Rainfall

How this works

NASA's GPM IMERG Early Run combines microwave and infrared satellite observations into a rainfall estimate every 30 minutes on a grid of 0.1° (about 11 km). We average the grid cells over each DS division and add up the last 24 hours, 3 days and 7 days.

Each total is then compared with the same product's daily record for that division since June 2000, over 9,600 days. "Top 1%" means that only 1 in 100 periods of the same length since 2000 was wetter. "Record" means none was.

Does it pick out real disasters?

We checked the record against documented floods and landslides. For each, the share of affected DS divisions whose 3-day rain was in their own top 1%:

Koslanda is the reminder that slopes can fail after a long wet spell without one extreme burst, and that one 11 km cell can miss local downpours.

Limits

• This is not a warning. Landslide warnings come from NBRO and are on Landslides. Rain never lowers or replaces them.
• The satellite reads low over the hills: at the 20 September 2026 peak it saw roughly half of what NBRO's warnings imply. That is why we show how unusual the rain is for each division, not millimetres.
• An 11 km cell smooths out local downpours, which are what set off many slips. "Not unusual" never means safe.
• The Early Run is not corrected with rain gauges. The record uses the same product, so its bias is largely shared by both sides of the comparison.