| Soil moisture is closely related to many basic agricultural activities and hydrological processes.Accurate soil moisture monitoring is significant in agricultural drought detection,crop yield estimation,climate change prediction,etc.Traditional methods of measuring soil moisture rely on observation stations to provide in situ measurement data.Still,the distance between stations varies from several to tens of kilometres and is affected by soil type,vegetation,terrain,climate and other factors.It is difficult to accurately characterize the region’s spatial distribution of soil moisture with conventional point measurement data.Remote sensing technology provides a powerful means for large-scale surface soil moisture monitoring.The data sources for retrieving soil moisture include optical,thermal infrared and microwave remote sensing.The physical mechanism of these three types of data sources is different,and the depth of soil moisture detection by various sensors is different,resulting in their respective advantages and disadvantages,and their application conditions are pretty other.The existing soil moisture retrieval methods can be divided into linear and nonlinear models according to the correlation between the band and the measured soil moisture data.The linear model is simple,but the retrieval accuracy is low.Although the nonlinear model(such as neural network)has high accuracy,the retrieval results are greatly affected by the data.In this paper,the existing widely used models are used to retrieve the surface soil moisture.Aiming at the shortcomings of these models,these soil moisture retrieval methods are improved.And the applicability of each retrieval model is studied.The main research results are as follows:(1)Improvement of soil moisture retrieval method based on optical remote sensing.Based on the relationship between optical satellite remote sensing surface reflectance and soil moisture,a band reflectance logarithmic model was constructed to retrieve soil moisture in bare soil/vegetation covered areas.This model directly shows the equation relationship between soil moisture and surface reflectance.I improved the Perpendicular Drought Index(PDI)model and proposed to build the Red-SWIR triangle index(RSTI)model using the red and short wave infrared(SWIR)band reflectance two-dimensional feature space of Landsat-8 remote sensing data.The correlation coefficient(r)between the estimated and measures soil moisture was around 0.5 with PDI model.The RSTI model estimates soil moisture with an accuracy of over 0.6 and the RSTI model is not limited by vegetation conditions.(2)Improvement of soil moisture retrieval method based on thermal infrared remote sensing.Based on MODIS data,the apparent thermal inertia(ATI)model was used to monitor the surface drought during the drought period.The r between the estimated and measures soil moisture was around 0.5 with ATI model.Because the ATI model is unsuitable for high evaporation areas,an ET-ATI triangle index(EATI)model is proposed based on ATI and evapotranspiration(ET)data.The r between the estimated and measures soil moisture was around 0.7 with EATI model.The EATI model considers the relationship between ATI,surface evaporation,vegetation transpiration and soil moisture simultaneously,which significantly reduces the impact of ET on the soil moisture retrieval model.(3)Improvement of soil moisture retrieval method based on microwave remote sensing.When using the single-polarization model(water cloud model)to retrieve soil moisture,the parameters of vegetation water content in the model are represented mainly by the vegetation index calculated by optical remote sensing.Calculate radar vegetation index(RVI)using GF-3 full-polarization data,and calculate dual-polarization radar vegetation index(Dp RVI)using Sentinel-1(VV/VH)dual polarization data.When no optical data is available due to weather,RVI or Dp RVI is used to characterize vegetation water content information.Moreover,the influence of surface roughness on backscattering is not considered in the single-polarization model.Therefore,in the full-polarization retrieval model,the influence of surface roughness on backscattering is removed through polarization ratio/combination.In the dual-polarization retrieval model(improved Chen model),the ratio of VH and VV polarization backscattering coefficients is used to characterize the surface roughness,and the effects of radar incidence angle and frequency on backscattering are considered.The r between estimated and measured soil moisture with the single/dual/full-polarization model were around 0.65,0.75 and 0.70,respectively.The full/dual-polarization model does not need optical data to obtain vegetation index or surface roughness measurement data.The retrieval accuracy is higher than that of the single-polarization retrieval model.(4)Applicability of soil moisture retrieval model.The effects of terrain factors(elevation,slope and aspect),vegetation biomass and ET on the soil moisture retrieval model were discussed.The terrain has a very obvious impact on the full/dual/singlepolarization retrieval model based on microwave remote sensing and the band reflectance logarithmic model based on optical remote sensing.In areas with severe topographic relief,the slope direction of the back-facing radar signal will appear radar shadow,and the terrain will block the optical signals received and reflected by the surface in the north aspect(shady slope)area.The retrieval accuracy in such areas is very low.The other retrieval models are less affected by the terrain.Vegetation biomass has specific restrictions on all retrieval models.Although most models can be used in vegetation-covered areas,they have higher retrieval accuracy in low vegetation areas,and the retrieval accuracy decreases with the increase of vegetation biomass.Except that the ATI model is not suitable for the high ET area,and the EATI model is almost not affected by ET.ET has a substantial impact on other retrieval models.The estimated soil moisture strongly correlates with the measured value in the high and low ET areas. |
PDF Full Text Request