Systematic errors of QPF in low-mountain regions as revealed by MM5 simulations

Abstract

Numerical simulations were performed with the mesoscale community model MM5 in order to investigate the simulation of orographically-influenced convective precipitation in southwestern Germany. With a representative set of 13 cases during Summer 2005, we performed sensitivity experiments with 7 km resolution and a nested model configuration with a horizontal grid spacing of 1 km. With parameterized deep convection at 7 km horizontal resolution, mainly three types of systematic errors were detected: Precipitation was strongly overestimated on the windward side and underestimated in the lee of low-mountain ranges, which is called the "windward/lee effect". Convection was triggered systematically too early by several hours and the simulated precipitation was distributed over larger areas with underestimated peak rain rates. In simulations with a grid size of 1 km with explicitly simulated deep convection, all these model errors were strongly reduced. However, a significant underestimation of precipitation appeared, and the timing error changed into a lag of two hours in the occurrence of the precipitation maximum in the afternoon. For a subset of the 13 cases, sensitivity studies using three different boundary layer and two land-surface schemes were performed and compared with observations of surface networks including Global Positioning System integrated water-vapor measurements. The land-surface scheme had a large impact on the simulated surface temperature and moisture fields. Most realistic results were obtained with the boundary layer scheme of the NCEP Medium Range Forecast Model (MRF) in combination with the MM5 5-layer soil model. Consequently, this configuration was selected for the development of an operational data assimilation and forecasting system during COPS.