Classification: false_positive Confidence: Model confidence in Bands 2 and 3 remains 'medium'. This event highlights a weakness in trusting QPE for these bands when gauge precipitation is zero. Continued conservative calibration is advised.
The model significantly overpredicted the river stage rise (26.1% error) on a day with zero gauge precipitation, largely due to an overestimation of the response efficiency of intermediate bands (2-3) under NORMAL moisture conditions.
| Metric | Predicted | Actual | Error |
|---|---|---|---|
| Peak height | 2.56 ft | 2.03 ft | +0.53 ft |
| Total rise | — | 0.16 ft | — |
| Band | Precip | Predicted Rise | Intensity | Moisture |
|---|---|---|---|---|
| 2 | 1.14" | 1.56 ft | HEAVY | NORMAL |
| 3 | 1.21" | 0.85 ft | HEAVY | NORMAL |
| 4 | 1.38" | 0.60 ft | HEAVY | NORMAL |
| 5 | 1.75" | 0.75 ft | INTENSE | NORMAL |
Headline: ~20% chance of rising to LOW in the next 6–12 h. Experimental. Settled outcome: None (reached None) LLM said headline was correct: None Notes: The empirical model predicted a ~20% chance of rising to LOW. The river did not rise to LOW (stayed at 2.03 ft, below 3.0 ft). The outcome is technically 'no rise to tier', which is consistent with a low-probability forecast. However, the empirical record for this specific day is not yet settled in the provided history, so null is returned.
The event was characterized by significant QPE rainfall across bands 2-5, but crucially, 0" of actual precipitation at the gauge. The actual river response was minimal (0.16 ft rise), peaking at 2.03 ft. The model predicted a peak of 2.56 ft, a substantial overestimation given the low baseline. This is classified as a false positive because the model predicted a 'meaningful' rise driven by QPE that did not materialize in the hydrograph, likely due to spatial mismatch between the radar estimate and the actual watershed infiltration/runoff potential on normally dry soil.
The primary driver of the error appears to be Bands 2 and 3. The model predicted 1.56 ft from Band 2 and 0.85 ft from Band 3, totaling ~2.4 ft of contribution on top of a 1.87 ft baseline. Since the actual rise was only 0.16 ft, the model vastly overestimated the efficiency of these intermediate bands under NORMAL moisture (1.0 multiplier). The lack of gauge precipitation suggests the rainfall was either spatially disjointed from the main channel or absorbed by the soil without generating runoff.
Bands 4 and 5 were also overestimated, but to a lesser degree relative to their input volumes. However, since the total error is dominated by the near-gauge contributions (Bands 2-3) in this prediction profile, conservative reductions are recommended for these bands to penalize the reliance on QPE when ground truth (gauge precip) is zero. Band 1 remains unchanged as it had minimal QPE and no gauge precip, behaving as expected (no contribution).
| Band | Change | Reason |
|---|---|---|
| 2 | -15% | Significant overprediction (1.56 ft predicted vs ~0.16 ft total rise). The model relied too heavily on QPE in Band 2 under NORMAL moisture conditions where runoff efficiency is likely lower than currently calibrated. |
| 3 | -10% | Overprediction (0.85 ft predicted). While lagged, the contribution was exaggerated relative to the actual hydrograph flatline. Small reduction to align with the low actual rise. |
| 4 | -5% | Minor overprediction. The contribution was likely negligible given the 0.16 ft total rise. Small conservative reduction. |
Model confidence in Bands 2 and 3 remains 'medium'. This event highlights a weakness in trusting QPE for these bands when gauge precipitation is zero. Continued conservative calibration is advised.