Fire Risk Analysis

EPRI publishes a failure rate normalized by deployed capacity: 0.2–0.3 failures per GW per year as of 2023–2024. For a 130 MW facility, that’s a 3.9% annual rate. The industry prefers a different number. EPRI’s Stephanie Shaw said 0.3% of projects had a fire with potential safety concerns in 2024, but that treats a 2 MW system and a 300 MW system as the same thing.

Both come from EPRI’s BESS Failure Incident Database. They measure different things: Shaw’s number is specifically “fires with potential safety concerns,” while the per-GW metric covers all failures. Neither is reassuring for a 130 MW facility next to thousands of homes.

“97% failure rate drop”

Jupiter Power’s open house presentation claimed a “97% drop” in global grid-scale BESS failure rates between 2018 and 2023, citing EPRI data. This is the per-GWh rate: failures divided by total deployed capacity. It drops because deployment is growing exponentially (the denominator gets huge). Their own chart tells a different story. The absolute number of failure incidents is not declining. EPRI’s database shows 8 incidents in 2024, which is among the highest annual counts on the chart. More batteries with a lower rate per GWh still means more fires.

The 97% number is also backward-looking. It’s dominated by older, smaller systems. The fleet is getting younger and larger. A 130 MW facility built in 2028 will have more cells, more potential failure points, and no long-term operating history to draw on.

The math

Probability of at least one incident over N years at annual rate r:

P = 1 - (1 - r)^N

Using the per-project rate (0.3%/yr)

The industry’s preferred number. Doesn’t account for facility size.

With aging (0.5%/yr years 1-10, 1.0%/yr years 11-20, 1.5%/yr years 21-30)

No utility-scale lithium-ion facility has operated long enough to measure aging-related failure rates. But the degradation mechanisms are well understood: internal resistance increases, dendrites form, thermal stability drops. This scenario models a modest increase in failure rate as the fleet ages. Jupiter Power’s Blackberry Grove application (Hillsboro, OR) estimates a 30-year project life.

Using the per-GW rate for a 130 MW facility

0.3 failures/GW/yr × 0.13 GW = 3.9% annual rate.

Side by side

Jupiter Power estimates a 30-year project life for its Blackberry Grove BESS in Hillsboro, Oregon (Case L2500161-SUDPLA). Using the industry’s own fire-specific rate over 30 years, the cumulative probability of at least one fire is approximately 9%. The per-project rate undercounts risk for large facilities. The per-GW rate includes all failures, not just fires. Accounting for aging over a 30-year life pushes the per-project estimate to roughly 26%, or 1 in 4. The actual risk for a specific facility is somewhere in this range.

For context

A 100-year flood has a ~26% chance of occurring in any 30-year window. The mid-range BESS fire estimate (9%) and the aging-adjusted estimate (26%) are in the range of risks that jurisdictions routinely plan for and regulate.

Why the real number is probably higher

Almost no utility-scale lithium-ion BESS has been running 10 years. Nobody knows what failure rates look like for aged systems.

Lithium-ion cells degrade: internal resistance goes up, dendrites form, thermal stability drops. These are well-understood mechanisms. The failure rate for a 20-year-old facility will not be the same as for a new one.

EPRI’s database has about 81 reported incidents with root cause data for roughly a quarter of them. Thermal events that didn’t make the news are probably underrepresented. EPRI itself notes it “cannot guarantee that the database captures every relevant BESS failure incident.”

Thermal runaway doesn’t always stay in one container. It can cascade. Neither metric distinguishes a contained single-container event from a Moss Landing-scale fire that destroyed 55–80% of the facility and evacuated the surrounding community.

Why the fleet-wide rate applies

No current King County code or state regulation requires a specific battery chemistry. The developer may install NMC, NCA, LFP, or any commercially available lithium-ion chemistry, and may change chemistry during augmentation over the facility’s lifetime. Because chemistry is not constrained, risk estimates cannot assume a favorable chemistry. The full fleet-wide incident rate applies.

