中文摘要：

为验证贝叶斯模式平均（BMA）在东江流域的适用性,基于TIGGE多模式集合预报资料对东江流域不同预报时段（1、3、5及10日）以及不同地域（上、中及下游）应用BMA,得到的主要结论为：1）BMA的预报能力在时空上都较为稳定,预报效果较好;2）在各预报时段中BMA极大提高了预报的精度,其中预报时段越短,误差减少的幅度越大;3）当95%分位数已超过警戒的雨量时,应做好相关的预警和预防工作,但BMA也会出现误报和虚报的情况,需要进行权衡和风险的评估,同时这也是需进一步研究和完善的地方;4）结合其他国内外的研究来看,BMA对天气要素预报的订正作用可能具有普适性,值得推广使用。

英文摘要：

To evaluate the performance of BMA（Bayesian Model Averaging） in forecasting precipitation in the Dongjiang basin and to provide the theoretical basis for the storm early warning system, the BMA-based prediction method is applied into precipitation forecast in the Dongjiang Basin. The evaluation is conducted for different lead time（i.e., 1-day, 3-days, 5-days, and 10-days） and different regions（i.e. up, mid and downstream） by using the TIGGE（The THORPEX Interactive Grand Global Ensemble） multi-model products. Conclusions are drawn as follows： 1） The forecasting skill of BMA is fairly good in forecasting accuracy for almost all regions and lead time. 2） The forecasting accuracy is improved by using BMA as compared with the original ensemble means, and the improvement is more significant for short lead time. 3） If the 95% percentile rainfall exceeds the disastrous rainfall amount, the early warning system and the preventive work should be initiated; however, the BMA forecasting results should be used with caution as the false warning would occasionally happen. 4） According to the results from our study, together with findings from other studies over different areas, we suggest that BMA would be suitable for precipitation forecasting and worth recommending.