A 1 km resolution dataset of Northern Hemisphere permafrost active layer thickness (2000–2024)

This dataset provides continuous annual maps of Active Layer Thickness (ALT) at 1 km spatial resolution across the Northern Hemisphere permafrost region from 2000 to 2024. The dataset was generated using an ensemble spatiotemporal machine-learning framework trained with 2,196 annual ALT observations. A spatiotemporal feature-matching strategy was employed to integrate multi-source dynamic climate drivers, land surface properties, and ecological information, enabling the reconstruction of spatially detailed and temporally continuous ALT estimates. The dataset includes annual ALT maps for 2000–2024 and accompanying pixel-wise uncertainty maps derived from repeated site-level resampling and model reconstruction. These uncertainty layers provide a transparent benchmark of prediction confidence and help users evaluate spatial variations in dataset reliability. This dataset provides a valuable resource for regional- to hemispheric-scale studies of permafrost change, carbon-cycle feedbacks, ecosystem responses, hydrological processes, and cold-region infrastructure assessments.