Soil Multifactor Radiative Transfer Model And Estimation Of Typical Soil Compositions From Hyperspectral Data

Food security is a critical foundation and component of national security,and soil is the basis for ensuring food security.The health level of soil,especially the quality of cultivated soil,directly affects the yield and quality of food production.However,according to survey results,the situation of cultivated soil quality in China is not optimistic,with a declining trend in the quality of some cultivated soil and severe soil heavy metal pollution in certain areas.Based on the current status and policy planning of cultivated soil quality in China,the construction of a monitoring system for cultivated soil quality is essential to controlling soil pollution and improving soil quality.Hyperspectral data provides rich spectral information that can quickly and accurately obtain the typical component distribution of soil in a large area with high precision during cultivated soil quality monitoring.However,most of the current hyperspectral soil composition estimation applications are carried out using data-driven models that cannot explain the spectral response mechanisms of soil compositions.The complex imaging environment of hyperspectral remote sensing imagery also makes it difficult to apply soil radiative transfer models in quantitative soil estimation.Additionally,due to the low content of soil heavy metal elements,traditional feature extraction methods cannot accurately extract spectral features,resulting in poor performance of soil heavy metal content estimation.To address the problem of the lack of spectral response mechanism in current estimation models,this dissertation proposes a Soil Multifactor Radiative Transfer(SMRT)model that explains the spectral response mechanism of soil components.Based on this,a Semi-Empirical Soil Multifactor Radiative Transfer(SESMRT)model is constructed by combining data-driven models,which ultimately achieves the estimation of soil organic matter based on aerial and satellite hyperspectral imagery.To address the issue of the poor stability of soil heavy metal spectral feature,a Binary Weight Symbiotic Organisms Search(BWSOS)algorithm is proposed to extract the spectral feature of heavy metals.Based on the chelation relationship between soil heavy metals and organic matter,the coupling of the organic matter response spectral feature in the SESMRT model with BWSOS algorithm achieves higher-precision estimation of soil heavy metal content,accurately evaluating the status of soil pollution in the study area.The main research contents of this article are summarized as follows:(1)In response to the problem of a single factor being considered in the current soil radiation transfer model,the SMRT model was proposed to describe the response mechanism of the main influencing factors on the spectra based on the analysis of the influence of multiple soil components on the spectra.The experimental results show that the~2 of the SMRT model spectral simulation is 0.9343 for the test set within the study area and 0.8428 for the test set in the ICRAF-ISRIC global soil spectral library.The SMRT model solves the problem that current soil radiative transfer models only consider a single factor and cannot be used in practical applications.Meanwhile,the absorption and scattering coefficients of the SMRT model in the visible band explain that black soil and laterite are colored by high content of organic matter and iron oxide,respectively.Compared with the traditional feature extraction methods,the SMRT model parameters can describe the influence of soil components on the spectrum from the radiative transfer theory,and the effective spectral features are extracted based on the absorption and scattering coefficients of soil components.(2)In response to the difficulty of applying soil radiative transfer models due to the complex imaging environment of imaging hyperspectral imagery,the SESMRT model,combined SMRT model and data-driven model,was proposed to enable the application of soil radiative transfer model in hyperspectral imagery.The spectra of soil organic matter calculated by SMRT model removes the interference of other soil compositions and significantly improves the correlation between spectra and organic matter.And,combined with the data-driven model,soil organic matter content could be estimated.The experimental results in GF5 satellite hyperspectral remote sensing images and Hy Map airborne hyperspectral remote sensing images show that the~2 of SESMRT model is 0.6842 and 0.7044,respectively.Compared with the estimation method based on the original spectrum,the SESMRT model has greatly improved the estimation accuracy.Moreover,by comparing multi-temporal estimation results,it can be found that increased the organic matter content of the cultivated soil increased,while the organic matter content of soil around the mining area decreased.(3)In response to the issue of poor stability in extracting spectral features of heavy metals in soil and the lack of explanation for the soil radiation transfer mechanism,the combined expression effect of spectral feature bands under differential weights was investigated,and the BWSOS algorithm was established to achieve the optimal band selection under differential weights.The Binary Weight Symbiotic Organisms Search based on Prior Knowledge(KBWSOS)algorithm was proposed using the adsorption mechanism between soil organic matter and heavy metals as a prior factor.The KBWSOS algorithm utilized the soil organic matter spectral feature bands in the SESMRT model to optimize the weight information of heavy metal spectral features for soil heavy metals content estimation.The BWSOS algorithm achieves superior accuracy in the estimation of several soil heavy metals.For As element adsorbed by organic matter,the KBWSOS algorithm has a better estimation accuracy than BWSOS algorithm,and spectral features selected by KBWSOS are more concentrated in the range of spectral features bands of organic matter.Finally,the soil pollution status of the study area was assessed through the soil heavy metal estimation results.The polluted soils in the study area are basically around the mine and most of the cultivated soils are not polluted.Therefore,the potential ecological risk index indicates that the study area is generally at a low risk level.