Index Methodology and Data Sources
1Purpose and Scope
The North American Environmental Health Index (NAEHI) maps ten environmental contamination and health risk indicators across US states and Canadian provinces, allowing users to compare regional burden side by side. The index is a public education resource. It is not a regulatory instrument, a clinical resource, or a peer-reviewed scientific product, and it should not be used as such.
Important limitation. Index scores represent normalised estimates of relative regional burden derived from published monitoring data. They are not measured contaminant concentrations, and they should not be used to assess individual exposure, make personal health decisions, or draw conclusions about specific communities within a state or province. Regional averages can substantially understate exposure for Indigenous communities, low-income populations, and those relying on private wells or aging infrastructure.
2Layer Selection Rationale
Ten indicators were selected by applying four criteria. First, documented human health impacts supported by peer-reviewed evidence. Second, monitoring data published at the state or provincial level by a government agency or independent research programme. Third, sufficient geographic variation across North America to produce meaningful regional differentiation. Fourth, applicability to both the US and Canadian contexts — indicators relevant only to one country were deprioritised. Indicators with contested health evidence, insufficient coverage, or substantial overlap with existing layers were excluded.
Included layers
| Layer | Primary health concern | Key data gap addressed |
|---|---|---|
| PFAS | Probable carcinogen; thyroid, immune, reproductive effects | Drinking water contamination from legacy industrial and military sites |
| Nitrates | Methemoglobinemia in infants; colorectal cancer (chronic) | Agricultural groundwater contamination affecting private well users |
| Heavy Metals | Neurodevelopmental harm (lead); carcinogen (arsenic, cadmium); neurotoxin (mercury) | Legacy industrial, mining, and urban infrastructure burden |
| Air Quality | Cardiovascular and respiratory mortality; 100,000+ US deaths per year | PM2.5 and ozone burden including wildfire smoke |
| Pesticides | Neurotoxicity; endocrine disruption; pollinator ecosystem harm | Agricultural chemical loading in surface and groundwater |
| Microplastics | Emerging — inflammatory and cardiovascular effects under investigation | Ubiquitous environmental contamination with no current regulatory framework |
| POPs | Carcinogen (PCBs, dioxins); neurodevelopmental and immune effects | Legacy industrial contamination and Arctic bioaccumulation |
| Lead in Water | Irreversible neurodevelopmental harm; no safe level in children | Lead service line prevalence and infrastructure legacy |
| Water Violations | Multi-contaminant; waterborne illness; 6,600+ US deaths per year | Tap water compliance failures in small and rural systems |
| Radon | Lung cancer; 21,000 US deaths per year; second only to smoking | Geological indoor exposure risk with no industrial origin |
Excluded indicators and reasons
Endocrine disruptors were excluded due to conceptual overlap with PFAS and Pesticides — the exposure pathways and affected populations are substantially the same. Electromagnetic fields were excluded given the contested state of the health evidence. Pharmaceutical compounds in water were under consideration but current monitoring coverage at the state and provincial level is insufficient to support meaningful regional differentiation. Wildfire smoke is not a separate layer but is incorporated into the Air Quality layer through its contribution to annual PM2.5 design values.
3Normalisation Methodology
Each layer is represented as an index score from 0 to 100. Higher scores consistently indicate greater burden or risk, with one exception: the Water Violations layer uses violation rate as its metric, which is directionally consistent with all other layers — higher score means higher violation burden.
Index values were derived through research synthesis rather than direct statistical transformation of raw monitoring datasets. The underlying data span multiple agencies, monitoring programmes, and countries in heterogeneous formats; building a direct normalisation pipeline from raw data would require dedicated data infrastructure that was not available for this version of the index.
Derivation process
- Primary source identification. For each layer, one or more primary government monitoring datasets were identified as the authoritative basis for regional rankings. These are documented in Section 4.
- Regional ranking from published data. Each region was assigned an initial relative ranking based on published summary statistics, regulatory violation records, monitoring summaries, and peer-reviewed literature. Rankings were informed by the specific metric documented for each layer — for example, nitrates use estimated percentage of groundwater samples exceeding 5 mg/L; air quality uses three-year average PM2.5 design values and ozone exceedance days.
- Conversion to 0–100 scale. Rankings were converted to index scores on a 0–100 scale with the highest-burden region anchoring the top of the scale and the lowest anchoring the bottom. Scores are assigned in bands rather than as continuous values to avoid implying false precision. The minimum band width is 5 points.
- Cross-validation. Initial values were cross-validated against at least one independent secondary source for each layer. Where primary and secondary sources disagreed, values were revised toward the more conservative estimate or toward the source with larger sample sizes and more recent data vintage. Corrections made during cross-validation are documented in the layer methodology notes accessible within the map interface.
