Research On Soil Spectral Classification For Digital Soil Mapping

Land is a natural complex consisting of many natural elements including geology,geomorphology,climate,hydrology,soil and vegetation.Soil is essential for human production and life.Soil is a natural historical complex,and like other objects,it has its own law of occurrence and evolution.With the development of remote sensing and computer science,digital soil mapping is becoming more and more mature,and the improvement of soil mapping accuracy is of great help to the division of land evaluation units,however,the determination of minimum unit of digital mapping and universality need to be solved.Soil spectra data are the basis of digital soil mapping.In this paper,soil spectra data were used as the research object,and based on the Chinese soil genetic classification system,and we did researches from two angles:(1)at present,the minimum level of soil spectral classification is at the great group level,and we will determine the further level of soil spectral classification.(2)In order to improve the universality of the soil classification model,soil samples with different soil moisture were mixed,and we built soil classification models with eliminating the effect of soil moisture.VNIR(400-2500 nm)spectra of four typical topsoil(Chernozems,Cambisols,Phaeozems,Arenosols)from the Songnen Plain were selected as objects and converted them to the firstderivative and continuum removal(CR),and we extracted spectral feature parameters(SFPs)with clear physical meaning.Principal components of reflectance(PCR)and principal components of first-derivative reflectance(PCFD)were used as inputs for building multinomial logistic regression MLR,multi-layer perception neural network,MLPNN,and decision trees(DT)at the genus and species levels.The results show the following:(1)the first and second absorption valleys of the topsoil spectral curves were substantially different between soils at the great group and genus levels but similar at the species level;(2)the topsoil samples can be allocated with SFPs,for which genus was the minimum level within the CSC for soil allocation with topsoil spectral reflectance;and(3)the DT model with SFPs as input was the most accurate at the genus level;the accuracy of allocation was 83.3% and the Kappa coefficient,an evaluation index,was 0.8.Air-dried soils can not be applied well to satellite,unmanned aerial vehicles or other platforms because of complex field environment.The field soil is affected by climate,and soil moisture is different,and the pure soil information is covered by crop residues or cow dung.Based on this point,we considered important soil moisture factor,and 53 soil samples with different moisture content were mixed with 135 air-dried soil samples to analyze the effect of soil moisture on different allocation indices,and some allocation indices that aren’t affected by soil moisture were selected to build soil allocation models.The results show the following:(1)the effect of soil moisture on absorption area(A),depth(H)and slope of different bands(SL)is large,and the effect of soil moisture on Principal components of reflectance(PCR)and principal components of first-derivative reflectance(PCFD)is larger than spectral feature parameters(SFPs);(2)the curves of continuum removal of Arenosols is less affected by soil moisture than other three classes.(3)The accuracy of DT model with considering differences of soil moisture is the highest,and the accuracy of soil allocation is 91.892% and the Kappa coefficient is 0.888,which is better than MLR model and MLPNN model with considering differences of soil moisture.Land management,soil surveys and land sustainable utilization all need to be based on different soil classes,and cultivated land of different soil classes may lead to differences on yield,so the quality of cultivated land also need to be based on soil classes.The results of this study can understand macroscopically the distribution characteristics of different soils,and RS technology and GIS auxiliary data will be used to delineate the protected area of Phaeozems,and have important guiding significance for land management and planning of land utilization in China.