This is an updated and extended record of the Global Fire Atlas introduced by Andela et al. (2019). Input data (burned area and land cover products) are updated to the MODIS Collection 6.1 (the original version featured in Andela et al. (2019) was based on collection 6.0 burned area and collection 5.1 land cover products, respectively). The timeseries is extended to cover the period 2002 to February 2026. Methodological Notes: The method employed to create the dataset precisely follows the approach described by Andela et al. (2019). The input burned area product is MCD64A1 Collection 6.1. It is described by Giglio et al. (2018) and available at: https://lpdaac.usgs.gov/products/mcd64a1v061/. The input land cover product is MCD12Q1 Collection 6.1. It is described by Sulla-Menashe et al. (2019) and available at: https://lpdaac.usgs.gov/products/mcd12q1v061/. Note that while the methods have remained the same compared to Andela et al. (2019), we do observe small differences between the Global Fire Atlas products originating from differences between the MCD64A1 collection 6.1 burned area data used here and the collection 6 data used in the original product. In addition, we observe more substantial differences in the dominant land cover class associated with each fire due to the differences between the MCD12Q1 collection 6.1 data used here and collection 5.1 data used in the original product. Please note that the year string in filenames refers to the locally-defined fire season in which the fire ignited, not the calendar year. For each MODIS tile, the fire season is defined as the twelve months centred on the month with peak burned area (see Andela et al., 2019). For example, for a MODIS tile with peak burned area in December, the 2023 fire season would be defined as the period from July 2023 to June 2024. This is particularly relevant in the Southern extratropics and the northern hemisphere subtropics, where the fire seasons often span the new year. The local definition of the fire season is based on climatological peak in burned area as described by Andela et al. (2019). Here we extended the time-series to include the fire season of 2002, and extended the time-series until February 2026. Usage Notes: Recommendation: Do Not Mix Dataset Versions We recommend downloading and using each version of the Global Fire Atlas as a complete dataset. Users should not extend an earlier version by adding only the final years from a newer release, as updates between versions may affect the full-time series. This is because each release includes a recalibration of model parameters, including fire persistence and fire season, which are derived on a per-tile basis using the complete available time series. Consequently, changes to the length of the input time series can alter these parameters for individual pixels, leading to minor differences throughout the dataset rather than only in the newly added years. Incomplete Observations for the Latest Fire Seasons: Please note that the year string in filenames refers to the locally-defined fire season in which the fire ignited, not the calendar year. As such, the time-series can be incomplete for the latest fire season (e.g. the "2026 fire season") and also for the penultimate fire season (e.g. the "2025 fire season") due to the way that fire seasons are defined (see above). For example, if the month with peak burned area for a tile is December, then full data covering the 2025 fire season in that tile are not available until midway through the 2026 calendar year. This contrasts with the original dataset from Andela et al. (2019), which only included the data for entire fire seasons between 2003 and 2016. Observational Outages and Time-series Consistency: For the purpose of time-series analysis, we note that the 2002 product may have been affected by outages of Terra-MODIS (most notably, June 15 2001 - July 3 2001 and March 19 2002 - March 28 2002), which affects the burn date estimates and Global Fire Atlas product. Following the launch of Aqua-MODIS in May 2002 burn date estimates are more reliable as estimated from both MODIS sensors onboard Terra and Aqua. We also note that, as Terra and Aqua approach the end of their operational lifetimes, the effects of orbital changes (e.g. changes in overpass time, sampling, and pixel size) may have affected the final years of the time series, although the extent of these impacts on the MODIS product suite remains under investigation. File Naming Convention: GFA_v{time-stamp}_{data-type}_{fire_season}.{file_type} {time-stamp} = Date that code was run. {data-type} = “ignitions” or “perimeters” for vector files; “day_of_burn”, “direction”, “fire_line”, or “speed” for raster files. {fire_season} = the locally-defined fire season in which the fire was ignited (see more below). {file_type} = ".shp" for vector files; ".tif" for raster files. Please note that the year string in filenames refers to the locally-defined fire season in which the fire ignited, not the calendar year. Hence the file GFA_v20260408_perimeters_2003.shp can include fires from the 2003 fire season that ignited in the calendar years 2002 or 2004. Coordinate systems (Map Projections): Vector data are provided on the WGS84 projection. Raster data are provided on the MODIS sinusoidal projection used in NASA tiled products. The WKT string defining this projection is: 'PROJCS["unnamed",GEOGCS["Unknown datum based upon the custom spheroid",DATUM["Not_specified_based_on_custom_spheroid",SPHEROID["Custom spheroid",6371007.181,0]],PRIMEM["Greenwich",0],UNIT["degree",0.0174532925199433,AUTHORITY["EPSG","9122"]]],PROJECTION["Sinusoidal"],PARAMETER["longitude_of_center",0],PARAMETER["false_easting",0],PARAMETER["false_northing",0],UNIT["metre",1,AUTHORITY["EPSG","9001"]],AXIS["Easting",EAST],AXIS["Northing",NORTH]]' Data Layers: Table 1: Overview of the Global Fire Atlas data layers. The shapefiles of ignition locations (point) and fire perimeters (polygon) contain attribute tables with summary information for each individual fire, while the underlying 500 m gridded layers reflect the day-to-day behavior of the individual fires. In addition, we provide aggregated monthly summary layers at a 0.25° resolution for regional and global analyses. File name Content SHP_ignitions.zip Shapefiles of ignition locations with attribute tables (see Table 2) SHP_perimeters.zip Shapefiles of final fire perimeters with attribute tables (see Table 2) GeoTIFF_direction.zip 500 m resolution daily gridded data on direction of spread (8 classes) GeoTIFF_day_of_burn.zip 500 m resolution daily gridded data on day of burn (day of year; 1-366) GeoTIFF_speed.zip 500 m resolution daily gridded data on speed (km/day) GeoTIFF_fire_line.zip 500 m resolution daily gridded data on the fire line (day of year; 1-366) GeoTIFF_monthly_summaries.zip Aggregated 0.25° resolution monthly summary layers. These files include the sum of ignitions, average size (km2), average duration (days), average daily fire line (km), average daily fire expansion (km2/day), average speed (km/day), and dominant direction of spread (8 classes). Table 2: Overview of the Global Fire Atlas shapefile attribute tables. The shapefiles of ignition locations (point) and fire perimeters (polygon) contain attribute tables with summary information for each individual fire. Attribute Explanation / units lat, lon Coordinates of ignition location (°) size Fire size (km2) perimeter Fire perimeter (km) start_date, start_DOY Start date (yyyy-mm-dd), start day of year (1-366) end_date, end_DOY End date (yyyy-mm-dd), end day of year (1-366) duration Duration (days) fire_line Average length of daily fire line (km) spread Average daily fire growth (km2/day) speed Average speed (km/day) direction, direc_frac Dominant direction of spread (N, NE, E, SE, S, SW, W, NW) and associated fraction MODIS_tile MODIS tile id landcover, landc_frac MCD12Q1 dominant land cover class and fraction (UMD classification), provided for 2002-2024 GFED_regio GFED region (van der Werf et al., 2017; available at https://www.globalfiredata.org/)