Local fire district: “cannot confirm preparedness”

On April 6, 2026, Chief Will Aho of Eastside Fire & Rescue (the district that would respond to a fire at this site) provided a written statement in response to questions about the district’s readiness. Chief Aho noted that several critical details “are not yet fully defined,” including battery chemistry, site design, fire protection systems, available water supply, and facility-specific emergency planning, and said:

“Because those elements are not yet fully defined, Eastside Fire & Rescue cannot confirm that we are fully prepared today to mitigate an incident of this type and scale.”

Chief Aho called resident concerns about water supply, site access, evacuation challenges, and the implications of final battery chemistry “valid and directly relevant to emergency response planning.” He added that “at this stage, it would be inaccurate to suggest that all fire and life safety questions have been answered.”

Eastside Fire deferred jurisdiction to King County Permitting and the Fire Marshal’s office for site-specific review. The statement is clear: the fire district that would actually respond to this facility says the safety questions remain open, and the information needed to resolve them has not yet been produced.

No tabletop exercise for a BESS fire in the Snoqualmie Valley has been disclosed. No BESS-specific mutual aid agreements have been made public.

Jupiter Power’s own fire-safety package is similarly thin. The Hazard Consequences Analysis filed with Commercial Site Development application CMST25-0005 models 20 cells in thermal runaway, which is about 0.001% of the facility, and does not model hydrogen fluoride or a fire scenario. Every NFPA 855 deliverable (Hazard Mitigation Analysis, Consequence/Risk Analysis, Emergency Response Plan, Fire Protection Facility Design, Commissioning Plan, End-of-Life Plan) is deferred to Q4 2026. The defensive water cooling strategy relies on a 32,000-gallon on-site tank and a municipal hydrant connection, and Eastside Fire & Rescue’s fire marshal asked for a self-reliant on-site supply that Jupiter declined. See What King County’s Records Show for source citations.

Who bears the risk

The risk distribution is asymmetric. The developer bears financial risk: insured, tax-advantaged, with limited liability structures. The adjacent residential community bears safety risk: uninsured for BESS-specific hazards, with no control over facility operations, chemistry selection, or maintenance practices. Local fire departments bear response risk: BESS fires require specialized protocols and equipment that may not be available.

Under current regulation, the community has no mechanism to:

• Require a specific battery chemistry
• Enforce ongoing compliance with voluntary safety standards
• Compel augmentation with equivalent-safety replacement batteries
• Prevent chemistry changes over the facility lifetime
• Mandate independent third-party safety audits

This is why we are asking King County to require a full Environmental Impact Statement before permitting this project. An EIS would require evaluation of these risks and the alternatives available to address them.

For this site

This facility would be adjacent to thousands of homes in the Snoqualmie Valley, next to fish-bearing streams that feed into the Snoqualmie River where ESA-listed Chinook salmon, steelhead, and bull trout are present, in a ridge-bounded valley where inversions trap smoke close to the ground.

No atmospheric dispersion modeling, hydrogeologic assessment, or seismic risk analysis has been conducted for this site. The fire risk estimates above, the HF quantities modeled below, the earthquake-to-fire chain, and the fire district’s statement that it “cannot confirm preparedness” all point to the same conclusion: a SEPA checklist cannot credibly evaluate these risks. A full Environmental Impact Statement is the process designed to answer them.

Jupiter Power’s fire safety studies

On April 6, 2026, Jupiter Power provided a four-study bibliography titled “Battery Energy Storage System (BESS) Fire Safety Impact Studies,” framed as evidence that containerized BESS fires produce “no harmful levels of contamination beyond the property line.” All four studies have significant gaps relevant to this site.

New York Inter-Agency Fire Safety Report (Dec 2023): Reviewed four lithium-ion BESS fires in New York. The largest was roughly 1/25th to 1/50th of Cascadia Ridge’s proposed capacity. At East Hampton, no outdoor sampling was conducted; the report itself says “no conclusions could be made” from wipe samples. At Warwick (adjacent to a school district), no soil or water samples were taken at all because no water was used in suppression. At Chaumont, soil directly under the fire was never tested because failed equipment had not been removed. The report is explicitly an interim product with ongoing investigation.

