Flood Predictor publishes every prediction it makes and logs every warning it issues. We do this for two reasons.
First, to learn. Every prediction is a testable hypothesis. When we get it wrong we want to know, so we can get better. Our model improves every time a flood happens that we predicted — or failed to predict.
Second, to hold power to account. When flood risk exceeds our warning threshold, we notify the relevant TD and local councillor by email, automatically. This is logged here publicly. If a community floods and the warning was sent, that is on the record. If the warning was not acted on, that is also on the record.
We believe that access to predictive information should not be limited to those who can afford it, and that public representatives should be held to the same standard of awareness as any citizen with a smartphone.
Every flood risk prediction made by the platform, by day. Showing predictions where flood probability exceeded 25% (Watch level or above).
| Date / Time | Station | River | County | Risk Level | Prob % | Ground Conditions | Coastal | Outcome |
|---|---|---|---|---|---|---|---|---|
| No predictions at Watch level or above in this period. | ||||||||
Every email sent to a public representative when flood risk exceeded our warning threshold.
| Date / Time | Station / Area | Representative | Risk Level | Probability | Outcome |
|---|---|---|---|---|---|
| Official notifications will appear here as warnings are issued. System operational and monitoring. | |||||
Accuracy improves over time as outcomes are confirmed. Pending predictions are those where the risk window has not yet closed. Accuracy rate = confirmed floods ÷ (confirmed + false alarms) × 100.