The following parts of this repository were developed with the help of an AI assistant (Claude, by Anthropic), working at the request of a project contributor who directed the work, reviewed it and ran the results:
harmonization/ : the MODIS/VIIRS harmonization pipeline (input cleaning, per-state threshold calibration, national build, CSV → SQLite export), its analysis, S3 helper, documentation notebook and READMErisk_model_weekly/ : the weekly hotspot-risk model (code, tests, configuration, README and the generated outputs)backend/ (parts):
app/services/nasa_fires.py: harmonized-archive reads (SQLite), the live in_calendar rule, national GIBS tiles with state tagging, history cachingapp/services/calendar_builder.py, app/services/open_meteo.py, app/services/mexico_states.py (+ app/data/mexico_states.json)app/routers/nasa.py, app/routers/fires.py, the region/scope fields in app/routers/model.py, and the removal of invented hotspots and seed stations (harmonizer.py, db/init.sql, app/routers/iot.py)tests/test_nasa_fires.pyweb/ (parts): lib/api.ts (live data only, no invented fallbacks), lib/fire-data.ts and lib/hero-data.ts (real harmonization numbers),
the harmonization/calendar text in app/page.tsx and components/ChihuahuaOsmMap.tsx, and the wording fixes in app/(landing)/about, app/(landing)/how-it-works and app/data
that mark the origin classifier as planned.github/workflows/pages.ymlEvery Python file in harmonization/ and risk_model_weekly/ starts with an “AI-assisted code” header, and the notebook has a note at the top.
AI-written code can contain mistakes: the numbers in the documentation come from running the code on the FIRMS data, but they should be reviewed before reuse.
Other code in the repository (the rest of web/ and backend/, esp32/, ml/, deploy/) was written by the team; its authors should add their own disclosure if relevant.