The Combination Of UAV Survey And Landsat Imagery For Karst Rocky Desertification Assessment

KRD(Karst Rocky Desertification)in southwest of China is one of the most serious environmental issues in China.It is also one important constraint for regional sustainability development which directly causes one of the major factors of poverty,economic backwardness,and degradation of the eco-environment in this region.At present,a series of KRD assessment methods have been developed based on satellite remote sensing technology.Because the image data is affected by the external environment(weather,cloud cover,terrain,etc.)and the sensor itself(resolution),the accuracy of KRD information extraction and evaluation is still low,and it is difficult to achieve quantitative research.In recent years,UAV remote sensing technology can quickly and easily capture high-resolution images,providing new ways and methods for KRD assessment research.The purpose of this research is to use the UAV survey combined with satellite remote sensing images to realize the KRD scale transition and develop a new KRD assessment method.The main results are listed below: 1)The overall accuracy of image of UAV classification is above 0.97,F1 is above 0.94;2)The regression relationship between the ratio of bare rock and image DN is shown as that the regression goodness of fit of DN value of band 2 is the highest,and R2 is 0.8608.Linear regression R2 and cross-validation from the measured data and model extraction results show that the R2 value is 0.71 and the RMSE value is 0.03;3)Based on Multiple data combination approach,the total area of rocky desertification in Shilin County is 6,047.94 ha,accounting for 3.6% of the total land area of Shilin County.The degree of KRD is mainly mild,with overall accuracy is 0.91 and Kappa value is 0.81;Based on vegetation coverage approach,the area of rocky desertification in Shilin County is 15949.98 ha,accounting for 9.48% of the land area of Shilin County.The degree of rocky desertification is mainly mild and moderate,with an evaluation accuracy of 0.65 and a Kappa value of 0.43.Conclusion as below: 1)Using high-resolution UAV images can extracting quantitative rocky desertification information and the vegetation is high;2)There is a good linear regression relationship between bare rock and satellite image DN values;3)The comparison of KRD assessment methods based on multiple data and KRD assessment methods based on vegetation coverage has higher accuracy and reliability;4)With the enlarge of the patch size of KRD,the information of KRD will be lost,and moderate rocky desertification and severe rocky desertification are susceptible to the patch scale of KRD assessment.