Classification: no_prediction Confidence: Model confidence remains medium. The failure to predict during a significant rainfall event (driven by gauge-precipitation mismatch) highlights a data-integration flaw rather than a hydrological parameter error. The coefficients are currently suppressed from the previous overprediction, which may be protective, but we cannot confirm if they are now under-responsive.
The prediction system failed to generate a forecast for 2026-06-14, likely due to zero gauge precipitation readings, despite a massive observed gauge rise to 9.25 ft driven by heavy upstream rainfall in both bands.
| Metric | Predicted | Actual | Error |
|---|---|---|---|
| Peak height | N/A | 9.25 ft | N/A |
| Total rise | — | 5.56 ft | — |
| Band | Zone | Precip | Predicted Rise | Intensity | Moisture |
|---|---|---|---|---|---|
The predictor returned no output (None) for today's event. The gauge precipitation sensor recorded 0.0 inches, which typically triggers a 'no rain, no prediction' logic in the upstream model. However, QPE data shows significant rainfall occurred in the watershed early in the day (03:13–05:13), with Band 1 receiving ~0.57 inches and Band 2 receiving ~0.80 inches. This disconnect between gauge rain and watershed rain is the root cause of the missing prediction.
The observed hydrograph shows a multi-pulse rise peaking at 9.25 ft at 00:15. Given the rainfall started around 03:13 the previous day (based on the daily aggregation window) or early morning, and the peak occurred at the start of the reporting day, the timing suggests the rainfall was intense and rapid. The antecedent moisture was SATURATED (4.889 in, multiplier 2.0), which explains the high runoff efficiency. The previous event on 2026-06-13 resulted in a massive overprediction (predicted 16.05 ft vs actual 8.35 ft), leading to a -20% reduction in coefficients. Today's event, had it been predicted, would have tested whether those reductions were sufficient or excessive. Since no prediction was made, no direct coefficient calibration can be applied.
Because the predictor did not engage, we cannot grade the accuracy of the peak magnitude or timing. The system needs to address why gauge-precipitation zeros result in no prediction when QPE indicates active storm cells. For now, the calibration coefficients remain unchanged from the previous adjustment, as we lack a predicted value to compare against the observed 9.25 ft peak. The confidence remains medium.
No changes made.
Model confidence remains medium. The failure to predict during a significant rainfall event (driven by gauge-precipitation mismatch) highlights a data-integration flaw rather than a hydrological parameter error. The coefficients are currently suppressed from the previous overprediction, which may be protective, but we cannot confirm if they are now under-responsive.