Research On Soil Classification Of Northern Songnen Plain Based On Remote Sensing

Land is particularly important for all humanity,and we can not live and develop without it.The most important substance in the land is soil which provides basic living conditions for human.Soil is the most fundamental carrier of existence for all living organisms,which plays an indispensable role in the development of human beings.With the rapid development of land planning and management and precision agriculture,the accuracy of soil map is higher.Therefore,higher requirements for soil classification are put forward.The traditional soil classification method depends on the field work and experience of soil experts.The classification process laborious and time consuming.Hence,more and more attention has been paid to soil mapping method with the higher accuracy and higher efficiency.The rapid development and application of remote sensing technology provide new technology and data for soil classification.At the same time,many soil scientists in the world have carried out a large number of soil classification research,and put forward some new classification methods.In this thesis,the northern Songnen Plain and Mingshui County as the research area were selected.The purpose of this study is to soil genesis theory as a guide,and select indoor soil hyperspectral data and Sentinel-2A images to classify soil great group.I measured the spectral reflectance in the visible and near-infrared regions(400~2500 nm)of 148 topsoil samples(0~20 cm)from 4 soil classes(Black soils,Chernozems,Blown soils,Meadow soils)collected in northern Songnen Plain.I extracted five spectral characteristic parameters with clear physiochemical meanings by continuum removal method,and compared these to the principle component,first spectral derivative and Continuum Removal of soil reflectance.Models were built using the K-means Clustering(K-mean),Multi-layer Perceptron Neural Network(MLPNN),Support Vector Machine(SVM),and Decision Tree(DT)methods.The DTs allocation model based on topsoil spectral characteristic parameters had the highest classification accuracy.Only the classification accuracy of Meadow soils was less than 85%,because the spectral curve of Meadow soils topsoil was similar to its adjacent soil due to soil erosion.The above research showed that topsoil reflectance spectral characteristics can be applied to soil classification,which provided a theoretical basis for the research of soil classification by remote sensing image.Therefore,Mingshui County as the study area was used to research soil classification by remote sensing image.Firstly,I used land use data in 2015 to remove the urban construction,rural housing and water areas.Secondly,atmospheric correction and resampling of Sentinel-2A remote sensing images were carried out.NDVI,EVIand NDWI indexes were extracted from preprocessed Sentinel-2A remote sensing images.Four terrain factors such as Slope,Aspect,Plan Curvature and Profile Curvature were extracted by DEM.The J-M distance was used to analyze the separability between soil classes.Thirdly,S2A+EVI+EL+SL was selected as the optimal combination parameter to construct maximum likelihood classification model.Finally,I made the soil map of Mingshui County.The results showed that the user accuracy of Meadow soils,Blown soils,Chernozems,Black soil and Swamp soils were 91.67%,69.23%,88.34%,88.61% and 90%,respectively.Meadow soils was the highest of 91.67% user accuracy,and Blown soils was the lowest 69.23%.The overall accuracy reached 89% and Kappa coefficient was 0.84,which showed that the classification results was better,and could meet the demand of soil mapping.Compared with traditional mapping methods,the results of this thesis can achieve detailed description of soil spatial and temporal changes.It not only ensures the quality of soil mapping,but also improves the efficiency of soil survey and renewal.At the same time,some suggestions are put forward for land use according to the characteristics of various soil types.The results of this study can be used to understand the characteristics of soil distribution and have important guiding significance for land management and land use planning in China.