PNNL

Abstract

Simulations using the Weather Research and Forecasting (WRF) model are conducted at grid spacings of 36 km, 12 km and 4 km for two summers in 2006 and 2007 over the conterminous United States and are compared to systematically evaluate the effects of model resolution and convective parameterizations across the gray zone resolutions. The convection permitting simulations at 4 km grid spacing are most skillful in reproducing the observed spatial distributions and diurnal variability. Notable differences are found between simulations with the Kain-Fritsch (KF) and the scale-aware Grell-Freitas (GF) convection schemes, with the latter being more skillful even at a grid spacing of 36 km in capturing the nocturnal timing that dominates in the Great Plains and North America monsoon region. The GF scheme also simulates a smoother transition from convective to large-scale precipitation as resolution increases, thus resulting in much reduced sensitivity to model resolution compared to the KF scheme. The GF scheme at 4 km is almost inactive, so precipitation is produced mainly by resolved processes and the cloud become shallower and more comparable to simulations at 4 km grid spacing with no convective parameterization. Non-hydrostatic dynamics has a positive impact on precipitation simulations over complex terrain even at 12 km and 36 km grid spacings. Using simulations where winds are nudged towards observations, we show that the conspicuous warm biases in the southern Great Plains is related to precipitation biases induced by large-scale circulation biases, which are insensitive to model resolution. Despite improving the summer season precipitation, neither nudging nor non-hydrostatic dynamics has any impact on the diurnal timing of precipitation, which is more dominantly controlled by the representation of convection that is sensitive to model resolution and convective parameterizations.