Skip to main content

California Freight Cleanup → Investigation 7-2

Will mid-century wildfire smoke swamp the program's health signal?

CEC program investments run for decades. Wildfire frequency is linked to vapor-pressure deficit, which every major climate model projects to rise through mid-century. We ran six climate models to quantify that uncertainty and compared it against the policy signal. The six-model median projects a 1.73× increase in wildfire PM2.5 by 2050. Any cost-effectiveness analysis that holds the 2023 baseline constant through mid-century is under-counting future burden by 60–90%.

Investigation 4-3 computes wildfire fuel-management cost-per-death-avoided at a stationary 2023 fire-year baseline. CEC program investments operate on decadal horizons, and wildfire frequency is mechanistically coupled to vapor-pressure deficit (VPD)—which all major climate models project to increase through mid-century. Does the Investigation 4-3 calculation hold across plausible 2040 and 2050 climate futures, or does climate forcing change the answer?

We answer structurally: we quantify the ratio of the CMIP6 climate uncertainty band to the policy signal (deaths avoided by a 10% fuel-management reduction), rather than replacing the Phase 1 baseline with a more accurate number. Climate-driven uncertainty dwarfs the policy effect—that is the most decision-relevant form of the result.

The investigation runs five fidelity rungs, all computing the same endpoint (wildfire deaths per year) but varying the climate forcing:

L1 — Stationary baseline. Burned area and smoke-day counts are held at the 2020-anchor observed values (4.4M acres, 35 smoke days). The wildfire PM2.5 multiplier is 1.00×. This is the Phase 1 assumption; it is the anchor, not a prediction.

L2 — VPD trend extrapolation. The Abatzoglou & Williams 2016 (PNAS) VPD-to-burned-area transfer function (βvpd = 0.33/hPa) is fed the observed 1980–2023 VPD trend (+0.48 hPa/decade), extrapolated to 2040 and 2050. Single-path, no ensemble uncertainty. PM2.5 multiplier at 2050: 1.58×.

L3 — CMIP6 six-GCM ensemble. Three SSP2-4.5 models (GFDL-ESM4, MPI-ESM1-2-HR, MRI-ESM2-0) and three SSP5-8.5 models (EC-Earth3, CESM2, ACCESS-CM2) provide downscaled VPD anomaly projections at 2030, 2040, and 2050. Each GCM drives the same Abatzoglou-Williams burned-area model. The six-member fan’s P10 and P90 define the uncertainty band. Median PM2.5 multiplier at 2050: 1.73×.

L4 — WRF-Chem SOA uplift. L3 CMIP6 median is retained, but PM2.5 exposure is bumped +8% to reflect secondary organic aerosol formation captured in WRF-Chem runs (from Inv 17’s L3-to-L4 bracketing). PM2.5 multiplier at 2050: 1.87×.

L5 — Earth-system reference. Full vegetation-aerosol-cloud-radiation feedback is not run inline. L5 applies a published ±7% CI widening to the L4 median; the median is unchanged. L5 is reported for completeness and upper-tail framing; it is not the operational estimate.

Fidelity ladder: wildfire deaths/yr projected at 2050 (anchor fire-year 2020)
Level Description Deaths 2050 PM2.5 mult P10 P90
L1 Stationary (Phase 1 anchor) 218 1.00×
L2 VPD trend extrapolation 343 1.58×
L3 CMIP6 six-GCM ensemble 376 1.73× 303 480
L4 WRF-Chem +8% SOA 406 1.87× 327 519
L5 ESM reference (CI widening only) 406 1.87× 304 555

The headline diagnostic: the CMIP6 P10–P90 fan at 2050 spans 178 deaths against a policy signal (10% fuel-management deaths avoided) of 21.8 deaths—fan-to-signal ratio 8.17×. The Phase 1 stationary baseline does not hold at 2040 under any of the six GCMs tested.

The L4/L5 collapse (identical 406-death median) is its own diagnostic: WRF-Chem SOA chemistry already captures the bulk of the climate-coupling effect. Full ESM vegetation feedback adds nothing to the median at 2050—only widens the upper tail (+15% at P90). L4 is the production estimate; L5 is reference-only.

Investigation 7-1 reads the mean L3 PM2.5 multiplier across the four per-year anchor scenarios: 1.86×. This scalar drives the 2050 PV-loss projection.

File Link Purpose
results.jsonFull L1–L5 ladder; per-year anchor results; GCM VPD projections; fan envelope
analysis.mdMechanical readout with diff table; fidelity-ladder summary
scenario.mdSticky methodology; key anchors; Investigation 7-1 / Investigation 4-3 downstream connections

Run provenance: generated 2026-05-01T21:29:00; results.json sha256 70b1647f66c7. Upstream: Investigation 4-3 (Di/Krewski 10% deaths-avoided, sha256 f7686f48f8df). NASA NEX-GDDP-CMIP6 downscaled VPD projections.

Key literature: Abatzoglou & Williams 2016 (PNAS); Williams et al. 2019 (Earth’s Future); Goss et al. 2020 (ERL); Aguilera et al. 2021 (Nature Comm); Hurteau et al. 2019 (Earth’s Future).