针对大渡河流域地形复杂，高差悬殊，气候垂直差异大等特点，构建基于四层嵌套网格的大渡河流域WRF模型：采用四维变分（4DVar）和集合卡曼滤波（EnKF）混合同化技术降低初始场误差，根据统计最优组合方法确定年内不同时期WRF模式参数方案集。将WRF输出作为预见期内新安江模型的输入，构建大渡河流域WRF新安江来水预报模型。选择2010—2017年汛期的20场、非汛期10场来水过程进行应用研究，结果表明：在预见期延长7 d的情况下，所建模型仍能够保证较高的预报精度。研究成果可为大渡河流域水力发电和防汛决策提供技术支撑，也可为同类研究提供借鉴。

Hydrometeorological forecasting is a major measure to improve the accuracy and extend the lead time of runoff prediction for a large watershed,and remains a challenge to hydrosciences.Although much progress has been made in the research of hydrological forecasting based on the combination of meteorology and hydrology in China,the results are mainly focused on theoretical methods,and there are relatively few highlights on the application level,especially for operational forecasts.Dadu River basin has the characteristics of complex topography,great height change and vertical climate variations,which brings extra difficulties for the hydrometeorological forecasting.Therefore,the Dadu River basin is taken as the research object,aiming to serve the scientific decision-making of flood control and power generation for the basin through scale conversion and the connection of meteorological and hydrological elements,and establishes a runoff forecast model for operational purpose based on the meteorologichydrologic coupling. A four-layer nested grid WRF model was used to solve the difficulties of microclimate simulation caused by the large changes in topography and severe vertical variability of the Dadu River basin.Assimilation technologies including four-dimensional variational (4DVAR) and Ensemble Kalman filter (EnKF) hybrid were used to reduce the initial meteorological field error.The statistical optimal combination was employed to determine the model parameters of WRF in different periods of the year.The Xin′anjiang model was employed as the catchment-scale hydrological model since the Dadu River basin is a wet region (although there is a small part of the semihumid area in the upper reaches).The outputs of future rainfall from WRF were used as the inputs of the Xin′anjiang model.Therefore,a WRF-Xin′anjiang model was established for runoff prediction for the Dadu River basin. The constructed WRF model based on the four-layer nested grid for Dadu River basin can forecast rainfall in the next 7 days,and the overall forecast accuracy (qualification rate) reached over 80%.For the Xin′anjiang model,the qualified rates in flood season and non-flood season were 100% and 86%,respectively,and the average determination coefficients were 090 and 086,respectively.In the verification period,the qualified rates in flood season and nonflood season were both 100%,and the average determination coefficients were 089 and 086,respectively.For the WRF-Xin′anjiang model,20 flood events were chosen in flood season and 10 flood events in nonflood season from 2010 to 2017 to verify the model.Results showed that the qualified rates of flood peak and flood volume in flood season both reached 100%,and the qualified rates of flood peak and flood volume in nonflood season reached 100% and 90%,respectively.It is indicated that with an extension of the lead time to extra 7 days,the WRFXin′anjiang model established could still guarantee a higher forecast accuracy. The WRF-Xin′anjiang model established can provide a reference for hydrological forecasting of the basin with complex terrain,wide height difference and significant vertical climate change,and provide technical support for hydropower generation and flood control decisionmaking of Dadu River basin.At the same time,with the accumulation of hydrometeorological data in the future,it is necessary to do indepth research on more refined WRF and hydrological model for the basin.