基于数据同化技术的延河流域绿水模拟研究

分别利用分布式时变增益水文模型（DTVGM）和分布式耗水过程模型（DEPM）对延河流域延安水文站以上区域进行水文过程模拟,并应用拓展卡尔曼滤波(EKF)算法对2个模型的绿水(实际蒸散发)模拟结果进行同化处理,从而优化了研究区的绿水量并得出绿水的空间分布规律。结果表明：在整个模拟期,DTVGM的月尺度效率系数（NSCE）达到了0.83,水量平衡相对误差为-1.97%,模型能够较好地模拟研究区的水文过程;DEPM的水量平衡相对误差为-1.81%,能较好地模拟流域的水量平衡;DTVGM和DEPM模拟的流域2010年平均绿水量分别为378.52 mm和375.55 mm,空间分布格局相似。与站点观测值比较,DTVGM和DEPM模拟绿水的NSCE分别是0.76和0.59,DEPM的结果具有更多的空间变化信息。同化结果表明EKF算法能综合优化2个模型的模拟结果,同化后DTVGM模拟研究区的平均绿水量为376.34 mm,NSCE为0.78;同化后研究区绿水标准差为40.37 mm,比同化前增加了7.79 mm,绿水空间分布体现了更多的空间变化信息,同时,空间分布时格局也更加合理。

The Simulation of Green Water in the Yanhe River Basin Based on Data Assimilation

The accurate green water simulation is very important for the crop growth, agricultural drought monitoring, food security and rational allocation of water resources. Now, there are three methods which can be used to get greenwater: field observation, remote sensing calculation and hydrological model simulation. The field observation can not get the accurate simulation result of catchment green water at large scale because of complicated spatial heterogeneity; the remote sensing calculation method can only get the instantaneous simulation result at a large scale region. Compared to the two methods above, the hydrological model can get the continuous green water simulation results in large scale catchment. But, because hydrological models have different structures and characteristics, different simulation results can be got when inputting the same simulation data in the same region. In order to comprehensively utilize the simulation results of different hydrological models, the data assimilation method is a good choice. In this study, the Distributed Time Variant Gain Model(DTVGM) and Distributed Evapotranspiration Process Model(DEPM) were used to simulate the hydrological process in the catchment controlled by the Yan’an station in the Yanhe River Basin based on their structures and characteristics, and the Extended Kalman Filter(EKF) data assimilation algorithm was used to assimilate the green water (actual evapotranspiration) simulated by the two models to optimize the green water in the study area. The statistical indexes show that the simulation result of DTVGM is applicable to simulate the hydrological processes in this study area, the Nash-Sutcliffe coefficient(NSCE) is 0.83 and the relative water balance is -1.97% in the whole simulation period. The water balance relative index of DEPM is -1.81% in the whole simulation period, which shows that DEPM can model the water balance well at the study area.Besides, the simulation results show that the green water in 2010 simulated by the two models is 378.52 mm and 375.55 mm, respectively. There are not obvious difference for the mean simulation results of two models and the spatial distribution pattern are also similar, but there is more spatial variable information for result simulated by DEPM model than that of DTVGM. Compared to the observed green water, the NSCE of green water simulated by DTVGM and DEPM is respectively 0.76 and 0.59, but the spatial distribution of green water simulated by DEPM has more change information. Then, the green water simulation result of DEPM was used as “observed value” to assimilate the green water simulated by DTVGM based on the EKF data assimilation method in order to optimize the green water simulation result.The average green water simulated by DTVGM after data assimilation is 376.42 mm and the NSCE is 0.78 when comparing to the observed green water. The Standard Deviation(SD) for the green water simulation becomes 40.37 mm after assimilation, which is significantly increased by 7.79 mm than the original modeling results. The green water in spatial distribution shows more spatial change information and is more reasonable in this study area.