Authors: Johanna Buchner*, University of Wisconsin-Madison, He Yin, University of Wisconsin-Madison, David Frantz, Humboldt University Berlin, Benjamin Bleyhl, Humboldt University Berlin, Tobias Kuemmerle, Humboldt University Berlin, Tamar Bakuradze, Geographic, GIS & RS Consulting Center , Anna Komarova, Greenpeace Russia, Afag Rizayeva, University of Wisconsin-Madison, Hovik Sayadyan, Yervan State University, Garik Tepanosyan, Armenian National Academy of Sciences, Volker C. Radeloff, University of Wisconsin-Madison
Topics: Land Use and Land Cover Change, Remote Sensing, Mountain Environments
Keywords: land cover and land use change, topographic correction, multi-temporal composites, Landsat, Caucasus
Session Type: Paper
Start / End Time: 1:10 PM / 2:50 PM
Room: Balcony B, Marriott, Mezzanine Level
Presentation File: No File Uploaded
Land-cover and land-use change is a major driver of global environmental change and monitoring these changes is important, because agriculture and forests experience strong modifications from anthropogenic activity. However, monitoring land-cover in mountainous regions with satellite imagery is challenging due to topographic effects. The Caucasus region encompasses high mountains and to date no comprehensive land-cover and land-use change assessment exists. Our goal was to examine the effect of topographic correction on land-cover classification and to assess land-cover and land-use changes in the Caucasus since 1987. First, we examined broad-scale topographic correction of Landsat composites on classification accuracy for one year. Second, we used topographically corrected Landsat composites covering six time periods to assess changes in forest and agriculture between 1987 and 2015 using class probabilities and post-classification comparison. We found that topographic correction of Landsat imagery increased our overall accuracy from 80% to 83%, but had great effects in mountain regions where approximately half of the coniferous forest in the topographically uncorrected map was coniferous forest in the corrected map, and only 65% of deciduous forest in the corrected map was deciduous forest in the uncorrected map. The results for our change assessment showed that agricultural abandonment was most prevalent between 1987 and 2000. We found very few forest losses, but forest gains on abandoned fields after 2000. Our results highlight the importance of topographic correction for land-cover and land-use classification to minimize classification errors among forest types and other classes and to ensure accurate results for large-area monitoring.