Land Cover Change Analysis and Spatial Variations in Southeast Asian Nations: Insights on Spatial Scale Dynamics

Land cover type is a fundamental aspect of studies using remote sensing for environmental analysis such as monitoring deforestation, quantifying wildland fire emissions, and more. There are many land cover products available for varying time periods and spatial resolutions, each with different land cover class definitions and number of land cover classes. These differences yield inherent variation in land cover estimates. In this study, we explore the spatial and areal variation between two major
datasets for their commonly available time of 2018 to 2022 (MODIS 500m Land Cover and Esri Sentinel-2 10m land cover) in the continental Southeast Asia countries of Cambodia, Laos, Myanmar, Thailand, and Vietnam. To enable comparison, we resampled the datasets and reclassified the land cover classes to a common scheme. Major disagreement was detected between Esri and MODIS, especially for the wetlands/flooded vegetation class in which only 3.6% of pixels were in agreement. In addition to quantifying variation between the datasets, we quantify land cover change at the decadal scale of 2001, 2011, and 2021 for each country and identified province-level hotspots of land cover change. Cambodia experienced the highest rate of land cover change with 67.8% of total land area changed, followed by Vietnam (58.6%), Laos (50.2%), Myanmar (46.9%), and Thailand (46.3%). The highest rate of land cover change was detected in Odtar Mean Chey, Cambodia, which had large swaths of forest area cleared for agricultural production, river damming, and more. Ultimately, many areas in the region experienced forest clearing for crop production (i.e. rice) which appeared to be classified as wetland in the MODIS product, and crop in the Esri product. Myanmar was the only country which did not experience net decline in forest area over the 2001 to 2021 time period.