Font Size: a A A

Study On The Wetland Remote Sensing Classification And Ecosystem Health Evaluation Based On Support Vector Machine

Posted on:2018-06-21Degree:MasterType:Thesis
Country:ChinaCandidate:H T XuFull Text:PDF
GTID:2321330515961526Subject:Hydraulic engineering
Abstract/Summary:PDF Full Text Request
Wetlands are important natural resources and ecosystems on Earth and play an irreplaceable role in regulating runoff,purifying the environment,controlling pollution,conserving water sources,resisting floods,conserving biodiversity and maintaining regional ecological balance.However,with the increasing human activities,the contradiction between human demand development and natural resources and environment is becoming increasingly acute,wetland resources have been seriously disturbed and destroyed,and led to a series of ecological and environmental problems.With a background in Linghekou estuarine wetlands this study uses remote sensing images as data source to explore the new methods of extracting wetlands cover information based on support vector machines theory and then depth study the ecosystem health status of wetlands.The main contents are given as follows:(1)Extract wetlands cover information based on support vector machines.Based on the remote sensing image and field GPS data,the remote sensing image of the study area was pretreated by ENVI and ArcGIS,this paper uses the support vector machine(SVM)classification method to classify and extract the image data.After the sample verification and precision evaluation,the cover information of Linghekou wetland in 1995-2014 is obtained.And provide data support for wetland landscape fragmentation analysis,wetland functional area division and ecosystem health assessment.(2)Landscape pattern change and division of function zones in Linghekou wetland.Taking data of TM images in 1995,2000,2005,2009 and 2014,choosing diversity index,dominance index,patch density index and shannon's diversity index,to study landscape pattern change and make division of functional zones of Linghekou Wetland Natural Reserve,Liaoning province in this paper for 5 periods.The information of 8 kinds of landscape types was extracted.The results showed that the extent of landscape fragmentation have augmented a lot with the effects of human activity from 1995 to 2005;and the extent of landscape fragmentation have declined relatively from 2005 to 2014.According to the rules put forward by Man and the Biosphere Programme of division of functional areas,areas of the Linghekou Wetland Natural Reserve were divided into,including central,buffer and experiment zones,their areas were 305.88 km2,268.28 km2 and 261.59 km2,respectively.(3)Ecosystem health assessment and warning.Based on the PSR model,the health evaluation index system of Linghekou wetland ecosystem was constructed.The weight of each index was determined by using the analytic hierarchy process.The single factor was evaluated by the logical growth model.Finally,the ecological health of the five periods was analyzed by comprehensive evaluation method.The research finally got the ecosystem health values that were respectively 0.642,0.617 in 1995 and 2000,which showed that the ecosystem health was in a healthier state;the ecosystem health values that were respectively 0.524,0.436 and 0.405 in 2005,2009 and 2014,which showed that the ecosystem health was in a subhealthy state and some necessary measures should to be immediately taken.Based on the gray system prediction model,the ecological health prediction model of Linghekou wetland is constructed,and the ecological health prediction of the study area is carried out.The result showed that the prediction for the Linghekou wetland ecosystem health were respectively 0.357,0.321,0.291 and 0.267.The wetland ecosystem health was in a general pathosis state and the trend of the situation continued to become more severe.Ecosystem health threat is very serious and this area needs more protection and management for the wetland ecosystem to survive.
Keywords/Search Tags:Support vector machines, Remote sensing classification, Division of function zones, Ecosystem health assessment, Gray forecasting
PDF Full Text Request
Related items