Land cover classification and change detection in drylands: an evaluation of remote sensing approaches

TitleLand cover classification and change detection in drylands: an evaluation of remote sensing approaches
Publication TypeThesis
Year of Publication2011
AuthorsHestir KLee
Academic DepartmentGeography
DegreeMaster of Applied Geography
Number of Pages156
UniversityNew Mexico State University
CityLas Cruces
Accession NumberJRN00570
Keywordsdrylands, land cover change, land cover classification, Mesilla Valley, remote sensing, technique, Yuma Valley
AbstractLand cover change is occurring at unprecedented rates worldwide and is believed to contribute to environmental concerns such as altered biogeochemical cycles, loss of wildlife habitat, decreases in biodiversity, and reduction of soil productivity. Of particular concern are dryland regions of the world because they are experiencing rapid population growth which is a driver of land cover change. The Southwestern United States is one dryland region which is experiencing rapid population growth and associated expansion of urban land cover. The Mesilla Valley of Southern New Mexico is no exception. In this region, which has experienced rapid growth during the last four decades, it is unclear at what rate land cover is changing and what effects this has had on desert ecosystems. Satellite remote sensing can be of great value for mapping land cover and assessing land cover change. However, accurate land cover mapping of arid and semi-arid environments is challenging due to the natural characteristics of these environments. Few researchers have systematically investigated which combinations of image bands, image derivatives, and hard classification algorithms address the difficulties of mapping land cover in arid environments. This is important because hard classifications are by far the most commonly used technique for mapping land cover and drylands are particularly vulnerable to land cover change due to human population pressures. This study examined a series of hard classification techniques and image feature stacks in order to determine which combinations may provide accurate land cover maps in the Mesilla Valley of New Mexico. The tests were repeated for a second dryland region (Yuma Valley, Arizona) to see if they were generalizable to another dryland region. The feature combination that gave the highest accuracy was used to provide land cover information and land cover change maps of the Mesilla Valley, New Mexico between 1985 and 2009.