Re-graded on November 4, 2026 under the new grading rule (changelog 2026.11.4.3). The original checked the empirical forecast against the wrong hour. Its grade did not change.
Classification: partial Confidence: Model confidence remains moderate; consistent over-prediction in Bands 2-3 on dry soil suggests a need for tighter calibration on runoff efficiency for these specific conditions.
The model over-predicted the peak by 26% and mis-timed the crest by 2.4 hours during a multi-band rainfall event on dry soil, attributed to an exaggerated response from Bands 2 and 3 relative to the actual runoff.
| Metric | Predicted | Actual | Error |
|---|---|---|---|
| Peak height | 3.97 ft | 3.15 ft | +0.82 ft |
| Total rise | — | 1.26 ft | — |
| Band | Precip | Predicted Rise | Intensity | Moisture |
|---|---|---|---|---|
| 1 | 1.62" | 0.25 ft | INTENSE | VERY DRY |
| 2 | 1.49" | 0.85 ft | INTENSE | VERY DRY |
| 3 | 1.34" | 0.38 ft | HEAVY | VERY DRY |
| 4 | 1.45" | 0.06 ft | HEAVY | VERY DRY |
| 5 | 1.80" | 0.18 ft | HEAVY | VERY DRY |
Headline: ~35% chance of rising to LOW within the next 12 h. Experimental. Forecast graded: the last forecast before the creek reached its crest, issued 14:18 at 2.53 ft (24 forecasts today) Outcome: LOW (3 ft) was reached; the gauge's high after this forecast was 3.15 ft today. A chance is not right or wrong on one day. Notes: The empirical forecast correctly identified a rise to the LOW tier (3.0 ft) with a 35% probability, which was realized as the gauge peaked at 3.15 ft.
The event involved significant rainfall across all bands (1.3-1.8"), resulting in a broad hydrograph with a 1.26 ft rise. The prediction of 3.97 ft against an actual 3.15 ft peak represents a 26% overestimation, which exceeds the 20% tolerance for a 'correct' grade, resulting in a 'partial' classification. While the direction of the rise was correct, the magnitude was inflated, particularly by Band 2 (0.85 ft contribution) and Band 3 (0.38 ft contribution). The antecedent moisture was classified as VERY DRY, which should suppress runoff, but the model's response coefficients for the middle bands appear too high for these conditions, leading to an over-prediction of the total volume reaching the gauge.
Predictions issued today (archive): 22 active issuance(s) from 2026-11-03T06:20:02.371355+00:00 to 2026-11-04T03:20:02.547422+00:00; issued peaks ranged 2.06 to 3.97 ft. The creek reached its crest at 2026-11-03T15:15:00-06:00. The prediction graded below is the HIGHEST peak issued before then (issued 2026-11-03T15:20:02.294689+00:00, 5.9 h ahead). 7 later issuance(s) don't count: the creek was already there.
| Band | Change | Reason |
|---|---|---|
| 2 | -10% | Band 2 provided the largest predicted contribution (0.85 ft) which likely drove the overprediction. Reducing sensitivity by 10% aligns better with the actual 1.26 ft total rise on dry soil. |
| 3 | -5% | Band 3 contribution (0.38 ft) was also likely overestimated relative to the actual hydrograph; a conservative 5% reduction improves balance without risking under-prediction in wetter conditions. |
Model confidence remains moderate; consistent over-prediction in Bands 2-3 on dry soil suggests a need for tighter calibration on runoff efficiency for these specific conditions.