Monitoring And Dynamic Study Of Desertification In Weigan And Kuqa River Delta Oasis,Xinjiang

Land desertification is one of the serious eco-environmental problems in the world.It not only threatens the living environment of the human beings,but also affects the social stability.In addition,it's an important factor that restricts the rapid development of the global economy.Therefore,timely and accurate grasping of the current situation and development trend of land desertification is of great significance for the research of global environmental change and desertification control.China is one of the countries that are seriously affected by desertification in the world,and Xinjiang is one of the areas where land desertification is more serious in China.In this study,the Landsat series data of TM,ETM+ and OLI remote sensing images are used to monitor and analyze the land desertification in the area of Weigan and Kuqa River Delta Oasis in Xinjiang from 2002 to 2016.The main contents and conclusions are as follows:(1)Taking account of the spectral features,spatial features and other auxiliary information of remote sensing data,and combine field surveys and sampling,a classification system and standard for land desertification in the study area are proposed in this study.The first class is divided into two categories: non-desertification lands and desertification lands.In the second class non-desertification lands are classified into vegetation,water and town,and desertification lands are classified into sandification land and salinization land.Then sandification land and salinization land are classified into four different levels in the third class.(2)The feature parameters of each class are extracted by original wave band,principal component analysis,minimum noise fraction transformation,vegetation/water/soil indexes,band math,spatial relationship of land type and other methods.And the abundance values of end-member are obtained by the decomposition of mixed pixels.Then extracting the classification information of vegetation,water,town,sandification land and salinization land by the method of hierarchical decision tree classification and random forest classification.The average accuracy of the hierarchical decision tree classification of remote sensing images is 93.42%,and the accuracy of the random forest classification is 92.31%.Therefore,the result of the hierarchical decision tree classification is better than that of the random forest classification in this study.(3)Extracting sandification and salinization information at different levels to improve the image classification and mapping.The extraction of salinization information is based on partial least squares regression(PLSR)algorithm.According to PLSR,taking the bands of Blue,Green,Red,Near Infrared(NIR),Short-wave Infrared 1(SWIR1),and Short-wave Infrared 2(SWIR2)in the original image as the independent variables,and the conductivity EC as the dependent variable to establish the prediction model of soil salinity based on laboratory spectra.And based on the laboratory spectral model,Normalized Difference Vegetation Index(NDVI)is added as independent variables to establish the modified prediction model of soil salinity based on image spectral index.The results show that the modified prediction model of soil salinity based on image spectral index is superior to the prediction model of soil salinity based on laboratory spectra.The modeling accuracy increases by an average of 31.43%,and the test accuracy increases by an average of 24.18%.(4)According to the results of land desertification classification in the study area,the dynamic changes of land desertification in the Weigan and Kuqa River Delta Oasis are analyzed through spatial statistical analysis,overlay analysis,transfer matrix calculation,and landscape pattern index calculation.The results show that(1)the proportion of desertification land in the study area from 2002 to 2016 is approximately 60%-70%.The conversion of different types of land objects is very frequent during various research periods,especially among the vegetation,sandification land and salinization land.(2)From 2002 to 2016,desertification in the Western Oasis has improved,and desertification lands have gradually changed to non-desertification lands;non-desertification lands have changed to desertification lands in the southeastern part of the study area,and the situation of desertification has become more severe.(3)the fragmentation degree of desertification land is larger,and the distribution of patches is irregular.In addition,the landscape shape of vegetation,sandification land and salinization land is more complex.The various types of land objects in the study area are relatively fragmented,the degree of reunion is moderate,and the land objects have a high level of landscape heterogeneity and complexity.There are no obvious dominant land types,and the distribution of landscape patches is uniform in the study area.