Wildfire Black Box · Colorado, United States · 2020
A fire across the Divide. An evidence trail across six years.
East Troublesome reached Rocky Mountain National Park during its October 2020 run and established fire on the eastern side of the Continental Divide. Satellite observations helped forecasters recognize that crossing. Later research examined a different set of questions: which homes were destroyed, where live trees remained, and whether forests could regenerate.
Record TB-WBB-US-2020-EAST-TROUBLESOME · Version 1.0 · Evidence reviewed 14 October 2026
Incident scope: keep the park and the whole fire separate
The fire was reported on 14 October 2020 in the Arapaho National Forest, outside the park. A Forest Service anniversary release places the initial smoke report north of Hot Sulphur Springs and records containment on 30 November. It reports two deaths and says the cause was still under investigation on 14 October 2021. That is a dated cause status, not evidence of a final finding available today. [1]
The NPS account describes an approximately 18-mile run before the fire entered the park and approximately 1.5-mile spotting across the Divide. Those are approximate, source-defined distances. They are not a surveyed continuous flame-front track, and no elapsed interval is provided for calculating a defensible rate of spread from them. [2]
| Source | Value | What the value measures |
|---|---|---|
| Forest Service joint release, 14 October 2021 | 193,892 acres; 589 homes and structures | Whole-incident figures in the anniversary narrative. [1] |
| Meldrum et al., February 2022 | 193,812 acres; 366 residences plus 214 outbuildings/commercial structures | Incident context for a separate parcel-level study; the structure categories total 580. [5] |
| Rodman et al., November 2022, Table 1 | 193,811.9 acres | Perimeter-based landscape assessment; its narrative rounds to 193,812. [6] |
| NPS recovery account, December 2021 | Approximately 22,668 acres | East Troublesome inside the park boundary. [8] |
| NPS overview, July 2021 | Approximately 30,000 acres, or 9% of the park | Combined park impacts of East Troublesome and Cameron Peak. [11] |
The two whole-fire rounded totals differ by 80 acres. The structure statements differ by nine after the paper’s categories are added. This review does not establish whether those gaps reflect mapping revisions, classification, reporting error or another explanation. The record preserves both statements. Subtracting the park figures would create another misleading result: the approximate combined figure is not a precise perimeter accounting from which to recover Cameron Peak’s area.
Origin boundary: the December 2021 NPS recovery article contains wording that places ignition in the park, then describes movement into the park a week later. For origin geography, this record relies on the Forest Service release and research study. It uses the NPS article for the recovery work it documents, retaining the wording discrepancy as a source limitation.
How satellite evidence reached emergency officials
The NPS release issued on 23 October 2020 described an eastern-side heat signature on the morning of 22 October, followed by a lookout’s confirmation of smoke in Spruce Creek. It also reported fallen trees blocking part of Trail Ridge Road during the previous night’s evacuations. Those are distinct observations: recognizing a new fire location and finding that an evacuation route is obstructed. [2]
Seaman and colleagues’ 2023 operational paper adds UTC timestamps. Their account explains that high terrain displaced the apparent location of features in the GOES-16 imagery by approximately five kilometres toward the northwest. Forecasters had to account for that parallax before identifying fire east of the Divide. A hot pixel’s apparent map position could not simply be treated as its ground location. [3]
| UTC | MDT, UTC−06:00 | Evidence object |
|---|---|---|
| 12:06 | 06:06 | First illustrated frame in the paper’s Figure 9. |
| 12:21 | 06:21 | Second illustrated frame. |
| 12:26 | 06:26 | Third illustrated frame. |
| 13:07 | 07:07 | Recorded first call to emergency officials in the authors’ account. |
| By 16:00 | By 10:00 | Fire visually confirmed; an upper limit, not an exact confirmation timestamp. |
The illustrated frames span 20 minutes. The time from the recorded first call to the stated confirmation deadline is 173 minutes. Because the paper says confirmation occurred by16:00, that calculation is an upper bound on the reported call-to-confirmation interval. It is not a measured evacuation delay, a resident’s warning receipt time or the amount of time saved by the satellite.
Evidence here does support the operational value of combining remotely sensed information, geolocation knowledge, communication and field confirmation. Estimating how many people would have been harmed without that sequence would require a separate counterfactual analysis. The archived release and later paper do not provide that calculation.
