Classification: false_positive Confidence: Model confidence remains high due to sufficient event history, but accuracy is inconsistent. The model tends to overpredict on smaller events and underpredict on massive saturated events. Further tuning of the response coefficient is needed to reduce variance.
The physics model predicted a peak of 128 CFS from 0.6 inches of rain, but the actual observed peak was only 91 CFS, indicating a significant overestimation of runoff response in wet conditions.
| Metric | Predicted | Actual | Error |
|---|---|---|---|
| Peak CFS | 128 CFS | 91 CFS | +37 CFS |
| Total rise | — | 16.5 CFS | — |
| Band | Zone | Precip | Predicted Rise | Intensity |
|---|---|---|---|---|
| 1 | boxley | 0.60" | 250 CFS | MODERATE |
The physics model generated a prediction for 0.603 inches of rainfall, estimating a peak flow of 128.0 CFS. However, the actual gauge peak was 91.0 CFS, resulting in a 40.7% positive error. This constitutes a false positive because the model predicted a meaningful rise that was substantially larger than what occurred. The timing error is also significant (22.9 hours), likely because the bulk of the rainfall occurred late in the day (UTC 02:05-04:05), causing the peak to be delayed relative to the model's standard lag assumption, or the peak simply didn't materialize as predicted due to the small total volume.
The overprediction suggests the current response coefficient (276 CFS/inch) combined with the WET multiplier (1.5) is too aggressive for this magnitude of rainfall. While the basin is wet (2.374" 7-day total), it may not be fully saturated enough to justify the full wet multiplier impact on such a small incremental rain event, or the base coefficient is still too high after recent increases. The model correctly identified that rain would cause a rise, but the magnitude was off by a wide margin.
No empirical forecast headline was issued for the Hailstone gauge today. Therefore, no assessment of empirical accuracy can be made for this specific gauge. Given the consistent trend of overprediction or high error on smaller events despite recent coefficient increases, the model appears to be oversensitive to rainfall input in this basin configuration.
| Band | Change | Reason |
|---|---|---|
| 1 | -15% | The model overpredicted the peak by 40.7% (128 CFS vs 91 CFS). To correct this bias, a -15% adjustment to the response coefficient is recommended to bring predictions closer to observed values for small-to-moderate events. |