Methodology¶
Research Approach¶
Port Health Watch produces site-specific health impact assessments for port-adjacent communities using peer-reviewed quantification frameworks and publicly available primary-source data. Every assessment follows the same four-step methodology the EPA uses in its Regulatory Impact Analyses for the Clean Air Act and National Ambient Air Quality Standards (NAAQS): emissions inventory, dispersion modeling, health impact calculation, and monetization.
All inputs, assumptions, and outputs are documented to ensure reproducibility. Any qualified researcher with access to the same public data sources should be able to replicate our results.
Data Sources¶
Input Data¶
| Source | Agency / Publisher | What We Use It For |
|---|---|---|
| Global Port Emissions Inventory Tool (goPEIT) | ICCT | Vessel-level emissions by port, vessel class, and operating mode |
| National Emissions Inventory | U.S. EPA | Port-area emissions from all mobile and stationary sources |
| American Community Survey | U.S. Census Bureau | Population demographics, income, poverty status by census tract |
| Social Vulnerability Index (SVI) | CDC/ATSDR | Community vulnerability to health stressors |
| EJScreen | U.S. EPA | Environmental justice screening indicators |
| Air Quality System (AQS) | U.S. EPA | Monitored ambient air quality data from regulatory monitors |
| PLACES — Local Data for Better Health | CDC | Census tract–level modeled health estimates (40 measures) |
| State/local health surveillance | State DOH / local health departments | Baseline rates for mortality, hospitalizations, ED visits |
| At-Berth Regulation compliance data | CARB | California-specific vessel compliance and technology performance |
| Toxics Release Inventory (TRI) | U.S. EPA | Facility-level toxic chemical release data for cumulative burden analysis |
| Climate Data Online (daily summaries) | NOAA NCEI | Wind speed, direction, and weather data for dispersion analysis |
Analytical Frameworks¶
| Framework | Publisher | Role in Our Methodology |
|---|---|---|
| Benefits Mapping and Analysis Program (BenMAP-CE) | U.S. EPA | Health impact assessment framework — provides concentration-response functions and monetization values |
| Intervention Model for Air Pollution (InMAP) | Open-source (Tessum et al.) | Reduced-complexity dispersion modeling — translates emissions changes into PM2.5 concentration changes at census tract resolution |
Health Impact Quantification¶
Step 1 — Emissions Inventory¶
Quantify at-berth emissions by pollutant (PM2.5, NOₓ, SOₓ, CO₂) using ICCT vessel activity data and EPA emission factors. Emissions are disaggregated by vessel class (container, tanker, bulk carrier, cruise, RoRo) and operating mode (at-berth, maneuvering, transit).
Step 2 — Dispersion Modeling¶
Model the change in ambient PM2.5 concentration in surrounding communities using InMAP or equivalent reduced-complexity model. Output is a grid of concentration changes at census tract resolution, calibrated against AQS monitor readings where available.
Step 3 — Health Impact Calculation¶
Apply peer-reviewed concentration-response functions to translate concentration changes into health outcomes:
ΔI = β × ΔX × P × I₀
Where:
- ΔI = change in health events (deaths, hospitalizations, ED visits)
- β = effect coefficient from epidemiological studies
- ΔX = change in pollutant concentration (μg/m³)
- P = exposed population
- I₀ = baseline rate of the health outcome in the affected population
Step 4 — Monetization¶
Convert health outcomes to economic values using EPA's Value of Statistical Life ($11.8M, 2024-adjusted) for mortality and standard willingness-to-pay estimates for morbidity endpoints (hospitalizations, emergency department visits, work-loss days, restricted activity days).
Key Concentration-Response Functions¶
| Health Endpoint | Source Study | Effect Estimate |
|---|---|---|
| All-cause mortality (adults 30+) | Krewski et al. 2009 (ACS CPS-II) | 6% increase per 10 μg/m³ PM2.5 |
| All-cause mortality (extended) | Lepeule et al. 2012 | 14% increase per 10 μg/m³ PM2.5 |
| Cardiovascular hospitalizations | Zanobetti et al. 2009 | Effect estimates from Medicare cohort |
| Respiratory hospitalizations | Zanobetti et al. 2009 | Effect estimates from Medicare cohort |
| Asthma ED visits | Mar et al. 2010 | Effect estimates from multi-city study |
Conservatism and Uncertainty¶
All estimates on this site are intentionally conservative. Where a methodological choice requires judgment, we choose the assumption that produces a lower health impact estimate. Specifically:
- We report ranges reflecting uncertainty in dispersion modeling and exposure assumptions.
- We exclude secondary PM2.5 formation from NOₓ and SOₓ precursors, which would increase total estimated health impacts.
- We use the lower-bound Krewski et al. (2009) function — 6% per 10 μg/m³ — as our primary mortality estimate rather than the higher Lepeule et al. (2012) estimate of 14%.
- We do not include chronic morbidity endpoints (cancer incidence, neurological effects) where epidemiological evidence is strong but quantification methods are less standardized.
This means our published figures are more likely to understate health impacts than overstate them. When we say a port's emissions contribute to a specific number of premature deaths or a specific dollar figure in health damages, the actual number is likely higher.
Source Documentation¶
Port Health Watch is an active research platform. Our source documentation standards are being applied progressively as pages are updated — not every page on the site yet meets the standard described below. We are publishing this standard transparently so readers know what to expect as the site matures, and so they can identify where existing pages fall short.
The Standard We Are Building Toward¶
Every quantitative claim should trace to a named primary source — the specific dataset, publication, or regulatory filing from which the number was derived. We distinguish between three categories of data:
- Measured data — values taken directly from a published dataset (e.g., AQS monitor readings, Census demographics, TRI releases). Cited with dataset name, year, and record identifier.
- Modeled estimates — values produced by applying an analytical framework to measured inputs (e.g., InMAP dispersion results, BenMAP health impact calculations). Cited with the model version, input parameters, and the measured data sources that feed the model.
- Derived figures — values calculated by combining measured or modeled data with published coefficients (e.g., monetized health damages using EPA VSL). Cited with the calculation method, the coefficient source, and all component inputs.
We do not cite secondary sources (news articles, summaries, or other analyses) as authority for quantitative claims. Where secondary sources informed our research direction, they are acknowledged but not treated as data sources.
Where We Are Now¶
Some pages on this site — particularly earlier port assessments — present quantitative estimates without full inline source documentation. The underlying data sources are listed on this page, and the analytical methods are described above, but the per-claim citation chain is not yet complete on every page.
Pages that meet the full documentation standard are marked with sourcing footnotes that identify the specific dataset, publication date, and record for each claim. The MERC Coal Terminal page is the current template for this standard. We are progressively updating other pages to match.
If you encounter a specific figure on this site and need to verify its source, contact research@porthealthwatch.org and we will provide the underlying data and calculation.
Corrections¶
If you identify an error in our data or analysis, contact research@porthealthwatch.org. We will investigate and correct any verified errors with a transparent correction notice documenting what changed, when, and why.