);
if(prevHasData == 0 && val !=0xFF){
dataStartAddr = rdAddr;
}
if(prevHasData == 1 &&
wrAddr = rdAddr;
rdAddr += 128;
if(val == OxFF)
prevHasData = 0;
else{
numRecords++;
prevHasData = 1;
val == OxFF)
uint32_t al, a2, a3;
al = analogRead(A6);
a2 analogRead(A7);
a3 = ((a2 & Cxl) << 16) 1 ((al &
((al & 0x2) << 13) 1 ((al & 0x4) << 12)
10) 1 ((al & 0x8) << 9)
0x1) << 15) 1 ((al & Ox2) << 14) 1
1 ((a2 & 0x4) << 11) 1 ((al & 0x8) <<
a3 = 0;
if(numRecords == MAX RECORDS){ //case where
don't know where the actual "start" is.
//TO DO - look for earliest time stamp
wrAddr = 0;
rdAddr = 0;
dataStartAddr = 0;
}else if(numRecords == 0){ //case where there
dataStartAddr = a3;
wrAddr = a3; //maybe make
rdAddr = a3;
}/**else if(wrAddr > dataStartAddr)
numRecords = (wrAddr - dataStartAddr) >>7;
else
memory is full we
is no data
this a random number
numRecords = ((0x00020000 - dataStartAddr) + wrAddr) >> 7;
return numRecords;
uint8 t BDStorage::saveBubbleEvent(struct eventinfo*
bubbledata * bdata)
{
//Check to make sure there is space to write
if(numRecords >= MAXRECORDS){
return 1;
}
if(wrAddr > OxFFFF){
Wire.beginTransmission(DEVADDRHIGH);
else {
Wire.beginTransmission(DEVADDRLOW);
eventInfo, struct
241
}
}
Wire.write((uint8_t) (wrAddr >> 8));
Wire.write((uint8_t) (wrAddr & OxFF));
Write8bit(eventInfo->errorVal);
Write32bit(eventInfo->eventTime);
Write8bit (min (eventInfo->numBubbles,
Writel6bit(eventInfo->det2avg);
Writel6bit(eventInfo->det2startval);
Writel6bit (eventInfo->det2endval);
Writel6bit(eventInfo->det3avg);
Writel6bit(eventInfo->det3startval);
Writel6bit(eventInfo->det3endval);
Wire.endTransmission((uint8_t)true);
to the chip
delay(5);
wrAddr += 18;
//total bytes written to chip == 18
//total bytes sent to wire buffer 21
//MSB = most significant bits
//LSB = least "
4));
//do not send a stop signal
Serial.println(" "); */
for (uint8 t i = 0; i < min(eventInfo->numBubbles,
if(wrAddr > OxFFFF)
Wire.beginTransmission(DEVADDRHIGH);
else
Wire.beginTransmission(DEVADDRLOW);
4); i ++){
Wire.write((uint8 t) (wrAddr >> 8));
bits
Wire.write((uint8 t) (wrAddr & OxFF)); //LSB
//MSB = most significant
least "
Write32bit
Write16bit
Write32bit
Writel6bit
Write32bit
Writel6bit
Write32bit
Write16bit
(bdata i]
(bdata [i
(bdata [i]
(bdata [i]
(bdata [i]
(bdata [i
(bdata [i]
(bdata [iI
.d2stime);
.d2sval);
.d2etime);
.d2eval);
.d3stime);
.d3sval);
.d3etime);
.d3eval);
if(i < min(eventInfo->numBubbles, 4)-1)
Wire.endTransmission((uint8_t)true);
else
Wire.endTransmission((uint8_t)true);
delay(5);
wrAddr += 24;
//24 bytes
wrAddr = ((wrAddr >> 7) + 1) << 7;
numRecords += 1;
//wrap the write address around to zero
if(wrAddr >=0x00020000)
wrAddr = 0;
242
}
"
if(eventInfo->numBubbles > 4){
//delay(5); //wait for page to write - can do with
eventInfo->numBubbles -= 4;
//save the remaining bubbles on new memory page
// add something here to manually shift data since fancy
doesn't work for some reason.
