多因子和多尺度合成中国夏季降水预测模型及预报试验

根据青藏高原60个站平均的月积雪深度、热带太平洋Nino 3区月海温和中国160个站月降水量等资料,用小波变换和相关分析,分析了1958～1998年秋冬季青藏高原异常雪盖与El Nino-南方涛动(ENSO)的关系、多时间尺度变化的特征及其与中国夏季降水的相关型式.并取青藏高原积雪和Nino 3区海温的年际变化、年代际变化和线性趋势三种不同时间尺度的小波分量作为预报因子,对我国夏季降水距平作线性回归,建立了相应的预测模型.最后,利用1999～2002年的独立资料进行了预报试验,并在2003年和2004年应用于实际预报.研究表明,青藏高原雪盖与ENSO这两个物理因子彼此具有一定的独立性.它们都是多时间尺度现象,并与中国夏季降水有较好的关系.在不同时间尺度上不仅有不同的相关型式,而且相对贡献也有变化.回归预测模型的拟合情况和预报试验表明,综合考虑前期秋冬季青藏高原雪盖和ENSO这两个物理因子的年际变化、年代际变化和线性趋势作为预报因子建立的预测我国夏季降水距平分布的模型,有一定的预报能力.

The Statistic Prediction Model and Prediction Experiments of the Summer Rain over China by Multiple Factors and Multi-Scale Variations

The roles of snow cover over the Qinghai-Xizang Plateau(QXP) and the sea surface temperature anomaly(SSTA) in the central and eastern Pacific Ocean in seasonal prediction of summer rain over China have been acknowledged for decades.The multi-scale variations of the snow depth over the QXP and SSTA in the Nio 3 region in the precedent autumn and winter are exploited for developing a statistical model to predict the summer rain over China.Monthly snow depth averaged at 60 stations over the QXP,the monthly SSTA and monthly rainfall at 160 stations in China from 1958 to 2004 are used.The snow cover and SSTA are independent of each other,for their correlation coefficient is only 0.05.Wavelet analysis is used to identify the interannual,interdecadal and linear trend components of the snow cover and SSTA.The summer precipitation over China has significant relationships with the multi-scale variations of snow cover and SSTA in precedent seasons with different correlation patterns and relative contributions at different time-scales.Taking the long-term variations of snow cover and SSTA into account can potentially contribute to additional forecast skill in the summer rain over China.The interannual,interdecadal and linear trend variations of snow cover and SSTA are taken as key predictors.The statistical model is established by using linear regression analysis.Three experiments are designed to select the best combination of predictors.The first set of predictors is the multi-scale variations of snow cover,the second is the multi-scale variations of the SSTA,and the third is the combination of the two sets.The data from 1958 to 1998 are used to assess the prediction skill of each model by the average probability of the same symbol(APSS) of the regressed and observed precipitation in each year.The mean APSSs in 31 years of the three experiments are 0.61,0.59 and 0.64 respectively.Thus,the prediction model by multiple-factors and multi-scale variations is effective for predicting the distribution of summer rain anomalies over China.The prediction experiments are carried out with the independent data from 1999 to 2002,and operational predictions in 2003 and 2004 are made by multiple factors and multi-scale variations.From 1999 to 2002,the predicted precipitation anomalies show that the statistical model predicts the spatial pattern of the observed anomalies reasonably well.In 2003,the model succeeded in predicting the precipitation in the Huaihe River valley,the Weihe River valley,Xinjiang Uygur Autonomous Region and the middle part of the Inner Mongolia Autonomous Region.The model also predicted the deficient rainfall in the eastern part of Northeast China and the northern part of North China,but missed the significant drought in the south of the Changjiang River valley.APSS in this year is 43.8%.In 2004,the main differences between the observation and the forecast are found in some areas of the Huanghe River-Huaihe River valley and in the south of Changjiang River valley.In spite of these differences,the distribution of predicted above- and below-normal rainfall agrees with the observation.APSS in 2004 is up to 57.5%,considerably better than that in 2003.Overall,the results suggest that multi-scale variations of the snow cover over the QXP and the SSTA in the central and eastern Pacific make significant contribution to seasonal climate prediction.Other statistical methods,rather than the linear regression,may increase the forecast skill.