WRF/WRF-Hydro气象-水文耦合模式的误差传递特征——以漳河流域径流预报为例

ERROR PROPAGATION CHARACTERISTECS OF WRF/WRF-HYDRO METEOROLOGICAL-HYDROLOGICAL COUPLING MODEL: TAKING THE RUNOFF FORECASTING IN ZHANGHE RIVER BASIN AS AN EXAMPLE

关键词:  气象-水文耦合  预见期  降雨-径流预报  漳河流域  误差传递

Key words:  meteorological-hydrological coupling  forecast period  rainfall-runoff forecasting

摘要:

在中小流域，有效的洪水预见期对调度决策、防洪减灾尤其重要。以漳河流域为研究区域，基于WRF模式和WRF-Hydro水文模型建立WRF/WRF-Hydro气象-水文耦合模式，探讨不同预见期WRF模式预报降雨对径流预报效果的影响，以及气象-水文模型之间的误差传递特征。（1）不同预见期下WRF模式预报的降雨峰值随预见期增长偏差越大。预见期为06 h和12 h时较预见期较长时（24 h、48 h），降雨量和峰值出现时间预报误差较小，降雨落区预报明显较好。（2）基于WRF/WRF-Hydro气象-水文耦合的径流预报结果表明，预见期越长洪峰流量及径流总量偏差越大。06 h和12 h预见期下的径流预报结果都较好，平均相关系数、平均纳什系数分别提高0.15、0.69。（3）模式之间的误差传递会因预见期长短和降雨预报效果有一定的差异。预见期为06 h和12 h时，数值天气预报误差传递给水文模型后有放大、有缩小。预见期较长时（24 h、48 h）随预见期加长，水文模型放大数值天气预报误差的程度越大，但是通过定量分析发现降雨误差和径流误差之间没有明显的线性关系。

Abstract:

Effective flood forecast periods are important for scheduling decisions making, flood prevention, and disaster mitigation in small and medium river basins. In this study, a WRF/WRF-Hydro meteorological-hydrological coupling model is established based on the WRF model and the WRF-Hydro model to explore the influence of WRF precipitation on runoff forecasting in Zhanghe River Basin under different forecast periods and the propagation of uncertainty from meteorological model to hydrological model. The results show that: (1) The more extended the forecast period, the more serious the overestimation of the rainfall peak predicted by the WRF model under different forecast periods. Under 06 h and 12 h forecast periods, the forecast error of rainfall amount and peak precipitation time is smaller, and the forecast of rainfall area is obviously better than that under longer forecast periods (24 h and 48 h). (2) The runoff forecast results of the meteorological-hydrological coupling model show that the longer the forecast period, the more serious the overestimation of runoff peak. The runoff forecast results under 06 h and 12 h forecast periods are better, and the average correlation coefficient and average Nash coefficient increase by 0.15 and 0.69 respectively. (3) The error propagation characteristics between models are different due to the length of forecast period and the effect of rainfall forecast. Under 06 h and 12 h forecast periods, the error of numerical weather forecast is magnified and reduced after propagation to the hydrological model. Under longer forecast period (24 h and 48 h), the longer the forecast period, the greater the hydrological model amplifies the error of numerical weather forecast.