Applying a support vector model to assess land cover changes in the Uvs Lake Basin ecoregion in Mongolia

The Uvs Lake Basin in western Mongolia is a natural world heritage site and is known for its diversity in landscape and wildlife. Recently, investigative research has shown that the protected pristine ecotone is suffering land degradation due to global warming. In order to obtain evidence of the changes over a long-term time scale, serial multi-temporal Landsat images obtained between 1995 and 2015 were used to classify land cover and land cover changes over the Basin ecoregion using a machine learning classification technique, support vector machine. Results showed that the forest land area in 1995 was 1888.48 km2 which was equivalent to 7.48% of the total area of the study site. The forest area showed considerable decrease by 301.36 km2 during the first decade (1995–2004) and 155.81 km2 during second decade (2004–2015). A total of 457.17 km2 or 24.21% of the forest land has been developed, most being changed into grassland. The major driver of such changes was illegal logging, forest fire, and pest damage. However grassland was changed primarily into bare land during the two decades. The area of glacier was decreased and primarily changed into water body. In contrast, the area of sand in the Basin ecoregion increased dramatically from 65.20 km2 in 1995 to 318.33 km2 in 2015 the increase being mostly from the transition of bare land. In summary, the drivers of the significant decrease of greenness coverage and increase of sand/bare land areas were the interaction of complicated disturbances in both anthropogenic and natural factors, in which logging, grazing, wind erosion, and global warming were the key causes. Keywords: Land suppression, Soil degradation, Forest reduction, Change analysis, Landscape ecolog