Fire is a major natural disturbance shaping global ecosystems, affecting the carbon cycle, biodiversity, and land-cover dynamics. Existing global fire datasets rarely link fire occurrences to subsequent land-cover changes or vegetation recovery. Here, we develop a global fire dataset for 2001-2024 that couples fire occurrence with land-cover trajectories and post-fire NDVI regreening, while quantifying key fire attributes (timing, duration, burned area, fireline length, spread speed, and shape index). Globally, fires averaged similar to 7.85 million occurrences and similar to 3.90 million km(2) burned area annually, showing a declining trend in both frequency and burned area. Spread speeds decreased, while fire duration slightly increased, and fireline length and shape index remained relatively stable. Clustering analyses identified three fire typologies-medium-scale complex, extremely rapid, and small stable-differing in burned area (152-3754 km(2)), spread speed (2.1-52.8 km(2)/day), and geometric complexity. Recurrent fires were concentrated in tropical and arid regions, accelerating land-cover transitions, particularly forest-to-grassland conversion. NDVI recovery was slower in repeatedly burned areas, especially in dry tropical biomes and deserts, where mean recovery exceeded 150 days and two-year recovery rates remained below 60%. This global dataset and analytical framework enable improved fire impact assessment and support adaptive ecosystem management.