A new vegetation type dataset in their distribution and changes across the circumpolar Arctic permafrost regions

Vegetation type serves as a crucial predictor of the Arctic land surface energy budget and its sensitivity to global warming. High-precision vegetation classification data is essential for understanding the vegetation-soil interactions, permafrost and ecosystem carbon cycle. Here, based on vegetation community structure and species composition, we offer a new dataset of vegetation type distribution in the circumpolar Arctic permafrost region from 1982 to 2020 using Google Earth Engine and machine learning. The results present an average modeling accuracy of 83%, which is greatly improved compared with the Arctic Vegetation Type Map. Our findings show that bare ground, graminoid, wetland and forest exhibit significant changes in the past four decades. The forest has the highest growth rate of 9.71 × 104 km2 per year, whereas bare land and graminoid tundra decreases at the rates of -5.21 × 104 km2 per year and -5.13 × 104 km2 per year, respectively. The improved mapping of vegetation types facilitates the quantification of the potential changes in land cover, serving as a key environmental indicator for assessing land carbon, water cycles and energy budgets.