Drought does not respect a region. The rain that fails over the Corn Belt, the Pampas and the Black Sea steppe in the same season is one circulation's story, and a reader comparing a dry spell in Kansas with one in Ukraine needs the two on the same map and the same colour scale. So every other site in this network draws a country or a region, and this one draws the world: one frame, 180°W to 180°E and 60°S to 75°N.

Two models, both from ECMWF, both open. HRES is the IFS high-resolution deterministic run — the physical model at the top of the world's verification tables — carried from its 00Z and 12Z cycles, the two that reach ten days on open data. AIFS is ECMWF's machine-learned global model, trained on four decades of reanalysis and started from the same analysis, published every six hours. Both are fetched at 0.25° under CC BY 4.0, by HTTP byte range against ECMWF's own index files, so a whole HRES cycle of rainfall costs about 70 MB rather than the 4.5 GB the files hold.

A scale built for the dry end. The run-total map spends six of its colours inside 25 mm and stops at 250, where a wet-climate ramp would spend them at 100 and 200. Below half a millimetre it draws nothing, so dry ground reads as ground. The frames are palettised PNGs whose colour table is published beside them, so hovering turns any pixel back into millimetres — to about 2 mm on the total and 0.4 mm on the period map.

What it is not, yet. This is a precipitation forecast, not a drought index. Soil moisture, evaporative demand, totals against climatology and the standardised indices drought monitors publish are the next layer, and they will be built on exactly these fields. For an official drought declaration, your national drought monitor or meteorological service is always the authority.