Roads, Roadless Areas, and Wildfire Across Ten Western States

The 2001 Roadless Area Conservation Rule restricts road construction across roughly a quarter of the National Forest System, and the debate over it turns in large part on fire. Advocates for road access argue that roads let crews reach and contain fires quickly. Advocates for roadless protection argue that roads bring more human ignitions. Both claims can be true at once, and federal fire data allow them to be tested directly.

This study uses the Interagency Fire Occurrence Reporting Modules (InFORM) Fire Occurrence Data Records — the authoritative interagency wildfire point dataset — to examine every recorded, sized wildfire that ignited on Forest Service land in ten Western states from 2001 through 2024: Alaska, Arizona, California, Montana, Nevada, New Mexico, Oregon, Utah, Washington, and Wyoming. That is 109,696 fires. We classify each ignition by its distance to the nearest road, by whether it fell inside an Inventoried Roadless Area or designated wilderness, and by its reported cause, then examine how fire size varies across those settings, with particular attention to fires caused by equipment and vehicle use.

Section 1

Background and Thesis

Roads play a documented role in fire management. They provide access for suppression crews and engines, serve as ready-made containment lines and fuel breaks, and give incident managers safe anchor points for burnout operations. A substantial body of research also finds the opposite effect on ignition: wildfire ignitions rise with proximity to roads, chiefly because roads concentrate the human activity — discarded cigarettes, dragging trailer chains, hot exhaust, escaped debris burns, arson — that starts most fires.

A 2026 study in Fire Ecology by Aplet, Hartger, and Dietz examined this across all eight contiguous-U.S. Forest Service regions using the same InFORM dataset. It reported that ignition density was far higher within 50 m of roads (7.99 fires per 1,000 hectares) than in Inventoried Roadless Areas (1.97) or designated wilderness (1.75), and that mean fire size ran in the opposite direction, largest in wilderness (239 ha) and smallest in the roadside zone (49 ha). The authors concluded that building roads into roadless areas would produce more fires without meaningfully reducing the likelihood of a large one.

Our aim is complementary. We reproduce the spatial classification at point level for ten Western states, and extend it in one direction the national study did not pursue: we isolate the equipment and vehicle use cause category, the ignitions most directly tied to motorized access. The thesis we test is threefold — that ignitions concentrate near roads; that fires igniting near roads are contained at smaller sizes than those in roadless and wilderness settings; and that equipment- and vehicle-caused fires, though road-associated, follow the same containment pattern where road access exists.

Section 2

Data Sources

• Wildfire ignitions: InFORM Fire Occurrence Data Records (National Interagency Fire Center), point locations with reported cause and final incident size.
• Roads: National Forest System Roads open to motor vehicles (operational maintenance levels 2–5; USDA Forest Service EDW), combined with U.S. Census Bureau TIGER/Line primary and secondary roads for 2024.
• Roadless areas: Inventoried Roadless Areas as defined under the 2001 Roadless Rule (USDA Forest Service EDW).
• Wilderness: National Wilderness Areas (USDA Forest Service EDW).

These are the same core layers used by Aplet et al. (2026), with one deliberate difference in the road network noted under Limitations.

Section 3

Methodology

Fire records

We queried InFORM for every fire with point-of-origin landowner category USFS, in one of the ten states, with a calendar year from 2001 through 2024 and a non-null incident size. The 2001 start aligns the window with the era of the Roadless Rule. Using point-of-origin ownership rather than a spatial forest-boundary intersection is a simplification discussed under Limitations.

Cause coding

We grouped fires by InFORM’s top-level cause field as Human, Natural (lightning and other natural sources), or Undetermined, and used the general-cause field to isolate Equipment and vehicle use as a distinct subset of human-caused fires. That category spans passenger vehicles, off-highway vehicles, heavy equipment, chainsaws and other small engines, exhaust particles, brakes, and related mechanical and friction sources.

Spatial classification

All layers were projected to an equal-area coordinate system in meters — NAD83 Conus Albers for the contiguous states, and NAD83 Alaska Albers for Alaska, so distances are computed accurately at Alaskan latitudes. For each fire we computed the straight-line distance to the nearest road in the combined network and tested whether the point fell inside an Inventoried Roadless Area or a wilderness polygon. Following the published method, we assigned each fire to one of four mutually exclusive zones:

• Road buffer — within 50 m of a road;
• Roadless / IRA — more than 50 m from a road and inside an Inventoried Roadless Area;
• Wilderness — more than 50 m from a road and inside designated wilderness;
• Other NFS land — more than 50 m from a road and outside both designations.

We also binned every fire into 250-meter distance-to-road classes out to 2,000 m, with a final class beyond 2,000 m.

Fire size and the “escape” measure

Fire size is the reported final incident size in acres. Because the distribution is extremely right-skewed, we report the median alongside the mean and, as the primary comparison, the escape rate: the share of fires reaching 300 acres or more (National Wildfire Coordinating Group size class E and above), a proxy for fires that escaped initial containment. The full point-level classified dataset is available on request so any figure below can be reproduced or re-cut.

Section 4

Results

4.1  Ignitions concentrate near roads

The distance gradient is immediate. The single largest group of fires ignited within 250 meters of a road, and counts decline at every step outward.

This is a statement about counts, not density: because the roadside zone is a small share of the land base, the ignitions-per-acre concentration near roads is steeper still. We report counts rather than density because we did not compute zone areas, a distinction noted under Limitations. The share of fires within 250 m of a road ranges from about 30% in the more remote forests of Montana, Nevada, and Wyoming to roughly 45–51% in California, Arizona, and Oregon.

