基于指数滤波法的中国区域深层土壤水分估算

Estimation of subsurface soil moisture in China based on exponential filtering method

微波遥感技术已经可以实现区域尺度表层土壤水分的观测,但是与生态、水文和农业等领域更密切相关的是深层土壤水分,如何实现由表层土壤水分估算深层土壤水分仍是亟待解决的科学问题。本研究利用中国区域2121个土壤水分观测站点的多层土壤水分数据(统计数据未含港、澳、台地区。下同),基于10 cm土壤水分,应用指数滤波法分别估算20、30、40和50 cm深度的土壤水分。结果表明:1)随着土壤深度的增加,指数滤波法的估算精度逐渐下降,NSE(纳什效率系数)中值对应4个深度分别为0.76、0.54、0.39和0.28,R^(2)(决定系数)中值分别为0.86、0.73、0.64和0.59,RMSE(均方根误差)中值分别为0.023、0.027、0.028和0.028 m^(3)/m^(3);2)由2121个土壤水分站点率定得到的站点最佳T值(表征土壤水分从表层运移至深层的时间)的中值估算深层土壤水分具有可行性,尤其估算浅层(如20 cm)土壤水分时,估算结果与用最佳T值相比,93%的站点NSE差值<0.05,82%的站点RMSE差值<0.001;3)表层和深层土壤水分的相关性越好,指数滤波法的估算精度越高。综上,指数滤波法在中国区域具有较好的适用性,中国区域T的中值,可以作为指数滤波法在中国区域应用的参考值。

Microwave remote sensing technology is able to realize the observation of regional-scale surface soil moisture,but the deep soil moisture is more closely related to ecological,hydrological and agricultural fields.How to estimate deep soil moisture from surface soil moisture is still a scientific problem to be solved.This study utilized multi-layer soil moisture data from 2121 soil moisture observation stations across China(Data do not include those of Hongkong,Macao and Taiwan region.The same below)and investigated the application of the exponential filtering method.Based on 10 cm soil moisture data,the exponential filtering method was applied to estimate soil moisture at depths of 20 cm,30 cm,40 cm,and 50 cm,respectively.The results show that:1)With the increase of soil depth,the estimation accuracy of exponential filtering method gradually decreases.The median values of NSE(Nash efficiency coefficient)are 0.76,0.54,0.39 and 0.28 corresponding to the four depths,respectively.The median values of R^(2)(determination coefficient)are 0.86,0.73,0.64 and 0.59,respectively,and the median values of RMSE(Root mean square error)are 0.023,0.027,0.028 and 0.028 m^(3)/m^(3),respectively;2)It is feasible to estimate subsurface soil moisture from the median value of the calibrated T,especially for shallow soil moisture(such as 20 cm)estimation.Compared with using calibrated T value,NSE difference is less than 0.05 at 93%sites and RMSE difference is less than 0.001 at 82%sites;3)The better the correlation between surface and subsurface soil moisture,the higher the estimation accuracy of exponential filtering method.The above results suggest that the EF method has good performance in China.Supported by a large number of measured data,T values of 2,5,10 and 14 d corresponding to the depths of 20,30,40 and 50 cm,respectively,can be used as references in China when using the EF method with 10 cm soil moisture.