基于多时相Sentinel-1SAR地表土壤水分反演的Alpha近似模型改进

Modification of Alpha Approximation Model for Retrieving Soil Moisture Data Based on Multi-temporal Sentinel-1 SAR

针对多时相SAR数据反演土壤水分的Alpha近似模型存在入射角变化影响模型适用性的问题,在小扰动模型SPM的理论基础上推导得出改进的Alpha模型,并使用时间序列Sentinel-1SAR数据结合地面观测网络在黑河和温尼伯两个实验区进行算法验证。对反演结果和实测数据进行相关分析,使用改进后模型反演得到的土壤水分与实测数据间的相关系数分别由0.473、0.601提高至0.851和0.821,在严格的α系数边界设定条件下,均方根误差(m^3·m^-3)分别由0.053、0.152减小至0.023和0.065;为进一步评价模型质量,引入水文领域中常用的模型评价指标纳什系数NSE,改进前后的NSE分别为-0.532、-1.243和0.713、0.588。结果表明,原始模型反演结果虽然和实测数据间呈现一定的相关性,但存在较大误差,而通过本文提出的改进模型可以很好地校正由于入射角变化带来的影响,得到较为可靠的结果,扩展了Alpha模型的应用。

The applicability of the Alpha approximation model for retrieving soil moisture data from multi-temporal SAR data will be affected when incidence angles change. Aiming at this problem, the Alpha model is modified based on the theory of Small Perturbation model, and verified it with the time series Sentinel-1 SAR data of Heihe and Winnipeg, the two experimental areas coupled with the ground observation network therein. A correlation analysis was done of the inversion with the observation data. It was found that the correlation coefficient increased from 0.473 and 0.601 to 0.851 and 0.821 respectively, while the root mean square error (m3·m-3) decreased from 0.053 and 0.152 to 0.023 and 0.065 respectively. In order to further evaluate quality of the model, the NSE coefficient which is commonly used in hydrology for evaluation of model was introduced for use. The NSE before and after the modification were -0.532 and -1.243, 0.713 and 0.588, respectively. The results show that though a certain relationship was found between the inversion using the original model and the measurement in data, a quite big error was too. However, once the model is modified as above mentioned, the impact of changes in incidence angle can be well corrected, making the inversion more reliable, which extends the application of the Alpha model.