Wildfire and Source Water: What the Evidence Says

Forested watersheds are the largest single source of surface drinking water in the United States. They cover about a third of the land but supply roughly half the nation's surface water, feeding the intakes that serve some 150 million people. So when one burns, it becomes a water-treatment problem: the canopy and litter layer are gone, the soil sheds water instead of absorbing it, and the next storm carries ash, sediment, nutrients, and organic matter toward the intake.

That much is well established. The harder question, the one a source-water analyst actually has to answer, is narrower: which of my intakes should I worry about, and what should I be sampling?

The literature answers that question in a way that is more useful, and less comfortable, than the headlines suggest.

Across the western United States, fewer than a third of burned watersheds show a detectable water-quality response.

• Rust et al. (2018) compiled 24,042 fires and analyzed 159 of them across 153 burned watersheds. Altered post-fire water quality was documented in about 29% of the analyzed watersheds in the first five years.
• Brucker et al. (2025), working across 245 burned basins, found the share of individual basins with statistically significant post-fire change ranged from 17% for total suspended solids to 45% for dissolved organic carbon, with roughly 20 to 30% typical for most constituents.
• A global review (Paul et al., 2022) found nitrogen increased in 77% of measurements but decreased in 7% and did not move in 16%. Phosphorus increased in 65% and decreased in 27%. Organic carbon increased in only about half of the studies that reported it.

This is not a reason to relax. It is the entire reason triage matters. If every burned watershed produced a treatment crisis, no one would need to prioritize. The response is real, it is sometimes enormous, and it is heterogeneous. That makes the operational problem not "does wildfire degrade water quality" but "does this fire, above this intake, on this ground."

Post-fire degradation is episodic, not a steady shift in the mean. In the Brucker et al. (2025) analysis, precipitation and runoff were the dominant explanatory covariates for every constituent in every region. The signal is delivered by storms.

Aggregated across burned basins, mean turbidity rose by as much as 4,420%. During peak response years, sediment ran 19 to 286 times pre-fire concentrations, and carbon, nitrogen, and phosphorus ran 3 to 103 times pre-fire levels.

Those are event-scale numbers. A monthly grab sample on the same Tuesday will not see them. The first significant rain on a fresh burn scar is a sampling event, not a weather event.

Brucker et al. tracked responses out to eight years, controlling for background hydroclimate by training basin-specific regression models on pre-fire data and reading the post-fire residuals. Averaged across burned basins, the elevated windows ran:

• Organic carbon: 2 to 8 years.
• Inorganic nitrogen: 5 to 8 years.
• Phosphorus: 2 to 3 years.
• Sedimentation: 4 to 8 years.
• Total dissolved solids and turbidity: 5 to 8 years.

Two things need saying plainly rather than smoothing away.

The first is the honest limit on those windows. For most constituents, responses declined and fell below the level of statistical significance after two to five years. Nitrogen and sediment are the exceptions: dissolved inorganic nitrogen stayed significantly elevated for five years and nitrate for the full eight, while sedimentation stayed elevated for four to eight years. The eight-year figures are also bounded by the study window rather than by an observed return to baseline. Only eight post-fire years could be analyzed, while other work suggests effects can persist for decades.

The second is that sediment and dissolved constituents run on different clocks. Rust et al. (2018) found that dissolved ions and metals tended to decline by year five while particulate matter kept climbing. Brucker et al. found total suspended solids peaked in year one, but suspended sediment concentration and discharge peaked in year four. The worst sediment year may not be the first one.

The long-nitrogen result is corroborated at single-fire scale. After the 2002 Hayman Fire in Colorado, Rhoades et al. (2019) found stream nitrate and total dissolved nitrogen elevated across 14 years of monitoring relative to both pre-burn conditions and unburned control catchments, with nitrate more than an order of magnitude higher in catchments that burned across more than 60% of their area, and mean annual nitrate export roughly 19 times higher than in unburned streams. A 2026 follow-on (Rhea et al.) sampled soil and vegetation 17 years after the fire and found the elevated export persists, driven by suppressed vegetation nitrogen demand rather than a rise in mineralization.

