Credits

Credits and licences

Software

• MapLibre GL JS - BSD-3-Clause.
• Inter and Space Grotesk typefaces - SIL Open Font License 1.1, self-hosted via Fontsource (MIT).
• Team portrait - provided by and used with permission of the pictured team member.
• PMTiles JavaScript library (pmtiles 4.5.0, Protomaps) - BSD-3-Clause; reads the layer tiles from static files.
• Vite - MIT.
• TypeScript - Apache-2.0.

Map tiles

• Basemap by OpenFreeMap (free, no API key; styles Positron and Dark), vector tiles from OpenMapTiles (schema BSD / design CC-BY 4.0), map data © OpenStreetMap contributors (ODbL). Attribution is shown on the map.

Fonts and images

• Fonts: Inter and Space Grotesk (SIL OFL 1.1), self-hosted, no font CDN. No external images. No NASA logo.

Data shown on the map

• Fire detections (real): NASA FIRMS archive, MODIS Collection 6.1 active fires (MOD14 from Terra and MYD14 from Aqua, 1 km; NASA MODIS Adaptive Processing System / LANCE) and VIIRS S-NPP 375 m active fires (VNP14IMG; NASA VIIRS Land Science team / LANCE), for Indonesia, Malaysia, Brunei, Singapore, Thailand, Myanmar, Laos, Cambodia, Vietnam, the Philippines and Timor-Leste, 2001–2025. NASA open data (CC0). "We acknowledge the use of data and/or imagery from NASA's FIRMS (https://earthdata.nasa.gov/firms), part of NASA's Earth Science Data and Information System (ESDIS)." We drop type 1–3 detections, MODIS confidence below 30 and VIIRS low confidence. The harmonized monthly counts are our own ratio baseline computed from these detections.
• Region shapes (first-level administrative units; 287 regions in 11 countries): geoBoundaries gbOpen ADM1 (Runfola et al. 2020, PLoS ONE 15(4):e0231866). The licence differs by country, as listed by geoBoundaries: Open Database License 1.0 (ODbL) for Indonesia (2017) and Malaysia (2017), source OpenStreetMap (Wambacher), © OpenStreetMap contributors; Singapore (2016, five regions), ODbL 1.0 as listed by geoBoundaries, source Urban Redevelopment Authority Master Plan 2014 subzone boundary via data.gov.sg; and Thailand, Cambodia, Laos and Timor-Leste, ODbL 1.0 as listed by geoBoundaries; Myanmar, CC BY 4.0; the Philippines, CC BY 3.0 IGO; Vietnam and Brunei (2011, four districts, source Wikimedia Commons, user Tachymetre), Public Domain. Our copy is simplified (the Philippine file to 5 m); the ODbL parts are offered under the same ODbL terms.
• Forest loss, 2001–2025 (real): Hansen, M.C. et al. 2013, High-resolution global maps of 21st-century forest cover change, Science 342:850, GFC v1.13 (University of Maryland, Google, USGS, NASA); CC BY 4.0. We recoloured and tiled it as share of area.
• Oil palm by planting year, 1989–2021 (real): Descals, A. et al. 2024, ESSD 16:5111, Zenodo 10.5281/zenodo.13379129; CC BY 4.0. Recoloured and tiled as share of area. The dataset has no tiles over about 20 % of Sarawak (interior highlands) and over parts of the new countries; the app says so instead of showing zero. For Thailand, Myanmar, Cambodia and the Philippines it is an oil-palm-only map; for Laos, Vietnam, Timor-Leste and Singapore the app shows no oil-palm data.
• Burned area (real): MCD64A1 v6.1 (Terra and Aqua MODIS), Giglio, L., Justice, C., Boschetti, L., Roy, D. (2021). MODIS/Terra+Aqua Burned Area Monthly L3 Global 500 m SIN Grid V061. NASA EOSDIS Land Processes DAAC. doi:10.5067/MODIS/MCD64A1.061 (no use restrictions; we recoloured and tiled it).
• Forest history and degradation, tree-plantation conversion (real): Tropical Moist Forests (TMF) v1.2025, European Commission Joint Research Centre (JRC), forobs.jrc.ec.europa.eu/TMF; Vancutsem et al. 2021, Science Advances 7, eabe1603, doi:10.1126/sciadv.abe1603. Free use with attribution: Source: EC JRC. Data aggregated and re-tiled by Abang Abang; this does not imply endorsement by the European Commission.
• Peat (real): Gumbricht, T. et al. (2017), Tropical and subtropical wetlands distribution v2.1, CIFOR, doi:10.17528/CIFOR/DATA.00058; Gumbricht et al. 2017, Global Change Biology 23, 3581–3599. CC BY 4.0. Clipped, rasterised to tiles and recoloured.
• What cleared forest became (real; layer “What cleared forest became”): Sims et al. (2025), Global drivers of forest loss at 1 km resolution, World Resources Institute / Google DeepMind, v1.3 (2001–2025), Environmental Research Letters 20(7) 074027, doi:10.1088/1748-9326/add606; data doi:10.5281/zenodo.19485190; CC BY 4.0. We recoloured and tiled it, one raster per class, and show a cell only where Hansen forest loss is recorded (at least 0.5 % of the cell). The map is built partly from fire data, so its “wildfire” class is not independent evidence about fire.
• Mines (real; layer “Mines (about 2019)”): Maus et al. (2024), Global mining deforestation footprint data from 2000 to 2019, Zenodo doi:10.5281/zenodo.7307210, ODbL 1.0. The polygons are the Maus et al. 2022 mining polygons (imagery of about 2019) merged with OpenStreetMap mining areas. Contains information from OpenStreetMap, © OpenStreetMap contributors. We kept 4,023 polygons in the 11 countries, simplified to about 2 m, with the commodity list from the same record.
• Forest loss to oil palm overlap layer: derived by us from Hansen and Descals (above); not a source dataset.
• Placeholder Forest alerts (synthetic blobs for two months) and pulpwood (synthetic; the planned source is the Spatial Database of Planted Trees v2, WRI / Global Forest Watch, Richter et al. 2024, CC BY 4.0). The layer panel labels both "Placeholder tiles". The gridded harmonized-fire layer before 2012 is an empty slot for the model's output and shows nothing real.

• Open data offered under ODbL 1.0: the simplified region file regions.geojson is a derivative of OpenStreetMap-based boundaries and stays publicly available under the Open Database License 1.0, © OpenStreetMap contributors.
• Mine outlines offered under ODbL 1.0: the simplified file mines.geojson is a derivative of the Maus et al. 2024 database (which includes OpenStreetMap data) and stays publicly available under the Open Database License 1.0, © OpenStreetMap contributors.

Models and AI

• No third-party pretrained AI model is used. The harmonized counts on the map come from a ratio baseline, not a neural model. Our own research model (a shared encoder for MODIS and VIIRS) is in progress and is not shown in the app.
• Built with AI assistance: the code, data processing, analysis and these pages were developed with Claude Code agents (Anthropic), directed and checked by the team.

No analytics, trackers or cookies. The only thing stored in your browser is your light/dark choice.