for (uint8 t 1 = 0; 1 < min(eventInfo->numBubbles, 8); 1
bdata[l].d2stime = bdata[l+4].d2stime;
bdata[l].d2sval = bdata[l+4].d2sval;
bdata[l].d2etime = bdata[l+4].d2etime;
bdata[l].d2eval = bdata[l+4].d2eval;
bdata[l].d3stime = bdata[l+4].d3stime;
bdata[l].d3sval = bdata[l+4].d3sval;
bdata[l].d3etime = bdata[l+4].d3etime;
bdata[l].d3eval = bdata[l+4].d3eval;
polling later
method below
++){
}
saveBubbleEvent(eventInfo, bdata);
//saveBubbleEvent(eventInfo, bdata+sizeof(bubbledata)*4);
}
return 0;
}
uint8 t BDStorage::readEvent(struct eventinfo* eventInfo, struct bubbledata *
bdata){
if(wrAddr > dataStartAddr)
if(rdAddr >= wrAddr)
return 1;
}else{
if(rdAddr < dataStartAddr && rdAddr >=wrAddr)
return 2;
if(rdAddr > OxFFFF){
Wire.beginTransmission(DEVADDRHIGH);
}else{
Wire.beginTransmission(DEVADDRLOW);
}
Wire.write((uint8_t) (rdAddr >> 8));
Wire.write((uint8 t) (rdAddr & OxFF));
Wire.endTransmission();
//MSB = most significant bits
//The value of rdAddr is used (but not changed) in
//So do not change the value of rdAddr until done
//reading from this page
eventInfo->errorVal = Read8bit((uint8_t)false);
eventInfo->eventTime = Read32bit((uint8_t)false);
eventInfo->numBubbles = Read8bit((uint8 t)false);
eventInfo->det2avg = Readl6bit((uint8_t)false);
Read-bit()
243
eventInfo->det2startval = Readl6bit((uint8_t)false);
eventInfo->det2endval = Readl6bit((uint8 t)false);
eventInfo->det3avg = Readl6bit((uint8_t)false);
eventInfo->det3startval = Readl6bit((uint8_t)false);
eventInfo->det3endval = Readl6bit((uint8_t)false);
//Check to make sure we can't overrun memory
if(eventlnfo->numBubbles > MAX NUMBERBUBBLES){
eventInfo->numBubbles = 4;
eventInfo->errorVal = eventInfo->errorVal I bdeMEMORY;
}
0; j < eventInfo->numBubbles; j++){
bdata[j].d2stime
bdata [j
bdata [j
bdata [j
bdata [j
bdata [j
bdata j]
bdata [j]
.d2sval =
.d2etime=
.d2eval =
.d3stime=
.d3sval =
.d3etime=
.d3eval =
Read32bit((uint8 t)false);
Read16bit
Read32bit
Read16bit
Read32bit
Readi6bit
Read32bit
Read16bit
(uint8_t)
(uint8_t)
(uint8_t)
(uint8_t)
(uint8_t)
(uint8_t)
(uint8_t)
false);
false) ;
false) ;
false) ;
false) ;
false);
false);
}
Read8bit((uint8_t)true); //just do this to send stop bit
rdAddr = rdAddr + 128; //increment
//wrap the write address around to
if(rdAddr >= 0x00020000){
rdAddr = 0;
}
return 0;
read address to next page
zero
uint32_t BDStorage::getRdAddro{
return rdAddr;
}
uint32_t BDStorage::getWrAddr(){
return wrAddr;
}
uint32_t BDStorage: :getDataStartAddr ()
return dataStartAddr;
I
uintl6_t BDStorage::getNumRecords()
return numRecords;
}
void BDStorage::deleteData()
uint8_t val;
rdAddr = 0;
while(rdAddr < Ox00020000){
if(rdAddr > OxFFFF)
Wire.beginTransmission(DEVADDRHIGH);
else
Wire.beginTransmission(DEVADDRLOW);
244
for(uint8_t j
}
bits
signif
Wire.write((uint8_t) (rdAddr >> 8)); //MSB = most significan
Wire.write((uint8_t) (rdAddr & OxFF));
Wire.endTransmission();
val = Read8bit((uint8_t) true);
if(val != OxFF){
if(rdAddr > OxFFFF)
Wire.beginTransmission(DEVADDRHIGH);
else
Wire.beginTransmission(DEVADDR_LOW);
Wire.write((uint8_t) (rdAddr >> 8)); //MSB = most
icant bits
Wire.write((uint8_t) (rdAddr & OxFF)); //LSB = least
Write8bit(OxFF);
Wire.endTransmission(true);
delay(5);
