Inversion Of The Clay Minerals Contents Of Soil Based On Short Wave Infrared Hyperspectral

In recent years, with the development of pedology, the study on thephysicochemical properties of soil is more deeply. The composition of the clayminerals in soil is closely related to the physicochemical properties of soil. It showsthat the enrichment of the clay minerals play an important role in the indicativefunction. The absorption of some clay minerals on the heavy metal ions andpesticides has great development potential and prospects. The combination of clayminerals and organic of the soil generates the organic mineral complex. It makes thesoil has great structure, thereby improves the fertility of soil that provides the sciencebasis for the studyon the crop yield of the precision agriculture.What advanced technology should be used to predict the content of the clayminerals becomes the significant and difficult problem for the study at present. Dueto the high spectral resolution The SWIR (spectrum range:1300～2500nm,interval:2nm)technology in lab could probe the soil spectral information which cannot be detected in the multispectral remote sensing. It has the capability ofidentifying the organic, water and clay minerals with the diagnostic spectral featuresin soil and so on. It can establish the prediction equation for the content of the soilphysicochemical parameters based on the relationship between spectral reflectanceinformation and the content of physicochemical parameters. The lab imagingspectrometers are portable and measure the spectral reflectance fast and simply.According to the principle of the clay minerals absorbing the SWIR, thereflectance features of the SWIR for measuring the soil sample were analyzed. Basedon the scheme of the comparison of the spectral absorption peak features of the pureminerals, spectral absorption peak feature of soil and the spectral reflectance of thewhole range, the method of the Multiple Linear Regrssion and Partial Least SquaresRegression were used to establish prediction model on the content of the kaolinite,illite and montmorillonite. Combined with the comparative analysis, the fitting degree and the stability of the mathematical model was established to determine theoptimal prediction equation.The research contents include:1. Studying the principle of the clay minerals absorbing the SWIR and analyzingthe reflectance feature of the SWIR for measuring the soil sample, the spectralabsorption and association parameters were gained by studying the continuumremoval2. Studying the relationship among the absorption feature peaks of pure minerals,spectral absorption peak feature of soil, the spectral reflectance of the whole rangeand the content of claymineral.3. Study on the method of establishing and testing for the content of claymineral.By researching the mathematical modeling method of MLR-Enter, PLSR,MSLR and MSLR-PLS, validity test, and stability test, which of these applied intoexperimental research, some conclusions was found out as followed:1. The result showed that, with decreasing of the soil granularity, the spectralabsorption peak feature of soil moved to left in turn, and all spectral reflectance ofthe sample increased systematically on the same soil layer, the spectral reflectance ofthe soil layer C was higher than layer B.2. The content prediction model of kaolinite based on the spectral absorptionpeak feature of soil and the SWIR spectral reflectance of the whole range had greatinternal fitting and external inspection suitable for predicting the content of kaolinite.For the prediction model of M5 based on pectral reflectance of the whole range, theindependent variables and dependent variable had a great fitting degree of 0.791 withthe RMSEP of 1.671 slightly larger than RMSEF. M5 had obvious advantagescompared with the other four models, therefore M5 was determined as the optimalprediction model for the kaolinite content.The expression of the optimal prediction model of the kaolinite content:Y=5.2+6085.9*R₁₄₃₂+7796.31*R₁₄₄₀+7704.68*R₁₉₂₂+4798.13*R₁₉₂₈+3986.83*R 2032+22182.3*R₂₀₃₆-1268.91*R₂₂₁₄+1119.89*R₂₂₂₀+12025.2*R₂₃₆₂3. It shows that for the content prediction models of kaolinite based on theSWIR spectral reflectance of the whole range, the fitting degree of the independentvariables and dependent variable was 0.616, and the value of the RMSEP had littledifference to the RMSEF is 1.671, and it had the obviously advantage of internalfitting and external prediction that compare to the spectral absorption peak features ofthe pure mineral. So this prediction equtation was determined as the optimalprediction model of the illite content.The expression of the optimal prediction model of the illite content:Y=2.95326-2715.91*R₁₃₈₄+7368.86*R₁₉₂₈-7448.27*R₂₁₆₂-1120.86*R₂₂₁₄-151.706*R₂₂₁₈4. The model based on the spectral absorption peak features of the pure mineral,the fitting degree was 0.707, with the value of the RMSEP slightly larger thanRMSEF. Thus the prediction model was determined as the the optimal predictionmodel for the montmorillonite content.The expression of the optimal prediction model of the illite content:Y=3.464+21607.561*R₁₄₁₄-2476.108*R₁₉₀₆-13938.897*R₂₁₆₄-7243.758*R₂₂₀₈-7719.877*R₂₂₃₄+1964.79*R₂₃₈₄+990.01*R₂₃₉₄+4529.677*R₂₄₀₄-2332.811*R₂₄₃₀-415.734*R₂₄₄₀...