PNW Post-Fire Source Water: The 2026 Burn Scars Ahead

This post has two halves. The first is for anyone who lives in the Pacific Northwest and wants to know whether the fires are going to affect their drinking water. The second half is the engineering analysis, written for the people who have to design a monitoring program before the autumn storms arrive.

The quick version: Seattle is not on fire. Neither is Portland. What you have been breathing is smoke that blew in from fires burning mostly east of the Cascades, plus some from British Columbia. Nothing is burning in the watersheds that supply Seattle, Tacoma, or Portland with drinking water.

The fires themselves are real and they are big. Oregon has burned roughly 2.3 million acres this year, a state record. Washington was at about 427,000 acres at the end of July, already past everything it burned in all of 2025, and it has climbed since. Most of it sits east of the crest: the Okanogan, Lake Chelan, the Yakima, the hills above Spokane, and just outside Boise.

Here is the part most coverage skips. The water problem has not started yet.

Fire by itself does not do much to a river. Rain on a burned hillside does. Burned soil loses much of its ability to soak up water, so the first serious storms wash ash, silt, and charred debris straight into creeks. If your town's water comes from one of those creeks, that means a slug of muddy, hard-to-treat water arriving all at once. The first real storms come in autumn, so we are two or three months out from finding out how bad this gets.

One place is already close. The Grasshopper Fire, on the far side of Mount Hood, reached the edge of the watershed that supplies The Dalles in early August. As of August 10 it had not crossed, and no advisory had been issued.

It also lasts longer than people expect. A 2025 study comparing 245 burned watersheds against 293 unburned ones found sediment and nitrogen still elevated up to eight years after a fire. Not weeks. Years.

Three things are worth reading if you want to go further.

• Wildfire effects on source-water quality, a USGS fact sheet on Colorado's Fourmile Canyon fire. Four pages, plain English, and it nails the key idea: gentle snowmelt is usually fine, hard storms are not.
• Wildfire and the future of water supply (Bladon and colleagues, 2014). Short, and the piece that pushed water utilities to treat fire as their problem too.
• Post-wildfire water quality in the Pacific Northwest (Wall and colleagues, 2026). Covers western Oregon and Washington, and is honest about how little we know here compared to Colorado or California.

Three things are worth checking, and the third is the one almost nobody thinks of.

• Where the fires are. NIFC National Fire News is the official daily national summary, updated each morning, listing every large fire by state with acreage and containment. For a map, NIFC's wildland fire open data site publishes current incident locations and perimeters. In Washington, DNR's wildfire portal is the state-level equivalent. Note that InciWeb, the site most news coverage links to, blocks traffic from outside the United States, so use the NIFC pages if you are reading from abroad.
• Smoke near you. The AirNow Fire and Smoke Map, or the agency smoke blogs for Washington and Oregon. They post plain-language forecasts, not just numbers.
• Your own water. Your utility publishes an annual report that names its source. Look yours up, then check whether a burn scar sits upstream. If one does, watch for notices this autumn. On a well or a small system, that is worth a phone call now rather than in November.

Short version: the smoke is the visible part, and it will clear. The water part is slower, quieter, and arrives with the first rain.

The rest of this post is the engineering version of the same argument, for the people who have to act on it. Wildfire and Source Water: What the Evidence Says covers what the literature says in general. This one is about the 2026 Pacific Northwest specifically.

The 2026 Pacific Northwest fire season is the region's largest in the modern record. As of August 10, the Northwest Geographic Area was at Preparedness Level 5 with 46 uncontained large fires out of 98 nationally, and the national year-to-date total stood at 46,064 fires and 6,354,141 acres. Oregon has burned roughly 2.34 million acres per the Northwest Interagency Coordination Center, passing its 2024 record of 1,797,796 and setting a new one. Washington had burned more than 1.7 times its entire 2025 season by July 31, with two months of fire season remaining.

Those numbers describe an emergency-response problem that is already well covered. They do not describe the engineering problem, which has not happened yet. Post-fire source-water impacts are driven by rain on burn scars, and the region's first significant runoff arrives with the autumn storm season, eight to sixteen weeks out. The window for designing a response closes before the first storm.

LYNXCE's position: 2026 has created a post-fire source-water exposure the region has not had to manage at this scale, and the tools for reasoning about it have never been tested here.

