Ensemble Forecasting Performance Studies On The Important Medium-Range Circulation Processes Affecting China

Based on the 500hPa geopotential height forecasts fields which are gotten from NCEP EPS, and a lot of ensemble products and methods on forecast verification and evaluation, the ensemble prediction performance on the summer important middle-range circulation influencing China are analyzed. Firstly the ensemble forecasting performance at the Asia-Europe region is compared with the Northern America, the ensemble performance on western Pacific subtropical high region and Asia-Europe middle-high latitude blocking region are analyzed and compared with the deterministic forecasts in the way of statistical average and case studies secondly, Moreover the low probability circulation events are discriminated and the corresponding ensemble probabilistic forecasts performance is analyzed. Following conclusions:1) As a whole, the NCEP-EPS circulation ensemble performance at the Asia-Europe region is nothing less than the prediction at the North America. It manifested in the following aspects: (1) ACC test shows that in addition to the winter, the ensemble mean forecast of the Asia-Europe region in the other three seasons is clearly superior to the North America region, The maximal lead time of the useable forecasts can be extended 0.6~1d more, even the differences in summer are greater. RMSE test showed that the Asia-Europe region forecast results in the four seasons are over North America region. (2) The seasonal differences about the reliability of the ensemble probability forecasting are not evident. When the lead time is short (long), the reliability in North America (Asia-Europe) region is better. The synoptic events identification scope on the Asia-Europe regions is relatively smaller, but it's relatively higher reliability forecasts, the North American region is on the contrary. (3) Summer ensemble probability prediction results in Asia-Europe region are clearly superior to the North America region. In winter, NorthAmerica region has better results. In autumn, the two regions' results are homological. In spring, when the lead time is less than 5d, rather, the North America region results are better, and after 5d, the Asia-Europe region has better forecasting results.2) In summer, the ensemble prediction performance on western Pacific subtropical high region is better than single deterministic forecasting. It manifested in the following aspects: (1) ACC tests indicate that <7.5d in the forecasting period, the ensemble mean forecast results are superior to a single deterministic forecasting. Compare with low-resolution single forecasts, ensemble mean can improve 1d on the maximal lead time of useable forecasting. The maximal lead time of ensemble mean useable forecasts in the subtropical region is lower than the Asia-Europe region near 1.5d. RMSE test showed that the results of ensemble mean were noticeably higher than the single deterministic forecasting. It is much better than the results of improving the model resolution. (2) The performance of ensemble probabilistic forecasts is clearly higher than the deterministic forecasting. The cause of this situation is that a probability forecast results in the lower probability threshold have higher hit rate and lower false alarm rate. Probability forecasting results also allow greater access to users within the higher economic value.3) In summer, the superiority of ensemble prediction in the Asia-Europe middle-high latitude region is come forth with the extension of the lead time. Manifested in the following aspects: (1) ACC test shows: when the lead time is greater than 5d, ensemble average forecast results are over high-, low-resolution single deterministic prediction. Compare with the low-resolution forecasts, they will extend the maximal lead time of useable forecasts to 12h later, reaching 7.5d. RMSE test results similar to ACC. (2) the ensemble probabilistic results are clearly higher than the deterministic prediction. It will enable more users to higher economic value. (3) The comparison for EV among Asia-Europe middle-high latitude region, western Pacific subtropical high region and all the Asia-Europe region indicate that in different lead time,the range of cost/lost ratio >0 is identical. When lead time is less (more) than9.5d, in Asia-Europe middle-high latitude region (all the Asia-Europe region) the probabilistic forecast results are the best, in western Pacific subtropical high region results are the worst. This is related with the lower predictability in the low latitude circulation system.4) The case studies about a subtropical high northward leap and a blocking system in summer, 2003 also confirm the performance of ensemble forecasting. It manifested in the following aspects: (1) NCEP-EPS can successfully forecast two typical medium-range circulation processes. The deviations of ensemble mean forecasts versus the observation value are less than high-resolution deterministic forecasts. The performance of the blocking is better than the subtropical high. (2) The distribution characteristic of forecast uncertainty, which is gotten from ensemble spread and RMOP, can reflect the forecast inconsistency about the oscillation phase (location) and amplitude (strength). (3) Analysis on the individual characteristics isograms can provide the inconsistent information among ensemble members, get the probability estimation of weather event, and improve the forecast performance.5) Firstly, a distinction method on the low probability events is proposed based on the climatological equally likely bins. Then, the corresponding forecasting evaluation about low probability circulation events is performed, based on the 2003 summer Asia-Europe region dataset. Following conclusions: (1) in the analyzed period and region, the frequency of the low probability events is slightly higher than the climatological average situation. For NCEP-EPS, the forecast frequency for smaller low probability events is easy to be higher, but relatively reliable, for larger low probability is decreased with the extension of lead time. When the lead time is less than 3d (more than 5d), the forecast frequency is higher (lower), with relative less reliable. (2) EV analysis showed that the performance of NCEP-EPS probabilistic forecast for the low probability events is higher than normal weather events. (3) Forecast hit rate analysis shows that the hit rate is lower in the high-probability threshold.Analysis pointed out that this is perhaps related with the ensemble spread. With the lower probability threshold, hit rate steadily increased which means the spread of ensemble members can make them to cover all possible region of observation value, thereby to get the better forecasting results.