Trained on Nelson's 30+ year project history & technical library
Nelson's AI team member
How the three contamination hubs are built, what is counted, what is left out, and where the figures should be read with care. Written so a reporter, regulator or analyst can check the work before quoting it.
The underlying records are public. Nelson selects the files, normalises field names and units, geocodes where coordinates are usable, classifies subgroups, joins records to jurisdictions, and flags quality problems. That compilation work is what makes the three hubs comparable across the United States and Canada.
The underlying government records remain in the public domain and are cited in the methodology. Nelson claims rights in the compilation only: selection, filtering, normalisation, geocoding, subgroup classification and jurisdiction joins.
Source
Hub
Included records
EPA UCMR 5 drinking water monitoring
PFAS
Public water systems with reported PFAS sample results, rolled up to system level.
PFAS
Stations with at least one PFAS analyte result and usable coordinates.
PFAS
Sites where PFAS appears in site contaminant records.
PFAS
Installations named in published PFAS assessment status reporting.
Hydrocarbon
Confirmed releases, with open and closed status carried through as published.
Canada Energy Regulator incident data
Hydrocarbon
Reported pipeline incidents involving a substance release.
USACE Formerly Used Defense Sites
Military
Properties in the FUDS inventory with usable coordinates, plus funded project records.
Federal Contaminated Sites Inventory
PFAS, Hydrocarbon, Military
Sites split by custodian and contaminant of concern: PFAS sites, petroleum sites, and National Defence custodial land.
A record is counted once, at the grain the agency publishes it.
Records without usable coordinates remain in the file but are not drawn, so plotted totals sit at or below published inventory totals.
Closure rate is closed records divided by assessed records. Unassessed Canadian sites are excluded from the denominator.
Subgroup counts can sum above or below the site count, because a site may carry several classifications or none.
Detections are reported as the agency published them. Nelson applies no threshold of its own.
The AFFF overlap layer pairs a military property with a PFAS record when the two sit within a set distance of each other and share meaningful name terms. Each pairing carries a confidence tier:
High: close proximity and a strong name match on a named installation.
Medium: proximity with a partial name match.
Low: proximity only, offered as a lead to check.
A pairing states that two published records describe the same ground. It is not a finding about foam use, release, or liability.
Agency reporting lags the field position, often by a year or more.
State and provincial programmes define release, closure and assessment differently, so benchmarks compare programme reporting rather than ground conditions.
Federal Canadian data covers federal custodial land only. Provincial, municipal and private land sits outside it.
Free text contaminant names are grouped by keyword, so an unusual name can go uncounted.
Source files are checked on a scheduled cycle. Each refresh publishes a dated release with record counts and content hashes, and the change log records every dataset that moved. Cite the release version so your figure stays checkable later.
If a figure looks wrong, tell us and we will check it against the source file and publish the correction in the next release. Send a correction.
Dataset compilation copyright 2026 Nelson Environmental Remediation Ltd. All rights reserved. Source attribution required. Reuse, republication, redistribution or model training on this compilation must credit Nelson Environmental Remediation Ltd. and link to nelsonremediation.com.