P54  QPF Verification of a WRF-ARW-based Storm-Scale Real-Time Forecasting System in Southern China for the 2011 Summer Months

Chen, Xunlai, Shenzhen Meteorological Bureau, China, Center for Analysis and Prediction of Storms, University of Oklahoma, USA, and Fanyou Kong, Tuanjie Hou, Center for Analysis and Prediction of Storms, University of Oklahoma, USA, and Qunfeng Zhen, Shenzhen Meteorological Bureau, China

The hourly assimilation and prediction system (HAPS), a WRF-ARW based realtime storm-scale forecasting system that was in operation in Shenzhen Meteorological Bureau since March 2010, has been verified for QPF during the 2011 summer months. The HAPS system was developed collaboratively by Shenzhen Meteorological Bureau (SZMB), the Center for Analysis and Prediction of Storms (CAPS) in the University of Oklahoma, and the Shenzhen Institute of Advanced Technology (SIAT) of Chinese Academy of Sciences. The HAPS system consists of an outer domain with 12-km horizontal grid spacing and a one-way nested high-resolution domain at 4-km grid spacing. Initial and boundary conditions for the 12-km horizontal resolution grid are provided by the fine-resolution European Center for Medium-Range Weather Forecasts (ECMWF) data initiated at 00 and 12 UTC each day. The radial velocity and reflectivity data from 7 operational WSR-98D radars in Guangdong province are analyzed into the initial condition at the 4-km domain using the ARPS 3DVAR and cloud analysis system every hour. Starting in March 2011, the HAPS system has been run operationally, producing 48h forecasts at 12-km domain twice daily and 12h forecasts at 4-km domain every hour. The precipitation forecasts at both 12- and 4-km domains are verified against rain gauge measurements from 1400 automatic weather stations (AWS) in the region. Bias score (BIAS), root-mean-square error (RMSE), and equitable threat scores (ETS) are computed during 2011 summer months to evaluate the ability of the HAPS system on QPF. The results show that the HAPS system is skillful throughout the entire forecast period. The skill deteriorates with increasing threshold and forecast lead time. There is a systematic improvement in terms of the accuracies and skill when the model horizontal resolution is increased from 12km to 4km.