中文摘要：

通过分析1965～1999年西北东部汛期降水与相关物理量年际增量的相关性，筛选出可作为西北东部汛期降水的预测因子，并研究其影响该区域汛期降水的物理机制。在此基础上，采用多元线性回归方法建立了汛期降水年际增量与预测因子的物理统计预测模型，并对2000—2014年的汛期降水进行预测。结果表明，5月Nifio3．4指数、1月北太平洋环流指数和西太副高强度指数、5月北半球极涡中心强度指数和东亚大槽强度指数的年际增量等5个变量可作为西北东部汛期降水的预测因子；该预测模型对2000—2014年汛期降水的预测准确率很高，预测与观测的汛期降水演变趋势非常一致，相对误差在±15％以内的有12a，距平同号率达到10／15，距平均方根误差为15％，可用于西北东部汛期降水的预测，能够提高其预测水平。

英文摘要：

Based on the monthly precipitation data during 1965 -1999, NCEP/NCAR reanalysis data with 2.5° × 2.5° spatial resolu- tion and the sea surface temperature from NOAA with 2 °× 2° spatial resolution, as well as circulation indexes and climatic indexes, the predictive factors of the year - to - year increment of precipitation in the east of Northwest China during the flood period were selected by using the correlation analysis, and the physical mechanism about the influence of predictive factors on precipitation was analyzed. On this basis the physically - based statistical forecast model between them was established by using multiple linear regression method, and the precipitation during the flood period from 2000 to 2014 were predicted. The results showed that the predictive factors were the year - to - year increments of the Nifio3.4 index in May, circulation index in the North Pacific and strength index of the Western Pacific subtropical high in January, central intensity index of the polar vortex in the Northern Hemisphere and intensity index of the East Asia deep trough in May. The forecast model exhibited a quite high accuracy in the predicted precipitation during the flood period from 2000 to 2014, and the evolution tendency of the predicted precipitation was very consistent with the observed. The relative error between the predicted precipitation and the observed in 12 years was within plus or minus 15%, the rate of precipitation anomalies with the same sign was 10/15, and the RMSE was 15%. As a result, this model could be used to predict the precipitation in the east of Northwest China during the flood period, and the forecast capability would also be improved.