Understanding Systematic Errors in Simulation of the Diurnal Cycle of Precipitation in Weather and Climate Models through a Multi-Model Intercomparison with a Hierarchical Modeling Framework 

This presentation reports the WCRP GEWEX GASS multi-model intercomparison study on diurnal cycle of precipitation, which is aimed to understand the processes that control the diurnal variation of precipitation over different climate regimes. The study focuses on the interaction between convection and its environments, afternoon and nocturnal convection over land, and convection transition. It used a hierarchy of models with different levels of complexity to diagnose and investigate the associated processes and model biases in simulation of the diurnal cycle of precipitation. Confronting models with detailed observations allowed to identify the deficiencies and missing physics in current weather and climate models and gain insights for further improving the parameterization of convection in General Circulation Models (GCMs). Results from the recently completed long-term single-column model intercomparison and GCM intercomparison studies indicate that most of the participating models share common model biases in simulating the diurnal cycle of precipitation, as already illustrated in previous studies, such as the precipitation peak occurring too early during the day, a lack of nocturnal precipitation and transition from shallow to deep convection. The issues are primarily related to deficiencies in cumulus parameterizations. Sensitivity tests with different cumulus parameterizations suggest that a unified treatment of shallow and deep convection could better capture the transition from shallow to deep convection and help delay the daytime precipitation peak to late afternoon over land. However, this does not improve the simulation of nocturnal precipitation that is often caused by elevated convection associated with the passage of meso-scale convective systems. The key to capture the observed nocturnal peak is to allow elevated convection to be captured by incorporating a mid-level convection parameterization or removing the restriction of the source layer for launching air parcel within the boundary layer. Including convective memory in cumulus parameterizations acts to suppress light-to-moderate rain and promote intense rainfall, however, it weakens the diurnal variability of precipitation and does not show an improvement in the simulation of the diurnal cycle. Results also suggest that simply increasing model resolution cannot fully resolve the biases the diurnal cycle of precipitation in low-resolution models as long as cumulus parameterizations are needed. The hierarchical modeling framework is useful in identifying missing physics in GCMs and testing new developments of model physical parameterizations over different convective regimes.