The most operationally direct evidence comes from EPA. Pennino et al. (2022) went to the Safe Drinking Water Information System and measured actual MCL violations at public water systems downstream of major wildfires. Post-fire, on average:

• Nitrate (surface-water sourced): 0.56 more violations per system, a 0.044 mg-N/L concentration increase, with more violations at 75% of sites.
• Total trihalomethanes: 0.58 more violations per system, a 10.4 µg/L increase, at 71% of sites.
• Haloacetic acids (HAA5): 0.82 more violations per system, an 8.5 µg/L increase, at 50% of sites.
• Arsenic (groundwater sourced): 1.08 more violations per system, a 0.92 µg/L increase, at 35% of sites.

TTHM and HAA5 concentrations stayed generally elevated for five years. These are violations in finished water at real systems, not modeled outputs.

The capital side is just as long-lived. The 1996 Buffalo Creek and 2002 Hayman fires together delivered more than a million cubic yards of sediment into Strontia Springs Reservoir, through which more than 80% of Denver Water's supply passes. Denver Water reports spending more than $27 million across the two events on water-quality treatment, sediment and debris removal, reclamation, and infrastructure. A single dredging campaign in 2010 recovered less than half its target volume at a cost of $18.5 million.

Severity, mostly, and where it sits.

• Rust et al. (2019) found nitrate and organic nitrogen correlated with the percent of basin area burned at moderate and high severity over the first five years post-fire. Total phosphorus was correlated with moderate-to-high severity burned area in years three through five.
• Beyene et al. (2023), assessing 54 wildfires against 11 prescribed fires, found that large, high-severity wildfires significantly raised spring trace-element concentrations, most consistently arsenic, with more variable responses in selenium and cadmium. Prescribed fires did not. Fire alone is not the driver. Severe fire is.
• Brucker et al. (2025) add a useful corrective: basin land cover explained more of the between-basin variability than burn extent or proximity did. Forest coverage alone explained up to 31% of the variance in the DOC response.

The through-line is that "10% of the basin burned" is a weak predictor on its own. What was burning, how severely, on what soil, how steep, and how well connected it is to the channel network above the intake all carry information that a burned-area percentage does not.

Most of the literature above was written from the interior West and California. A 2026 multi-agency review of the western Pacific Northwest (Wall et al., with authors from the US Forest Service, EPA, USGS, Oregon DEQ, Oregon State, Washington State, NOAA, the Eugene Water and Electric Board, and industry) states directly that the region's uniquely wet, highly seasonal climate means findings from other, more frequently burned regions may not be directly applicable.

Three findings from that review deserve a wider audience.

The fire regime changed almost overnight. The 2020 Labor Day fires burned roughly 334,500 hectares in a few weeks, about 83% of the total area burned in the western PNW over the previous 30 years combined. Fires between 2021 and 2024 burned a further 1.8 million hectares. Historical fire return intervals in the Coast Range and western Cascades ran 150 to over 400 years, with some evidence suggesting up to 1,000.

The treatment infrastructure was built for the old regime. Of PNW utilities serving more than 10,000 people from surface water, 67% have surface water only and no groundwater fallback. Many PNW sources have historically carried so little sediment and organic matter that they need only minimal treatment, including unfiltered and slow-sand-filtration systems. Those systems have lower thresholds for source-water turbidity and DOM and are less resilient to exactly the fluctuations a fire produces. As the authors put it, many of these utilities were simply not designed to treat the impacts of wildfire on source water.