Earlier, NOAA’s smoke narrative through 16:10 UTC on 21 October described two plumes from East Troublesome and Cameron Peak, partly obscured by clouds. Its product limitations exclude quantitative particulate concentration and vertical-distribution estimates. A visible plume is evidence of transported smoke; it is insufficient to assign an AQI, breathing-zone exposure or indoor dose to a particular home. [4]
Home losses: a denominator audit of Columbine Lake
The parcel study is valuable because the rapid assessments preceded the fire. Researchers used assessments of 461 homes from summer 2019, rather than reconstructing every pre-fire attribute from debris afterward. Missing roadside observations reduced the complete-data sample to 352 structures; 23 of those were destroyed. Across all assessed structures, 26 were destroyed. The paper also reports 30 destroyed structures in the community, including properties outside the assessment dataset. [5]
| Dataset | Structures | Destroyed | Derived fraction |
|---|---|---|---|
| All assessed properties | 461 | 26 | 26 ÷ 461 = 5.64% |
| Complete assessments | 352 | 23 | 23 ÷ 352 = 6.53% |
| Incomplete assessments, derived by subtraction | 109 | 3 | 3 ÷ 109 = 2.75% |
| Community losses outside the assessment dataset | Unknown denominator | 4, derived from 30 − 26 | No valid percentage calculated |
Changing from all assessed properties to complete assessments raises the displayed fraction by approximately 0.89 percentage points. The fire did not become more destructive when the sample changed. A different numerator and denominator were selected. The excluded 109 properties represent 23.64% of assessments; their missing values cannot safely be interpreted as “no hazardous feature.”
The authors fitted separate spatial models and tested alternative treatments of incomplete data. This page does not refit those models or turn their coefficients into a universal probability for a homeowner. It also avoids treating the complete-loss classification as a measure of all damage: a structure outside the destroyed category may still have experienced smoke damage, burns or other losses.
A useful comparison must establish exposure as well as outcome. Treating the final perimeter as an independent treatment boundary is particularly problematic where building ignitions and defensive actions can help determine that boundary. The study’s observational design and local setting remain attached to its findings. A higher aggregate risk score is not a guarantee of destruction, and a lower score is not a guarantee of survival.
Forest recovery: mapped seed access and later field evidence
Rodman and colleagues classified high severity using Relativized Burn Ratio (RBR) ≥283. Their live-tree proxy combined that classification with pre-fire canopy height ≥5 metres, at approximately 100-foot pixel spacing. The report gives East Troublesome a 57.4% high-severity share and divides that area by distance to a mapped live tree. These are vegetation and seed-access indicators, not a soil-burn-severity classification. [6]
| Distance class | Share of high-severity area | Derived share of whole mapped fire area |
|---|---|---|
| Less than 300 feet | 37.9% | 57.4% × 37.9% = 21.75% |
| 300–650 feet | 32.2% | 57.4% × 32.2% = 18.48% |
| More than 650 feet | 29.9% | 57.4% × 29.9% = 17.16% |
| Outside the high-severity class | Not part of that denominator | 100% − 57.4% = 42.60% |
The three distance shares sum to 100% of high-severity area. They do not mean all burned land lies in those classes. Conversely, the 42.60% complement does not mean untouched forest: it is simply the mapped area outside that high-severity class. Products above use rounded source percentages and are displayed to two decimals for reproducibility, not to claim extra mapping precision.
A live tree within a distance band is a proxy for a potential seed source. That alone does not identify its species, viable seed production, dispersal to a particular point or seedling establishment. The 2022 report’s species-specific recruitment predictions are model outputs. They are useful planning evidence, and they should remain distinct from subsequent field measurements.
A Forest Service webinar abstract dated 11 April 2024 describes field work in 88 lodgepole-pine-dominated stands across both Mullen and East Troublesome. It reports low recruitment and reduced seed viability associated with previous beetle mortality. This is later evidence about sampled stands, not an East Troublesome-wide census. The abstract does not identify which fire corresponds to each of its two recent-harvest seedling-absence percentages, so this page does not assign them to East Troublesome. [7]
The model and field account can guide the next research question without being forced into a verdict that one is “wrong.” A responsible comparison would match the stands, survey dates, species, thresholds and spatial predictions. That matched validation was not performed here. Raw imagery, canopy rasters and stand-level field data were not reprocessed for this record.
Watersheds and infrastructure: treatment is not a measured outcome
The NPS December 2021 recovery account describes 42 issues identified in a BAER scoping session and a decision to relocate ten wilderness campsites. Assessments were constrained by damage, safety and winter conditions. These numbers describe identified concerns and management actions; neither is a count of hazards eliminated or acres that had recovered. [8]
A USGS data release listed on 2 January 2024 models post-fire dNBR from pre-fire variables for East Troublesome and Grizzly Creek. It uses random-forest modelling to support burn-severity and debris-flow-hazard work. “Pre-fire prediction” describes the model’s inputs and intended application; this later release is not proof that its forecasts were issued to responders before October 2020. [9]
Grand County’s recovery page connects the incident to watershed-restoration work and links to a USGS debris-flow map. The map endpoint could not be retrieved during this review. No basin-specific probabilities, rainfall thresholds or validation scores are therefore claimed. A project description, a conditional hazard model and measured sediment delivery are different evidence objects. [10]
The anniversary release also documents seeding, mulching and infrastructure-protection work. Such activity is evidence of a response. Demonstrating effectiveness would require measurements and a comparison capable of separating treatment effects from rainfall, terrain, burn severity and other influences. This study does not manufacture an effectiveness percentage from the area treated.
