Study On The Effect Of Human Activities On The Spatial Pattern Of Soil Organic Matter

As an important component of soil,soil organic matter is an important factor,the soil fertility and soil quality in the ecosystem and the carbon cycle plays an important role,is an important source of carbon cycle in terrestrial ecosystem and sinks.The change of organic matter plays a key role in the environmental problems such as desertification,grassland degradation and regional eco-environmental degradation.The development of hyperspectral remote sensing technology provides a convenient and efficient means for obtaining soil physical and chemical parameters.Compared with traditional methods,it saves time and effort and reduces costs.Compared with conventional multispectral remote sensing,the inversion precision can be improved.Its high spectral resolution for analysis of the intrinsic attributes such as soil conditions,provide detailed data,has been to estimate soil organic carbon,water,nitrogen,phosphorus,potassium and other related elements content has become more widely used.For the contrast of the different degree of human disturbance under the same background prediction model and the spatial pattern of soil organic matter content of desert,distribution,based on the Tianshan Fukang soil as the research object,using a variety of means and methods,the forecast analysis on the content of soil organic matter.Using field measured spectral data and imaging spectral data,through the spectrum transform?original,first order differential and second order differential from bottom,bottom,first order differential,reciprocal second order differential,inverse logarithms,inverse logarithms first-order differential,inverse logarithms second-order differential?choose sensitive wave bands,salt index,vegetation index calculation,improve the vegetation index as the independent variable,respectively in the left area,human disturbance area set up organic matter content of the estimate of multivariate linear regression model and multiple stepwise regression model,principal component regression model and partial least squares regression model,artificial neural network model,comparing different model precision,choose the best models for predicting the organic content.Secondly,the spatial pattern of soil organic matter in the study area was analyzed by using Landsat8 remote sensing image and geostatistical method,and its distribution law was analyzed.The main results and conclusions of this study are as follows:?1?After removing organic matter quality score is greater than 2%of the sample,the sample data in the area more tending to normal distribution,the precision of the prediction model has improved,with no interference region and human disturbance zone prediction effect are the best first order differential and second order differential multiple stepwise regression model,R² is 0.78,0.54,RMSE is 1.41 g/Kg,2.26 g/Kg,RPD is 2.14 and 2.09.?2?In combination with spectral composite index model,based on the reflectance image combined with the measured spectral reflectance of vegetation index,the index of salt as the independent variables of multiple linear regression model accuracy is higher,after spectral transformation,the model accuracy is improved,the use of the measured spectra with salt index,vegetation index of unmanned interference area of first order differential inverse of multivariate linear regression model and artificial interference zone logarithmic multiple linear regression model of first order differential,R² of 0.76 and 0.73 respectively.?3?Through correlation analysis,there is no interference area of narrow band SI₂,SI₃ and RVI and NDVI,and wide range of SI₁,SI₂ and RVI and NDVI,human disturbance of the narrow and wide band SI₁ salt index,SI₃ and SI1,SI₂,vegetation index RVI and NDVI and RVI,RDVI has good correlation with organic matter content and use the multiple linear regression model and partial least squares regression model.In the model established,the narrow-band partial least-squares regression model with the undisturbed area and the human interference area is the best,with R² reaching 0.75,0.82,and RPD as 2.01 and 2.14 respectively.?4?Based on vegetation index calculation formula of B7 band is added to improve the correlation between vegetation index and organic matter is improved after the former have improved,at the same time as the independent variable of ANNs and the precision of the partial least-squares regression model has promoted,no interference area,human interference area the best models are based on the improved ANNs model of vegetation index,R² is 0.83,0.76 respectively,RPD is 2.15 and 2.06.?5?Soil organic matter content in the study area,using ordinary kriging interpolation and remote sensing inversion two inversion methods,two methods to present the distribution of organic matter content is relatively consistent,no interference area,human disturbance area are all present among low,high and low southwest,northeast,north high-trends,among them there is no interference in organic matter content is more concentrated 15 to 20 g/Kg,human disturbance area mainly concentrated in the 10 to 15 g/Kg.Human activities have a great influence on the content and distribution of organic matter.For example,the content of organic matter in undisturbed area is higher than that of human disturbance,and the higher the degree of human disturbance is,the lower the organic matter content.At the same time,when the model is established,the accuracy of the model is always higher than that of the artificial interference zone.