TransCom model simulations of methane: Comparison of vertical profiles with aircraft measurements

Abstract

To assess horizontal and vertical transports of methane (CH4) concentrations at different heights within the troposphere, we analyzed simulations by 12 chemistry transport models (CTMs) that participated in the TransCom-CH4 intercomparison experiment. Model results are compared with aircraft measurements at 13 sites in Amazon/Brazil, Mongolia, Pacific Ocean, Siberia/Russia, and United States during the period of 2001–2007. The simulations generally show good agreement with observations for seasonal cycles and vertical gradients. The correlation coefficients of the daily averaged model and observed CH4 time series for the analyzed years are generally larger than 0.5, and the observed seasonal cycle amplitudes are simulated well at most sites, considering the between-model variances. However, larger deviations show up below 2 km for the model-observation differences in vertical profiles at some locations, e.g., at Santarem, Brazil, and in the upper troposphere, e.g., at Surgut, Russia. Vertical gradients and concentrations are underestimated at Southern Great Planes, United States, and Santarem and overestimated at Surgut. Systematic overestimation and underestimation of vertical gradients are mainly attributed to inaccurate emission and only partly to the transport uncertainties. However, large differences in model simulations are found over the regions/seasons of strong convection, which is poorly represented in the models. Overall, the zonal and latitudinal variations in CH4 are controlled by surface emissions below 2.5 km and transport patterns in the middle and upper troposphere. We show that the models with larger vertical gradients, coupled with slower horizontal transport, exhibit greater CH4 interhemispheric gradients in the lower troposphere. These findings have significant implications for the future development of more accurate CTMs with the possibility of reducing biases in estimated surface fluxes by inverse modeling.