Sensitivity of simulated tropospheric CO to subgrid physics parameterization: A case study of Indonesian biomass burning emissions in 2006

Recent cumulus and turbulence parameterization changes to the NASA GISS ModelE2 have improved representation of the Madden-Julian Oscillation and low cloud distribution, but their effect on composition-related quantities is not known. In this study, we simulate the vertical transport of carbon monoxide (CO) from uncontrolled biomass burning in Indonesia in late 2006, during which uniquely high CO was detected in the upper troposphere. Two configurations of ModelE2, one without the changes (AR5) and one with the changes (AR5′), are used for an ensemble simulation of the transport of CO from the biomass burning. The simulation results are evaluated against new CO profiles retrieved jointly from the Aura Tropospheric Emission Spectrometer and the Microwave Limb Sounder. Modeled upper tropospheric CO using the AR5 physics was unrealistically high. The AR5′ physics suppress deep convection that reaches near the tropopause, reducing vertical transport of CO to the upper troposphere and bringing the model into better agreement with satellite CO. In this regard, the most important changes were related to the strength of entrainment of environmental air into the convective column, the strength of re-evaporation above cloud base, and a negative plume buoyancy threshold based on density temperature. This study illustrates how individual, noncomposition model changes can lead to significantly different modeled composition, which in this case improved agreement with satellite retrievals. This study also illuminates the potential usefulness of CO satellite observations in constraining unobservable processes in general circulation models.