In the spring of 2021 the Colorado Department of Transportation put a temporary rockfall barrier and two new rockfall fences into Glenwood Canyon, at an estimated fencing cost of $2,692,000, after its own modelling showed the work would pay for itself several times over. That July, intense rainfall on the burn scar produced a debris flow that FHWA’s case study describes as “a greater magnitude than the fencing was designed to absorb.” It destroyed some of the fencing.
The sequence is worth following because it is the clearest documented account of what fire does to a highway, and of the order in which it does it.
The fire, then the slope
FHWA’s Emergency Relief case study records that on 10 August 2020 a fire broke out in the median of Interstate 70 in Glenwood Canyon. The Grizzly Creek fire “grew out of control, burning over 32,000 acres,” the canyon was evacuated, and the Interstate closed for two weeks while the fire continued to burn beside the road. Motorists needing to cross had “a several hundred-mile, 4-hour detour.”
The inventory of what burned is more varied than the phrase fire damage suggests. Flames caused significant damage to highway, bridges, fiber optic cable and signage. Then the slope answered. The fire “triggered extensive rockfall activity that damaged additional CDOT assets, including rockfall fencing, guardrails, roadway surfaces, utility conduit, bike path and pedestrian facilities, and drainage structures.” Damage from the fire and the debris flows that followed “required over $11 million in repairs.”
Roadside hardware is the category agencies most often underprice, because fire consumes it wholesale along a corridor rather than damaging it at a point, which makes the response a replacement program and not a repair. The design question behind that hardware is treated in guardrails and barriers.
Drainage is underpriced for a different reason, which is that the damage is buried. A peer-reviewed accounting of Oregon’s 2020 Labor Day fires documents, at Bad Banks Bridge on Oregon Highway 22, structural damage to the concrete inlet, wing walls and apron, to the culvert liner and to a retaining wall, along with galvanized coating melted off a manhole cover and melted HDPE diversion pipes. What that failure costs later is set by how highway drainage systems work.
Why the hazard peaks after the fire is out
The US Geological Survey states the mechanism briefly: “Vegetation is removed and soil properties change, reducing the ground’s ability to absorb rainfall.” What follows are “fast-moving mixtures of water, mud, rocks, and vegetation that can surge downslope like flowing concrete,” and USGS notes that these flows “often move faster than a person can run.” The trigger threshold is low. “Even modest rainstorms can trigger dangerous flash floods and debris flows in steep burned areas,” typically from “short bursts of intense rainfall,” and the hazard “can occur during the first storm following a wildfire.”
Glenwood Canyon then became one of the better-instrumented tests of how well that hazard can be forecast. A 2024 study in Natural Hazards and Earth System Sciences documented 40 debris flows in 25 drainages during the summer of 2021 and evaluated the USGS rainfall thresholds against them. The fire-wide 15-minute intensity threshold at 50 percent likelihood was 25.9 millimeters per hour in the first year after the fire. In year one the model held up: 89 percent of observed debris flows “were triggered by rainfall rates higher” than the threshold. For the second year, the operational threshold shifted to the USGS value at 75 percent likelihood, 33.7 millimeters per hour, and that threshold did not hold: eight storms exceeded it and produced no debris flows. The paper’s discussion attributes this to the 75 percent threshold, which it says “may have been too conservative for the second monsoon season.” An updated estimate for year two, using revised dNBR values, put the 50 percent threshold at 40 millimeters per hour. No debris flows were reported in 2022 at all, despite substantial rainfall. The companion volume model, which predicts how much material arrives, “overestimates for this region by a median value of 4.4 times.”
A forecasting tool that is right about timing in year one, over-cautious in year two, and off by a factor of four on volume is still the best available basis for closing a road. It is a poor basis for sizing a structure, which is where the fences come in.
A benefit-cost case that was correct and insufficient
CDOT worked its decision through its own Risk and Resilience Analysis Procedure, a manual for calculating risk to state assets “from flooding, rockfall, and fire debris flow.” Having a standard method already in place is what let the department produce an economic justification quickly, and it is the practice most worth copying from anything in the climate and resilience section. FHWA’s case study also notes one boundary in that manual without comment: “The manual does not consider how climate change may affect geohazard impacts.”
