Study On Decision Tree Classification Of Aquatic Vegetation In Lake Ulansuhai Based On Multi-source Remote Sensing

As an important part of inland lakes,aquatic vegetation has special ecological functions.The change of aquatic vegetation can effectively reflect the ecological status of lakes,which is of great significance for the protection and management of lake environment.Taking Wuliangsuhai as the research area,based on the analysis of the measured spectral data of aquatic vegetation,the decision tree model based on Python is used to classify the measured spectral data and multi-source remote sensing data of aquatic vegetation;Using GF-1 WFV remote sensing data with the best classification effect,the temporal and spatial distribution characteristics of marine vegetation and its response to temperature in Wuliangsu from 2014 to 2020 were analyzed.The main conclusions are as follows:(1)The results show that the remote sensing reflectance of emergent plants is higher than that of other three kinds of ground objects in the range of 710-1400nm;In the range of 500-1350 nm,the remote sensing reflectance of floating yellow moss is higher than that of submerged plants and water(including clear water and turbid water);According to the band combination calculation,the floating yellow moss can also be distinguished from the submerged plants above the water surface;The remote sensing reflectance of submerged plants above water surface is higher than that of water body in the range of 700-1000nm;Although submerged plants and water bodies can not be completely distinguished,the accuracy of discrimination can also meet the practical application;The reciprocal logarithm of the spectrum,the position of "six sides" and the slope are easy to distinguish.In addition,there are obvious changes in the spectral growth process of aquatic vegetation,and the spectral characteristics used to distinguish different types of aquatic vegetation also change.(2)The results of GF-1,landsat-8 and sentinel-2 aquatic vegetation decision tree classification models based on field measured data show that sentinel-2 classification results are significantly better than GF-1 and landsat-8,so reasonably increasing the sensor band can effectively improve the accuracy of decision tree classification model,and increasing the number of layers of decision tree classification model will also improve the classification effect.(3)GF-1 WFV,GF-6 WFV,Landsat-8,zhuhai-1(ZH-1)and sentinel-2 were selected as satellite data sources to construct decision tree classification models of aquatic vegetation with different satellite sensors.The comparison of classification effect and accuracy showed that GF-1 WFV had the highest overall classification accuracy(OA = 93.45%,kappa coefficient = 0.9194),followed by GF-6 WFV(OA =93.10%,kappa coefficient = 0.9194),Kappa coefficient = 0.9150)and sentinel-2(OA= 92.76%,kappa coefficient = 0.9054).Although there are many bands in ZH-1hyperspectral data,the overall classification accuracy is not the highest(OA =91.54%,kappa coefficient = 0.8960),and the accuracy of Landsat-8 is the lowest.In order to apply the existing decision tree classification model to remote sensing images of different time phases,a relative radiometric correction method for GF-1WFV data is established,and the classification effect can meet the requirements of practical application.(4)Based on the GF-1 WFV satellite data from 2014 to 2020,the spatial and temporal distribution characteristics of marine vegetation in Wuliangsu were studied;The area of submerged plants increased before 2018,and then decreased;In 2014 and2017,the floating yellow moss broke out on a large scale,with a large area and little difference in other years.Based on the analysis of the spatial and temporal distribution characteristics of aquatic vegetation in 2017 and 2020,it is found that the change of aquatic vegetation area is related to the growth period,showing a trend of first increase and then decrease;The maximum area of emergent vegetation and submerged vegetation appeared between June and July,while that of floating yellow moss was the largest in July.(5)The large-scale fish culture in the haihao area of Southern Wuliangsu resulted in the decrease of submerged plant area and the disappearance of floating yellow moss;Temperature is an important factor affecting the growth of aquatic vegetation.Emergent vegetation and floating yellow moss have a positive correlation with air temperature,while the correlation between submerged vegetation and air temperature is relatively weak.