Classification: false_positive Confidence: Model confidence for Bands 2, 3, and 4 should remain 'medium' or potentially degrade if QPE continues to outpace actual runoff. The discrepancy highlights a need to verify QPE accuracy for mid-watershed bands.
The model predicted a significant peak of 4.68 ft based on high QPE inputs in Bands 2-4, but the actual gauge rise was negligible (0.06 ft) despite wet antecedent conditions, indicating a major overestimation of runoff efficiency.
| Metric | Predicted | Actual | Error |
|---|---|---|---|
| Peak height | 4.68 ft | 2.74 ft | +1.94 ft |
| Total rise | — | 0.06 ft | — |
| Band | Precip | Predicted Rise | Intensity | Moisture |
|---|---|---|---|---|
| 2 | 1.06" | 2.61 ft | HEAVY | WET |
| 3 | 1.54" | 1.96 ft | HEAVY | WET |
| 4 | 1.53" | 1.32 ft | INTENSE | WET |
| 5 | 0.78" | 0.40 ft | HEAVY | WET |
Headline: WATCH — Recent rainfall is within the typical range of historical rises to MEDIUM Settled outcome: None (reached None) LLM said headline was correct: None Notes: The empirical model predicted a 'MEDIUM' tier rise (Watch), but the actual outcome remained at 'below_low' (2.74 ft < 3.0 ft threshold). The empirical forecast was also incorrect, suggesting widespread overestimation of runoff potential by both physics and statistical models for this event.
Today's event represents a significant false positive in magnitude prediction. The physics model forecasted a peak of 4.68 ft, primarily driven by substantial QPE rainfall estimates in Bands 2 (1.06"), 3 (1.54"), and 4 (1.53") during the evening hours of June 6. However, the USGS gauge recorded only a 0.06 ft rise, peaking at 2.74 ft, which is effectively baseline noise given the preceding flat hydrograph shape. This indicates that either the QPE data severely overestimated actual precipitation in the watershed, or the current coefficients for Bands 2, 3, and 4 are generating too much runoff response for the actual ground conditions.
The discrepancy is particularly notable because the antecedent moisture was classified as WET (2.85" 7-day avg), which typically enhances runoff. However, the lack of any corresponding rise suggests that the 'WET' multiplier (1.5) combined with the Band 2-4 response coefficients is currently too aggressive for this specific spatial distribution of rain. The rainfall was heavily concentrated in the middle bands (2-4), which have lag times of 3-5 hours. If this rainfall had occurred, we would expect a clear rise starting late on June 6 and peaking early June 7. The absence of this rise implies the effective rainfall was much lower than QPE suggests, or infiltration exceeded model assumptions despite the wet tier.
Conservative calibration adjustments are recommended to reduce the responsiveness of the middle bands (2, 3, 4) which were the primary drivers of the erroneous prediction. Band 1 and 5 were less significant in the prediction and saw less QPE, so they will remain unchanged. No adjustments are made to moisture or intensity tiers, as the error likely stems from band-specific response coefficients or QPE accuracy rather than the global multipliers.
| Band | Change | Reason |
|---|---|---|
| 2 | -15% | Band 2 predicted 2.61 ft of rise from 1.06" QPE but contributed to zero actual rise. Reducing coefficient by 15% to penalize this significant overestimation. |
| 3 | -20% | Band 3 predicted 1.96 ft from 1.54" QPE with zero actual rise. Applying maximum allowable reduction to curb overprediction from this dominant source. |
| 4 | -20% | Band 4 predicted 1.32 ft from 1.53" QPE with zero actual rise. Applying maximum allowable reduction to curb overprediction from this dominant source. |
Model confidence for Bands 2, 3, and 4 should remain 'medium' or potentially degrade if QPE continues to outpace actual runoff. The discrepancy highlights a need to verify QPE accuracy for mid-watershed bands.