Land Cover in the Caucasus

Mountainous regions are changing rapidly across the world due to both land-use change and climate change. Given the importance of mountainous regions for ecosystem services and endemic biodiversity, monitoring these changes is essential. Satellite data provide a great resource to map land-cover change in mountainous regions, however mapping is especially challenging there because topographic complexity affects reflectance. Here, we used Landsat data to perform a long-term land-cover change assessment for a large mountainous region, i.e., the Caucasus Mountains with topographic correction. Additionally, we used historical Corona spy satellite data to map land cover in the Caucasus in years 1964-65.

The land cover maps for 1987-2015 each consist of 10 land cover classes (1: coniferous forest, 2: mixed forest, 3: deciduous forest, 4: barren, 5: rangeland, 6: cropland, 7: built-up, 8: wetlands, 9: water, 10: snow and ice).

Landsat-based land-cover change maps exist for cropland (cropland = 1, non-cropland = 0, transitional cropland = 2) and forest (forest = 1, non-forest = 0, transitional forest = 2) for 6 time steps (1987, 1995, 2000, 2005, 2010, 2015). All Caucasus land-cover and land-cover change maps are stored as GeoTIFF format at 30 meter spatial resolution. (Buchner et al., 2020)

The land cover map for 1965 consists of 11 land cover classes (1: forest, 4: barren, 5: grassland, 6: cropland, 7: urban, 8: wetland, 9: lake, 10: snow-and-ice, 11: river, 13: clouds, 15: cloud shadows). (Rizayeva et al., 2023)

The Corona-to-Landsat change maps depict changes between the following land cover classes: forest (1), barren (4), grassland (5), cropland (6), urban (7), wetland (8), and water (9). For example, a pixel value of 11 indicates that the land remained forested (1) in both the initial and the subsequent year. A pixel value of 16, represents forest loss to cropland, as the land was classified as forest (1) in the first year and cropland (6) in the second year. Similar interpretation apply to other pixel values. (Rizayeva et al., 2025)