}
rdAddr += 128;
}
dataStartAddr = wrAddr;
numRecords = 0;
void BDStorage::resetRead(){
rdAddr = dataStartAddr;
}
// Private Methods
void BDStorage::Write8bit(uint8_t data)
{
Wire.write(data);
}
void BDStorage: :Write16bit (uint16_t data)
I
Wire.write(data >> 8);
Wire.write(data);
void BDStorage::Write32bit(uint32 t data)
{
Wire.write(data >> 24);
Wire.write(data >> 16);
Wire.write(data > 8);
Wire.write (data);
}
uint8_t BDStorage::Read8bit(uint8_t sendStop)
{
uint8 t mydata = 0;
uint8_t bytesReceived = 0;
uint8 t devAddr;
if(rdAddr > OxFFFF)
devAddr = DEVADDRHIGH;
245
t
else
devAddr = DEVADDRLOW;
bytesReceived = Wire.requestFrom(devAddr, (uint8 t) 1, (sendStop));
while (Wire.available()
{
mydata = Wire.read(;
} //end of while loop
return mydata;
uintl6_t BDStorage::Readl6bit(uint8_t sendStop)
uintl6_t mydata = 0;
uint8_t bytesReceived = 0;
uint8_t c = 0;
uint8 t devAddr;
if(rdAddr > OxFFFF)
devAddr = DEVADDRHIGH;
else
devAddr = DEV ADDRLOW;
//read the 2 bytes from eeprom memory
bytesReceived = Wire.requestFrom(devAddr, (uint8 t)2, sendStop);
while (Wire.available()
{
c = Wire.read();
mydata = (mydata << 8) + c;
} //end of while loop
return mydata;
uint32_t BDStorage::Read32bit(uint8_t sendStop)
uint32 t mydata = 0;
uint8_t bytesReceived = 0;
uint8 t c = 0;
uint8 t devAddr;
if(rdAddr > OxFFFF)
devAddr = DEVADDRHIGH;
else
devAddr = DEVADDRLOW;
//read the 4 bytes from eeprom memory
bytesReceived = Wire.requestFrom(devAddr, (uint8 t)4, sendStop);
while (Wire.available()
{
c = Wire.read(;
mydata = (mydata << 8) + c;
I //end of while loop
return mydata;
}
246
A14: Arduino library files for bubble size sensor: BDStorage.h
Please contact Kyle Delwiche (k.delwichea alum.mit.edu) for latest version.
/* *
Author: Schuyler Senft-Grupp
Version: 1.1
License: This work is licensed under the Creative Commons Attribution-NonCommercial-
ShareAlike 4.0 International License. To view a copy of this license, visit
http://creativecommons.org/licenses/by-nc-sa/4.0/.
**
#ifndef BDSTORAGE H
#define BDSTORAGE H_
#include <stdio.h>
//#include "pins arduino.h"
#include <Wire.h>
#include <BubbleDetector.h>
//constants up here
const uint8_t DEV_ ADDRLOW= b01010000; //this is the 7 bit 12C address - the wire
library adds in the 8th bit depending on read or write
const uint8_t DEVADDR HIGH = ObO1010100;
const uint8 t EMPTY-PAGE = 0xFF;
const uint16_t MAXRECORDS = 1024;
class BDStorage{
public:
uintI6_t begino;
uint8 t saveBubbleEvent(struct eventinfo* eventInfo, struct bubbledata * bdata);
uint8_t readEvent(struct eventinfo* eventInfo, struct bubbledata * bdata);
uint32_t getRdAddr();
uint32_t getWrAddro;
uint32_t getDataStartAddro;
uint 16t getNumRecordso;
void deleteDataO;
void resetReadO;
private:
uint32_t dataStartAddr;
uint32_t rdAddr;
uint32_t wrAddr;
uintI 6_t numRecords;
void Write8bit(uint8_t data);
247
void Write I6bit(uintI6_t data);
void Write32bit(uint32 t data);
uint8_t Read8bit(uint8_t sendStop);
uint16_t Read I6bit(uint8_t sendStop);
uint32_t Read32bit(uint8_t sendStop);
extern BDStorage bdstorage;
#endif // BDSTORAGEH
248
A15: Arduino library files for bubble size sensor: BubbleDetector.h
Please contact Kyle Delwiche at k.delwichealalun.mit.edu for latest version.
