单 位：
陕西省土地整治重点实验室;长安大学地球科学与资源学院;内蒙古农业大学水利与土木建筑工程学院
关键词：
土壤;水分;模型;地表粗糙度;水云模型;冠层含水率;反演精度
摘 要：
土壤含水率是农业、环境、气象等领域进行建模的重要参数。该研究将微波遥感与光学遥感相结合,利用Sentinel-1数据交叉极化比及变换土壤调节植被指数对地表粗糙度进行估计,构建了一种改进的水云模型(modifiedwatercloud model, MWCM)。分析将NDVI、NDWI和NDWI1725,2200等植被指数作为植被冠层含水率时,水云模型(water cloud model,WCM)及MWCM农田地表土壤含水率的反演精度。结果表明:从总体精度上来看,MWCM的反演精度优于WCM。在不同植被覆盖度情况下:当植被覆盖度为中、低程度(NDVI<0.5),MWCM具有较高的反演精度;在较高的植被覆盖度情况下(NDVI≥0.5),WCM与MWCM的反演精度较为接近。MWCM可有效的建立微波后向散射系数与地表土壤水分的关系,提高土壤含水率反演精度,为各种地表覆盖类型的土壤含水率反演提供研究思路及理论支持。
译 名：
Inversion of surface soil moisture content of Spanish farmland using modified water cloud model
作 者：
Ma Teng;Han Ling;Liu Quanming;School of Earth Science and Resources, Chang'an University;Water Conservancy and Civil Engineering college, Inner Mongolia Agricultural University;Shaanxi Key Laboratory of Land Consolidation;
关键词：
soil;;moisture;;models;;surface roughness;;water cloud model;;vegetation canopy water content;;inversion accuracy
摘 要：
Soil moisture content is an important parameter for constructing models in the agricultural, environmental, meteorological fields. The main method to retrieve soil moisture is SAR and optical remote sensing. To improve the soil moisture content inversion accuracy of farmland using remote sensing, the Duero basin of Spain was selected as a representative region(41 °06'N-41 °32'N,5°01'W-5°45'W). There were 23 automatic soil monitoring stations in the area. The soil moisture content of 19 observation stations was collected from ISMN, Sentinel-1 and Sentinel-2 were selected as remote sensing sources. The data for solving model parameters and verifying models were selected on January 16, April 16, June 15, August 8 and November 6 in 2018. The Sentinel-1 data with different incident angles on the above date was obtained. Sentinel-2 data obtained January 17, April 16, June 16, August 8 and November 6 in 2018. The orbit correction, radiation correction, improved LEE Sigma filter and geocoding were performed on Sentinel-1 images. Atmospheric correction was performed on Sentinel-2 images. Sentinel-2 images were used to produce vegetation index such as NDVI, NDWI and NDWI1725, 2200. Taking the 24-hours average of the above date as the soil water content. A Modified Water Cloud Model(MWCM) was established. In the MWCM, ground surface roughness was regarded as a variable related to the cross-polarization ratio and Transformed Soil Adjusted Vegetation Index(TSAVI). Three vegetation indexes(NDVI, NDWI and NDWI1725, 2200) were calculated and took into WCM and MWCM which were the indicator of vegetation water content. The overall RMSE of retrieved soil moisture of WCM using NDVI, NDWI, and NDWI1725, 2200 were 0.106, 0.118 and 0.113 m3/m3. The vegetation reflection parameters of three WCM were all equal to 0. It meant that under the condition of VV polarization in the C band, vegetation reflected energy could be ignored. The result also meant that the inversion accuracy of soil moisture content using WCM with NDVI, NDWI, and NDWI1725, 2200 were low when surface roughness was not considered. The MWCM was established where the backscatter coefficient of vertical polarization was expressed as decibel and vegetation canopy water content was substituted by NDVI, NDWI, and NDWI1725, 2200. The RMSE of retrieved soil moisture was 0.082, 0.094 and 0.077 m3/m3 using MWCM. It meant the WCM in which surface roughness was added had the higher inversion accuracy. The cross-polarization ratio and TSAVI are fine indicators of ground surface roughness. The MWCM with NDWI1725, 2200 had the highest inversion accuracy, which meant NDWI1725, 2200 was a good index to the reflection of surface vegetation. The model had lower inversion accuracy when the vegetation water content was substituted by NDVI than the model with NDWI1725, 2200. The result also showed that NDWI was not a fine index to reflect vegetation water content. Different surface vegetation coverage was represented by NDVI equal to 0-0.2, 0.2-0.3, 0.3-0.4, 0.4-0.5 and 0.5-0.7. Overall, the inversion accuracy of MWCM gradual decreased with increasing of surface vegetation coverage. In the condition of NDVI equals 0-0.5, the MWCM had a higher inversion accuracy than WCM. Because the ground surface was covered by vegetation, the influenced of surface roughness was reduced, when NDVI equaled 0.5-0.7. The WCM and MWCM had similar accuracy. Therefore, the MWCM could get higher accuracy in vegetation coverage land than WCM. NDWI1725, 2200 was a good vegetation index using in the MWCM under different vegetation cover conditions. It provided ideal and theoretical support for such research. The crop type and other land cover types were not considered in this study which might influent the reflection parameter and the model accuracy. In the future study, the MWCM should be further modified to accommodate different crop type cover condition.
