Skip to main content

California Freight Cleanup → Investigation M-1

Which mix of interventions is actually on the frontier?

66.8 deaths avoided / $2B transport vs 11.5 wildfire

This is the portfolio-assembly stage of the California Freight Cleanup cascade. It pulls the per-sector results from transport, building, wildfire, and biomass investigations, scores roughly 120 combinations on a cost-versus-deaths-avoided basis, and names six representative portfolios that every downstream investigation then operates on. Those six portfolios are the source of truth for all subsequent California Freight Cleanup portfolio analysis.

the CEC freight solicitation asks applicants to show how a limited California air-quality budget should be split across competing interventions. Each earlier investigation sized one lever in isolation — transport, building, wildfire fuel management, biomass retirement — answering “is this lever worth pulling?” but not “which combination dominates?” This investigation answers the combination question directly: if you have $2B and must spend it on exactly one sector, which sector wins?

Intervention menu. Five sectoral options are pulled from upstream results.json files. Transport offers five sub-options (T1 baseline through T5 heavy-duty-first). Building offers four (B1 through B4 cooking-first). Wildfire offers four prescribed-burn intensity rungs (5/10/20/30% area treated). Biomass provides one binary option: retire the DTE Stockton 50 MW wood-fired plant (NPV-positive once solar/storage replacement economics are netted out). A CRF resolution study at $2M is reported alongside the action portfolios as an information option.

Efficient frontier. The Cartesian product of the option sets yields approximately 120 combinations, each scored at the mixed-CRF posterior (β = 0.02439, Investigation 6-3 hierarchical Bayesian). The Pareto front—the convex envelope in (cost, deaths-avoided) space—yields 12 non-dominated portfolios spanning $0 to $13.9B.

Named portfolios. Six labeled portfolios are pulled off the frontier for narrative clarity and re-used by every subsequent California Freight Cleanup portfolio investigation. A 500-draw joint Gaussian-copula Monte Carlo (over βPM2.5, βO3, emissions_scale, surrogate_σ, and VSL) produces P5/P50/P95 bands on deaths-avoided and net benefit for each. All deaths-avoided figures are also reported under Di et al. (2017) and Krewski et al. (2009) CRF anchors separately, producing a 3.2× spread (Di scale factor 0.023, Krewski 1.992)—the dominant narrative-uncertainty band throughout the cascade.

SectorOptionDeaths avoided/yrCost ($B)DAC share
TransportT1_baseline13.10.000.208
TransportT2_accelerated79.92.000.208
TransportT4_equity48.41.000.207
BuildingB1_baseline37.00.000.208
BuildingB2_accelerated82.42.000.208
BuildingB4_cooking_first39.40.500.208
Wildfire5% PM2.5 reduction5.81.650.091
Wildfire10% PM2.5 reduction11.53.300.091
BiomassRetire DTE Stockton3.730.00 (NPV+)0.356
InformationCRF resolution studyEVSI $0.25B0.002ROI 125×
LabelDeaths avoided/yr (P50)Cost ($B)P(NB > 0)Composition
A_free_lunch53.80.001.000T1 + B1 + none + DTE retire
B_transport_2B120.62.000.194T2 + B1 + none + DTE retire
C_wildfire_instead59.61.650.014T1 + B1 + wildfire 5% + DTE retire
D_all_in_4B166.04.000.036T2 + B2 + none + DTE retire
E_smart_2B123.02.500.090T2 + B4 (cooking-first) + none + DTE retire
F_maximum_impact200.613.900.000T2 + B2 + wildfire 30% + DTE retire

Transport wins the $2B head-to-head by 5.8× over wildfire

At the $2B marginal: transport electrification (T2 over T1 baseline) avoids 66.8 deaths/yr; the same $2B on wildfire treatment (10% PM2.5 reduction) avoids 11.5; building electrification (B2 over B1) avoids 45.4. Transport wins by 5.8× over wildfire. This single comparison drives the CEC narrative throughout the cascade.

B4 (cooking-first building) is the single best deaths-per-dollar option at 78.8/B

B4_cooking_first ranks first at 78.8 deaths/$B, ahead of T4_equity (48.4) and B2_accelerated (41.2). At only $0.5B, it is the best marginal return in the menu. E_smart_2B (T2 + B4) earns its label by pairing the two most cost-effective paid interventions.

The CRF bracket spans a 3.2× deaths range (Di vs Krewski)

Di et al. (2017) scale factor: 0.023; Krewski: 1.992. Under Di, A_free_lunch yields only 1 death/yr avoided; under Krewski it yields 107. This 3.2× spread—not cost uncertainty—is the dominant narrative-uncertainty band throughout the cascade.

CRF resolution study: EVSI $0.25B at $0.002B cost (ROI 125×)

A $2M CRF resolution study carries EVSI of $250M—a 125× return. This is the “buy more knowledge instead” option that sits alongside all action portfolios on the frontier.

This investigation is the most upstream portfolio stage. Its six named portfolios feed directly into five downstream investigations: the robustness analysis (Investigation 6-4), the CRF-conditional regret surface (Investigation 6-6), the NSGA-II Pareto frontier (Investigation M-3), the cascade-wide sensitivity analysis (Investigation 3-10), and the ISRM validation (Investigation 6-8/58). If the frontier composition shifts when transport, building, wildfire, or biomass inputs are updated, all five must re-run.

ItemSHA-256 (12-char)
results.jsone99f6cb65d76
analysis.md
scenario.md
Upstream: Investigation 1-1 (transport) investigations/5_transport-electrification/latest/results.json 3326583e4445
Upstream: Investigation 1-2 (building) investigations/6_building-electrification/latest/results.json 99dcff1b4dfb
Upstream: Investigation 6-3 (CRF posterior) investigations/21_crf-hierarchical-bayes/latest/results.json 3104ba850408
Run timestamp 2026-05-04T07:46:18   N_MC = 500   seed = 20260429   Pareto points = 12