Every forecast the University of Reading has issued since August 2022 is re-scored with the method our dashboard uses today, then compared with the soil moisture that actually followed. We publish the result whether it flatters us or not.
Loading the latest scores…
The forecast: the chance that soil moisture over the next 91 days is in the driest third of past seasons, averaged over each province's land, from TAMSAT-ALERT (University of Reading, CC BY 4.0).
What happened: the share of each province that was in its driest third, from TAMSAT soil moisture driven by observed satellite rainfall. Field sensors and farmer reports are being added as a second check.
The score: the Brier skill score against simply saying "one in three" every time. A verdict of proven skill needs the whole 95% range above zero, found by re-sampling entire seasons, so a few lucky forecasts cannot pass. Each season is left out of its own idea of "normal".
Limits: four seasons is an early record and the ranges are wide. The archived forecasts are scored without the seasonal rainfall outlook the dashboard now adds, because those outlooks were not archived.
More on sources and limitations: data methodology.
Every scored forecast is available through the institutional API, case by case, so your own risk team can check it.