基于GF-3和Landsat 8遥感数据的土壤水分反演研究

基于我国首颗全极化雷达卫星高分三号（GF-3）和Landsat8数据,研究浓密植被覆盖地表土壤水分反演方法。为了提高浓密植被覆盖地表土壤水分反演精度,首先利用PROSAIL模型、实测植被参数及Landsat8光学数据分析了8种植被指数与植被冠层含水量的相关性,从中优选出归一化差异水指数（NDWI₅）用于反演植被冠层含水量,并通过分析植被含水量和植被冠层含水量的关系,构建植被含水量模型;然后结合植被含水量反演模型和简化MIMICS模型校正了植被对雷达后向散射系数的影响,最后基于AIEM建立裸土后向散射系数模拟数据集,发展一种主动微波和光学数据协同反演浓密植被覆盖地表土壤水分模型,并以山东省禹城市为研究区,实现了玉米覆盖下HH、VV和HH+VV 3种模式土壤水分反演。实验结果表明： ① NDWI₅为最佳植被指数,对于去除植被影响有较好效果;② 基于此方法,利用GF-3和Landsat8卫星数据反演得到的土壤水分具有较高的精度;③ 相比HH和VV两种极化模式,HH+VV双通道模式对土壤水分反演结果更好,决定系数（R2）为0.4037,均方根误差（RMSE）为0.0667 m ³m ^-3。

Soil Moisture Retrieval Study based on GF-3 and Landsat8 Remote Sensing Data

As an important component of soil, soil moisture plays an important role in crop growth. The GaoFen-3(GF-3) satellite, as the first C-band full-polarization Synthetic-Aperture Radar (SAR) satellite of China, provides a valuable data source for soil moisture monitoring. In this study, a soil moisture retrieval algorithm was developed over densely-vegetated areas based on GF-3 and Landsat8 data. To improve the accuracy of the soil moisture retrieval, this paper firstly analyzed the correlation between eight vegetation indices and Vegetation Canopy Water Content (VCWC) based on the PROSAIL model, measured vegetation parameters and the Landsat8 optical data. The Normalized Difference Water Index (NDWI₅), which was identified as the optimal index from these indexes, was used to obtain the VCWC. The inversion model of Vegetation Water Content (VWC) was established by analyzing the relationship between measured VWC and the VCWC. Secondly, the model was integrated with simplified Michigan Microwave Canopy Scattering (MIMICS) model to correct the effects of vegetation on the radar backscattering coefficient. Finally, the backscattering coefficient simulation dataset of bare soil was established based on the Advanced Integrated Equation Model (AIEM) for developing the soil moisture retrieval model over densely-vegetated areas by combining active microwave and optical remote sensing data. The soil moisture retrieval algorithm was validated in a region of corn in Yucheng city, Shandong province, with soil moisture retrievals obtained at HH, VV and HH+VV combination, respectively. Results show: ① NDWI₅ had the best fit with measured VCWC values among the eight vegetation indices, with the coefficient of determination (R ²) reaching 0.7433, and the Root Mean Square Error (RMSE) being 0.5146 kg/m ². Thus, it was adopted to correct the effects of vegetation. ② The proposed algorithm based on GF-3 and Landsat8 satellite data performed well in soil moisture retrieval that resulted in improved accuracy in soil moisture monitoring. ③ Compared with the HH and VV polarization, the HH+VV dual-channel mode exhibited the highest accuracy, with a R ² of 0.4037 and a RMSE of 0.0667 m ³m ^-3, followed by the HH polarization (R ²=0.2894, RMSE=0.0692 m ³m ^-3) and the VV polarization (R ²=0.3577, RMSE=0.0675 m ³m ^-3). Our findings suggest that the proposed algorithm has good potential for operationally estimating soil moisture from the new GF-3 satellite data with high accuracy.