Wildfire Black Box · Colorado · Historical case
The High Park Fire’s watershed record continues long after containment. This case follows the evidence from a mapped burn area to storm-driven sediment transport and drinking-water operations. Its central question is how to distinguish a predicted hazard, a measured river response and a treatment outcome.
Incident record: keep the source and denominator attached
The 17 July 2012 interagency BAER report gives an incident start of 9 June, containment on 1 July, and 259 homes and cabins destroyed. Its watershed table uses 87,487 acres, including 14,072 classified as unburned inside that mapping footprint. [1]
A Forest Service bulletin published in May/June 2017 describes a lightning origin and reports 87,415 acres. These are source-defined figures; this record preserves them rather than averaging them into a new incident total. [2]
| Quantity | Calculation | Result |
|---|---|---|
| High soil severity / whole mapped footprint | 5,714 ÷ 87,487 × 100 | 6.53% |
| High soil severity / mapped burned classes | 5,714 ÷ (87,487 − 14,072) × 100 | 7.78% |
Tecnobosque calculation from the report’s figures, not a new field measurement. The high-severity acreage is unchanged; only the denominator changes. These percentages concern the report’s soil-severity classification, not building destruction or a universal vegetation-mortality threshold.
The 2012 debris-flow assessment was a scenario model
USGS assessed the drainage network and 44 selected basins along State Highway 14 and the burn perimeter. The models used terrain, soils, burn severity and specified rainfall. Published probability ranges describe variation across the assessed locations under each storm scenario. [3]
| Rainfall in one hour | Assigned rainfall recurrence interval | Modelled probability range |
|---|---|---|
| 25 mm | 2 years | 1–84% |
| 43 mm | 10 years | 2–95% |
| 51 mm | 25 years | 3–97% |
The rainfall recurrence interval is not a debris-flow recurrence interval. Nor does the largest probability become a single probability for the entire fire. This was a preliminary emergency assessment first posted 23 July 2012, not a count of debris flows subsequently observed.
To read any such map, retain the storm scenario and basin identity alongside the colored value. A number detached from those conditions answers a different question.
Source-water changes and drinking-water treatment are separate evidence
The openly available abstract of Writer and colleagues (2014) reports higher dissolved organic carbon (DOC) and disinfection-by-product formation in samples associated with thunderstorms and spring snowmelt than under base-flow conditions. Researchers chlorinated samples under controlled formation conditions and evaluated alum coagulation. This was a treatment experiment, not a direct measurement of every customer’s tap water. [4]
The authors report that Fort Collins expanded monitoring, used multiple water supplies and built a presedimentation basin, maintaining high-quality drinking-water delivery in the year after the fire. A disturbed river therefore does not, by itself, establish that unsafe water was delivered.
Two 2015 storms moved sediment differently
Wilson and colleagues monitored nested hillslopes and catchments in Skin and Hill Gulch. Their paper, published online in 2020 and in a 2021 journal issue, analyzes 8 July and 16 August 2015 events, the third summer after the fire. [5]
The first storm averaged roughly 11 hours at low intensity, following wet antecedent conditions. Hillslope sediment delivery was limited and channels commonly incised. The second averaged roughly one hour at high intensity, producing stronger hillslope-to-stream connections and deposition at many channel sites. Unconfined reaches reduced sediment delivery during the intense event.
These observations explain why “more rain” is an incomplete description: duration, intensity, earlier rainfall and channel geometry differed. The study also identifies measurement uncertainty at peak flows, including extrapolated rating curves and differences between sampling methods. Two events are not a universal response rule for every burned watershed.
Where did the fire-derived carbon go?
Cotrufo and colleagues (2016) tracked pyrogenic carbon through soils, suspended material and riverbank deposits using chemical markers. Storm transport moved material into low-density deposits; the major September 2013 rainfall event removed or buried deposits formed earlier. The study documents redistribution between storage locations, rather than equating every soil loss with immediate export from the whole watershed. [6]
The 2017 Forest Service bulletin also describes improved local Poudre water quality after the 2013 flood scoured problematic channel deposits. That local recovery observation belongs alongside the flood’s broader destructive effects. It does not establish that flooding generally improves water supplies or that displaced sediment disappears. [2]
For an evidence record, define the boundary of “recovery.” A cleaner reach, a stabilized hillslope and a treatment plant operating normally are different endpoints. Movement beyond one monitored reach cannot establish the fate of material farther downstream.
