中文摘要：

以2005年8月开始运行的、8个成员的上海区域降水集合预报系统为基础，设计2个对比试验方案，进行了3个月（2005年9—11月）的平行对比试验。对比试验将成员从8个增加至12个，系统的8个成员与试验一增加的成员都从预报模式的不确定性出发形成，试验二增加的成员考虑了模式初始条件的不确定性。对试验结果进行了检验与分析，并与控制试验结果进行比较。结果显示：增加集合成员数可以增大系统发散度，但对比试验仍存在系统发散度偏小的问题；同时考虑预报模式与初始条件不确定性的试验二的降水集合平均预报效果与降水概率预报效果都好于只考虑预报模式不确定性的试验一，也好于控制试验，试验一的降水集合平均预报效果总体上则比控制试验差，降水概率预报效果也不理想。采用试验二方案对系统进行改进后的整体预报效果较改进前有提高。

英文摘要：

Two comparing test schemes are designed based on the operational ensemble prediction system for Shanghai regional precipitation. The system has eight members and started operating at Aug. 1, 2005. According to the test schemes, parallel comparing tests were made during Sept. and Nov. 2005. In comparing tests, members are added from eight to twelve. The eight members in the system and the added members in test 1 are created according to the model uncertainty. The added members in test 2 are created according to the initial condition uncertainty. Comparing test results are verified and compared with the control test results. The results show although adding members can enlarge the spread, comparing tests still have the problem of too small spread. The ensemble mean forecasts and probability forecasts of test 2 in which both model uncertainty and initial condition uncertainty is considered are better than that of test 1 in which only model uncertainty is considered. The ensemblemean forecasts and probability forecasts of test 2 are also better than that of the control test. On the contrary, the ensemble mean forecasts of test 1 are worse than that of the control test. The probability forecasts of test 2 are not good either. The system is improved according to the scheme of test 2. The new system can offer a better precipitation forecasting than the old system.