The Capacity Constraint
Margin = available warning time minus total evacuation time (clearance + 1-hour mobilization delay). Negative means the fire arrives before everyone is out, even with a perfect MC-informed trigger.
Evacuation Margin by Community
Positive margin (green) means the community clears before fire arrival. Negative margin (red) means people are still on the road when the fire hits. Sorted worst to best.
Which Communities Flip?
Contraflow (reversing inbound lanes for outbound traffic) doubles road capacity. None of the four failing communities flip with contraflow alone — Paradise, Centerville, Indian Falls, and Broomfield all remain stuck. The deficits are too large for a capacity doubling to close. These communities need phased evacuation or shelter-in-place plans.
Contraflow factor: 2.0x capacity. Mobilization delay: 60 min (baseline). Even the best-case scenario (contraflow + 30-min mobilization) cannot rescue any of these four.
Q4 assumed infinite road capacity — everyone who's told to leave, leaves. Q7 drops that assumption. For four communities with limited egress, an earlier trigger doesn't help if the roads can't handle the load. These places need shelter-in-place plans, not better forecasts.
The answer is yes or no, not 2.45 vs. 2.31 hours. A traffic microsimulation would refine Paradise's margin from −2.45h to maybe −2.31h. Same answer: they can't make it. A simple throughput model (population / road capacity) gets the binary right, validated against NIST Camp Fire data. When the question is pass/fail, precision past the decimal point is wasted effort.
Broomfield is the extreme case. With 74,112 people and only 2,400 veh/hr (baseline, 3 roads), clearance takes 12.35 hours. Even with a trigger at hour 40 and fire arrival at hour 42 (P10), the margin is −10.35 hours. Contraflow helps but cannot close a 10-hour gap. This community needs phased evacuation or shelter-in-place — not a better forecast.
Reality Check: Paradise Gridlock
NIST TN 2252 documented the Camp Fire evacuation in 2,600+ observations. Our model predicts Paradise cannot evacuate. What actually happened?
| Metric | Our Model | Actual (NIST TN 2252) |
|---|---|---|
| Paradise can evacuate? | No (−2.45h) | 85 deaths |
| Road bottleneck? | Yes (limited capacity) | Skyway: ~1,200 cars/hr |
| Contraflow helps? | Still fails (−1.23h) | One-way conversion planned; fire blocked roads |
| All roads available? | Not modeled | All 5 arteries blocked at least once |
| Gridlock duration | Not modeled | 4–5 hours |
| Shelter-in-place needed? | Yes (model says can’t clear) | 31 refuge areas, 1,200+ civilians sheltered |
NIST confirms the finding. Our throughput model says Paradise can’t clear in the available window. What actually happened: Skyway gridlocked within 30 minutes, 40,000 people spent 4–5 hours escaping, 1,200 had to shelter in place, and 85 died. The death pattern maps to both failure modes: 83% died at home (compliance, Q8) and 10% died in vehicles (road capacity, Q7). A traffic microsimulation would give tighter clearance estimates. It wouldn't change the conclusion: this community cannot evacuate fast enough.
Model: vehicles = population / 2.5 persons. Road capacity: 800 vph (small/1 road), 1,600 vph (medium/2 roads), 2,400 vph (large/3 roads). Mobilization delay: 60 min. Margin = (MC P10 arrival − MC trigger) − total evacuation time. Reality data: NIST TN 2252 (Maranghides et al. 2023), PBS Frontline (2019).