Development and Evaluation of Contrail Models

Abstract

Condensation trails (contrails) are aircraft-induced ice clouds that are estimated to account for up to 50% of aviation’s climate impacts. Uncertainties in the impact of individual contrails have motivated the development of contrail models, such as CoCiP, a 0-D rapid assessment model, and APCEMM, a 2-D model with detailed ice microphysics. However, there are gaps within the current contrail modeling literature. There is no model both sufficiently fast for rapid assessment of contrail impacts and detailed in its ice microphysics modeling. There are few studies calibrating and validating the performance of contrail models on individual f lights. The absolute and relative magnitudes of errors due to weather data uncertainty and errors due to modeling assumptions have not been extensively studied, despite many studies relying on the CoCiP model and the ERA5 weather data for their analyses. This thesis addresses these gaps. The APCEMM model is optimized to achieve a decrease in runtime by 95% and is improved with depth estimation, vertical advection, and atmospheric turbulence modules. A set of 152 flight-attributed LIDAR cross sections is assembled to compare APCEMM and CoCiP results against individual contrail observations on metrics such as contrail width, depth, and optical depth. A method dubbed “ambient parameter inference”, where contrail models infer the meteorological conditions necessary to reproduce a contrail observation, is developed to produce estimated distributions of ambient parameters. These distributions are used to analyze model sensitivities, biases in the weather data, and errors due to weather data uncertainty and modeling assumptions. I find that the distributions of the wind shear and vertical humidity profile as inferred by APCEMM have means and medians within the range of radiosonde measurements of these quantities, suggesting that the model adequately accounts for the sensitivities of contrail properties to these parameters. Compared to the APCEMM-inferred parameters, the ERA5 weather data predicts a 3.8 times higher average supersaturated layer depth and a 56% lower wind shear, suggesting systematic biases. CoCiP infers on average a 39% lower supersaturated layer depth and a 3.0 times higher ice supersaturation level compared to APCEMM. Due to the APCEMM-inferred parameters’ closer agreement with radiosonde measurements, this suggests that there may be modeling errors due to CoCiP’s inability to resolve the contrail’s vertical profile and its lower sensitivity to relative humidity. Errors in the ambient humidity data are found to possibly account for an over 100% average absolute error in optical depth when using APCEMM, greater than the 72.5% attributable to CoCiP modeling limitations. APCEMM is found to predict contrails with a 29.3% longer average lifetime and a 4.34-5.92 times average higher energy forcing compared to CoCiP when using the ERA5 weather data. This suggests that inter-model disagreement is on the same order of magnitude as the already known errors resulting from meteorological data gaps.