California Freight Cleanup → Investigation 1-1

How much does freight electrification help — and what does ozone take back?

Five California transport scenarios evaluated at four target years, from baseline ACC II compliance to aggressive heavy-duty NOx controls. The finding that runs through all of them: cutting freight NOx raises ozone in VOC-limited basins, erasing anywhere from half to nearly all of the PM_2.5 benefit.

The decision

California faces a concrete near-term policy menu: comply with ACC II as written (T1), accelerate with $2B in incentives (T2), absorb a federal waiver loss and revert to market adoption (T3), redirect effort toward disadvantaged communities (T4), or layer aggressive heavy-duty NOx controls on top of the LDV trajectory (T5). These bracket the decision space CEC and CARB actually face. Three questions follow: how many deaths does each scenario avoid relative to a frozen-2011 status quo? What fraction of the PM_2.5 benefit is erased by the ozone disbenefit from lower NOx? And which scenario is optimal in expectation—and what is further certainty worth before committing?

Methodology

We map 32M California residents across 21,164 grid cells (22.9% DAC share) with spatially-varying baseline concentrations from AQS regional means: LA Basin 14 µg/m³, SJV 16, Bay Area 10, Sacramento 9, rest-CA 8. O₃ baseline is 42 ppb uniform; NO₂ ranges from 25 ppb (LA Basin) to 12 ppb elsewhere.

Each scenario produces a 16-dimensional emission-fraction vector at four target years (2025, 2030, 2035, 2040). A vehicle fleet model translates ZEV new-sales share into on-road emissions via a seven-year fleet-turnover lag and region-specific adoption multipliers (LA Basin and Bay Area adopt faster; SJV and rural CA slower). ISRM sector marginals translate emission fractions into PM_2.5 concentration deltas. Ozone changes use a regime-aware NOx-sensitivity function: VOC-limited cells versus NOx-limited cells are classified by region and a directional disbenefit is computed at each cell.

10,000 shared draws (seed 42) run across all five scenarios simultaneously. “Shared draws” means CRF and VSL variates are identical across scenarios within each draw—so scenario ranking uncertainty is cleanly separated from absolute-magnitude uncertainty. CRF options are Di et al. 2017 (HR 1.073/10 µg/m³) and Krewski et al. 2009 (HR 1.056), sampled as a structural choice. VSL is drawn from Triangular($5M–$11.6M–$20M) per EPA 2024 guidance. EVPI and EVPPI (CRF, VSL separately) are computed at 2035. Sobol first- and total-order indices are computed at N = 1024 for each scenario at 2035 under both CRF specifications.

Findings

Transport scenarios at 2035: PM_2.5, ozone, and net deaths avoided

Scenario	PM2.5 avoided	O3 caused	Net avoided	O3 offset %	Net-neg. cells (pop.)	Cost ($B)
T1 — ACC II as written	103.1	51.9	63.8	50.4%	2,306 (4.3M)	$0
T2 — Accelerated, $2B	141.5	74.9	87.0	52.9%	2,652 (4.9M)	$2.0
T3 — Delayed (no waiver)	43.7	22.1	26.9	50.5%	2,303 (4.3M)	−$0.5
T4 — Equity (DAC-focused)	130.7	73.6	65.8	56.3%	2,312 (4.3M)	$1.0
T5 — Heavy-duty NOx	103.1	95.6	39.8	92.8%	6,216 (12.2M)	$1.5

The VOI analysis at 2035 finds T1 optimal in 55.8% of MC draws and T3 optimal in 30.5%—driven by the asymmetric cost structure: T3’s −$0.5B cost credit makes it competitive when health-benefit differences are small. T2 is optimal in only 1.8% of draws despite its higher raw benefit; the $2B cost penalty is large relative to the marginal mortality advantage over T1. EVPI is $125M—meaningful residual uncertainty, but not enough to halt action.

The BAU baseline yields 14,504 deaths/yr at mean (P5–P95: 7,616–22,830), driven by CRF and VSL uncertainty rather than population or concentration uncertainty. Scenario differences—13 to 80 deaths avoided—are small against that spread. That gap is the core argument for resolving CRF uncertainty before committing accelerated spending.

Caveats

• Fleet-turnover lag is a uniform seven-year approximation. Real turnover is income-stratified and regionally heterogeneous. The uniform lag compresses cross-scenario differentiation at 2025 and 2030; T1 and T2 appear closer than they will be in practice.
• ISRM sector marginals are statewide, not facility-resolved. Intra-region source mix is held constant; T5 (heavy-duty NOx) cannot be distinguished from LDV NOx within the on-road slot at sub-regional resolution.
• Ozone chemistry is regime-aware but not photochemically calibrated. The scenario_o3 function applies a regional NOx-sensitivity factor without re-running CAMx or CMAQ. Critically, it does not yet incorporate the Sillman (1995) VOC–NOx indicator thresholds that sharpen the regime boundary. Ozone offset percentages are directionally correct; PM_2.5 mortality figures are unaffected by this gap.
• Baseline concentrations are frozen at 2011 AQS regional means. California ambient PM_2.5 has been declining; the frozen baseline makes absolute health burden slightly conservative and compresses scenario differentiation at later years.
• T3’s −$0.5B cost is foregone mandate spending, not revenue. The negative-cost entry inflates T3’s competitiveness in VOI net-benefit comparisons. Read the 30.5% T3-optimal probability with this accounting convention in mind.
• Baseline mortality rates are 2013 vintage. Post-2013 all-cause mortality trends are not reflected; death counts are slightly conservative relative to a more recent rate schedule.

Provenance

File	Link	Purpose
results.json	Full MC summary, ozone disbenefit, VOI, all scenarios × years
analysis.md	Mechanical readout, diff-from-previous-run table, stubs for human interpretation
scenario.md	Sticky methodology, key anchors, downstream dependency map

Run provenance: generated 2026-05-04T07:23:27; results.json sha256 3326583e4445. Seven downstream investigations read this investigation’s outputs via upstream_value: Investigation 1-3 (combined), Inv 11 (CRF-conditional), Investigation 4-3 (wildfire vs. electrification), Inv 13 (biomass anomaly), Investigation 6-2 (CRF residual VOI), Investigation M-1 (portfolio frontier), Inv 36 (EVPPI–GAM).