Editorial and data standard

Methodology

Our standard is not “trust us.” It is: identify the version, inspect the record, understand each transformation and follow the controlling source.

CURRENT DATASET VERSION

Virginia issued air permits

Source snapshot dated July 13, 2026. The public table contains 198 rows and seven exported fields per row.

Publisher
Virginia Department of Environmental Quality
Source date
2026-07-13
Rows
198
Export
CSV · UTF-8
01

Acquire and preserve

A scheduled source check retrieves Virginia DEQ’s issued-data-center permit table once each day. A new snapshot is accepted only when the agency date, table shape, record count and required fields pass structural validation. If the source cannot be retrieved, the last verified snapshot is retained and the run records an availability warning; an unavailable page is never treated as an empty dataset.

02

Normalize without guessing

The export preserves the agency’s facility name, permit identifier, issue date, program type, locality, regional office and direct document URL. Dates are converted to a consistent display format, county abbreviations are expanded for navigation and insecure source URLs are upgraded to HTTPS. Operator pages use disclosed name patterns; there is no hidden fuzzy ownership matching.

03

Validate each record

Every accepted row must have all seven fields, a valid issue date and a Virginia DEQ document link. Facility-and-permit pairs must be unique. The current source floor is 150 rows, which prevents a malformed or partial agency response from replacing the verified collection.

04

Detect changes

Each verified table receives a fingerprint. New fingerprints are compared with the prior snapshot to identify added rows, removed rows and changed fields. Up to 52 distinct snapshots are retained for the public source change tracker. Routine checks that find no source change do not create artificial activity.

05

Bound every claim

A permit is evidence of a regulatory record, not proof of construction, routine operation, actual emissions, compliance, exposure or harm. A research model is evidence of its assumptions and estimates, not an observed outcome. Evidence reviews label source type, scenario and unit of analysis before interpretation.

06

Keep outputs inspectable

The current CSV, individual agency links, dataset date, row count and change history are public. Evidence reviews place citations beside the claims they support. The processing repository is not currently presented as a public artifact; no claim of fully open-source reproduction is made.

07

Correct in public

Corrections should identify the page, disputed sentence or field and the strongest available primary source. Material factual or methodological corrections are dated in the public corrections log. Routine agency data refreshes remain in the separate permit change tracker.

08

Attribute carefully

Citing a researcher does not make that person a reviewer or supporter. External roles are published only with permission, a defined scope and a review date. Source checks are labeled as source checks—not as expert review.

KNOWN LIMITATIONS

What this version does not cover

  • The index is not a complete registry of every U.S. data center or every Virginia environmental approval.
  • The state table can change without notice; the displayed source date controls.
  • Facility names do not necessarily identify ultimate ownership.
  • Permit records do not measure operating hours, ambient concentrations or health outcomes.
  • Some agency documents can later move or be replaced.