PRECIP_EVENT_TASK.txt

------

Task: PRECIP_EVENT_TASK

Inputs:
- Regional precipitation graphics (e.g., 6-hour QPF / precipitation type / 1000-500 mb thickness).
- Large-scale background fields (e.g., 500 mb height or anomaly) if necessary.
- Multiple models may be provided, with uneven forecast length coverage across UTC times.
- Recent satellite imagery may be provided for the past few hours as an observational consistency check.

Goal:
Assess the short-term precipitation event itself, including timing, intensity evolution, persistence, thermal/phase risk, and whether the event structure supports or undermines ski value.
Produce a concise, decision-oriented assessment of the event's practical significance and whether it warrants continued monitoring.
Decide how closely the situation should be monitored.
Evaluate whether multi-model snowfall trends support or challenge RWDI-provided snowfall numbers, and express a confidence interval for those numbers.

Analysis focus:
1. Snowfall intensity evolution
   - Is snowfall front-loaded, back-loaded, or steady?
   - Any signals that intensity could ramp up or collapse?

2. Thermal and phase risk
   - Thickness / freezing-level trend during the event.
   - Risk of snow turning wetter, mixed, or rain at Whistler elevations (1600m+).
   - Is temperature trend improving or degrading with time?

3. Model convergence vs divergence
   - Do models agree on timing, intensity envelope, and thermal structure?
   - Identify whether differences are minor (timing/placement noise) or structural (pattern-level disagreement).

4. Satellite consistency check (if recent satellite imagery is provided)
   - Use past ~3h satellite evolution only to assess whether the observed cloud shield / frontal band is moving in a way that is broadly consistent with the forecast.
   - Focus on whether the storm system appears FASTER, SLOWER, or DISPLACED relative to the short-term forecast signal.
   - Use this only to refine short-horizon precipitation persistence and timing at Whistler area over the next ~6-12h.
   - Do NOT use satellite imagery alone to infer precise snowfall totals or rain-snow phase.

5. Decision relevance
   - Is this shaping into a meaningful refresh, a low-value dusting, or a potentially underrated precipitation event?
   - Do not confuse raw snowfall amount with powder quality; a warm, wet, or mixed-phase refresh is still a downgrade.
   - Explicitly state what would need to change for this event to upgrade or downgrade.

6. RWDI snowfall number consistency check
   - Extract RWDI snowfall amounts for the relevant period (if provided in synopsis/day cards).
   - Compare RWDI values against multi-model trend consensus, not single-frame peaks.
   - Rate RWDI number consistency using:
     - HIGH: RWDI numbers are broadly consistent with the multi-model trend envelope
     - MED: RWDI numbers are plausible but sensitive to timing/thermal uncertainty
     - LOW: RWDI numbers are outside or weakly supported by current multi-model envelope
   - Provide a practical snowfall confidence interval (low/high range) for the decision window,
     based on model spread and trend persistence.
   - If RWDI numbers appear high/low relative to model consensus, explain the likely reason
     (timing offset, intensity concentration, phase uncertainty, or model divergence).

Model confidence:
- Assign a qualitative convergence level (HIGH / MED / LOW).
- Briefly justify the rating based on cross-model agreement.
- Explicitly relate convergence level to the width of the snowfall confidence interval
  (HIGH -> narrower range, LOW -> wider range).

Guidance:
- Focus on trends and relative changes, not precise snowfall totals.
- Do not issue explicit GO / GO_WITH_RISK / NO_GO labels here, and do not turn this task into a substitute for DECISION_TASK. Leave final tactical powder recommendations to DECISION_TASK when that task is selected.
- Do NOT overstate confidence when lead time is short but models disagree.
- If the already-provided evidence is sufficient to show a warm, wet, or mixed-phase storm is unlikely to deliver worthwhile powder quality in the decision window, do NOT request extra bridge frames merely to refine the exact downgrade timing.
- If key uncertainty cannot be resolved with current inputs, request the minimum additional images needed.
- Keep precipitation assessment within the fixed decision window:
  tomorrow daytime, tomorrow night, and up to daytime of the day after tomorrow.
- Do not request late-period precipitation frames beyond day-after-tomorrow daytime unless explicitly requested by the user.