Vegetation Type and Degredation Classification on the Mongolian Steppe Using Random Forests

Random Forest;Steppe;Mongolia;Classification;Natural Resources and Environment

Liu, WeiBrown, Daniel ;

University of Michigan

关键词: Random Forest;    Steppe;    Mongolia;    Classification;    Natural Resources and Environment;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/99530/Wei%20Liu%20Thesis%20August%202013.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

Creating consistent supervised vegetation classifications in different countrieswhere training data are of different levels of quality and detail is challenging butimportant. For example, mapping steppe types and degradation of Mongolia and InnerMongolia Autonomous Region (IMAR), China, using the same classification schemewould be helpful for doing comparative studies between the two regions andacquiring a better understanding of how country level differences affect vegetation onthe Mongolian Plateau.Steppe and degradation maps, created through on-screen digitizing that combinedimage and ground information as input, were available in IMAR but not inMongolia. We explored supervised classification using Random Forests (RF) toidentify a reasonable sampling and training strategy and applied identical methodsto classify remotely sensed images (Landsat Thematic Mapper 5) in IMAR andMongolia using the same classification systems for the two countries in threeecological regions (meadow steppe, typical steppe and desert steppe). A number ofchallenges limit our ability to extend classifications trained in IMAR to Mongolia forcreating consistent vegetation maps.