SDG&E Gateway Water Runoff Report (Sep 2024): Two water samples collected five minutes apart, six hours after ignition. No follow-up sampling. No pre-fire baseline. No testing for HF (hydrogen fluoride), the primary toxic concern from lithium-ion thermal runaway. No fluoride, VOC, SVOC, or lithium testing. Copper was detected at 2-3 times freshwater aquatic life thresholds, but the report compared results only against drinking water standards. For Cascadia Ridge, where runoff flows toward Fisher Creek and the Snoqualmie River (both fish-bearing, with ESA-listed species), aquatic life criteria are the relevant benchmark. SDG&E’s own contractor conducted the analysis.

SDG&E Gateway Air Quality Report (Sep 2024): The fire burned for 11 days. Actual air monitoring covered roughly 2 of those 11 days. The first 2.5 hours of thermal runaway (the most intense emissions phase) went completely unmonitored. HF testing used low-sensitivity paper strips, not continuous instruments. No PM2.5 or PM10 monitoring. No weather data was recorded, making readings uninterpretable for exposure assessment. No plume tracking; monitors were at fixed locations that may not have been downwind. SDG&E’s own contractor conducted the monitoring.

ACP Literature Review (Aug 2025): Commissioned and funded by the American Clean Power Association, the clean energy industry’s trade group. Authored by Fire & Risk Alliance, a consulting firm that works for BESS developers. The most significant US BESS fire (Moss Landing, 1,200 MWh, January 2025, with ongoing EPA/CERCLA investigation) was excluded from the main analysis and relegated to an appendix as “unique, globally.” All seven detailed case studies involved fires in 1-2 containers. The report itself acknowledges that “much research is still needed,” emissions data is “still somewhat limited,” and waterborne contamination from BESS fires “is not well studied.”

What none of these studies address

Every study in Jupiter’s bibliography shares the same structural gaps:

• No complex terrain or valley analysis. All sites were flat, open terrain. The Snoqualmie Valley regularly traps emissions close to the ground during temperature inversions.
• No aquatic life standards applied. Fisher Creek and the Snoqualmie River support ESA-listed Chinook salmon, steelhead, and bull trout. Aquatic life criteria are more stringent than drinking water MCLs.
• No continuous HF monitoring. HF is the primary toxic product of LFP thermal runaway. Paper strips and excluded instruments are not adequate.
• Small-scale fires only. A 130 MW facility contains dozens of containers. No study addresses multi-container fire propagation.
• No private well groundwater modeling. Residences adjacent to the Cascadia Ridge site rely on private wells.
• No PM2.5/PM10 monitoring. Particulate matter is the primary health-relevant emission from battery fires and was not measured in any air study.

Under WAC 197-11-330(d), the SEPA responsible official must weigh uncertainty about environmental impacts toward a finding of significance. Jupiter’s own sources acknowledge the research gaps. When the applicant’s best evidence contains those admissions, a Determination of Non-Significance is difficult to defend.

Safety standards

Three overlapping standards govern BESS fire safety. None of them sets a national siting standard for where utility-scale facilities can be built relative to homes.

NFPA 855

The National Fire Protection Association’s Standard for the Installation of Stationary Energy Storage Systems. Covers design, installation, commissioning, operation, and decommissioning. The 2023 edition mandates fire suppression for nearly all ESS installations and requires battery management systems monitoring temperature, voltage, and state of charge. Referenced by the International Fire Code and many local jurisdictions.

UL 9540A

Underwriters Laboratories’ test method for evaluating thermal runaway fire propagation in BESS. The only consensus standard in the U.S. and Canada explicitly cited in NFPA 855 for large-scale fire testing. Tests progress from individual cell through module, unit, and full installation levels. Jupiter Power says their batteries are “fire safety tested according to strict Underwriters Laboratory protocols” — this likely refers to UL 9540A.

