| Under the double background of global climate change and anthropogenic activities,a forest ecosystem has remarkable response and change into the vertical elevation.The Himalayan region is an ideal place to study forest change because of its large vertical differences in elevation and a wide range of anthropogenic activities and management practices across the territories of five sovereign states.In order to accurately to identify the Himalayan region's forest area,distribution,change and influence factors,this article was based on the field investigation into point data,using the CART,random forest,naive bayesian classifier respectively to extract of Himalayan region forest information,to select the optimal classifier--random forest classifier to forest information extraction.On this basis,the use of Google Earth remote sensing image point data onto HF data and existing seven different resolution data(JAXA Forest data and Globe Land30 land cover data,the FROM GLC-land cover data,Global Forest trees Watch cover data,GLCF VCF trees cover data,CCI-LC land cover data,MCD12Q1 land cover data)of the overall accuracy and Kappa coefficient,Forest mapping accuracy,users of precision evaluation;The consistency of forest area and space between HF forest data and several existing forest data was analyzed at national scale and pixel scale.Finally,information was extracted from the selected optimal forest to analyze the temporal and spatial changes of the forest in the effective years(14 years)from 1984 to 2018 and the influencing factors of the changes,so as to provide reference and basis for the scientific evaluation and management of the forests in the Himalayan region.The results show that:(1)In Google Earth Engine(GEE)cloud platform based on optical remote sensing data(Landsat)and some auxiliary data,using the random forest,CART,naive bayesian three classifier respectively for the Himalayan region in 2018 forest information extraction,the results showed that the random forest classifier in the extraction of the Himalayan region in overall accuracy(0.9767)and Kappa coefficient(0.9530)are superior to other two kinds of classifier.(2)HF forest data in 2010 were compared with other seven forest data.The areas of HF forest data and other seven forest data fluctuated around 200×103km2.In the comparison of accuracy,HF forest data had the highest overall accuracy(0.982)and Kappa coefficient(0.963).On the national scale,the comparison between HF forest data and the seven forest data shows that,except for the MCD12Q1 data,the other seven data have a high consistency.On the pixel scale,the uncertainty of HF forest data and seven forest data is relatively large.(3)In 2018,the spatial distribution of forests in the Himalayan region was analyzed and the north-south slope was compared.It was found that the forests in the Himalayan region were mainly distributed in the south slope.Among the three regions,the forest area of the central Himalayas is the largest(90465.38 km2).Among countries,India has the largest forest area(64.77 × 103km2).From the perspective of the temporal and spatial changes of forests in the Himalayan region during the effective years from 1984 to 2018(14 years),the temporal changes of forests fluctuated within a certain range,while the spatial changes had certain differences.Forest change factor,the influence of altitude,gradient,slope direction and the distance from the road,with the change of the forest,all have a certain correlation on the elevation,the forest changes mainly in low altitude and high area,on the slope,the area of the forest changes up to a certain slope,slope direction,slope to the area of the slope changes were similar,mainly in sunny slope,the farther the distance from the road,the smaller changes in forest area;Logistic regression law shows that among the influencing factors of forest change,slope,altitude,b2?soilcarbon,b5?soilcarbon,b3?soilp H and distance from water system are more significant than other factors. |
PDF Full Text Request