Classification: correct Confidence: Model confidence remains high. The sequence of events (false positive -> correct -> correct) indicates the calibration is converging on the true physical response of the Boxley basin. The coefficient is no longer overpredicting small events.
The physics model accurately predicted the small peak flow rise (28.7 CFS predicted vs 30.1 CFS actual, -4.7% error) resulting from a night-time rainfall event in wet antecedent conditions, though timing was off by 2.4 hours.
| Metric | Predicted | Actual | Error |
|---|---|---|---|
| Peak CFS | 29 CFS | 30 CFS | -1 CFS |
| Total rise | — | 24.6 CFS | — |
| Band | Zone | Precip | Predicted Rise | Intensity |
|---|---|---|---|---|
The physics model performed well in magnitude prediction for this event. With a WET antecedent moisture multiplier (1.5) applied to 0.732 inches of rainfall, the model predicted a peak of 28.7 CFS. The actual peak was 30.1 CFS, resulting in a very small error of -4.7%, which falls well within the 'correct' classification threshold of ±15-20%. This confirms that the significant -20% coefficient adjustment made after the August 6th false positive was appropriate; the previous coefficient was far too high for this basin's response to moderate rainfall. The model now correctly scales the response to the observed precipitation.
The timing error of 2.4 hours (predicted arrival ~04:51 UTC Aug 8 vs actual peak ~02:30 UTC Aug 8, noting the timestamp format difference) is consistent with previous events where the model's fixed lag/dispersion parameters struggled with the exact hydrograph shape of quick-response storms. However, given the small magnitude of the event and the excellent magnitude accuracy, no further aggressive tuning is warranted at this stage. The model is stabilizing.
No empirical forecast headline was issued today, so no assessment of empirical accuracy can be made. The absence of an headline for a sub-500 CFS event is consistent with typical whitewater forecasting thresholds where only significant rises are flagged.
No changes made.