Worked example: lower concentration can accompany a larger load
The daily mass transported is concentration multiplied by water discharge and elapsed time. With concentration C in mg/L and discharge Q in m³/s:
Daily suspended-sediment load (kg/day) = 86.4 × C × Q
The conversion uses 1,000 liters per cubic meter, 86,400 seconds per day and 1,000,000 milligrams per kilogram.
| Quantity | Scenario A | Scenario B |
|---|---|---|
| Concentration | 10 mg/L | 6 mg/L |
| Discharge | 1 m³/s | 2 m³/s |
| Calculation | 86.4 × 10 × 1 | 86.4 × 6 × 2 |
| Daily load | 864 kg/day | 1,036.8 kg/day |
Scenario B has 40% lower concentration but 20% greater daily mass transport. The change in flow explains the difference. A concentration decrease alone therefore cannot establish a reduction in the amount transported.
Real storm concentrations and flows change through time. A single sample cannot be multiplied across an entire day without explicitly assuming it represents that period; estimating an event load requires observations through the event and accounting for their timing. The example predicts neither treatment performance nor the safety of finished drinking water.
Known At and evolution of knowledge
- 17 July 2012: BAER stabilization priorities and source-defined mapped classes. [1]
- 23 July 2012: USGS publishes conditional hazard estimates. [3]
- April 2014: the water-treatment study reports first-year responses. [4]
- 2016–2017: carbon-transport research and a five-year agency review expand the recovery record. [6] [2]
- 2020–2021: publication of nested monitoring interprets storms observed in 2015. [5]
Publication dates tell us when this evidence entered these documents. They do not establish when every responder, researcher or resident first knew it. A later explanation belongs in the retrospective record, with its observation dates retained.
FireDNA: a watershed-to-intake evidence chain
This profile organizes the case without assigning a severity rank:
- Landscape: keep soil condition, unburned areas and the mapping denominator explicit.
- Transfer: follow material from its source through channel storage to the monitored outlet.
- Timing: distinguish the fire, individual rain events, sampling campaigns and publication.
- Service: separate river conditions, laboratory treatment tests and utility delivery outcomes.
A useful monitoring record identifies the station, sample type, units, time, flow condition and treatment stage. Mark what was measured, modelled or calculated. That record supports comparisons between events without forcing unrelated quantities into one “recovery” score.
Limits of this reconstruction
This source-based historical study was reviewed on 9 October 2026. Tecnobosque did not collect field samples, rerun the USGS hazard models or reproduce treatment experiments. Claims about Writer et al. are limited to the publisher’s open abstract. Other research findings retain their sampling locations and incident period; they do not describe present-day water quality.
The two calculations are labeled separately: one changes the denominator of reported mapped classes, while the other uses hypothetical constant-flow conditions. Neither is a drinking-water compliance assessment, a treatment recipe or a new field dataset.
Official records and original research
- Interagency High Park BAER report (PDF). 17 July 2012. Summary and watershed-condition table, printed page 4.
- US Forest Service: Learn from the Burn (PDF). Science You Can Use Bulletin, May/June 2017, issue 25.
- USGS: High Park debris-flow scenario assessment. Verdin, Dupree and Elliott (2012), OFR 2012–1148. First posted 23 July 2012.
- Writer et al.: High Park water-treatment study. Journal AWWA 106(4), E189–E199 (2014). DOI: 10.5942/jawwa.2014.106.0055. Open abstract.
- Wilson et al.: nested runoff and sediment monitoring (PDF). Hydrological Processes 35, e13975 (2021; online 2020). DOI: 10.1002/hyp.13975.
- Cotrufo et al.: pyrogenic-carbon redistribution (PDF). Global Biogeochemical Cycles 30, 1348–1355 (2016). DOI: 10.1002/2016GB005467.
Sources accessed 9 October 2026. The published studies, agency reports and Tecnobosque calculations serve different evidence roles, identified above.