Classification: false_negative Confidence: Major prediction failure significantly reduces confidence in all band coefficients, particularly Band 1 which appears severely underestimated
The model completely missed a significant 1.06 ft rise caused by light rainfall across all bands, severely underpredicting the response.
| Metric | Predicted | Actual | Error |
|---|---|---|---|
| Peak height | 4.06 ft | 5.11 ft | -1.05 ft |
| Total rise | — | 1.06 ft | — |
| Band | Precip | Predicted Rise | Intensity | Moisture |
|---|---|---|---|---|
| 5 | 0.15" | 0.02 ft | LIGHT | DRY |
This event represents a major prediction failure where the model predicted only a 0.02 ft rise but observed a 1.06 ft rise - an error of over 5000%. The rainfall was light but distributed across all bands (0.18" Band 1, 0.16" other bands), yet the model only processed Band 5 rainfall in its prediction. The sharp hydrograph shape and immediate rise at midnight suggests near-gauge rainfall was the primary driver, but the model failed to capture this. The gauge precipitation sensor recorded only 0.05" compared to QPE Band 1's 0.18", indicating potential QPE overestimation, but even accounting for this discrepancy, the model's response coefficients appear severely underestimated. The DRY antecedent conditions (0.5x multiplier) may have been incorrectly applied, as the actual response suggests much more responsive conditions than the model assumes.
| Band | Change | Reason |
|---|---|---|
| 1 | +20% | Severe underprediction of near-gauge response - increase maximum allowed to begin addressing the 5000% error |
| 2 | +15% | Multi-band event showed significant response from intermediate bands |
| 3 | +15% | Multi-band event showed significant response from intermediate bands |
| 4 | +10% | Modest increase as headwater bands also contributed to the response |
| 5 | +10% | Modest increase as headwater bands also contributed to the response |
Major prediction failure significantly reduces confidence in all band coefficients, particularly Band 1 which appears severely underestimated