Key takeaways
- A six-year study (2020–2025) compares Veriground ground stations with CMORPH satellite rainfall estimates across Côte d'Ivoire, Ghana, and Ecuador.
- Both data sets follow the gamma-type distribution typical of tropical rainfall, so they look similar overall.
- CMORPH reports ~18% more rainfall on average (nearly 47% more in Ecuador) and more frequent rain days: 3.1 vs 2.2 per week.
- Of the 20 heaviest rainfall events per source, only one date/location appears on both lists.
- Across 28 regression runs the correlation stayed weak (R² under 37%) — the two sources cannot be reliably combined.
This study compares daily rainfall from two very different sources across cocoa-growing regions of Côte d'Ivoire, Ghana, and Ecuador, from January 2020 through December 2025. Veriground measures rain directly at ground-level weather stations, while CMORPH — NOAA's Climate Prediction Center Morphing Technique — estimates rainfall from satellite microwave and infrared data on a global 8 km grid. The goal was to see how closely the two agree, and whether they could be combined to give cocoa regions faster, more accurate rainfall information.
Across 32,870 matched daily observations, the two sources look broadly similar on the surface: comparable maximum values and skewness, and both following the gamma-type distribution typical of tropical rainfall. But they diverge in the details. CMORPH reports about 18% more rainfall on average than Veriground — and nearly 47% more in Ecuador — and it registers rain far more often, roughly 3.1 rain days per week versus 2.2 for Veriground. Everyday experience of rainfall in West Africa lines up more closely with the lower Veriground frequency.
Where the two sources part ways most clearly is at the level of individual days and places. Of each source's twenty heaviest rainfall events, only one date-and-location combination appears on both lists. Across 28 regression runs — split by year, country, season, and El Niño/La Niña phase — the relationship stayed weak, with R² never exceeding about 37% and large standard errors throughout.
The conclusion is that there is no reliable way to use one source to correct or improve the other. The gap is not a matter of rescaling or fine-tuning; it reflects two fundamentally different approaches to measuring rainfall — local, ground-level observation versus wide-area satellite modeling. The findings reinforce the value of direct, on-the-ground measurement in regions where rainfall can vary sharply over just a few kilometers, and point to a need for further research into why the differences are so large.