次季节多模式集合预测在陕西汛期降水的应用

Application Study on Subseasonal Multi-Model Ensemble Prediction of Precipitation during Flood-Season in Shaanxi

摘要: 为提升陕西汛期降水预测准确率，本研究基于多模式集合(MME)预测数据，构建时空投影模型的动力–统计相结合的预测方法(DSTPM)，开展延伸期降水预报试验。采用“模式误差订正 + 物理因子约束”的释用技术，融合不同气候区特征差异优化模型。结果表明：(1) 通过时空尺度检验评估，多模式集合预测效果明显高于单一模式，且随着候数增加下降缓慢。(2) 长波辐射(OLR)在6种预报因子中预报技巧最高且在不同区域均表现稳定。(3) DSTPM对陕北、关中及陕南中东部的降水预报效果较好，时间相关系数(TCC)技巧较MME本身提升3~6%，但陕南西部的预报能力仍有待提升。基于次季节–季节(S2S)多模式预报产品探索构建动力–统计降尺度预报方法，可为抗旱防涝调度、区域水资源管理提供关键预报信息，助力提升应对汛期极端降水事件的前瞻性防灾减灾能力。

Abstract: To improve the accuracy of flood-season precipitation prediction in Shaanxi, this study develops a dynamic-statistical forecasting method based on a spatiotemporal projection model (DSTPM), using Multi-Model Ensemble forecast data, and conducts extended-range precipitation prediction experiments. A model post-processing approach that combines model bias correction with constraints from physical predictors is adopted, while differences among climatic regions are incorporated to optimize the model. The results show that: (1) Evaluations across temporal and spatial scales indicate that the MME markedly outperforms individual models, and its fore-cast skill decreases only gradually as the lead time in pentads increases. (2) Among the six fore-cast predictors, Outgoing Longwave Radiation (OLR) exhibits the highest forecasting skill and performs consistently across different regions. (3) The DSTPM performs well in northern Shaanxi, central and eastern parts of Guanzhong, and southern Shaanxi, and the Time Correlation Coefficient skill is 3%~6% higher than that of the MME alone, although its predictive capability in the western southern Shaanxi still requires improvement. The development of a dynamic-statistical downscaling forecasting method based on Subseasonal-to-Seasonal (S2S) multi-model forecast products can provide critical forecast information for drought and flood mitigation operations and regional water resource management, thereby strengthening anticipatory disaster prevention and risk-reduction capacity for extreme precipitation events during the flood season.