/* *
Description: Header file for all constants and definitions for
bubble detector code.
Author: Schuyler Senft-Grupp
Version: 1.1
Date: 5/15/2013
License: This work is licensed under the Creative Commons Attribution-NonCommercial-
ShareAlike 4.0 International License. To view a copy of this license, visit
http://creativecommons.org/licenses/by-nc-sa/4.0/.
**
#ifndef BUBBLEONEDETECTOR
#define BUBBLEONEDETECTOR
#include <Arduino.h>
const uint8_t MAX _NUMBERBUBBLES = 12; //The maximum number of bubbles that can
be stored at the same time.
const uint8_t NUMBKGDPOINTS = 8; //the number of background points to store
const uint8_t NUMADCPOINTS = 8; //the number of background points to store
//bde stands for bubble detector error
const uint8_t bdeTIMEOUT = 0b00000001;
const uint8_t bdeDET2MOREBUBBLES = Ob000010;
const uint8 t bdeDET3MOREBUBBLES =Ob00000 100;
const uint8_t bdeDETINOEND = Ob00001000;
const uint8_t bdeDET2NOEND = ObOO010000;
const uint8_t bdeDET3NOEND = Ob00100000;
const uint8_t bdeFALSESTART = ObOl000000;
const uint8_t bdeMEMORY OblO000000;
/*********** ** ************** ***** *** * ******* ***
Declare structs to hold background measurements for each sensor
These use continuous ring buffers
struct backgrounddata{
uint16_t rb [NUMBKGDPOINTS]; //ring buffer for data
uint8_t pos; //pos in the ring buffer - 0 to 7
uint I6_t minvalue; //minimum value in the ring buffer
249
uint 16 t maxvalue; /maximum value in the ring buffer
uint32 t total; //the sum of all values in the ring buffer - divide by
NUMBKGDPOINTS to get avg value
uint 16_t startdetvalue; //the value required to signify a bubble start event
uint I6_t enddetvalue; /the value required to signify a bubble end event
Declare structs to hold ADC measurements for each sensor
during a bubble detection event
struct ADCdata{
uint 16_t rbADC [NUMADCPOINTS]; /ring buffer for data
uint8_t posADC; //pos in the ring buffer - 0 to 2
//uint16 t minvalueADC; /minimum value in the ring buffer
//uint 16 t maxvalueADC; //maximum value in the ring buffer
uint32_t totalADC; /the sum of all values in the ring buffer - divide by
NUMBKGDPOINTS to get avg value
//uint16_t startdetvalueADC; //the value required to signify a bubble start event
//uintI 6_t enddetvalueADC; /the value required to signify a bubble end event
Declare structs to hold event information
*************** *** ** * ****** ***********
struct eventinfo{
uint32_ t eventTime;
uint8_t numBubbles;
uint8 t errorVal;
uint16_t detIavg;
uint16_ t detl startval;
uint16 t detlendval;
uint 16 t det2avg;
uint16 t det2startval;
uint16__t det2endval;
uint I6_t det3avg;
uint 16_ t det3startval;
uintI 6_t det3endval;
/****************** ** *********** ** ***** ***
Declare structs to hold bubble information
struct bubbledata{
uint32_t dl stime;
250
uint16_t dl sval;
uint32_t dletime;
uint16_t dleval;
uint32_t d2stime;
uint16_t d2sval;
uint32 t d2etime;
uint16_t d2eval;
uint32_t d3stime;
uint16_t d3sval;
uint32__t d3etime;
uintI6_t d3eval;
#endif
251