Using T639 model products and the precipitation observations from 25 national automatic weather stations in Ningxia between May and September from 2010 to 2013, we have established the precipitation probability forecast equations of light rain, moderate rain and heavy rain within forecast time length of 168 h between May and September in Ningxia based on probability regression and cross-validation method, and compared and validated the forecast skills of objective T639-MOS products, T639 simulations and NMC forecasts on precipita-tion in Ningxia between May and September in 2014. Results show that comparing to T639 and NMC, T639-MOS presents much lower false alarm ratio, a little higher miss ratio and much higher forecast accuracy, and its forecast skill increases by an average of about 10%. T639-MOS has a good forecast result to large-scale precipitation and severe precipitation, especially with the approach of forecast time and the function adjustment of numerical forecast model. Representativeness and physical meaning of predictors selected are definite based on the objective forecast method. Water vapor, dynamic and thermodynamic factors are important for precipitation forecast in Ningxia. A key forecast factor of light rain is water vapor flux in the mid-and low-level, and those for moderate rain and heavy rain are moist potential vorticity and meridional wind. Additionally, the configuration of critical dynamic factors in the upper-and low-level is considered for heavy rain. Samples of precipitation in Ningxia are relatively low, and most of them are light rain;precipitation samples for moderate rain are only a few. So, the higher the precipitation intensity is, the higher the miss ratio is, and the lower the forecast accuracy is. Nevertheless, the T639-MOS has the highest forecast accuracy to the different levels of precipitation.%利用2010—2013年5—9月T639模式产品和同时段宁夏25个国家级气象站降水实况资料，应用概率回归和交叉验证方法，建立宁夏汛期5—9月小雨、中雨、大雨3个级别的168 h分级降水概率预报方程，对比检验客观产品T639-MOS、数值模式T639和上级指导产品NMC对2014年5—9月降水预报效果。检验结果表明：相较T639和NMC， T639-MOS空报率明显降低，漏报率略有增加，但预报准确率明显提高，预报技巧较T639和NMC平均提高10%左右；T639-MOS对大范围降水和强降水预报效果较好，且随着预报时效临近，数值模式预报性能调整，预报能力显著提高；客观方法选取的预报因子代表性和物理意义明确，水汽、动力和热力因子是宁夏降水预报的关键因子，其中，小雨预报关键因子是中低层水汽通量，中雨和大雨预报关键因子是湿位涡和经向风，大雨预报还考虑了关键动力因子的高低层配置；宁夏降水样本数相对较少，且主要以小雨为主，中雨以上降水是小概率天气，所以，降水强度越大，漏报率越高，预报准确率明显下降，但不同级别降水都是T639-MOS预报准确率最高。
展开▼