The numbers were not marginal. Owner consequences over one year were estimated at $7,478,838 without rockfall fencing and $1,000,923 with it, and over five years at $18,949,679 against $2,777,494, giving benefit-to-cost ratios of 2.41 at one year and 6.01 at five against fencing costed at $2,692,000. The assumptions are stated, including “an assumed 20% reduction every year in rockfall likelihood as slopes re-establish stability and vegetation over a 5-year mitigation period,” and one exclusion is stated too: “The costs reflect owner costs only, and do not consider user costs, such as detour impacts.” Given a four-hour detour, that exclusion is not small, and it runs the analysis against the department’s interest rather than for it.
CDOT deliberately narrowed its scope. It focused on small and medium rockfall events, “for which the agency determined mitigation strategies had a high benefit-to cost ratio,” and concluded that mitigating large events “would be very costly and have minimal benefits.” A temporary seven-foot barrier fence went in for the recovery period, plus a 1-kilojoule fence at milepost 121 and a 3-kilojoule fence at milepost 123. Then July arrived.
The totals do not agree, and both are honest
Anyone quoting a cost for Glenwood Canyon should say which cost. FHWA’s Emergency Relief case study puts the repairs at over $11 million. The NHESS study reports that “Initial repair costs for the highway were more than USD 50 million” with overall road infrastructure repair costs estimated at “USD 116 million,” figures the paper itself attributes to news reporting rather than to its own analysis. These are not competing estimates of the same quantity. They are different boundaries drawn around a multi-year recovery, and the gap between them is where user costs, later phases and non-federal work sit.
Oregon shows the same pattern at state scale. The Oregon Department of Transportation’s wildfire fact sheet lists the 2015 Canyon Complex Fire at 110,000 acres burned and “Approximately $5M in costs to ODOT,” the 2017 Eagle Creek Fire at 48,000 acres and “Approximately $20 Million in costs to ODOT,” and the September 2020 fires at over 1 million acres burned with “Costs could exceed $1 Billion.” The peer-reviewed reconstruction of the highway repair bill for four routes affected in 2020, covering Highways 224, 22, 126 and 138 across 307 kilometers of closures, arrives at $44,894,471 in combined temporary and permanent repair, with hazard tree removal at $5,000,000, slope and rock scaling at $2,112,200, structural damage at $6,216,640, traffic control at $3,172,310, and pavement damage at $428,660. The authors note that missing data constrained their comparisons across corridors.
The contemporaneous agency figure and the later route-level accounting are an order of magnitude apart because they count different things, and the smallest line in the peer-reviewed table is the pavement. On these fires the road surface was the cheap part. Trees, slopes and structures were the bill.
What fire-hardening can and cannot reach
ODOT’s own list of physical measures is short and it opens with a concession: “Fortifying a highway against a large wildfire is difficult.” What it offers is fire-resistant materials for hardware such as traffic control devices and guardrail posts, more vegetation management and erosion control, and proactive tree removal along certain routes. The barriers it names are structural rather than technical. Unlike a landslide, “large portions of the highway system can be affected by a single wildfire event,” which means measures have to be applied network-wide, “some highway segments could be closed for months,” and “Federal aid often comes after the fact.”
Pavement itself is more resistant than intuition suggests, with one exception that cuts against current practice. FHWA’s 2023 review of pavement resilience reports work showing that asphalt pavements can burn in extreme fire scenarios but that common asphalt types “can be considered ‘intrinsically flame-resistant materials,’” while polymer-modified asphalt “shows a ‘pronouncedly greater fire risk.’” Polymer modification is exactly what a warming climate argues for on the heat side of the ledger. The same review also flags a loading problem that has nothing to do with flame: one of the most noted impacts on rural local roads is the weight of firefighting traffic and post-fire debris removal, which “could potentially reduce pavement life.”
Which brings the argument back to the destroyed fences, and to a judgement worth defending rather than mocking. CDOT sized its structures for the events its own manual could price, using a procedure that excluded climate change and excluded user cost, and it declined to build for large events on explicit benefit-cost grounds. A flow larger than any of that arrived within months. The department then set out to repair the fencing for the small and medium events anyway, which remains the right answer on its own numbers and is not protection against the thing that broke it.