Research On Joint Probability Prediction Of Severe Convective Weather Based On Ensemble Forecast

By using the U.S. NCEP global Ensemble Forecast System （GEFS） products, summer daily precipitation data in northwest China in recent 50 years, and convective weather case information in northwest China in nearly 5 years, we firstly statistical analyzed the nonlinear evolution and forecast jump features of NCEP ensemble forecast products, also studied summer heavy precipitation events frequency variation characteristics and atmospheric physical conditions in northwest China. Based on the above researches, we improved "two-step" and the joint probability forecasting method, using case-based reasoning （CBR） and combined programming technology to generate strong convective weather precipitation prediction based on ensemble forecast in certainty, level and probabilistic forecasts. We got following conclusions:（1） Nonlinear evolution of the GEFS products in Asia is less than that in North America, and also a small non-linear change showed in the summer in both Asia and North America.（2） Ensemble mean forecast has better prediction consistency than corresponding control forecast. When the forecast lead-time is greater than or equal to 240 h, ensemble mean forecasts’time-averaged predict jumping index is typically only 25 to 50% of corresponding control forecasts. Therefore, using ensemble mean forecast can effectively reduce the incidence of "forecast jump". During the summer, the frequency of "forecast jump" phenomenon fluctuates around 17.5%, does not show substantial growth compared with the annual frequency ranged 10～20%.（3） Low level multi-year average temperature and humidity significantly increased by 0.2℃ and 3% respectively, and high level multi-year average temperature and humidity declined by 0.2℃ and 6% respectively in northwest China in recent 10 years. It causes atmosphere instability increased, summer average K index increased higher than 2.5℃ over the previous 10 years, atmospheric convection frequented, and the multi-year average frequency of heavy rainfalls and strong convective weather extremes increased by 0.6 d over the previous 10 years.（4） By using ensemble forecast products and technology in strong convective weather forecasting, we can improve the accuracy forecast by "two-step" method which improves Ts and Bs score by 0.5 and 0.1 respectively in one experiment.（5） We also can produce severe convective weather probabilistic forecast by the joint probability forecasting and other methods. After comparative analysis, the joint probability forecast usually chose CAPE180₀mb,850hPa vertical velocity,850hPa relative humidity and whole layer precipitable water as atmospheric physical indicators. The use of artificial intelligence technology （e.g. CBR） and joint programming techniques can automatically determine the threshold value of these indecies, and quickly calculate the probability of severe convective weather in the target area.