Font Size: a A A

Estimation Of Soil Organic Matter Content In Cultivated Land Based On GF-5 Hyperspectral Remote Sensing Image

Posted on:2022-10-30Degree:MasterType:Thesis
Country:ChinaCandidate:X Z YanFull Text:PDF
GTID:2493306326488014Subject:Agricultural remote sensing
Abstract/Summary:PDF Full Text Request
Soil organic matter(SOM)content and its spatial distribution characteristics are important soil attributes for cultivated land quality evaluation and soil carbon and nitrogen cycle research.Rapid and accurate monitoring of SOM content is also of great significance for precision fertilization and fertility improvement.Hyperspectral satellite remote sensing has high spectral resolution and has the advantage of detecting subtle differences in soil surface properties.In order to evaluate the potential of GF-5hyperspectral image to estimate soil organic matter(SOM)content in cultivated land,and the influence of different soil types on the accuracy of SOM content spectral estimation,this paper took Jiansanjiang Reclamation Area in Heilongjiang Province as the research object,and obtained the hyperspectral image of GF-5 satellite borne visible short wave infrared hyperspectral camera(AHSI)sensor In this paper,the reflectance data of 193 sampling points in the study area were extracted,and the reflectance of soil samples was measured by non imaging hyperspectral spectrometer indoor;based on the reflectance reciprocal,logarithm,first-order differential and other spectral mathematical transformation of the spectral reflectance data of the sample points,the spectral sensitive band of SOM content was determined by correlation coefficient method;and then based on all the samples in the study area Based on the indoor and GF-5 image reflectance data,the SOM content estimation models of multiple stepwise regression(MLSR)and partial least squares regression(PLSR)were constructed,and the effects of GF-5 image data and modeling methods on SOM content estimation accuracy were compared and analyzed;based on the GF-5 image reflectance data of different types of soil(meadow soil,swamp soil and black soil)in the study area,the principal component analysis(PC)model was constructed A)The effects of different soil types on SOM estimation accuracy were compared and analyzed by using MLSR model,PLSR model and MLSR model.The conclusions are as follows:(1)Using GF-5 full band and sensitive band spectral data,based on MLSR and PLSR linear statistical model,the estimation accuracy of SOM content of all samples in the study area is not ideal.The model of log reciprocal first-order differential(1/(Ln R))’of MLSR reflectance in the whole band has relatively high accuracy in estimating SOM content,with modeling accuracy R2 = 0.538,RMSE = 3.602,verification accuracy R2 = 0.383,RMSE = 5.009.The sensitive band has no advantage over the whole band in the estimation accuracy of GF-5 SOM content.(2)In terms of SOM content estimation of different soil types using GF-5 spectral data,the estimation accuracy of black soil is much higher than that of meadow soil and swamp soil.Among them,the black soil PLSR full band(LNR)’SOM content estimation accuracy is the highest,modeling accuracy R2 = 0.967,RMSE = 0.505,verification accuracy R2 = 0.729,RMSE = 1.065;meadow soil MLSR full band(1 / R)’ SOM estimation model accuracy is the highest,modeling accuracy R2 = 0.352,RMSE =5.019,verification accuracy R2 = 0.067,RMSE =9.073;the optimal model for predicting SOM content of swamp soil is the(Ln R)’ model of PLSR sensitive band,with modeling accuracy R2 = 0.594,RMSE= 3.107,verification accuracy R2 = 0.376,RMSE = 8.543.(3)The factors influencing the accuracy of SOM estimation model in the study area include atmosphere,instrument signal noise,soil properties(soil moisture,surface roughness,soil iron oxide,etc.),data preprocessing and modeling methods.In the future,we need to improve and study the quality of hyperspectral data,data preprocessing method and modeling method,so as to improve the accuracy of the Hyperspectral Estimation of SOM content.
Keywords/Search Tags:GF-5 hyperspectral satellite imagery, Soil organic matter, Soil type, Multiple linear stepwise regression, Partial least squares regression
PDF Full Text Request
Related items