Calibration of Probalistic Quantitative Precipitation Forecasts From The Rsm Ensemble Forecasts Over Hydrologic Regions

Abstract

Cool season precipitation plays an important role in freshwater supply over the Southwest United States, which is marked by heterogeneous topographic and hydrologic scenarios. More accurate precipitation prediction is highly desirable for both the public and hydrological model users. Numerous studies indicate that ensemble forecasting provides more skillful weather forecasts than a single deterministic forecast run. The National Centers Environmental Prediction (NCEP) Regional Spectral Model (RSM, Juang and Kanamitsu 1994) ensemble system was performed to forecast daily precipitation during winter 2002-2003 over the southwest US (Yuan et al. 2004). Probabilistic quantitative precipitation forecasts (PQPF) from 11 ensemble members are good at discriminating precipitation events in terms of high relative characteristic curve (Wilks 1995) areas. The forecast skill presents large spatial variation, with the highest skills over the California region. However, significant wet biases in the RSM forecasts result in some low statistical scores and unskillful forecast indices, in particular, over the Colorado Basin River Forecast Center (CBRFC) and the Great Basin Region. It is indispensable to calibrate such biases to increase the accuracy of PQPF and quantify the real atmospheric uncertainties in weather forecasts. An artificial neural network is applied to conduct this post-processing over four US Geological Survey (USGS) hydrologic Unit Regions: the Upper Colorado Region, the Lower Colorado Region, the Great Basin Region, and the California Region (see Fig. 1, Yuan et al. 2004).