Soil Moisture Retrieval In Vegetation Covered Agricultural Areas Based On Optical And Microwave Remote Sensing Data

Soil moisture is an important component of terrestrial ecosystem.Soil moisture has three states:solid,liquid and gas,which can be converted into each other.It plays a great role in the transfer of water and heat,the material exchange and the energy balance between land and atmosphere.All kinds of chemical,physical and biological processes in soil need the support of soil water.Meanwhile,soil water is the main source of water for many vegetation,such as crops.It affects the growth and yield of vegetation by changing the growth conditions,such as soil fertility,soil temperature and ventilation.Traditional methods of soil moisture monitoring based on monitoring sites are time-consuming,laborious and inefficient.Remote sensing technology has become the mainstream approach to obtain soil moisture information due to its advantages of macroscopic,strong timeliness and wide monitoring area.Since both surface morphology and vegetation types have complex effects on the propagation of electromagnetic waves,the land surface soil moisture inversion in the vegetation covered areas is still a challenging task.In this paper,we will focus on how to combine the advantages of optical and microwave remote sensing to improve the accuracy of surface soil moisture inversion in vegetation covered areas.The study area was in Jingxian County,Hengshui City,Hebei Province.Using Lansat-7 ETM+,Setninel-2 optical remote sensing data and Setninel-1 active microwave remote sensing data,and combining with the field measured data,we realized the soil moisture retrieval in the vegetation covered agricultural areas based on the optical remote sensing inversion algorithm and the optical and microwave remote sensing cooperative inversion algorithm,respectively.The feasibility of these two inversion algorithms was verified and the inversion accuracy was compared.This article mainly achieved the following results:（1）On the one hand,the visible and near-infrared data of Landsat-7 ETM+were used to calculate 6 vegetation indices（VIs）,including NDVI,DVI,ARVI,RVI,SAVI and MSAVI.On the other hand,the land surface temperature（LST）obtained by mono-window algorithm based on the thermal infrared data.And then,the feature spaces were constructed by VIs and LST,which included NDVI-LST,DVI-LST,ARVI-LST,RVI-LST,SAVI-LST and MSAVI-LST.The corresponding temperature vegetation drought indices（TVDIs）were calculated from the feature spaces.R-TVDI,calculated from RVI-LST,had the best correlation with the measured soil moisture data,with the determination coefficient R²,RMSE and MAE being 0.5212,0.0476 cm³/cm³and 0.0391 cm³/cm³,respectively.（2）The improved water cloud model considered three vegetation factors,including VWC（vegetation water content）,VFC（vegetation fractional coverage）and vegetation height.The backscattering coefficients from bare soil could be obtained more accurately by the improved water cloud model.The retrieved VWC was inversion from the normalized difference water index(NDWI₁₆₁₀),and there was a good correlation between NDWI₁₆₁₀and measured VWC,with R²of 0.8343 and RMSE of 0.7377 kg/m².VFC was calculated by the pixel dichotomy model,and the NDVI_minand NDVI_maxwere determined to be 0.2397 and0.8482,respectively.The vegetation height was calculated according to the fitting relation between radar polarization ratio（VH/VV）and measured maize height,with R²of 0.4609.（3）The Sentinel-1 VH polarization mode is easily affected by the vegetation layer,while VV polarization mode has a strong penetrability,which can better monitor the change of the backscattering coefficients from the soil under the vegetation layer.The combination method based on the improved water cloud model and the Oh model had a high inversion accuracy in the VV polarization mode,with R²of 0.6530,the RMSE（Root Mean Square Error）of 0.0401cm³/cm³,and the MAE（Mean Absolute Error）of 0.0327 cm³/cm³.The three evaluation indexes of VV polarization mode were superior to VH polarization mode.（4）The comparison of the inversion accuracy between the optical remote sensing inversion algorithm based on R-TVDI and the optical and microwave remote sensing cooperative inversion algorithm based on the combination method,showed that the co-inversion algorithm performed better in three evaluation indexes,namely R²,RMSE and MAE,and it could more accurately reflect the surface soil moisture in the vegetation covered farmland.