IFC 1207

The International Fire Code Section 1207, adopted via Washington State Building Code Council. For remote outdoor installations, it requires a minimum 100-foot setback from other buildings, lot lines, public ways, and exposure hazards. It requires evaluation of thermal runaway, mechanical failures, management system failures, and failures of external protection systems.

Detection and suppression

Current detection options include VESDA (Very Early Smoke Detection Apparatus, laser-based air sampling), gas sensors for off-gases that precede thermal runaway (CO, H2, VOCs), and thermal imaging. Suppression options include water-based systems (effective at cooling but don’t stop the electrochemical reaction), water mist, aerosol systems, and hybrid approaches.

All of these have limitations. The Moss Landing fire began after its fire suppression system failed. Water application can cause electrical shorts and temporarily increases the peak HF production rate by 35%, though total HF produced is unchanged. Aerosol systems work best before ignition in tight enclosures. No suppression technology has been proven to reliably stop cascade propagation once thermal runaway is deeply established at scale.

How much HF?

Larsson et al. (2017) (PDF), published in Scientific Reports (Nature portfolio), measured HF emissions from lithium-ion battery fires across 7 cell types and 39 fire tests using two independent measurement methods. They found 20 to 200 mg of HF per Wh of battery capacity. The paper extrapolates: a 1,000 kWh system could produce 20-200 kg of HF. The proposed Cascadia Ridge facility is approximately 520,000 kWh, 520 times larger than that example. Scaling the paper’s findings: a full-facility fire could produce 10 to 104 metric tons of HF. Even a fire involving just 1% of the facility’s capacity could produce 100 kg to over a metric ton. HF is immediately dangerous to life and health (IDLH) at just 30 ppm, roughly 25 milligrams per cubic meter of air. To put that in physical terms: under a valley temperature inversion with a 30-meter mixing height, about 91 kg of HF distributed across a 30-acre area would bring that volume to IDLH. A 1% facility fire produces 100 kg to over a metric ton of HF, enough on its own to push that 30-acre layer to roughly 1× to 11× IDLH. A full-facility fire would produce concentrations on the order of 100× to 1,000× IDLH across the same volume. HF is colorless, and symptoms of skin and lung exposure can be delayed by hours.

No atmospheric dispersion modeling has been done for this site. Nobody has studied what HF concentrations would look like at Cascade View Elementary (approximately half a mile away), along school bus routes on Snoqualmie Parkway (0.1 miles away), or in the surrounding neighborhoods during a lithium-ion battery fire. The Snoqualmie Valley’s topography traps airborne emissions close to the ground during temperature inversions rather than dispersing them. These quantities are large enough that dispersion modeling should be required before this project is approved.

Isolation and evacuation distances

Two federal sources reference isolation distances for lithium battery incidents, but neither was designed for a facility of this scale.

The EPA’s BESS guidance says to “set an isolation zone for large commercial BESS that is at least 330 feet, depending on the site.” This is an emergency response isolation zone, the area first responders should clear during an active fire. It is not a siting standard or a permanent setback from homes and schools.

The DOT’s Emergency Response Guidebook (ERG), Guide 147, covers lithium batteries in transportation. The 2020 edition recommended “initial downwind evacuation for at least 100 meters (330 feet)” for large spills. The 2024 edition removed that specific distance and replaced it with “increase the immediate precautionary measure distance, in the downwind direction, as necessary.” For fires involving a rail car or trailer of lithium batteries, the ERG recommends evacuating 500 meters (1/3 mile) in all directions.

The ERG is explicitly designed for transportation incidents involving quantities many orders of magnitude smaller than a 130 MW BESS. Its modeled scenarios are a trailer or rail car of batteries, not a fixed installation with 520,000 kWh of capacity. The 2024 edition also reflects a shift in fire response strategy: for large lithium battery fires, the guidance now says to let the fire burn itself out and protect surroundings rather than attempt suppression.

No federal agency has published isolation or evacuation distance guidance calibrated to utility-scale BESS facilities. The distances that exist are either emergency response minimums for much smaller quantities or have been removed from current guidance altogether.

