OCHA Centre for Humanitarian Data

Raster Statistics

The pipeline behind the team's zonal statistics: SEAS5, ERA5, IMERG, and FloodScan rasters summarised per COD admin unit, on schedule, into the rasterstats database that monitoring apps and anticipatory-action frameworks read from.

Analyses

Zonal-stats speed benchmarks

Measurements behind the weighted zonal-stats rewrite (PR #49): the exactextract coverage-weight method vs the legacy upsample-and-rasterize method on real prod data, scaling with admin count, and a reproduction of the 2024 methodology benchmarks.

/benchmarks/

What the method change does to the values

How the exactextract coverage-fraction rewrite changes the numbers in the database: differences by dataset, admin level and polygon size, what stays identical, and the sub-pixel admin units that gain values for the first time.

/method-change/

Pipeline sourceGitHub

The ds-raster-stats repository: per-dataset configs, the zonal-stats implementations, and the scheduled jobs that keep the database current.

github.com/OCHA-DAP/ds-raster-stats

Access. Everything listed here is public — no sign-in, no passphrase.

Adding a page? This site is served straight from the gh-pages branch: create a directory with an index.html and add a card above. One repo has one Pages site — new products get their own path, never the root.