The Very Short Range High Resolution Time-lag Ensemble Forecasting Based On Smb-warr

As a result of the accuracy of extrapolation radar echoes or satellite imagery of clouds decreases rapidly with time,particularly during the first hour and numerical forecast system that routinely updated every six hours or long is unsuited for the nowcasting,so the rapid refresh numerical model forecast has played important roles in the nowcasting. However,the high resolution numerical model and high frequency data assimilation cause rising uncertainty,we research the very short range high resolution ensemble forecast (VSRHREF) system. The VSRHREF is based on East China regional operational numerical forecast model SMB-WARR (Shanghai Meterological Bureau WRF ADAS Rapid Refresh),runs using a time-lag ensembling technique. The VSRHREF is routinely updated every hour with hourly output through6hr forecast length. With the method of arithmetic average, it can provide area temperature, rainfall, probability forecast and so on, to improve the ability of the city’s fining forecast.Based on data from AWS (automatic weather station).evaluation of the hourly area temperature and rainfall from model output of Shanghai in flood season shows that:1) Observational data: we controlled the quality of data with Barnes method and analyzed the characteristic of representative areas. The temperature of land stations and sea stations has significant day-change characters without precipitation. The difference of them is minimum at8am and11pm, but in the daytime (night),the temperature of land stations (sea stations) is higher than sea stations (land stations),besides, the difference is owe to the sunlight and the speed and direction of wind. Further, the AWS is sensitive to the changes in the weather event and can reflect urban heat island effect.2) Rainfall:The ensemble mean is better than any other members from light to heavy rain grades, but worse than some members in rainstorm grade. Meanwhile, the most close to forecast, the skill is not the best and the little to heavy rain could be well forecasted6hours in ahead. Furthermore, the probability forecast has an advantage over the ensemble mean, and has good directions for the happening of rain in the very short range, and the big(small)probability is more useful from light to moderate rain grades (heavy rain to rainstorm grades).3) Temperature:Remove the forecast bias by the way of the adaptive (kalman filter type) algorithm. The uncorrected ensemble member ml has better performance in forecasting temperature of the whole Shanghai. It’s worth noting that the equally weighted ensembles (EWE) has good directions than all ensemble members for forecasting temperature of the central city, but after correction all ensemble members and EWE’s skill could be improved,meanwhile, the EWE has better performance than any other members. Besides, the rank distribution changes from the adverse "L" type to "U" type, and the reliability of probabilistic also prove the improvement of the corrected ensemble forecast, and could be well forecasted6hours in ahead. Then, the unequally weighted ensemble (UEWE) performs better than EWE.