Evaluating the diurnal cycle of convection and clouds over land simulated by NCAR CAM4 and CAM5 in a weather forecasting framework using ARM SGP data

Description

It is very challenging for global climate models to simulate the correct diurnal cycle of convection and clouds over land. For the observed late-afternoon precipitation peak, models usually report a quicker onset around noon. Previous studies suggest that this model deficiency is usually due to the lack of an intermediate stage involving shallow convective clouds and their gradual moistening of free troposphere. Recently NCAR released its community atmospheric model (CAM) version 4 and 5 in which significant modification is made to the boundary-layer turbulence scheme and shallow convection scheme. It is worthwhile to evaluate the effects of these modifications focusing on two aspects: (1) the development of shallow cumulus clouds over land and (2) the diurnal evolution of late-afternoon deep convection over land.

The models are run at 0.9° by 1.25° horizontal resolution with 26 vertical levels for June 2008 to March 2010 in the weather forecasting framework. We evaluate the second-day forecast by the models, which are initialized with ECMWF analysis data every day at 0 UTC and forecast for the six following days. This framework facilitates the identification of model parameterization deficiency in the short range forecast before other compensating errors could potentially mask such deficiency as they do in the long-term climate simulations (Phillips et al. 2004). Specifically, we compare model simulations with the observed diurnal cycle statistics we developed for different convection regimes: forced and active fair-weather shallow cumulus and the shallow cumulus that transits to late-afternoon deep convective clouds with heavy precipitation using ARM long-term observations (Zhang and Klein 2010, 2012). We also will work on a few case studies in the warm season of 2008 and 2009, targeting different convection regimes and will make detailed exploration of the parameterization performance of the two CAM versions in simulating clouds, precipitation, and convection-related variables. Such practice serves as a prelude to provide working directions for our following work, which is to set up typical-convection composite-cases over land based on ARM SGP observations for large-eddy simulations and single-column CAM so as to improve and develop cloud and convection parameterizations.