Mesoscale convective systems (MCSs) are the largest type of convective storms that develop when convection aggregates and induces mesoscale circulation features. They range in size between one hundred to several hundred kilometers. These storms are responsible for producing well over half of the spring and summer rainfall in the central U.S., and account for the majority of the 100-year extreme rainfall events. Better simulation of MCSs in climate models would greatly improve our ability in projecting future changes in precipitation under global warming.

In this study, we systematically evaluate MCSs simulated by the Model for Prediction Across Scales (MPAS) coupled with the physics package of the Community Atmosphere Model (CAM5) in the Community Earth System Model (CESM) framework. MPAS-CAM5 is configured for global variable resolution climate simulations with regional refinements at 50 km and 25 km grid spacings over North America. Decade-long simulations have been conducted with prescribed sea surface temperature and sea ice cover. A recently developed MCS tracking algorithm (FLEXTRKR) has been adapted to identify and track MCSs consistently in observations and model simulations at different resolutions. The model underestimates the frequency and precipitation of warm season MCSs east of the Rocky Mountains, with smaller bias during spring when the baroclinic forcing is stronger compared to summer. The model reasonably captures the observed patterns of MCS favorable large-scale environment identified by Self Organizing Map, but the frequency of occurrence is underestimated due to a low moisture bias. Further, the MCS-like precipitation under favorable large-scale environments is too weak and persists for too long compared to observations, likely related to deficiencies of the cumulus parameterization. Further process-oriented analysis and sensitivity experiments will better inform model development to improve simulations of MCSs and their associated hydrological extremes.