Classification: correct Confidence: Model confidence is stable. The ability to correctly classify 'fizzle' events (where QPE rain does not translate to gauge rise) is a key strength. The 36% magnitude error is acceptable for near-baseline predictions and does not indicate a systematic bias requiring adjustment.
The model correctly predicted a minimal stage rise of 0.34 ft, closely matching the observed 0.25 ft peak during a quiet, dry-day rainfall event that was not detected by the gauge's precipitation sensor.
| Metric | Predicted | Actual | Error |
|---|---|---|---|
| Peak height | 0.34 ft | 0.25 ft | +0.09 ft |
| Total rise | — | 0.06 ft | — |
| Band | Zone | Precip | Predicted Rise | Intensity | Moisture |
|---|---|---|---|---|---|
| 1 | upper_richland | 0.28" | 0.12 ft | MODERATE | DRY |
| 2 | falling_water | 0.16" | 0.03 ft | LIGHT | DRY |
Headline: No rise indicated by recent rainfall Settled outcome: None (reached None) LLM said headline was correct: True Notes: The empirical headline 'No rise indicated by recent rainfall' was correct. The gauge remained near baseline (0.25 ft peak) with no rise to the 'low' threshold (3.0 ft). The model's 0% probability of rise was borne out by the actual outcome.
This event represents a 'fizzle' scenario where QPE sensors detected light rainfall (0.28" Band 1, 0.16" Band 2) that produced a negligible stage rise of 0.06 ft, peaking at 0.25 ft. The model predicted a peak of 0.34 ft, resulting in a 36% magnitude error. However, given that both values are near the baseline of the gauge (which has a 'too_low' threshold of 3.0 ft) and the predicted rise is less than 0.1 ft, this falls within the range of acceptable noise for a low-magnitude event. The classification is 'correct' because the model correctly identified the lack of a significant hydrologic response, avoiding the 'false positive' misclassification seen in previous high-rainfall/fizzles events.
The discrepancy in magnitude (0.34 ft vs 0.25 ft) is likely due to the baseflow contribution in the prediction model versus the actual flat hydrograph. The gauge precipitation sensor recorded 0.0" for the day, confirming that the QPE rainfall was either not reaching the ground or was too minor to generate measurable runoff beyond existing baseflow. The timing error of 1.6 hours is also consistent with the model's lag parameters for such small signals.
Predictions issued today (archive): 2 active issuance(s) from 2026-09-06T21:20:02.871517+00:00 to 2026-09-06T22:20:02.060314+00:00; issued peaks ranged 0.34 to 0.34 ft. The prediction graded below is the HIGHEST issued peak of the day (issued 2026-09-06T21:20:02.871517+00:00), not the last card of the day.
Grade-only mode: no adjustments suggested; calibration is fit offline.
Model confidence is stable. The ability to correctly classify 'fizzle' events (where QPE rain does not translate to gauge rise) is a key strength. The 36% magnitude error is acceptable for near-baseline predictions and does not indicate a systematic bias requiring adjustment.