Skip to content

Wildfire risk score and open source datasets

Latest
Compare
Choose a tag to compare
@frgfm frgfm released this 24 Dec 16:05
· 41 commits to master since this release
8a66bd1

This first release provides models and datasets for wildfire risk estimation. Among others, this introduces a D-5 wildfire risk inference for French regional areas.

Note: pyro_risks mainly requires pandas, geopandas and scikit-learn , while the API only requires fastapi and uvicorn

Highlights

Datasets

Various datasets with entries indexed by location and time of measurement.
New

  • Added NOAA & BDIFF (#6), NASAFIRMS (#7), FWI (#10), ERA5Land (#13), NASAVIIRS (#18), MergedEraFwiViirs (#22), ERA5T (#24, #29) datasets
  • Added dataset merging utilities (#9, #13), and closest neighbour aggregation (#13)
  • Added data source URL download for modis and ghcn (#4)
  • Added live API data fetching (#25)

Models

Trainable models to estimate wildfire risk
New

  • Added Random Forest and XGBoost risk estimation models (#22, #25, #29)

Documentation

The documentation of the python library
New

  • Added Sphinx documentation autobuild (#1)
  • Added datasets page (#6)
  • Updated README (#1, #16, #27) and added CONTRIBUTING (#16)

Tests

Unittests for the python package
New

Web Server

Web server made to expose some of the python library features
New

  • Added basic FastAPI web server to expose wildfire risk inference (#27)

Others

New

  • Added package setup (#1, #6) and CI job (#1, #17)
  • Added CI jobs for lint checking, doc building and unittesting (#1, #5, #14)
  • Fixed geopandas installation in CI (#6)
  • Renamed repo and package (#12)
  • Added example scripts (#13, #18, #22, #24)
  • Added org funding option
  • Added docker orchestration for web server (#27)
  • Added Heroku deployment of web server (#27)