Distribution Information Extraction And Dynamic Change Monitoring Of Torreya Grandis Based On Multi-source Data

Rapid expansion of Torreya(Torreya grandis ‘Merrillii')forests in the mountainous region in Zhejiang province in the past three decades has produced many environmental problems such as soil erosion and poor water quality,requiring update of its spatial distribution at timely ways.However,to date there are no suitable approaches available for mapping Torreya forest distribution,especially the new Torreya plantations due to their complex landscape patterns.Firstly,this research used high spatial resolution satellite images(Chinese satellite GF-1 and ZY-3 images)and digital elevation model(DEM)data to extract old Torreya forests and new Torreya plantations using a newly proposed expert rules based approach.Then the spatial distribution of Torreya in the high resolution imagery was developed as a reference data by scaling up for extraction of old Torreya forests and new Torreya plantations in large area using random forest algorithm.The 30 m classification result in 2015 and the accuracy assessment were made.An binary mask of new Torreya plantations based on the predict results in 2015 was made to mask the image in 2004,and then the distribution of new Torreya plantations and other land covers in 2004 were interpreted using ISODATA unsupervised method combined visual interpretation.The classification of Landsat imagery in 1995 and 1985 were then mapped as the same way of classification of imagery in 2004.The dynamic monitoring analysis was then made and the land covers changed to new Torreya plantations were assigned.The study mainly gets the following conclusions:(1)The newly proposed expert rules approach can effectively distinguish both old Torreya forests and new Torreya plantations from other land covers with producer's accuracies of 84% and 92%,and user's accuracies of 77% and 85%.Comparing with traditional supervised classification approach-maximum likelihood classifier,the new approach considerably improved the classification accuracy.(2)The combination of high resolution imagery into the Landsat data classification as a reference data is a cost-effective method.The random forest classification algorithm achieved good classification result,in which,the producer's accuracy of old Torreya forests and new Torreya plantations was 93% and 85%,and the user's accuracies was 87% and 88%.(3)The main old Torreya forest in the study area was mainly distributed in the border area of Zhuji City,Shaoxing County and Shengzhou City,and the border area of Zhuji City and Shengzhou City in the central mountainous area of Kuaiji Mountain.While new Torreya plantations in the Kuaiji mountain near old Torreya forest was more concentrated distribution,but also scattered in the south of the Kuaiji Mountain,the border area of Dongyang and Pan'an at the Dapanshan mountain,and the junction area of Pujiang,Lanxi,Jiande at the Longmen mountain,and a small Part of Fuyang.(4)In 1985,the new Torreya plantation areas were mainly concentrated in the vicinity of the growth area of old Torreya forest,which is mainly concentrated in the border area of Zhuji,Shengzhou and Shaoxing,and the junction of Dongyang,Zhuji and Shengzhou,and from this center area,new Torreya plantations gradually expanded outwardly,from Kuaiji mountain area to the nearby Longmen mountain and Dapanshan mountain,namely including Pan'an,Pujiang,Jiande,Lanxi and other regions.The maximum increase rate of new Torreya plantations happened between in 2004-2015.During the period from 2004 to 2015,the most important types of conversion were shrub and coniferous forest.