Research On The Evaluation Of The Medium-term Extreme Temperature Forecast In My Country Based On The Ensemble Forecast Data

In this paper,we first use the historical data of T639 real-time ensemble forecast system to estimate and diagnose the distribution of the model climate,which is the key link of the extreme temperature forecasting,and further to verify it by using the observations in China during 2016.Studies have shown that the T639 ensemble system has certain ability for the extreme temperature prediction in China,but the forecasting skill varies greatly in different regions of China.From the national mean TS score,the use of the ensemble maximum method can significantly improve the prediction skills of extreme high temperatures,while the ensemble mode method significantly improves the prediction skills of extreme low temperatures.However,since the current perturbation method is difficult to form fully discretized ensemble members,and too few ensemble members in NCEP-GEFS,those factors may have an impact on predictive skills in extreme temperatures.At present,only the ensemble mean method,the ensemble maximum or minimum method and the ensemble mode method are used to examine the prediction performance of the extreme temperature.Our results suggest that the correct and appropriate method is essential for the extraction of extreme information in ensemble prediction,which can significantly improve the ability of the model to predict extreme events.Secondly,according to the reforecasts of NCEP-GEFS,the variation and frequency distribution of extreme temperature during winter and summer in China are analyzed,and the variation of extreme temperature frequency and its relationship with the mean state are further discussed.Based on the information of the ensemble members,the historical extreme low/high temperature of China's winter/summer is also verified.The results showed that the amplitude and frequency distribution of extreme temperature in winter and summer in different regions of China are quite different,showing strong topographic dependence.Therefore,the study of extreme temperature in China needs to be carried out in different regions.Among them,the average winter temperature(the frequency of extreme low temperature occurrence)in various regions of China shows a clear rising(decreasing)trend(excluding northeast China),and the rising(declining)trend in southern China is more obvious.For probabilistic forecasting,this model has a certain predictive performance for six sub-regions in China's winter,and the highest forecasting skill is over the Yangtze River Basin.Comparing the prediction methods of the three ensemble prediction methods,the method of ensemble mean and minimum can effectively improve the forecasting skill of extreme low temperature in winter in China when the lead time is less than 5 days and more than 5 days,respectively.Unlike the extreme low temperature forecast in winter,the regional average temperature and extreme high temperature frequency and their trend show large differences in the six sub-regions of China in the summer before 1993 and after 1994,indicating that extreme high temperatures are closely related to background temperature.For probabilistic prediction,the model has a certain predictive performance for the extreme high temperatures of six sub-regions in China and has the highest predictive skill for the Northeast,North China and Yangtze River basins.Comparing the prediction methods of the three ensemble prediction methods,the method of ensemble mean and ensemble maximum can effectively improve the forecasting skill of extreme high temperature in China when the lead time is less than 4 days and more than 5 days,respectively.Finally,based on NCEP-GEFS historical reforecast data and uniform grid observational data,the variation characteristics of extreme temperature in winter(summer)season in China over the past 30 years were verified.The results showed that the NCEP-GEFS historical reforecasts can reproduce the interannual variation and trend of the average temperature and extreme temperature in winter and summer in China.In addition,the probability density function distribution of the NCEP-GEFS forecast field is significantly different from that of the observation,especially in the high-end tail portion.Since this model has systematic deviations in the identification of extreme temperatures,the study of extreme temperatures predicted by the model cannot simply be applied to the analysis of extreme temperatures in observation.It is recommended to define extreme temperatures by using relative thresholds based on climate percentiles when studying extreme temperatures,which can effectively correct systematic bias in NCEP-GEFS reforecasts.