The most widely reported water angle this season has been Seattle smoke. It is the wrong angle, and correcting it is the first step in scoping any post-fire work.

Cross-checking the National Interagency Fire Center national fire report against the Washington Department of Natural Resources wildfire portal: no large fire is burning in the Puget Sound lowland counties, and none in any westside municipal source watershed. Seattle Public Utilities' Cedar River Municipal Watershed and the South Fork Tolt are unburned, as is Tacoma's Green River and Howard Hanson supply. Portland's Bull Run, which burned in the 2023 Camp Creek Fire, is unburned in 2026; the Grasshopper Fire's growth has been north and northeast, away from it.

The map below makes the pattern hard to argue with. Plotting every 2026 perimeter over 400 acres against the Cascade crest puts the scars almost entirely on the east side, while the four westside supply basins sit in unburned terrain. The one place the two come close is Mount Hood, shown in the inset.

Two exceptions are easy to misread. Border 2 (10,582 acres) and Luna (3,583 acres) are burning in northern Whatcom County inside North Cascades National Park, in the upper Skagit, a basin that feeds Seattle City Light's hydropower dams and not Seattle's drinking water. Three Queens, near Snoqualmie Pass, sits east of the crest in the Cle Elum drainage, not the Cedar River basin.

The systems that do have scars upstream are smaller and have less technical capacity. Acreage and containment below are from the NIFC national fire report of August 10, 2026.

• Okanogan River: Sinlahekin, 141,411 acres, 42 percent contained.
• Wenatchee, Entiat, and Lake Chelan: Little Giant, 111,429 acres, 2 percent contained.
• The Dalles Municipal Watershed: Grasshopper, 57,087 acres, 23 percent contained.
• Methow: Ptarmigan, 22,682 acres, 8 percent contained.
• Upper Yakima at Cle Elum: Three Queens, 2,000 acres, 0 percent contained.
• Spokane River and the Spokane Valley-Rathdrum Prairie aquifer: Autumn Lane, 5,776 acres, 73 percent; Old Trails, 3,176 acres, 53 percent; Fairview, 992 acres, 53 percent.
• Boise River and the Boise Foothills: Claremont, 6,606 acres, contained and off the active list.
• Brownlee Reservoir: Tartar, 158,027 acres, 89 percent contained.
• Owyhee: Big Grass, 541,079 acres, 41 percent contained.

One fire post-dates the inventory above and deserves watching rather than tabulating. Wrights Spring, in the Fremont-Winema National Forest about two miles west of Sprague River, Oregon, reached 33,970 acres at 0 percent containment on August 10 with extreme fire behavior, evacuations, and road closures reported. Its contributing area has not been assessed here, and assigning it to a water system before that work is done would be exactly the mistake this section is about.

Grasshopper deserves naming. It reached the boundary of The Dalles Municipal Watershed on August 5, and its subsequent growth ran north into the adjacent Fivemile and Fifteenmile Creek headwaters. The city's public works director described it then as not inside the watershed but right at its edge. No advisory has been issued, and no source confirms it has entered. The Spokane Complex, meanwhile, destroyed 846 structures over a sole-source aquifer.

Burn-scar geography, not acreage, determines exposure. Two basins can burn the same fraction and produce entirely different consequences depending on whether the severe burn sits on steep, connected, upstream terrain or on flat ground near the outlet. Any screen ranking by percent-of-basin-burned gets this wrong.

Oregon's April 1, 2026 statewide snow water equivalent was the lowest on record for the SNOTEL network back to 1981 and for the 98-year snow-course record. Basin values ran from 1 percent of median in the Malheur to 9 in the John Day, and 61 of 73 forecast points projected April to September runoff in the lowest 15 percent on record. Washington's was 52 percent of normal.

The important detail is what did not happen. Oregon's water-year precipitation was 85 percent of normal, and Washington's near or above. This was a temperature-driven snow drought, not a precipitation drought. Precipitation largely arrived. It arrived as rain, ran off early, and left soils and fuels drying from a much earlier start date. That is a different antecedent condition from a dry year, and a different soil-moisture state going into a burn.

