A Regional Model Simulation of the 1991 Severe Precipitation Event over the Yangtze–Huai River Valley. Part II: Model Bias

Abstract

This is the second part of a study investigating the 1991 severe precipitation event over the Yangtze?Huai River valley (YHRV) in China using both observations and regional model simulations. While Part I reported on the Mei-yu front and its association with large-scale circulation, this study documents the biases associated with the treatment of the lateral boundary in the regional model. Two aspects of the biases were studied: the driving field, which provides large-scale boundary forcing, and the coupling scheme, which specifies how the forcing is adopted by the model. The former bias is defined as model uncertainty because it is not related to the model itself, while the latter bias (as well as those biases attributed to other sources) is referred to as model error. These two aspects were examined by analyzing the regional model simulations of the 1991 summer severe precipitation event over YHRV using different driving fields (ECMWF?TOGA objective analysis, ECMWF reanalysis, and NCEP?NCAR reanalysis) and coupling scheme (distribution function of the nudging coefficient and width of the buffer zone). Spectral analysis was also used to study the frequency distribution of the bias. The analyses suggest that the 200-hPa winds, 500-hPa geopotential height, and 850-hPa winds and water vapor mixing ratio, which have dominant influences on Mei-yu evolution, are sensitive to large-scale boundary forcing. In particular the 500-hPa geopotential height, and 850-hPa water vapor mixing ratio near the Tibetan Plateau and over the western Pacific Oceans are highly dependent on the driving field. On the other hand, the water vapor in the lower troposphere, wind at all levels, and precipitation pattern are much more affected by the treatment of nudging in the coupling scheme. It is interesting to find that the two commonly used coupling schemes, the lateral boundary coupling and the spectral coupling, provide similar large-scale information to the simulation domain when the former scheme used a wider buffer zone and stronger nudging coefficient. Systematical model errors, existing in the north of the simulation domain, are caused by the overprediction of low-level inversion stratiform clouds. The analyses further indicate that the model mesoscale signal is not significantly influenced by the different treatments of the nudging procedure. However, it is also shown that the model performance, especially the monthly mean precipitation and its spatial pattern, is substantially improved with the increase of buffer zone width and nudging coefficient.