面向寒区水文模拟的区域气候模式降水订正方法对比

Comparison of RCM Precipitation Correction Methods for Hydrologic Simulation in Cold Region

使用区域气候模式降水驱动水文模型是提高站点稀少地区水文模拟精度的一种重要思路。如何根据地面台站观测对区域气候模式降水数据进行误差订正,一直都是寒区水文研究的热点问题。然而,目前尚无系统研究对比不同降水订正方案并评估其对寒区水文模拟的影响。因此,对比了两种主流的区域气候模式降水订正方法:概率密度匹配法和最优插值法,以2004~2009年为研究时间段,评估了两者在玛纳斯河流域的表现。结果发现:在统计意义上两种方法各有优劣。概率密度匹配法对中低值降水的订正结果良好,年平均降水的空间分布也更合理,但对高降水值的订正结果不稳定,且相关系数和均方根误差没有改善。订正前,模式降水与台站观测降水的相关系数为0.37,均方根误差2.80 mm/d,订正后的相关系数为0.36,均方根误差为2.70mm/d。最优插值法对相关系数和均方根误差改善显著,订正后分别为0.85和1.46mm/d,但空间分布相比订正前改善不明显,且会得到更多的微小降水。将订正前后的降水数据分别驱动水文模型,在使用同一套率定参数的情况下,概率密度匹配法订正的降水对径流模拟的改善微弱,纳什效率系数仅从0.63提高至0.65,而最优插值法则对降水的订正更加有效,订正后的降水模拟的径流纳什效率系数提高至0.71。本研究有助于解决寒区水文模拟中降水数据的质量优化问题,提高寒区水文的模拟精度。

It is a significant way to improve the accuracy of hydrologic simulation in the area having sparse observation sites by using the regional climate model precipitation driving hydrological model.As a result,a critical issue arises,which is how to correct the precipitation data coming from Regional Climate Model(RCM) based on the observation in the hydrology research in cold region.However,no systematic studies have been conducted to compare different precipitation correction methods and evaluate their impact on the hydrologic simulation in the cold region yet.Due to this,two kinds of mainstream regional climate model precipitation correction methods,Quantile Mapping(QM) and the Optimal Interpolation(OI),have been compared and evaluated in Manas River basin between 2004 to 2009.The results show that both methods have their own advantages and disadvantages in statistical significance.The correction result of QM is good in low precipitation value and the average annual precipitation is more reasonable at spatial distribution.But when comes the high precipitation value,the result is not stable.Moreover,the correlation coefficient(R) and Root Mean Square Error(RMSE) doesn’t improve.Compared to R 0.37 and RMSE 2.80 mm/d,the modified R is 0.36 and RMSE is 2.70 mm/d;OI can improve R and RMSE significantly,one increases to 0.85 and the other reduces to 1.46 mm/d after the correction.Despite that,OI also has its limitations.It gets more tiny precipitation relative to observation and the improvement of spatial distribution is not obvious.Using precipitation data before and after the correction to drive the hydrological models in a same set of model parameters.The results show that QM improves the simulation slightly,Nash-Sutcliffe efficiency coefficient(NSE) changes from 0.63 to 0.65 compared before,while OI is comparatively better,NSE increases to 0.71.This study is helpful to solve the problem of quality optimization in the preparation of hydrological simulation precipitation data and improve the precision of hydrologic simulation in cold region.