By early August, drought coverage in the D1 through D4 categories reached 63.7 percent of Washington, 91.0 percent of Oregon, and 77.5 percent of Idaho, with Oregon registering its first exceptional drought of the season. Lightning on July 15 and July 22 ignited more than 600 fires within ten days, and a dry cold front on July 25 drove more than 300,000 acres of additional growth. Above-normal fire potential is forecast through September, so the inventory above is a floor.

The literature has moved substantially in the last two years, and the direction of movement is toward longer effects than older reviews implied.

The most important recent result is Brucker et al. (2025), which compared 245 burned against 293 unburned western US watersheds from 1984 to 2021 across twelve constituents. During peak response years, organic carbon, nitrogen, and phosphorus reached 3 to 103 times pre-fire concentrations and sediment 19 to 286 times, with turbidity up to 4,420 percent higher in the most affected basins. Carbon and phosphorus stayed significantly elevated for one to five years; nitrogen and sediment for up to eight. Earlier literature implied recovery within one to three years, and monitoring scoped on that assumption closes before the signal does.

Shakesby and Doerr (2006), still the field's anchor review, documents post-fire increases in peak hydrologic response of up to two orders of magnitude. Ebel and Moody (2017) sharpened the mechanism in a way that should change modeling practice. Across a compiled dataset, burned soils showed significantly reduced sorptivity and wetting-front potential, while field-saturated hydraulic conductivity was not significantly different from unburned soils. Reducing saturated conductivity to represent a burn is adjusting the wrong parameter. Capillarity is what changes, and ponding begins sooner.

The treatment consequence is concrete. Hohner et al. (2016), on Colorado's High Park Fire, found post-wildfire intake water required a median of 7.5 milligrams per liter more alum than reference water, and post-rainstorm samples achieved under 10 percent dissolved organic carbon removal even above 65 milligrams per liter. Uzun et al. (2020) adds the byproduct dimension: elevated bromide shifted disinfection-byproduct speciation toward brominated species for up to two years.

Writer and Murphy (2012) stated the asymmetry in four pages: snowmelt-driven increases stayed within the treatment capacity of most plants, while high-intensity storms drove turbidity, carbon, nitrate, and some metals up by one to four orders of magnitude. That is the design constraint. Post-fire risk lives in peaks and first-flush timing. Medians barely move, and calendar-based sampling cannot catch it.

When a fire is contained, the instrument agencies and consultants reach for is the USGS post-fire debris-flow hazard assessment, built on Staley et al. (2016) and the rainfall thresholds in Staley et al. (2017). It is the correct tool, with a provenance worth stating out loud.

The model was fit to 1,550 records across 34 fires: 939 training records from southern California, and 611 test records from Montana, New Mexico, Colorado, Arizona, and Utah. No Pacific Northwest fire appears in either dataset. Every post-fire threshold product issued for a 2026 Oregon, Washington, or Idaho burn scar is an extrapolation beyond the model's validated domain.

That is not an argument for discarding it. It remains the best-calibrated instrument available. It is an argument for stating the extrapolation in a deliverable rather than letting a reviewer assume it was tested here.

The regional literature complicates this rather than resolving it. Wall et al. (2026), a 22-author synthesis, argues that findings from more frequently burned regions may not apply here, and that post-fire water quality in the region remains understudied. But that paper's study region is bounded to forested areas west of the Cascade crest, and every 2026 scar above is east of it.

The eastside may in fact behave more like the interior West. Wall, Roering and Rengers (2020) groups the Interior Northwest with southern California and the Rockies as regions where runoff-initiated post-fire debris flows are widely documented, attributing their rarity west of the crest to long-duration, low-intensity rainfall that rarely exceeds soil infiltration rates.

So the honest position is narrower than the one the region usually states. The problem is not that interior-West science fails here. It is that nobody has established which Pacific Northwest basins it applies to, because the region is absent from the calibration record on one side of the crest and from the published literature on the other.

Stating what the literature does not cover is more useful than overstating what it does.

Irrigation and agricultural water infrastructure is essentially unstudied. No peer-reviewed work was found on post-fire sedimentation of irrigation canals or diversions, on ash-laden water effects on pumps, screens, or emitters, on livestock water quality, or on irrigation-district delivery reliability. A USDA-funded Colorado State project on the question is active, so this may change. Barnard et al. (2023), a 16-author USDA-led communication, argues the link between mountain source-water hydrology and agricultural supply is a knowledge gap, not a solved problem. That matters in 2026: Oregon projected Owyhee April-to-September runoff at 17 to 29 percent of median, and Big Grass has burned more than half a million acres across that terrain.