The monitoring network is pointed at the wrong places. Monitoring is concentrated in the low-elevation Puget Lowlands and Willamette Valley, where the people are. Fire burns higher and further east. Streamgages overrepresent large rivers and underrepresent the small headwater streams most likely to burn, and to burn hot. Fewer than 3% of continuous USGS gages in Oregon and Washington measure anything beyond hydrology and the "Big 5" of temperature, turbidity, specific conductance, dissolved oxygen, and pH. No published studies were found examining post-fire changes in conductivity, metals, dissolved oxygen, or pH anywhere in the region.

And one finding cuts against the easy story: bulk dissolved organic carbon may not spike here the way it does elsewhere. The review reports minimal DOC shifts in western PNW watersheds after the 2020 fires, and in one case a decrease with increasing burn severity. What did shift with severity was DOM composition, toward more aromatic, polycondensed, nitrogen-containing compounds. For a plant, that distinction matters: the disinfection-byproduct question survives even when the DOC number does not move.

Here is the finding that reframes everything above.

In the first two years after the 2020 Labor Day fires, the hydrologic and geomorphic response in western Oregon was muted. Busby and Wilcox (2024) attribute this to below-average precipitation, an absence of extreme storms, and rainfall intensities that stayed below debris-flow-triggering thresholds. Peak flows in the first two post-fire years stayed below the two-year recurrence interval in both the Holiday Farm and Archie Creek study areas.

Severely burned, steep, connected watersheds sat directly above drinking-water intakes, and the storm that would have converted that exposure into a treatment crisis did not arrive.

That is luck, not resilience. And it is precisely why an observed record is a poor place to start. A quiet monitoring record after a fire tells you what the weather did. It does not tell you what the watershed is capable of doing when the weather is less kind.

This is also why we are careful with our own flagship numbers. After a fire it is tempting to point at a single dramatic turbidity peak, as the North Umpqua produced after the 2020 Archie Creek Fire, and treat it as proof. But a lone peak cannot by itself separate a fire effect from a large storm, and the most extreme readings in a burned watershed usually come from small headwater tributaries rather than the regulated intake downstream. Attribution needs flow-normalization or the kind of multi-basin, hydroclimate-controlled design Brucker et al. used. An observed peak is a reason to look, not a proof on its own.

Pulling it together, the evidence points at four properties.

• Event-triggered, not calendar-triggered. Sample the first flush and the first significant storms.
• Multi-year, not single-season. Brucker et al. recommend utilities prepare for one to eight years of elevated loads, with potential increases up to roughly 300 times pre-fire levels. They note that sedimentation basins may be needed to manage eight or more years of post-fire sediment, while increased coagulant dosing may address the DOM response.
• Constituent-prioritized. Turbidity, DOM and disinfection-byproduct precursors, nutrients, and metals do not matter equally in every burn. Which ones matter here depends on the drivers in this basin.
• Prioritized across intakes. Because most burned basins will not respond, and a few will respond dramatically, the scarce resource is not sampling capacity. It is knowing where to point it.

Brucker et al. reach the same conclusion from the data: the variability in post-fire response underscores the importance of localized, basin-by-basin assessment.

The teams carrying this risk are frequently the least equipped to model it: small surface-water systems, watershed councils, and rural districts with thin budgets and no in-house GIS or hydrology capacity. The evidence tells them they need site-specific, event-timed, multi-year monitoring. It does not tell them which intake to look at first.

That gap is why we built PWSW. It takes a fire perimeter and an intake or assessment point and returns a fire-specific monitoring brief: which constituents to expect, and where and when to sample, plus an at-a-glance map and GIS layers. It is free, open source under Apache-2.0, and runs locally on free US public data.

To be exact about what it is: PWSW is a screening and monitoring-design aid. It reports a relative vulnerability rank, calibrated to rank-agree with the USGS Staley et al. (2017) M1 post-fire debris-flow method. It does not predict turbidity, DOC, or any concentration, and it is not a regulatory or compliance model. It screens for exposure before the storm. It does not tell you what the water will do.

See a full run, from a fire perimeter to a monitoring brief, in the PWSW walkthrough.