Wetland Information Extraction Of The Yellow River Source Based On Decision Tree Model

As one of the three major ecosystems in the world,wetlands play an important role in protecting the ecological environment and species diversity.In recent years,under the dual influence of human activities and climate change,the ecological environment in the source region of the Yellow River in the hinterland of the Qinghai-Tibet Plateau has experienced varying degrees of degradation,including severe shrinkage of glaciers,sharp decline in species diversity,increased soil erosion and grassland degradation.Brought severe pressure on the local wetland ecological environment.Obtaining accurate wetland information(type and area)has become one of the prerequisites for the study of wetland ecological environment changes in the source region of the Yellow River.The traditional wetland remote sensing extraction method only uses the spectral information of the image,and it is difficult to combine the spatial information between the multi-source remote sensing data and the feature type,which makes the wetland extraction accuracy limited.Aiming at the shortcomings of wetland remote sensing extraction accuracy,on the basis of fully considering different wetland spectral characteristics,combined with multi-source remote sensing data and GIS spatial analysis method,a decision tree model based on expert knowledge is used to extract the Yellow River source wetland.The main results of the paper are as follows:(1)On the basis of field surveys,combining the Ramsar Convention and the national standard for wetland classification,construct the Yellow River source wetland interpretation mark,and propose a wetland remote sensing classification scheme suitable for the Yellow River source area.(2)According to the characteristics of the wetland types in the study area that are complex and unevenly distributed,a decision tree extraction model based on expert knowledge is constructed,using NDWI combined with shape index to complete the extraction of river and lake wetlands,and for swamp wetlands,it is divided at the NDWI threshold On the basis of this,TVDI(Temperature Vegetation Drought Index),DEM,NDVI and GIS spatial analysis methods are introduced to comprehensively extract.The decision tree extraction model comprehensively considers the different characteristics of wetlands,makes full use of spectral and geometric features,while reducing manual operations,and improving the accuracy of information interpretation.(3)The accuracy of the extraction results of the decision tree model based on expert knowledge is evaluated.The overall accuracy of the decision tree method is 89.69%,the Kappa coefficient is 0.87,and the same accuracy evaluation method is used to evaluate the results of the random forest classification.The overall accuracy is 81.64 %,Kappa coefficient 0.77.It shows that the decision tree model has higher extraction accuracy and is more suitable for wetland information extraction in the source region of the Yellow River.(4)Wetland extraction results show that the total area of the Yellow River source wetland is 2126.33km2,and lakes and swamp wetlands are the main types of wetlands in the study area,accounting for 65.81% and 33.58% of the total wetland area;On the banks of rivers,around lakes and low-lying shade areas.