Research On Karst Rocky Desertification Information Extraction Based On Red-Nir-Swir1 Spectral Feature Spac

The complex topography,severe surface fragmentation and landscape heterogeneity of the karst region in southwest China make it extremely difficult to extract information on stone desertification in the region.In order to establish a simple,rapid and direct index model to indicate the degree of stone desertification development,this study firstly used the lithological vector and land use type data of the study area to mask non-karst areas and areas where stone desertification is unlikely to occur,and used the red band(Red),near infrared band(Nir)and shortwave infrared 1 band(Swir1)reflectance of Landsat 8 OLI remote sensing images as the feature variables to establish a two-dimensional Swir1-Nir1 index.The reflectance spectral feature spaces of Swir1-Nir,Red-Nir and Swir1-Red were established based on the reflectance of Landsat 8 OLI images.The three index models were then constructed based on the variation of the degree of stone desertification in the three reflectance spectral feature spaces,namely the vertical stone desertification index 1(PRDI1),vertical stone desertification index 2(PRDI2)and vertical stone desertification index 3(PRDI3).Finally,the accuracy of the stone desertification extracted by these three index models is verified and compared with existing index models constructed based on surface parameter feature space(KRDI,RSDDI)and linear spectral decomposition-based methods(SMA).The main findings are as follows:(1)In the comparison of the index models PRDI1,PRDI2 and PRDI3 constructed based on the reflectance spectral feature space,the overall accuracy and Kappa coefficient of PRDI1 are0.859 and 0.821 respectively,which are higher than those of PRDI2 and PRDI3,and its mapping accuracy and user accuracy are higher than those of PRDI2 and PRDI3,showing the best applicability and stability.(2)The comparison between PRDI1 and models based on surface parameter feature space(KRDI1,RSDDI)and mixed image element decomposition method(SMA)shows that the extraction accuracy of PRDI1 is higher than that of KRDI1,RSDDI and SMA,and the overall accuracy is improved by 0.9%and 1.1%and 16%respectively,which is more accurate and effective in the extraction of spatial pattern and intensity information of stone desertification.Compared with KRDI,RSDDI and SMA,the parameters of PRDI1 model are easier to obtain,and the model construction process is simple and has higher accuracy,which has certain application potential and promotion value in stone desertification information extraction.(3)Through the analysis of the evolution characteristics,evolution rate,evolution direction and scale of stone desertification,the stone desertification in Huaxi District mainly shifted from low level to high level during 1995-2000,and the degree of deterioration intensified,mainly influenced by unreasonable human activities;during 2000-2020,due to the government’s measures of returning farmland to forest and comprehensive management of stone desertification,the stone desertification in some areas deteriorated,but in general,the stone desertification in the study area developed in a benign direction,and the ecological environment gradually recovered.(4)Based on GIS technology,the factors affecting the formation and development of rock desertification in Huaxi District were analysed in terms of natural factors(including carbonate rock lithology,slope and elevation)and human factors(population density),and the analysis results showed that rock desertification in the study area mainly occurred in limestone areas and higher elevation areas,and the occurrence of rock desertification was correlated with slope.For anthropogenic factors,human activities are an important factor influencing the occurrence of stone desertification,which mainly occurs in areas with a population density of 400-700people/km~2.