The Research of Retrieving Exposed Soil Organic Matter and Soil Mechanical Composition by Remote Sensing in North Xinjiang Province, China
|School||Xinjiang Normal University|
|Course||Cartography and Geographic Information Systems|
|Keywords||soil organic matter soil mechanical composition hyperspectral remote sensing MODIS the north region of Xinjiang Province|
Xinjiang Province is one of the typical ecologically fragile regions in China, and the problems of land degradation in this region such as sandy desertification, soil erosion and soil salinization are so serious that limit the development of economy and society in Xinjiang Province. The monitoring and assessment of land degradation is the fundament of sustainably using regional land resources, and the key to monitor and assess the land degradation is to dynamically timely and quantitatively monitor the core indicators of land degradation such as soil organic matter and soil mechanical composition by remote sensing. So, it is very important to build a method to retrieve soil organic matter and soil mechanical composition based on remote sensing to support the research of land degradation and its rehabilitation.With the development of remote sensing and especially for the hyperspectral remote sensing, it is possible to use the hyperspectral information of soil to quantitatively gain the soil properties parameters. In this study, we selected the north Xinjiang as the research region, and used ASD-FieldSpec Portable Spectroradiometer to measure the field spectrum of soils, and analyzed the soil organic matter and soil mechanical composition in the laboratory. Based on transferring the original soil reflectance spectrum into several forms, we selected the best band which had the largest correlation coefficient and the best simulation method, and then the models for retrieving soil organic matter and soil mechanical composition (the percentage of sand particle from 0.05 to 2mm, and we use“sand percentage”for short) in research regions were built. Finally, we used MODIS images and the models built above to retrieve the soil organic matter and soil mechanical composition in research region. Through overall analysis and study, the following main conclusions were drawn:(1)In the research of estimation of soil organic matter based on field spectrum data, we found that the correlation between soil organic matter and soil original spectrum and its transformation of logarithm’s reciprocal could reach the maximum value at 430nm, and the correlation coefficient were 0.68 and -0.68 respectively. Based on the statistic and comparing the estimated value by model and observed value, we found that the logistic regression model using the spectrum data at 370nm transferred by reciprocal as independent variable could get the highest accuracy when estimating soil organic matter.(2)In the research of estimation of soil sand percentage based on field spectrum data, we found that the correlation coefficient among soil sand percentage and soil spectrum transferred were higher than that for soil organic matter, and the correlation between soil sand percentage and the soil spectrum data transferred by logarithm’s reciprocal could reach the maximum value at 430nm, and the correlation coefficient was 0.76. The accuracy of soil sand percentage estimation models were higher as total, which nearly reached 90%, and the one-dimensional linear regressions model using spectrum data at 370nm transferred by reciprocal and spectrum data at transferred by one-order differential as independent variable could get the highest accuracy when estimating soil sand percentage.(3)Based on systemically analyzed the band spectral and spatial resolution of MODIS, the correlation among soil organic matter and soil sandy percentage with soil spectrum data and estimation accuracy, we found the MODIS band3 could be used to retrieve soil organic matter and soil sand percentage at regional scale. We had built models for retrieving soil organic matter and soil sand percentage by using spectrum data at 460nm to meet MODIS bands. The logistic regression model using the spectrum data transferred by reciprocal as independent variable could get the highest accuracy when retrieving soil organic matter, and retrieving results showed that the soil organic matter of bare soils in research region was always between 0 to 3.24g/kg; The one-dimension three-order regression model using the original spectrum data as independent variable could get the highest accuracy when retrieving soil sand percentage, and the retrieving results showed that the soil sand percentage of bare soil was always near 95%. The retrieving results could match the actual conditions well.