基于数据同化的地下水埋深插值研究

以西北干旱区典型县域磴口县为研究区,以2015年8月份40个地下水采样点的样品数据为基础,引入集合卡尔曼滤波(En KF)数据同化将其优化作为主变量,以蒸散发量反演结果以及归一化植被指数(NDVI)数据为协变量,进行协同克里金插值,同时与未采用同化的协同克里金插值结果以及经同化采用普通克里金插值结果进行交叉验证。结果表明:三者在较大空间尺度上对地下水埋深空间分布趋势的模拟基本一致,南部沙漠地区整体较高,在空间分布上表现为明显的地理规律性。同化后的数据进行协同克里金插值的结果改善最显著,平均误差、均方根误差、平均标准误差均优于未同化插值结果,其中平均误差仅为0.270 5 m。与普通克里金插值方法相比,协同克里金插值考虑蒸散发与NDVI的协同作用,精度明显提高,平均误差减小0.409 7 m,均方根误差减小0.078 4 m,平均标准误差减小1.016 7 m。

Interpolation of Groundwater Depth Based on Data Assimilation

Groundwater monitoring is limited by practical conditions, and only limited monitoring results can be obtained when it is observed. As a kind of geostatistical interpolation method, cooperative Kriging (co-Kriging) method can effectively represent the transformation of discrete point-like information to planar continuous information. Dengkou County, a typical county in the arid region of Northwest China, was selected as the study area. The sampled data from 40 groundwater sampling sites in 2015 was selected as the main variable. And this data optimized by EnKF was used as the basic data of co-Kriging interpolation. The evapotranspiration results and NDVI data were selected as the covariates. Co-Kriging interpolation was carried out by using the sampled data from 40 groundwater sampling sites in August, 2015, as the main variable, which were optimized by EnKF, and the evapotranspiration results and NDVI data were used as the covariates. Meanwhile, the results of co Kriging interpolation without using EnKF model and Kriging interpolation optimized by EnKF model were used to verify the accuracy. The results showed that the spatial distribution trend of groundwater depth was basically the same at large scale, the value in the southern desert region was higher, and the spatial distribution showed obvious geography regularity. The most significant improvement was achieved with EnKF model. Based on this improvement, the mean error, root mean square error and mean standard error were all better than those without assimilation, with the mean error of 0.2705m. Compared with the ordinary Kriging interpolation method, co-Kriging model took the synergistic effect of evapotranspiration and NDVI into consideration, and the precision was obviously improved. The mean error was decreased by 0.4097m, the root mean square error was decreased by 0.0784m and the mean standard error was decreased by 1.0167m. This study can provide a scientific basis for spatial visualization simulation and reasonable management of water resources in arid areas.