← 2026-10-01  |  2026-10-14 →

Hailstone (Upper Buffalo) — Daily Analysis: 2026-10-02

Classification: false_positive Confidence: The physics model's confidence is currently 'high', but this event suggests that confidence may be overestimated for dry, low-intensity rainfall events. The recent history shows the model performing well in 'VERY DRY' conditions, but this event, despite being labeled 'DRY', behaved similarly to 'VERY DRY' in terms of runoff response. A -20% adjustment to the response coefficient will help improve accuracy in this regime.

Event Summary

The physics model significantly overpredicted the peak flow (66.5 CFS vs 5.49 CFS actual) in dry conditions, while the empirical model correctly predicted no meaningful rise.

Prediction vs Actual

Metric Predicted Actual Error
Peak CFS 66 CFS 5 CFS +61 CFS
Total rise — 1.1 CFS —

Empirical Forecast

Headline: No rise indicated by recent rainfall LLM said headline was correct: True Notes: The empirical model correctly predicted 'No rise indicated by recent rainfall'. The actual peak of 5.49 CFS was near baseline (min 4.36 CFS) and far below the low threshold of 500 CFS. The model's low probability of rise (2%) was well-calibrated.

Band Contributions

Band Zone Precip Predicted Rise Intensity
1 ? 1.09" 62 CFS MODERATE

Analysis

The physics prediction was a clear false positive. The model predicted a peak of 66.5 CFS, which is over 10x the observed peak of 5.49 CFS. This massive error stems from the model failing to adequately attenuate the rainfall input due to the antecedent 'DRY' moisture tier. Despite a 7-day total of 2.301" (which is relatively wet) the system is currently in a leaf-off season (October), and the 'DRY' tier multiplier (0.5) combined with the SCS curve number logic likely did not account for the high infiltration capacity of the bare soil in this specific event. The rainfall was also spread out over 12 hours (max 0.246"/hr), making it a 'MODERATE' intensity event that produced very little runoff on dry ground.

The empirical forecast was correct. It labeled the event as 'No rise indicated by recent rainfall' and correctly identified that the gauge value (5.25 CFS) was well below the low threshold (500 CFS) and that no significant rise occurred. The actual peak of 5.49 CFS is consistent with baseflow plus a minor infiltration-driven response, confirming the empirical model's assessment that this was not a whitewater event.

Graded Prediction

Predictions issued today (archive): 22 active issuance(s) from 2026-10-02T05:21:02.174418+00:00 to 2026-10-03T02:21:01.673500+00:00; issued peaks ranged 3.9 to 66.5 CFS. The prediction graded below is the HIGHEST issued peak of the day (issued 2026-10-02T17:21:01.972870+00:00), not the last card of the day.

Coefficient Adjustments (SUGGESTED by the analyzer, NOT applied - grade-only mode)

Band Change Reason
1 -20% Severe overprediction (66.5 vs 5.49 CFS) in dry conditions. Reducing the response coefficient by the maximum allowed 20% helps dampen the model's reaction to rainfall when antecedent moisture is low, aligning better with the observed low runoff.
← 2026-10-01  |  2026-10-14 →