MC Warning Pays Off
Fifteen communities are analyzed (Louisville excluded — no Q3 uncertainty data). The MC ensemble triggers earlier, at the P10 arrival — when 10% of draws already show fire at the community. That shift provides 1–7 hours more warning, which pushes compliance from the steep part of the S-curve into the saturating region.
Warning Hours vs. People at Risk
The compliance S-curve means the first hours of warning are worth dramatically more than later hours. These four communities show the knee of the curve most clearly — risk drops steeply from 0–4 hours, then saturates.
Compliance model grounded in Camp Fire research: Grajdura et al. 2021, Wong et al. 2022, NIST TN 2252. At 0h: 5% comply. At 1h: 13%. At 2h: 32%. At 4h: 77%. At 8h: 90%.
Lives Saved by MC — By Community
Sorted by impact. Broomfield dominates because of its enormous population (74,112). Windsor (24,390 saved) and Paradise (15,448 saved) round out the top three. Louisville is excluded (no Q3 uncertainty data).
The S-curve makes the first hours disproportionately valuable. From Camp Fire research (Grajdura 2021, Wong 2022, NIST TN 2252): at 0 hours warning, only 5% of people evacuate. At 1 hour, 13%. At 2 hours, 32%. At 4 hours, 77%. At 8 hours, 90%. MC's 1–7 hour shift lands squarely on the steep part of that curve. Past 8 hours, you're in the flat region where more warning barely moves the needle.
Compliance S-curve limitation. The S-curve is calibrated to Camp Fire departure data (NIST TN 2252). Different populations — rural vs. urban, repeat-evacuees vs. first-time, elderly vs. mobile — will have different curve shapes. The 83% at-home death rate from Camp Fire may not generalize to all communities. For site-specific planning, local compliance data should replace this generic curve. Communities with prior wildfire experience (e.g., Kincade-area residents who evacuated in 2017) likely have steeper curves; communities with no fire history may have flatter ones.
4 minutes of compute, 146,643 fewer people at risk. A 200-draw Monte Carlo runs in 4 minutes. Across four real fires, the MC trigger reduces the exposed population from 170,515 to 23,872. That's the cost of uncertainty quantification: 4 minutes. The payoff is not abstract.
Without MC, the deterministic trigger leaves 170,515 people exposed. With MC, that drops to 23,872 — an 86% reduction. The reason is the S-curve shape: those 1–7 extra warning hours land right where each additional hour converts thousands of holdouts into evacuees. A single-point forecast gives you false precision. The ensemble gives you the margin you need to account for the worst case.
At-risk = population × (1 − compliance(warning_hours)) × road capacity constraint. Deterministic trigger = hour when single-point forecast reaches community. MC trigger = hour when ≥10% of ensemble draws reach community. Warning hours = median arrival − trigger hour.