Permafrost is one of the most sensitive components of the geosphere under global change, and its thermal dynamics directly influence surface energy and water balance, ecosystem processes, and carbon cycle feedbacks. Accurately characterizing the spatial distribution of permafrost, mean annual ground temperature (MAGT), and active layer thickness (ALT), as well as their long-term evolution, is essential for understanding the response of high-latitude regions to climate warming and for predicting future environmental risks. This dataset is derived from permafrost monitoring records provided by the Global Terrestrial Network for Permafrost (GTN-P) and the Circumpolar Active Layer Monitoring (CALM) program, using MAGT and ALT as core indicators. Multiple environmental factors, including topography, climate, soil, and vegetation, were integrated, and several machine learning algorithms were tested to construct predictive models, with the best-performing model selected through cross-validation for spatial mapping. The dataset covers the pan-Arctic region north of 45°N during 1980–2019, reconstructed at 5-year intervals (1980–1984, 1985–1989, 1990–1994, 1995–1999, 2000–2004, 2005–2009, 2010–2014, 2015–2019) with a spatial resolution of 1 km.
This dataset is based on permafrost and active layer thickness monitoring records provided by the GTN-P (Global Terrestrial Network for Permafrost) and CALM (Circumpolar Active Layer Monitoring) programs, supplemented with multiple environmental predictors for model development. Topographic factors were derived from the Shuttle Radar Topography Mission (SRTM) DEM provided by the United States Geological Survey (USGS) at a spatial resolution of 1 km × 1 km. Using the SAGA GIS software, 11 terrain-related indices were extracted, including elevation (H), slope (Slope), plan curvature (PlanC), profile curvature (ProC), topographic wetness index (TWI), total catchment area (TCA), relative slope position (RSP), slope length factor (LS), convergence index (CI), valley depth (VD), and closed depressions (CD). Climatic predictors were obtained from the WorldClim v2.1 global climate database, while vegetation information was derived from the NASA MOD13A3 product, providing the normalized difference vegetation index (NDVI). Soil properties, including sand, silt, and clay content at multiple depth intervals, were taken from the SoilGrids global soil database.
This dataset is based on permafrost and active layer thickness monitoring records from GTN-P (Global Terrestrial Network for Permafrost) and CALM (Circumpolar Active Layer Monitoring), with corresponding environmental factors extracted and matched. The predictors include topographic variables (elevation, slope, aspect), climatic factors (air temperature, precipitation, freezing and thawing indices), soil properties (sand, silt, and clay content), and vegetation information (NDVI). During model development and evaluation, several common machine learning classification algorithms were tested, and 10-fold cross-validation was applied for performance comparison. The results showed that the random forest model achieved the best accuracy and stability and was therefore selected as the final approach. Based on the feature importance ranking and feature combination testing, the optimal set of predictors was identified, and the final model was trained using all monitoring sites. The resulting data products cover the period 1980–2019, reconstructed at 5-year intervals (1980–1984, 1985–1989, 1990–1994, 1995–1999, 2000–2004, 2005–2009, 2010–2014, 2015–2019). Each time slice is provided as a GeoTIFF raster file at a spatial resolution of 1 km.
The model evaluation results show that the predicted MAGT achieved an RMSE of 1.23 ℃ with an R² of 0.85, while the ALT predictions yielded an RMSE of 55.6 cm with an R² of 0.84, indicating that the dataset reliably captures the spatiotemporal variability and evolutionary trends of permafrost thermal conditions.
| # | number | name | type |
| 1 | 2020YFA0608501 | Research on Arctic Terrestrial Environmental Change and Its Effects | National key R & D plan |
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.ARCTIC-CHANGE.DB7133.2026
10.12072/ncdc.arctic-change.db7133.2026