Regulatory gaps

• No federal siting standards. BESS siting is governed by local land use and building codes, which vary enormously. Some jurisdictions have no BESS-specific ordinances.
• Inconsistent setbacks. Ranges from 10 feet (IFC egress minimum) to 150 feet (some county ordinances) with no national consensus for utility-scale facilities near residential areas.
• No standard air monitoring requirement. At Moss Landing, EPA deployed monitoring stations reactively after the fire started. No BESS facility is required to have permanent fenceline air quality monitoring.
• Post-incident cleanup is reactive. EPA had to order cleanups at both Gateway and Moss Landing. Whether operators carry adequate financial assurance for worst-case cleanup is unclear.

Earthquake risk

The Snoqualmie Valley sits at the intersection of three fault systems:

• Southern Whidbey Island Fault (SWIF), mapped directly through the valley, capable of M6.5–7.4 shallow earthquakes
• Seattle Fault, M7.0–7.5, very shallow (<25 km), highest shaking intensity
• Cascadia Subduction Zone, M9.0+, minutes of sustained shaking

King County’s BESS ordinance requires IEEE 693 seismic qualification for structural and nonstructural components, which keeps equipment on its mounts during an earthquake. But it doesn’t address what happens inside the cells.

Lab testing shows lithium-ion cells can develop internal short circuits, the initiating event for thermal runaway, from small mechanical deformations. The faults threatening this valley are capable of M6.5–7.5 shallow earthquakes that produce significant ground acceleration. No study has tested this chain at utility scale, and nobody has tested whether utility-scale BESS enclosures protect cells the way crash-tested EV packs do.

Beyond cell damage, there’s no requirement to model the full chain for this site: earthquake shaking damages cells → thermal runaway → fire → contaminated runoff into Fisher Creek, less than 10 vertical feet below the development area. The SWIF runs through this valley, but the BESS ordinance doesn’t cross-reference KCC 21A.24.290’s seismic hazard area provisions. And in a major earthquake, fire department access, water supply, and evacuation routes may all be compromised, exactly when a BESS fire would be most dangerous.

No utility-scale BESS has been hit by a major earthquake. That’s because the fleet is young and mostly deployed in low-seismicity areas, not because these systems are proven earthquake-safe. An EIS alternatives analysis would require evaluation of battery chemistries that do not exhibit thermal runaway, which would eliminate the earthquake-to-fire chain entirely.

Evacuation

The Snoqualmie Valley has limited evacuation options. Snoqualmie Ridge, the closest residential area, has two primary egress routes: Snoqualmie Parkway and SR-18. Residents in unincorporated areas downhill of the site have even fewer options. Moss Landing evacuated 1,200 to 1,500 people from a smaller community with more ways out. The affected population here is larger and the routes are more constrained.

The City of Snoqualmie’s own Comprehensive Emergency Management Plan (ESF #16, November 2017) identifies several constraints:

• The police department has very limited on-duty personnel for a large-scale evacuation, and may need to rely on statewide mutual aid from areas not affected by the event.
• The city is heavily dependent on roads, bridges, and assets it doesn’t own or operate, including King County and WSDOT infrastructure.
• In a major earthquake, “such severe damage to critical infrastructure that elements of evacuation pre-planning is impractical to determine prior to an actual evacuation.”
• The plan assumes the city will have enough accurate and adequate notice to implement an evacuation. “There may be times were there is too little or too late of notice to successfully evacuate all or certain parts of the population.”
• Private ambulance services would not be available to Snoqualmie due to commitments to higher-population areas like Seattle.
• School buses and Metro buses are in the city but may not have staffing to operate them during an evacuation.

The plan lists trigger events: earthquakes, mudslides, floods, volcanic activity, fires, dam failure, terrorism. It does not contemplate an industrial toxic release from a battery storage facility. The plan predates Ordinance 19824 by seven years. No BESS-specific evacuation scenario has been studied for the Snoqualmie Valley.

The population numbers in the plan (11,000+ residents, 3,200 students, plus thousands of casino and hotel visitors) are from 2017. The Snoqualmie Valley has grown since then.