Spatial Identification Research Of Soil Types In Typical Black Soil Area Based On Different Temporal Responses

Land is the carrier for the survival and development of terrestrial organisms on the earth,while soil is an important part of land,and soil plays a crucial role in human life and development.With the constant change of the global environment and the sustainable development of science and technology,planning,utilization and protection of land resources are becoming increasingly important for the development of future society,as a result,the fields and disciplines which are closely related to land are also developing rapidly.As the data reference of many fields or disciplines such as land use planning,precision agriculture development and sustainable land use,soil data plays an important role in the development of land field and disciplines.Therefore,accurate soil data has also become an important prerequisite for the development of land disciplines.Soil classification is one of the classical problems in soil science.Traditional methods of soil classification rely more on field operations and subjective analysis,which requires a large amount of cost to be invested in the collection and analysis of soil profile,including labor,funds,time and other costs.Therefore,it is urgent to identify soil types in a fast and accurate quantitative way.Based on the classification system of soil genesis in China and remote sensing technology,this study took Mingshui County,a typical black soil area in Songnen plain,as the research area.In order to accomplish spatial identification and analysis of three soil types(Phaeozems,Chernozems and Cambisols)in Mingshui County,a total of 20 scenes Landsat satellite remote sensing image data of bare soil period from 1984 to 2018,the second soil census data,SRTM DEM data with 30 meters spatial resolution and land use type data were used in the study,we introduced 4 temporal responses information(mono-temporal,bi-temporal,multi-temporal and hyper-temporal);2classification methods(random forest model,maximum likelihood classifier)and 2 data compression dimensionality reduction methods(principal component analysis,tasseled cap transformation);3 data sets(6 bands,3 indexes after tasseled cap transformation,6 bands + 3indexes after tasseled cap transformation);5 terrain attributes(slope,aspect,curvature,relief degree of land surface,elevation).In addition,this paper analyzed the reasons for the differences of interannual soil types identification results,we attempted to reveal the relationship between interannual soil types identification changes and soil salinization changes.The main research contents and results of this paper are as follows:(1)Abundant temporal responses can improve the accuracy and stability of the prediction models.The highest overall accuracy and Kappa coefficient were obtained by using six bands of information of reflectance and three indexes of tasseled cap transformation,combined with hyper-temporal data,with an overall accuracy of 80.56% and a Kappa coefficient of 0.704.Compared with the monotemporal,bi-temporal,and multi-temporal data,the overall accuracy of using hyper-temporal data was increased by 17.78%,12.23%,and 8.89%,and the Kappa coefficient of using hyper-temporal data was increased by 0.265,0.185,and 0.136,respectively.(2)Tasseled cap transformation is an effective method of data compression and dimensionality reduction.The prediction accuracy of the three indexes after tasseled cap transformation(test2)is higher than that of the six bands of Landsat images(test1).Through the using of random forest model,we found that the sensitivity of near-infrared(NIR)and shortwave infrared 2 band(SWIR2)in Landsat remote sensing image data was higher than that of other 4 bands.The sensitivity of greenness index was higher than the other two indexes.(3)The introduction of terrain attributes is beneficial to the space identification of soil types in black soil area.In this paper,the maximum likelihood classifier and five terrain attributes are used.Relief degree of land surface(RDLS)is helpful to improve the identification accuracy of Phaeozems and Chernozems,and elevation(ELE)is helpful to improve the identification accuracy of Cambisols.After the introduction of relief degree of land surface(RDLS),the overall accuracy and Kappa coefficient of hyper-temporal mapping were 88.22% and 0.818,higher than the accuracy of other terrain factors.(4)There was a significant correlation between the results of interannual soil type identification with differences and soil salinization.The correlation coefficient between the salinization area of cultivated land and the area of Phaeozems and Chernozems was-0.529*,the correlation coefficient between the salinization area of non-cultivated land and the area of Cambisols was 0.568**.It was further explained that soil salinization is one of the reasons for the change of soil type area.This study accomplished the soil types space identification combined with remote sensing technology,we discussed the influence of the temporal responses for the space identification,improved the efficiency and precision of the soil type identification,and provided technical support for updating soil data.Through soil types data revealed the interannual change trend of soil quality,the identification results of soil types and the potential relationship between land degradation was explored by us.It enriched the application of soil type data in land resource management and provided support for the research of land use planning and sustainable land use.