基于Sentinel多源遥感数据的农田地表土壤水分反演  

Surface soil moisture inversion of farmland based on Sentinel multi-source remote sensing data

作　　者：李万涛 杨明龙[1,2] 唐秀娟 夏永华 杨赈[1,2] 严正飞 LI Wan-tao;YANG Ming-long;TANG Xiu-juan;XIA Yong-hua;YANG Zhen;YAN Zheng-fei(School of Land Resources Engineering,Kunming University of Science and Technology,Kunming,Yunnan 650093,China;Application Engineering Research Center of Spatial Information Surveying and Mapping Technology in Plateau and Mountainous Areas Set by Universities in Yunnan,Kunming,Yunnan 650093,China;Kunming Institute of Surveying and Mapping,Kunming,Yunnan 650091,China)

机构地区：[1]昆明理工大学国土资源工程学院,云南昆明650093 [2]云南省高校高原山区空间信息测绘技术应用工程研究中心,云南昆明650093 [3]昆明市测绘研究院,云南昆明650091

出　　处：《南方农业学报》2025年第1期87-96,共10页Journal of Southern Agriculture

基　　金：国家自然科学基金项目(62266026)。

摘　　要：【目的】通过多源遥感数据协同作用分析滇中地区姚安灌区的农田地表土壤含水率,为后续对滇中高原地区的地表土壤水分研究提供参考。【方法】选择Landsat 8、Sentinel遥感数据为数据源,构建土壤水分与特征参数关系式,比较线性回归模型、BP神经网络模型、粒子群优化(PSO)的BP(PSO-BP)神经网络模型、随机森林(RF)算法预测土壤含水率的精度,选择最佳方法反演姚安灌区农田地表土壤含水率。【结果】协同Sentinel-1微波数据和Sentinel-2光学数据,水云模型作用下VV后向散射系数减少0.1~0.4 dB、VH后向散射系数减少0~0.05 dB;加入特征参数,对比线性回归模型,BP神经网络模型的决定系数(R^(2))提高0.4589、PSO-BP神经网络模型的R^(2)提高0.3811、RF算法的R^(2)提高0.4544,其中,BP神经网络模型的R^(2)和均方根误差(RMSE)较优。依据BP神经网络模型反演的土壤含水率与监督分类的土地利用分类进行叠加分析,可知姚安灌区土壤含水率集中在20%~30%,位置主要集中在姚安灌区中部,土壤含水率10%~20%区域主要集中在姚安灌区北部,而土壤含水率30%~40%区域覆盖面积少且分散。姚安灌区的土壤类型根据土壤墒情的划分标准主要属于褐墒(合墒)和黑墒(饱墒)。【建议】优化模型及算法,增加土壤含水率实测数据量,提高反演精度;针对水资源分布不均的问题,融合无人机遥感数据,对土壤含水分进行实时监测,动态分配水资源,形成土壤水分评价机制与监测机制,实现水资源的合理分配。【Objective】The synergistic effect of multi-source remote sensing data was used to analyze the surface soil moisture rate of farmland in Yao’an irrigation area in central Yunnan,which could provide reference for surface soil moisture research in central Yunnan Plateau.【Method】Landsat 8 data,Sentinel microwave data were selected as data sources to construct the relationship between soil moisture and feature parameters,and the accuracy of linear regression model,BP neural network model and particle swarm optimization(PSO)BP(PSO-BP)neural network model and random forest(RF)algorithm in predicting soil moisture content was compared,the best method was selected to analyze the surface soil moisture content of farmland in Yao’an irrigation area.【Result】Combined with Sentinel-1 microwave data and Sentinel-2 optical data,the VV backscatter coefficient was reduced by 0.1-0.4 dB and VH backscatter coefficient was reduced by 0-0.05 dB under the action of water cloud model.Adding feature parameters,compared with the linear regression model,the coefficient of determination(R^(2))of the BP neural network model was increased by 0.4589,R^(2) of PSO-BP neural network model was increased by 0.3811,and R^(2) of RF algorithm was improved by 0.4544,among which the root mean square error(RMSE)of the BP neural network model was better.According to the superposition analysis of soil moisture content and land use classification of supervised classification inverted by BP neural network model,it could be found that the soil moisture content in Yao’an irrigation area was concentrated in 20%-30%,and the location was mainly concentrated in the middle of Yao’an irrigation area,the area with soil moisture content of 10%-20%was mainly concentrated in the northern part of Yao’an irrigation area,and the area with soil moisture content of 30%-40%covered a small and scattered area.According to the classification criteria of soil moisture,the soil types in Yao’an irrigation area mainly belonged to brown moisture(combined moistu

关 键 词：水云模型 Sentinel数据 线性回归模型 BP神经网络模型 土壤水分反演 

分 类 号：S152.7[农业科学—土壤学]