Quantifying Vegetation Canopy Eco-water Using Remote Sensing In Maoxian,Upper Reaches Of Mingjiang River

Eco-water?layer?of vegetation refers to the water closely related to the plants on the land surface,specifically the vegetation layer on the surface of earth,containing the water intercepted and stocked by vegetation leaves,roots,humus,the plant itself and the part of the soil where root reaches.The existing forms and moving mechanisms of vegetation eco-water are different from other water resource.The eco-water reserves can not be only used for evapotranspiration,but also regulate surface water and ground water.In addition,it directly effects the distribution of water resources in hydrological cycle and the retention time in storages.As an important storage in terrestrial ecosystem,eco-water?layer?is difficult to accurately quantify with usual methods,which may hinder the effective exploration of its spatial-temporal distribution and evolution,and may not beneficial to comprehend and master the particularity and complexity.However,relying on the advantages of remote sensing?RS?qualifying vegetation eco-water?layer?not only enriches the theoretical study of hydrology and ecological environment,but also contributes to provide scientific theory and technical support for ecological environment improvement and hydrological cycle.Moreover,it has scientific significance and chiefly potential socio-economic benefits for the study of ecological environment in the western region of China and even the whole country.The upper reaches of Minjiang is located in the western Sichuan Plateau with complex topography.Therefore,its ecological environment is sensitive and fragile with specially ecological status.Besides,the upper reaches of Minjiang are rich in forest resources,with many types of forest and extensive distribution of eco-water.General methods have obvious limitation on those areas with complicated internal structure of forest and special geological and geomorphological conditions,such as the study area of this paper.It is difficult to accurately quantify vegetation canopy eco-water reserves.In view of this,on the basis of new ideas and methods for quantifying eco-water in project team for many years,and the summary of researches at home and abroad,this paper utilizes Landsat8 OLI image to extract a variety of vegetation information?such as vegetation index,vegetation types,reflectance and texture index?by means of 3S?RS,GIS,GPS?integration technology.Multiple linear and Model Tree?a machine learning approach?in this paper were applied,which effectively compensated the limitations caused by merely using empirical models and vegetation index.The main results are:?1?The overall accuracy of the Landsat8 OLI supervised classification was85.15%,with Kappa coefficient 0.81.The forest area is about three times of shrub area.The main vegetation types are coniferous forest,broad forest,mix forest and shrub,accounting for 42.11%,9.73%,1.89%and 18.49%of the total area of study area,respectively.?2?The multivariate linear model and Model Tree model of vegetation canopy eco-water reserves in the study area were constructed.Comprehensively comparing model assessment indexes and variable types,the results indicate that the prediction accuracy of Model Tree model is higher than multivariate linear model.The multivariate linear regression model?including 7 variables?constructed by band reflection,vegetation indexes and texture indexes,with the model efficiency?R²?and root mean square error?RMSE?of the 10 fold cross-validation were 0.66?p<0.001?and 16 Mg ha^-1?1 Mg=10⁶ g?,respectively.While,the Model Tree model?including5 variables?with the R² of 0.65?p<0.001?and RMSE of 11 Mg ha^-1.Consequently,the prediction accuracy of Model Tree model was better than multivariate linear model.?3?The spatial distribution characteristics of canopy eco-water stock among different vegetation types in the study area are obvious.The total stock and average stock of vegetation canopy eco-water are 7.51×10⁶ Mg and 27.02 Mg ha^-1,respectively.The canopy eco-water stock of each vegetation type was mainly coniferous forest?accounting for 69.30%of the total stock in the study area?,followed by broad forest and shrub,mixed forest was the least.As the important component of eco-water?layer?,vegetation canopy eco-water is crucial to the accuracy of quantitative inversion of the eco-water?layer?.In this study,RS technology was applied with a variety of vegetation information.Different modeling algorithms are used to realize the remote sensing quantitative inversion of vegetation canopy eco-water.The research will contribute to accurately quantify the eco-water?layer?and reveal its dynamic inversion process.Furthermore,it can further reveal the primary situation of the vegetation eco-water on land surface in western Sichuan Plateau,and also can provide scientific theory and technical support for ecological environment protection and restoration.