UCLA

Recent studies have shown that spring land surface temperature (LST) and subsurface temperature (SUBT) over the high elevation areas in the western US (WUS) have significant impacts on the downstream summer droughts/floods in North America. In this paper, both the National Centers for Environmental Prediction—Global Forecast System (NCEP-GFS) general circulation model (GCM) and the weather research and forecasting (WRF) regional climate model (RCM) are employed, where RCM scenarios utilized initial and lateral boundary conditions derived from the corresponding NCEP-GFS scenarios. Here we use a late spring flood in the US Southern Great Plains (SGP) case to examine whether simulation of the LST/SUBT downstream effects is sensitive to the domain size choice, change in dynamical cores within the same model, as well as to the representation of surface processes parameterizations. Although all RCM experiments with different settings simulate reasonably geographical patterns of observed LST and precipitation anomalies, we found that the choice of the domain size is crucial for proper downscaling the LST/SUBT downstream effects to accurately produce the observed precipitation/LST anomalies over the SGP/WUS, respectively, along with the associated large-scale features. The southern boundary location has been identified to be crucial in producing the SGP Low Level Jet strength, which in turn brings more moisture from the Gulf of Mexico to the SGP and thereby resulting in a better simulation of the precipitation anomaly in that area. The sensitivity of the simulation of the LST/SUBT downstream effect to dynamical cores is assessed by inter-comparing the Non-hydrostatic Mesoscale Model (NMM) and the Advanced Research WRF dynamic cores. We find NMM was better at generating the large-scale eastward wave train, a crucial process associated with the LST/SUBT downstream effect. Meanwhile, this study also shows that the LST/SUBT downstream effects were not significantly dependent on the surface process parameterizations, although the Simplified Simple Biosphere model version 3 (SSiB3) highlighted a better performance over SSiB2.