OCHA Centre for Humanitarian Data

Precipitation Intercomparison

Our triggers read rainfall from one product at a time. This work asks how much the choice matters: a global, monthly, land-only comparison of CHIRPS, IMERG, ERA5, the JRC ASAP blend and six gauge analyses — biases, agreement on anomalies and dry terciles, and trends, split by how many rain gauges stand behind each number.

Analysis

How much do observed rainfall products disagree?

Relative bias against gauges and against each other, monthly anomaly agreement, triple collocation, dry-tercile detection (the anticipatory-action view), trends 1983–2020 and 2001–2025, drift and known breaks, and a per-country explorer.

/report/October 2026

Correlation explorer: any two products, any period

Pick two products and a period — calendar year, rainy season, or any 3-month window. A zoomable map shows their year-to-year correlation at every 0.5° pixel; click a pixel, or pick an admin-1 unit or country, for a year-by-year scatter.

/explorer/Interactive

Code

Source repositoryGitHub

Ingest (Databricks), 0.5° regridding, metrics and this site. The regridded cubes live on the team's dev blob; the README lists every source.

OCHA-DAP/ds-precip-intercomparison

Access. Everything listed here is public. All inputs are openly published products; CRU TS (ODbL) appears only as derived statistics.