Classification: no_prediction Confidence: Confidence remains low due to repeated 'no_prediction' events. The model is missing minor to moderate events where QPE detects rain but gauge sensors do not, or where the prediction engine fails to trigger. Until the prediction engine generates outputs for these QPE-positive days, calibration coefficients cannot be refined.
The prediction system failed to generate a forecast for 2026-06-22, recording a null predicted peak, despite observed QPE inputs and a minor 0.9 ft gauge rise.
| Metric | Predicted | Actual | Error |
|---|---|---|---|
| Peak height | N/A | 2.21 ft | N/A |
| Total rise | — | 0.9 ft | — |
| Band | Zone | Precip | Predicted Rise | Intensity | Moisture |
|---|---|---|---|---|---|
Headline: Watch — Recent rainfall is in the lower-bound range preceding rises to LOW Settled outcome: no_change (reached below_low) LLM said headline was correct: True Notes: The empirical headline predicted 'no_change' and the gauge remained in the 'below_low' tier. The forecast was accurate.
The model did not produce a prediction output for this event, classifying it as a 'no_prediction' event. This continues a pattern seen in recent history (June 8, 14, 15, 16) where the system fails to generate forecasts, likely due to discrepancies between gauge-precipitation sensors and QPE data or internal thresholding issues. The gauge recorded 0.0 inches of precipitation, while QPE showed scattered rainfall in both bands, particularly a burst around 07:00-08:00 CST.
The actual hydrograph showed a broad rise of 0.9 ft, peaking at 2.21 ft, which is below the 'low_floatable' threshold of 4.0 ft. The rise started at 07:30, roughly 2-3 hours after the peak QPE rainfall at 07:13-08:13, consistent with the 2.7-hour lag. The broad shape suggests runoff from both bands, though the magnitude was small. The antecedent moisture was WET (1.768"), which typically amplifies response, but the rainfall totals were relatively low and scattered.
The empirical model correctly assessed the situation as 'no_change' within the 'below_low' tier, noting that while recent rainfall was in a range that could precede a rise to LOW, it did not actually reach that threshold. Since no prediction was generated, no coefficient adjustments can be made based on prediction error. The system's inability to trigger a prediction on days with measurable (via QPE) but gauge-zero precipitation remains a blind spot.
No changes made.
Confidence remains low due to repeated 'no_prediction' events. The model is missing minor to moderate events where QPE detects rain but gauge sensors do not, or where the prediction engine fails to trigger. Until the prediction engine generates outputs for these QPE-positive days, calibration coefficients cannot be refined.