南方农业学报2025，Vol.56Issue(1)：87-96,10.DOI:10.3969/j.issn.2095-1191.2025.01.008
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
摘要
Abstract
[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 mois-ture 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(R2)of the BP neural network model was increased by 0.4589,R2 of PSO-BP neural network model was increased by 0.3811,and R2 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.Ac-cording to the classification criteria of soil moisture,the soil types in Yao'an irrigation area mainly belonged to brown moisture(combined moisture)and black moisture(full moisture).[Suggestion]Optimize the model and algorithm,in-crease the amount of measured data of soil moisture content,improve the accuracy of inversion;and integrate unmanned aerial vehicle(UAV)remote sensing data to monitor soil water rate in real time and dynamically allocate water resources in view of the problem of uneven distribution of water resources,so as to form a soil moisture evaluation mechanism and monitoring mechanism to achieve reasonable allocation of water resources.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Key words
water cloud model/Sentinel data/linear regression model/BP neural network/soil moisture inversion分类
农业科技引用本文复制引用
李万涛,杨明龙,唐秀娟,夏永华,杨赈,严正飞..基于Sentinel多源遥感数据的农田地表土壤水分反演[J].南方农业学报,2025,56(1):87-96,10.基金项目
国家自然科学基金项目(62266026) National Natural Science Foundation of China(62266026) （62266026）
