California Freight Cleanup → Investigation 4-3
When does wildfire prevention beat electrification per dollar saved?
Head-to-head cost-per-death comparison using the same population baseline, the same health calculator, and the same annual PM2.5 data for both interventions. The comparison is conditional on fire year: in a quiet year like 2023, wildfire prevention costs 7–19× more per death avoided than transport electrification. In a catastrophic year like 2020, the gap closes to near parity. Note: transport and building electrification inputs carry a stale upstream flag at last run — see Provenance below.
The decision
Where does a finite public health dollar go furthest? Two narratives compete in California air-quality policy: wildfire fuel management as a health co-benefit, and vehicle and building electrification as the long-term intervention. We put both on the same accounting sheet — same population baseline, same health calculator, same 2023 annual-mean PM2.5 data. The answer depends heavily on which fire year you are budgeting against: wildfire’s share of the statewide burden ranges from 3.7% in a quiet year to roughly 25% in a 2020-type catastrophe.
Methodology
Phase 1 — Sector decomposition. ISRM sector PM2.5 arrays are decomposed into five sectors: area sources (77.3%), on-road transport (12.4%), wildfire (7.0%), residential (3.4%), and EGU (0.0% via ISRM gas/coal/oil pathway; biomass EGUs route through the area sector). The controllable fraction (on-road + residential + EGU) is 15.7% of annual-mean statewide PM2.5.
Phase 2 — Wildfire PM2.5 reduction scenarios. Four reduction levels (5%, 10%, 20%, 30%) are modeled as linear scalings of the ISRM wildfire sector array. Burned-acreage cost benchmarks from USFS: $500/acre (prescribed burn) to $1,500/acre (mechanical thinning). Efficacy anchored to Schweizer & Cisneros 2017 (30–60% PM2.5 reduction in treated areas) and Fann et al. 2018 (USFS/EPA). The statewide scaling is conservative: it assumes uniform geographic treatment, whereas real programs concentrate in the wildland-urban interface.
Phase 3 — Health impact computation. Di et al. 2017 (HR 1.073/10 µg/m³) is primary; Krewski et al. 2009 (HR 1.056) is sensitivity. The Krewski result is approximately 3× larger than Di because Krewski's cohort covers all adults ≥30 while Di's Medicare-linked cohort covers only ≥65 — a ~4× larger at-risk population, not a baseline-concentration effect. DAC share is 9.1% (Di) and 12.0% (Krewski) across wildfire scenarios.
Phase 4 — Multi-year analysis. GFED5.1 fire-emissions data for 2010, 2018, 2020, 2021, and 2023 anchor per-year wildfire PM2.5 burden ratios. Deaths avoided per scenario scale with the GFED5 California PM2.5 burden ratio relative to the 2023 anchor. The 2023 anchor is a near-minimum fire year (332k acres vs. 4.4M in 2020), so the 2023 figures are conservative lower bounds.
Phase 5 — Electrification reference. Investigation 1-1/6 ozone-disbenefit JSON subfiles (stale sha256 flag at last run—see Provenance) are read for 2035 net deaths avoided by T1 (ACC II baseline), T2 ($2B accelerated), T5 (heavy-duty), and B2 ($2B building retrofit). Cost-per-death-avoided is computed on the same scale as the wildfire scenarios.
Findings
| Intervention | Deaths avoided | $/death low ($M) | $/death high ($M) |
|---|---|---|---|
| T1 baseline electrification | 63.8 | $0 | $0 (no added cost) |
| T2 accelerated EV ($2B) | 87.0 | $23 | $23 |
| B2 building retrofit ($2B) | 34.9 | $57 | $57 |
| Wildfire 10% reduction (2023) | 11.5 | $143 | $430 |
| Wildfire 30% reduction (2023) | 34.6 | $143 | $430 |
| Wildfire 10% reduction (2020) | 89.3 | $18 | $55 |
The cost-per-death ratio is approximately constant across wildfire reduction levels (linear CRF in the relevant PM2.5 range) but varies dramatically with fire year. In the catastrophic 2020 season, a 10% wildfire reduction avoids 89 deaths at $18–$55M/death—competitive with T2 electrification. In the 2023 near-minimum year, the same intervention avoids only 11.5 deaths at $143–$430M/death. The decision-relevant question: which fire-year scenario is the CEC program implicitly budgeting against?
A secondary finding from the sector decomposition: on-road transport is 12.4% of statewide PM2.5 and wildfire is 7.0%, but the controllable fraction (on-road + residential + EGU) is only 15.7% of total PM2.5. Area sources at 77.3% are not addressable through the electrification program; they are the background against which both wildfire and electrification interventions operate.
Cross-element relevance
Investigation 4-3 feeds two elements simultaneously. For Element 7 (wildfire-PV), the wildfire PM2.5 deltas are consumed by Investigation 7-1 to compute the PV-preservation co-benefit of wildfire portfolios. For Element 4 (co-benefits / disbenefits), Investigation 4-3 establishes the quantitative frame: is wildfire prevention a co-benefit of the electrification portfolio, or a competing program? In a typical fire year, electrification wins by 7–19×. In a 2020-type year, the gap closes. The answer depends on which fire year the CEC treats as the planning horizon. Each answer built on the last.
Caveats
- 2023 is a near-minimum fire year. The statewide wildfire PM2.5 fraction of 7.0% is a low-fire-year figure. In a 2020-type year this fraction reaches 15–25% in fire-adjacent regions, and the $/death comparison inverts in favor of wildfire prevention. No multi-year expected value is modeled.
- Linear statewide scaling overstates geographic uniformity. The delta is applied as a uniform fractional reduction of the wildfire array. Real prescribed-burn programs concentrate in the WUI; Bay Area (11.6% wildfire fraction) and Sacramento (9.2%) benefit more per treated acre than LA Basin (3.0%).
- Time-horizon mismatch. Wildfire $/death is computed against an implicit single-year treatment program; electrification $/death reads Investigation 1-1/6 2035 multi-year program costs. The two are not annualized on a common basis.
- Krewski sensitivity is 3× Di. The Di/Krewski spread is the dominant uncertainty band, reported as two columns rather than a posterior distribution. Neither CRF is propagated as a Monte Carlo here.
- Investigation 1-1/6 upstream inputs: stale sha256 flag at last run. The ozone-disbenefit subfiles (T1/T2/T5/B2) used for the electrification reference carried a sha256 drift warning at the time of the 2026-05-04 run. The electrification cost-per-death numbers are directionally stable; a re-run against fresh Investigation 1-1/6 outputs is needed before quoting these numbers in formal submission.
Provenance
| File | Link | Purpose |
|---|---|---|
results.json | Full per-year and per-scenario cost tables; sector decomposition; electrification reference | |
analysis.md | Mechanical readout, diff table, stale upstream flags | |
scenario.md | Sticky methodology, key anchors, Element 4 / Element 7 cross-reference |
Run provenance: generated 2026-05-04T07:44:19; results.json
sha256 115912c9a3a3. Upstream inputs via upstream_artifact:
Investigation 1-1 ozone-disbenefit subfiles (T1/T2/T5), Investigation 1-2 ozone-disbenefit
subfile (B2). GFED5.1 raw files for 2010, 2018, 2020, 2021, 2023 (sha256-tracked).
Key literature: Schweizer & Cisneros 2017; Fann et al. 2018 (USFS/EPA); Di et al. 2017 (NEJM 376:2513); Krewski et al. 2009. GFED5.1 final + beta-2023 (Binte-Shahid 2024 emission factor table).