Reservoir sedimentation has physics but no economics. Sankey et al. (2017) projects that nearly nine tenths of western US watersheds will see post-fire sedimentation rise more than 10 percent by the 2041 to 2050 decade. The cost side has no peer-reviewed equivalent, and Denver Water's widely cited 27 million dollar figure is a utility self-report.

Hillslope treatment effectiveness is better established than local practice reflects. Robichaud et al. (2010) found straw mulch produced roughly 63 to 95 percent reductions in first and second-year sediment yield across six study fires, but no measurable effect from contour-felled log erosion barriers for larger return-period events, which produce most of the runoff. Contour-felled barriers still appear in local post-fire plans.

For any system with a 2026 scar upstream, four things should happen before the first significant autumn storm rather than after it.

1. Determine whether a scar is actually upstream. This is a contributing-basin question, not a proximity question. Delineate to the intake or assessment point and check whether the severe burn falls inside it. A fire visible from the plant may drain elsewhere entirely.
2. Rank by where the burn sits, not by how much burned. Steep, high-severity, connected, upstream terrain drives the response.
3. Get a baseline now. Pre-storm data is the only thing that makes post-storm data interpretable, and the window closes with the first significant rain. This is the highest-value action available in August and September.
4. Design event-based sampling, not calendar-based sampling. Murphy et al. (2023) functions as a monitoring-design specification, with six site-selection criteria and tiered parameter lists. It recommends monitoring continue for many years, notes that few studies have run past five, and observes that debris-flow risk can rise again five to ten years out as roots decay. Pair it with a storm trigger. Turbidity, dissolved organic carbon, nitrate, and manganese are the priority constituents for most surface-water systems.

Steps one, two and four are the reason LYNXCE built PWSW. It delineates the contributing basin to the intake, ranks where the severe burn sits inside it, and returns a prioritized sampling design. Step three is fieldwork that no tool can stand in for, and it is the step whose window closes first. See a full PWSW run, from fire perimeter to monitoring brief. The cost of doing all four in August is a fraction of the cost of interpreting a November turbidity excursion with no pre-storm record to compare it against.

One transition worth planning around. Design-storm rainfall for Oregon and Washington still rests on NOAA Atlas 2, published in 1973; Idaho moved to NOAA Atlas 14 Volume 12 in September 2024. NOAA Atlas 15 closes the gap, with preliminary CONUS estimates due September 2026, changing the rainfall inputs to every post-fire threshold product in use here. Expect to revisit design-storm assumptions mid-program, a transition covered in NOAA Atlas 15 CONUS Preview.

LYNXCE developed PWSW for the screening problem above. It is a free, open-source tool that turns a wildfire perimeter and an intake or assessment point into a fire-specific source-water monitoring brief: which constituents to expect, where to sample, and when. It runs locally on public data from USGS, USFS RAVG and MTBS, NOAA and PRISM, and USDA STATSGO, with no API keys and nothing sent anywhere. Its Intake Vulnerability Index is a relative screening rank, calibrated to agree in ordering with the USGS Staley et al. (2017) M1 model at a Spearman rho of 0.853 across a twelve-site Pacific Northwest and California panel. The headline band assumes full hillslope-to-channel connectivity, a conservative upper bound.

It screens and helps design monitoring. It does not predict concentrations, and it is not a regulatory model. Post-fire response is too nonlinear to forecast reliably, which is why the output is a prioritized monitoring design rather than a number.

In the North Umpqua pilot, the Archie Creek scar above the Roseburg intake screens HIGH. The observed post-fire record supports the concern the screen raised, and it also shows why the screening question is where rather than whether. Turbidity at Rock Creek near Glide, a burned tributary, peaked near 1,310 FNU, while the regulated Roseburg intake on the mainstem peaked near 162 NTU. Those are observed field values from different points in the same basin, not outputs the tool produced, and the gap between them is the whole argument for delineating before sampling.

See a full run, from a fire perimeter to a monitoring brief, in the PWSW walkthrough. Teams working a 2026 scar this autumn can reach us at [email protected].