Evaluation of the Effect of Cumulus Parameterization Physics on Convective Updraft Speeds

Description

Cumulus parameterizations in general circulation models (GCMs) have traditionally predicted the mass flux (density x vertical velocity x area fraction) rising in convective updrafts but not the actual strength of the convection. This is sufficient to calculate convective heat and moisture tendencies but does not provide information on how much condensate is lofted and detrained to produce stratiform rain and anvil clouds and whether lightning is generated. For these reasons many GCMs have begun to diagnose the convective updraft speed as well. Until recently few direct observational constraints on the updraft speed existed beyond anecdotal information from a few field experiments. ARM’s development of radar wind profiler algorithms to retrieve updraft speed profiles in deep convection has changed the situation dramatically. Statistics now exist for a maritime tropical warm pool environment (the TWP-ICE IOP active period), a midlatitude continental regime (the SGP), and a tropical continental region (the GoAmazon 2014/15 IOP). These data sets have shown that both GCMs and cloud-resolving models overpredict updraft speeds, particularly at high levels. We have compiled statistics of convective events in the major deep convective regions of the world from the GISS GCM with several versions of our cumulus parameterization: The CMIP5 model, which is known to produce overly weak entrainment; a restrictive version of our cold-pool parameterization, which rarely allows weakly entraining convection; and a modified cold-pool scheme with stronger downdrafts and more frequent triggering of weakly entraining convection. For events that go no higher than the 300 mb level, all model versions correctly predict the strongest convection in the US Great Plains, slightly weaker convection over the Amazon, and the weakest convection over the Tropical Western Pacific. The CMIP5 model convection is somewhat too strong relative to ARM retrievals, especially at the higher percentiles, while the two cold-pool models are more consistent with the retrievals. When deeper penetrating convective events are included, though, all model versions can be seen to have excessive updraft speeds above the 250 mb level, approaching 50 m/s over the US, while over the Tropical Western Pacific the results vary from one model to another, with the restrictive cold-pool version’s peak speeds only 20 m/s. This behavior is associated with a dramatic decrease in entrainment rates above the 300 mb level.