4.2  Fire size and escape rise with distance from roads

The size pattern runs opposite to the ignition pattern. Fires that began near roads were the most likely to be held small; fires that began in roadless and wilderness country escaped to large size markedly more often. Pooled across the ten states, the escape rate climbs steadily across the four zones.

That the median is identical at 0.10 acres across every zone — matching the roughly 0.04-hectare median Aplet et al. found nationally — is the clearest single confirmation that the great majority of fires are extinguished or self-extinguish while very small, wherever they start. What changes with setting is the rate at which the exceptions get away.

The gradient holds when the states are examined one at a time. The escape rate is lowest in the roadside zone and highest in roadless or wilderness settings in nearly every state, and is most pronounced in the large mountain wilderness of Montana, Wyoming, and Arizona.

4.3  Equipment- and vehicle-caused fires

These ignitions are the most road-bound of all. Across the ten states, 63.7% of the 10,688 equipment/vehicle fires started within 250 meters of a road. Where they started near roads, they were contained far more effectively than the minority that started deeper in the backcountry.

The cause most tightly bound to motorized access is also, where that access exists, among the most reliably contained.

Section 5

Conclusions We Can Draw

Two findings hold consistently across ten states and nearly 110,000 fires. First, wildfire ignitions are concentrated near roads, and human causes — including equipment and vehicle use — drive that concentration. Second, fires that ignite near roads are held to smaller sizes and escape initial containment substantially less often than fires that begin in roadless and wilderness areas, where the escape rate runs roughly two to four times higher.

Roads are associated with more fires but smaller ones; roadless and wilderness settings with fewer fires but a higher share that grow large. Neither pattern alone settles the policy question, and honest analysis has to hold both.

For the equipment- and vehicle-caused fires most directly tied to motorized recreation and access, the containment signal is especially clear. These ignitions overwhelmingly occur near roads, and near roads they are caught small; it is the smaller share starting away from road access that account for the larger average sizes and higher escape rates. This is consistent with roads functioning as both detection corridors and suppression-access infrastructure.

These results align closely with the published national analysis and with earlier work by Johnston et al. (2021), which likewise found roadless areas experiencing fewer ignitions but a comparable number of large fires that ran roughly a third larger on average. Our contribution is to confirm the size gradient at point level across ten Western states and to show that it holds for the equipment/vehicle cause specifically.

The interpretation requires care

The association between road access and smaller fires does not, by itself, prove that roads cause containment. At least two mechanisms are at work and this analysis cannot fully separate them. One is physical: roads provide access for engines and crews and serve as fuel breaks and control lines, so fires near them are reached and stopped sooner. The other is managerial: incident managers deliberately allow many fires in remote, low-value roadless and wilderness terrain to burn for resource benefit or when suppression resources are stretched, which inflates average sizes in exactly those settings independent of any access effect. The larger fires in roadless and wilderness zones therefore reflect a blend of harder access and intentional management choice, and the InFORM fields used here do not distinguish the two.

The honest reading is that the data support a nuanced, zone-based view of fire and access rather than a blanket case for or against roads. Where values at risk are high, road access is associated with keeping fires small, and expanding or retaining that access has a defensible fire-management rationale. Where the objective is to let fire play its ecological role at low risk to people and infrastructure, roadless and wilderness areas are where larger fires already occur, partly by design. Any decision to build or retire roads carries a real ignition cost on one side and a real containment and access benefit on the other, and the balance is properly struck locally rather than by a single national presumption.

Section 6

Limitations and Weaknesses

We state these plainly, because a study that names its own weak points is more useful than one that hides them.

• Land selection is attribute-based. We identified Forest Service fires using the InFORM point-of-origin ownership field rather than spatially intersecting each point with the National Forest System Lands boundary. This is close to, but not identical with, the published method, and ownership is coded inconsistently on a share of records.
• The road network omits minor local roads. We used Forest Service maintenance-level 2–5 roads plus TIGER primary and secondary roads, but not the full set of TIGER local and residential roads that the national study included. Some fires we classify as far from a road may sit near an unmapped minor road, which would tend to overstate distances in the remote zones.
• Counts, not density. We measured each fire’s distance to the nearest road but did not compute the area of each zone, so we report fire counts rather than ignitions per unit area. Our results confirm the direction of the national density finding but do not reproduce its per-hectare rates.
• The escape threshold is a convention. Defining escape as 300 acres or more is a reasonable but arbitrary cutoff; a different threshold would shift the absolute rates, though the ordering across zones is stable.
• Some cells are small or noisy. Alaska contributes only 302 fires and is effectively a null case; Nevada, Wyoming, and several equipment/vehicle cells are small enough that individual rates carry wide uncertainty, and single large fires can move a state’s mean. The very high escape rates that appear for “undetermined” fires elsewhere in the data reflect selection bias — undetermined fires are more likely to be investigated when they grow large — and we do not treat them as a real property of that category.
• Border effects. Roads were gathered by state bounding box; a fire near a state line is measured against roads in the same box, which is accurate in practice but not a formal buffer around each point.
• Association is not causation. As discussed above, the road–size relationship is confounded by deliberate management decisions to let remote fires burn, which this analysis cannot isolate.
• Ten states, not the nation. These states span very different vegetation, terrain, and fire regimes; results should not be generalized beyond them without repeating the analysis elsewhere.