TECNOBOSQUE · CROSS-INCIDENT ANALYSIS · 3 October 2026
A warning can exist without reaching a household. A household can understand a warning yet face too little time to move. Research on Mati and Gatlinburg illuminates these different parts of wildfire evacuation. This comparison explains what each approach measures and why their results should not be turned into a league table of community performance.
Two studies, two questions
The 2025 Mati study by Kalogeropoulos and colleagues compares uncertain fire-arrival and evacuation times using trigger boundaries. Its application identifies a low evacuation safety margin from the time of detection. This is a probabilistic reconstruction under model assumptions, not a measurement of an individual’s survival probability. [1]
NIST’s Gatlinburg research used a retrospective online survey conducted from October 2018 to May 2019, with 323 participants in the final selected sample. It investigated information sources, preparation and evacuation decisions during the 2016 Chimney Tops 2 fire. Respondents commonly relied on media, social information and environmental cues; most of the sample evacuated. [2]
| Question | Mati modelling [1] | Gatlinburg survey [2] |
|---|---|---|
| Primary evidence | Probabilistic fire and evacuation timing | Recalled household experiences |
| Main contribution | Whether the available physical window accommodates movement | How people learned about danger and decided to respond |
| Key limitation | Results depend on model inputs and assumptions | Selection and recall constrain population-wide inference |
| Cannot establish alone | Every household’s actual understanding | The fire-arrival time on every possible route |
The clocks that must not be combined casually
Tecnobosque analytical framework: distinguish detection, warning issuance, warning receipt, preparation, departure and completion of movement. These are useful event markers, not a claim that both studies measured every stage in the same way.
A figure described as “evacuation time” might begin at an official order or at vehicle departure. Those definitions exclude different parts of the process. Before comparing two durations, identify both their starting event and their endpoint. Otherwise one community can appear faster simply because its clock started later.
Similarly, ignition time is not necessarily detection time. A long interval between ignition and community impact does not establish that residents had that entire interval available for an informed decision.
Worked example: the missing 20 minutes
Illustrative only—not data from either fire: suppose a household receives a warning at 14:00, departs at 14:20 and reaches the defined destination at 14:45. Travel lasted 25 minutes; the interval from warning receipt to arrival was 45 minutes. Neither figure is inherently wrong, but they answer different questions.
If a comparison uses 25 minutes for one community and 45 minutes for another without stating the definitions, it creates a misleading difference. No assumption about road speed or human behaviour is needed to see the problem: the measurement boundaries alone change the result.
What combining the approaches could add
A physical model can explore how much movement time a scenario permits. Household evidence can help investigate the delays, information gaps and practical circumstances that precede movement. A future integrated analysis would need compatible locations, timestamps and population definitions before bringing those datasets together.
This article does not merge the published datasets or calculate a new safety factor. It identifies the evidence that would be needed: route-specific conditions, dated communications, departure observations and clearly defined destinations. Missing values should remain missing rather than being replaced by a convenient average.
Why there is no “better evacuation” score here
Fatality counts alone cannot isolate warning effectiveness. Exposure, population, fire progression and available movement options differ. Likewise, a retrospective association between a warning and evacuation does not prove that sending the same message elsewhere would reproduce the outcome.
The defensible comparison is between questions and evidence: what constrained the physical window, what shaped the decision to move, and where the record cannot connect them. This is historical analysis, not a live evacuation instruction or a tool for choosing when to leave.
Sources and further reading
- Kalogeropoulos et al. (2025), Quantifying dire evacuations using trigger boundaries; institutional research abstract.
- Walpole et al. (2020), NIST TN 2103: household evacuation survey.
Read the full Tecnobosque case studies: Mati 2018 · Chimney Tops 2–Gatlinburg 2016 · Fire Archive.
Method: editorial comparison of published research scopes. No original field investigation, pooled statistical analysis or incident ranking. The worked example is hypothetical.