- Regional plausibility review. Final values were reviewed for internal consistency — for example, verifying that states with well-documented contamination histories (Parkersburg WV for PFAS, Chicago IL for lead) received appropriately elevated scores, and that remote regions with minimal industrial activity received appropriately low scores for industrial-source layers.
Recommended future improvement. The most significant methodological upgrade would be direct percentile ranking from raw monitoring datasets — for example, computing each state's nitrate index score as its actual percentile rank among all 63 regions based on measured groundwater concentrations from USGS NAWQA well-sampling data. This would make every score reproducible from primary source data and eliminate judgment calls in the conversion step. Implementing this requires programmatic access to raw EPA, USGS, and Health Canada datasets.
Severity classification
Index scores are grouped into five severity bands for display purposes. These bands are a communication device and do not correspond to regulatory thresholds or clinical risk levels.
| Band | Score range | Interpretation |
|---|---|---|
| Minimal | 0–20 | Among the lowest burden regions on the continent for this indicator. Does not mean zero contamination. |
| Low | 21–40 | Below-average regional burden. Contamination is present but at lower relative levels. |
| Moderate | 41–60 | Near-average continental burden. Consistent with typical conditions for this contaminant across North America. |
| Elevated | 61–80 | Above-average burden relative to other North American regions for this contaminant. |
| High | 81–100 | Among the highest burden regions on the continent. Not necessarily above regulatory limits — reflects relative position. |
4Data Sources by Layer
Listed below are the primary and secondary sources for each layer, the specific metric used, and the data vintage.
PFAS
Nitrates
Heavy Metals
Air Quality
Pesticides
Microplastics
Persistent Organic Pollutants (POPs)
Lead in Drinking Water
Water Violations
Radon
5Confidence Classification
Each region carries a single confidence classification applied uniformly across all layers, based on the overall density and consistency of monitoring data available for that region. This reflects monitoring infrastructure broadly rather than data quality for any specific contaminant.
| Classification | Criteria | Typical regions |
|---|---|---|
| High confidence | Extensive federal monitoring network coverage; large well sample sizes across multiple independent programmes; consistent multi-year records enabling trend validation; multiple layers with direct measured data rather than modelled estimates. | California, Texas, New York, Florida, Ohio, Pennsylvania, Illinois, Michigan, New Jersey, North Carolina, Georgia, Virginia, Washington, Colorado, Arizona, Minnesota, Massachusetts, Wisconsin, Maryland, Indiana, Tennessee, Missouri, Oregon, Connecticut, Iowa, Kansas, Nebraska, Oklahoma, Louisiana, Kentucky, Alabama, South Carolina, Arkansas, Mississippi, Utah. |
| Moderate confidence | State or provincial monitoring data available but with meaningful gaps in geographic coverage or contaminant scope; shorter or less consistent record periods; some layers rely on modelled estimates rather than direct measurement; cross-border data integration challenges for Canadian provinces. | Alaska, Hawaii, Idaho, Montana, Nevada, New Mexico, West Virginia, Maine, New Hampshire, Vermont, Rhode Island, Delaware, South Dakota, North Dakota, Wyoming, DC; Ontario, Quebec, British Columbia, Alberta, Manitoba, Saskatchewan. |
| Estimated | Sparse direct sampling; values primarily extrapolated from regional geological, land-use, and industrial patterns; limited dedicated monitoring infrastructure; multiple layers rely on proxy indicators rather than measured data. | Northwest Territories, Nunavut, Yukon, Newfoundland and Labrador, New Brunswick, Nova Scotia, Prince Edward Island. |
Note on confidence and accuracy. A High confidence classification reflects the density of available data, not certainty that the index value is precisely correct. High confidence regions have more data against which values can be validated and corrected. Estimated regions have genuinely fewer data points and wider uncertainty ranges. For all regions, index scores should be interpreted as order-of-magnitude estimates of relative burden, not precise measurements.
6Overall Burden Index
The Overall Burden layer is the unweighted arithmetic mean of all ten individual layer scores for each region. Its purpose is to identify regions with elevated contamination across the widest range of categories simultaneously. Multi-contaminant exposure is well established in the literature as producing health effects that single-contaminant assessments do not predict — this layer provides a first approximation of that cumulative picture.
Rationale for equal weighting
Equal weighting was chosen deliberately. A weighted composite — assigning higher weight to contaminants with stronger dose-response evidence or higher mortality burden — would be more analytically sophisticated, but implementing defensible weights requires expert consensus on relative risk that is outside the scope of this project at its current stage. The WHO, EPA, and IARC publish risk frameworks that could support weighting in a future version, but applying them without formal expert review would introduce unjustified precision. An honest unweighted average is preferable to a weighted one where the weights are not independently validated.
