Soil Moisture Estimation With The Optical Trapezoid Model And Its Improved Model Based On Sentinel-2A Data

Soil moisture plays an important role in terrestrial production,meteorological changes and hydrological processes,and it’s also the key to drought monitoring and management.Remote sensing technology provides a powerful tool for large-scale,multi-temporal monitoring of soil moisture.with the development of remote sensing technology,scholars have begun to estimate soil moisture and other related research by remote sensing.In this paper,the northern area of Luhe District of Nanjing City as the research area,estimate soil moisture with the optical trapezoid model(OPTRAM)based on the Sentinel-2 image.And improve the original optical trapezoidal model by replacing NDVI with four vegetation indexes.Retrieval effect of OPTRAM for soil moisture and the four improved models are compared and analyzed.The main conclusions reached are as follows:(1)The characteristic space of OPTRAM conforms to the basic distribution conforming to the trapezoidal model,and its variation is seasonal.The STR-NDVI feature space is trapezoidal,and the wet and dry edges fit well which the STR have a significant negative linear correlation of the dry edges with the NDVI.Due to the different degree of drought in the research area in different seasons,the effect of soil water content on wet edges and dry edges are also different.Because the uniform distribution of vegetation coverage in the study area in May,the construction of the STR-NDVI feature space is the most stable with the Sentinel-2 image on May 22.(2)The OPTRAM is feasible to estimate soil moisture in Luhe District of Nanjing.The regression analysis of the OPTRAM and TVDI with measured soil moisture data showed a significant negative correlation(P<0.01),and the overall fitting effect of the TVDI was slightly better than the OPTRAM.The sampling points were divided into forest land,bare land and agricultural land according to the ground cover.The results showed that the effect of OPTRAM in different land use types was better than TVDI on soil moisture retrieval.Among them,the OPTRAM model has the best inversion effect for agricultural land,and the worst effect for bare land inversion.Therefore,it is scientific and reasonable to apply the OPTRAM to retrieve soil moisture in Nanjing,especially on agricultural land.(3)The improved OPTRAM soil moisture inversion accuracy is better based on four vegetation indices.The characteristic spaces of EVI_RE1,EVI_RE2,PVI,Fpv,with short-wave infrared conversion reflectance(STR)were constructed,and he the model was used to invert soil moisture in the study area.The STR-NDVI feature space distribution of the OPTRAM model and the four improved models are obvious trapezoidal shapes.The fitting effect of wet and dry edges is significant,and the dry edges are negatively correlated and wet edges are positively correlated.The fitting effect of STR-PVI is better,which just considered the fitting coefficients the linear fit between the dry and wet edges of the STR and the vegetation index as the evaluation criteria.However,on the whole,STR-Fpv has more advantages in all the models.The simulated values calculated with the model and the measured soil moisture data of the corresponding samples were analyzed,and it was found that the remote sensing simulation values showed a significant negative linear relationship with the soil moisture.Among the 5 models,the improved STR-Fpv fits best and the model is the most stable.(4)The drought in the study area was divided into five levels:drought,partial drought,light drought,normal and humid.The statistics result showed that the soil moisture retrieved by STR-NDVI and STR-Fpv model were mainly distribute in drought and light drought region.The STR-Fpv model is more capable of showing boundaries.Most of the land in drought or drought conditions is agricultural land or bare land.Woodland is generally in a normal or moist state.