Evolution of knowledge
| Evidence date | What enters the record | Boundary |
|---|---|---|
| 21–23 October 2020 | Smoke observations, ridge crossing, evacuation activity and route obstruction | Operational observations; the NPS release retained incomplete damage assessment. |
| 2021 | Anniversary totals and documented park stabilization work | Later reporting; no final cause finding established by this review. |
| 2022 | Parcel models and a landscape assessment | Retrospective research with defined samples and classifications. |
| 2023 | Detailed satellite-to-dispatch chronology | Published reconstruction of a 2020 operational sequence. |
| 2024 | Model data release and a field-research webinar abstract | Different products; neither is a complete 2026 recovery census. |
Method and unresolved questions
Reading evacuation evidence: our guide to wildfire evacuation statistics explains how alert records, reported departures and mobile population changes use different counting units and timelines. Its examples are synthetic and do not add observations to this incident.
Tecnobosque transcribed the cited values, checked the relevant PDF pages visually, calculated the displayed fractions and UTC conversions, and drew the two original figures. The tables preserve source-defined inputs separately from derived values. This is a documentary and arithmetic audit, not a new ignition investigation, satellite-image classification or statistical model fit.
Important unresolved items include the final ignition mechanism, the explanation for conflicting whole-incident totals, exact warning receipt by individual residents, a counterfactual evacuation benefit and spatially matched validation of forest-recruitment predictions. The 14 October 2026 review date identifies this review; it does not convert older webpages into a current incident bulletin.
For related evidence problems, continue with Fourmile Canyon’s home-loss study, High Park’s watershed-to-intake record and Riverside’s soil-severity raster comparison. The cases address different settings and measures; they form a reading path, not a ranking.
Primary sources and review scope
- Forest Service and partners: One year later—East Troublesome recovery. Release 14 October 2021. Origin setting, source-defined totals, containment and recovery actions; final-cause status limited to that date.
- NPS: East Troublesome on both sides of the Continental Divide. Release 23 October 2020; page updated 26 October. Archived operational narrative, route obstruction and incomplete damage assessment.
- Seaman et al.: Multispectral Satellite Imagery Products for Fire Weather Applications. June 2023, Journal of Atmospheric and Oceanic Technology. Reviewed title, East Troublesome section and Figure 9 on PDF pages 11–12; selected concluding methods. Original satellite datasets and videos not downloaded or reanalysed.
- NOAA satellite smoke narrative through 16:10 UTC, 21 October 2020. Reviewed Colorado observations and product limitations. Qualitative plume evidence, not local AQI or measured PM2.5.
- Meldrum et al.: Parcel-level risk affects wildfire outcomes. Published 12 February 2022; Fire 5, 24; DOI. Reviewed selected context, methods, tables and discussion, especially PDF pages 10–12. Counts transcribed; no parcel-level model refit or copying of its licensed basemap.
- Rodman et al.: The historic 2020 fire year. November 2022, Southwest Ecological Restoration Institutes, 29 pages. Reviewed Table 1, mapping methods and East Troublesome subsection, especially printed pages 8 and 14–15. Original GIS inputs not processed.
- Forest Service / Charles Rhoades: Low seed viability and other obstacles to post-fire tree regeneration. Webinar 11 April 2024; page updated 3 December 2025. Reviewed posted abstract only; presentation recording and underlying field data not reviewed.
- NPS: Rocky Mountain National Park rallies to recover. Updated 20 December 2021. Park footprint and stabilization narrative; origin wording discrepancy retained.
- USGS: Burn-severity modelling data for East Troublesome and Grizzly Creek. Catalogue date 2 January 2024; DOI 10.5066/P9T077XK. Reviewed catalogue and model description; data files and predictive accuracy not independently evaluated.
- Grand County: Fire and Watershed Recovery. Retrieved 14 October 2026; publication date not established. Reviewed recovery description and official resource links. Linked USGS map retrieval timed out; no map estimates transcribed.
- NPS: Cameron Peak and East Troublesome fires. Updated 9 July 2021. Combined park impact, date-level evacuation context and weather transition; not a current access or trail-status bulletin.
Suggested citation: Tecnobosque. 2026. East Troublesome Fire 2020: Satellite Warnings and Recovery. Wildfire Black Box, TB-WBB-US-2020-EAST-TROUBLESOME, version 1.0. Evidence review 14 October 2026.