Water Violations layer treatment
The Water Violations layer measures violation burden — higher scores indicate more violations. This is directionally consistent with all other layers, so no adjustment is made when computing the Overall Burden average. A high Water Violations score correctly adds to a region's cumulative burden.
Recommended future improvement. Differential risk weighting based on peer-reviewed relative risk estimates, implemented following formal consultation with environmental health scientists. A reasonable starting framework would assign higher weights to lead in drinking water, PFAS, nitrates, radon, and air quality, given their stronger dose-response evidence and higher mortality burden relative to contaminants such as microplastics where the human health literature remains at an early stage.
7Limitations
Methodological limitations
Index values are estimates derived from published monitoring data, not direct statistical transformations of raw datasets. Converting published summary statistics to 0–100 scores requires judgment calls at several steps — which region anchors the top of the scale, how to weight composite metrics, and how to handle missing data. A future version implementing direct percentile ranking from raw agency datasets would eliminate most of these judgment calls and make every score reproducible from primary source data.
The index operates at the state and provincial level. Within any given state or province, contamination can vary enormously — a state average masks localised hotspots and communities with significantly higher or lower burden. The index should not be used for sub-regional comparisons.
Equal weighting in the Overall Burden layer treats all contaminants as equally significant, which is a simplification. Lead in drinking water and radon, with strong dose-response evidence and high mortality burden, plausibly deserve higher weight than microplastics, where the human health evidence remains preliminary.
Data limitations
Private wells are not covered by public water system monitoring programmes in either country. An estimated 40 million Americans rely on private wells; comparable national data for Canada are not systematically collected, though the proportion is significant in rural and remote areas. Nitrate and PFAS contamination from private well water is likely underrepresented in this index as a result.
Canadian provincial monitoring data are generally less comprehensive than US federal monitoring data, particularly for PFAS, pesticides, and microplastics. Canadian index values carry wider uncertainty ranges than their US counterparts for most layers.
Indigenous community exposure is systematically underrepresented by population-wide regional averages for several layers — particularly POPs, heavy metals, microplastics, and radon in northern territories — where traditional country food pathways and older housing infrastructure produce substantially higher exposure than regional means suggest. Regional blurbs within the map interface flag this disparity where documented, but the index scores do not quantify it.
Temporal limitations
Data vintage spans 2019–2024 across all layers. The index does not update automatically. Recent contamination events, regulatory changes, infrastructure improvements, or new monitoring data are not reflected.
Scope limitations
The index covers environmental contamination and health risk at the population level. It does not assess occupational exposure, indoor air quality beyond radon, consumer product chemical exposure, food safety, or climate-related health risks except where these intersect with the covered contaminants (wildfire smoke is captured within the Air Quality layer).
8Recommended Future Improvements
The following are listed in priority order by expected impact on credibility and accuracy.
- Percentile ranking from raw datasets. Replace research-informed estimates with direct percentile ranks computed from raw monitoring data — USGS NAWQA for nitrates, EPA UCMR5 for PFAS, EPA AQS for air quality, state radon programme databases for radon. This requires programmatic data access and processing infrastructure but would make every score reproducible from primary sources.
- Expert review of index values. Having the methodology and index values reviewed by one or more environmental health scientists or public health researchers would substantially strengthen the index's credibility and likely surface corrections to specific regional values. This does not require formal peer review — a structured informal assessment would be a meaningful improvement at this stage.
- Differential risk weighting for Overall Burden. The cumulative index should use a formally justified weighting scheme rather than equal weights. WHO burden of disease estimates, IARC carcinogen classifications, and EPA relative risk guidance all offer frameworks that could support this, but applying them requires expert consultation to avoid introducing unjustified assumptions.
- Sub-regional granularity. Expanding to county or census tract level for US data — where monitoring density permits — would capture within-state variation that state averages currently obscure, and would make the index substantially more useful for environmental justice research.
- Annual update cycle. Index values should be refreshed as new monitoring data become available. PFAS in particular is evolving rapidly — UCMR5 data collection continues through 2025 and the final dataset will be substantially larger than what informed this version. Lead service line inventories are also being updated as states comply with the 2021 Lead and Copper Rule Improvements.
- Indigenous community exposure layer. A dedicated indicator is needed for disproportionate exposure in Indigenous communities through country food pathways and aging housing infrastructure. This disparity is currently flagged in qualitative regional descriptions within the map but is not quantified in any index score — a gap that meaningfully limits the index's usefulness for environmental justice analysis.