Live PM2.5 operational screening method
Live observations are provisional unless explicitly stated otherwise and may subsequently be revised or invalidated by the original data provider.
The live screen is an operational early-warning filter for PM2.5 station assessments. It is not a validated anomaly detector, does not identify a source and is not a source-attribution model, health assessment or regulatory determination.
Eligibility
- The newest usable observation is no more than four hours old.
- At least 18 valid hourly observations are available in the preceding 24 hours.
- At least 48 valid hourly observations are available in the preceding 72 hours.
- No unresolved provider or pipeline quality condition designated as exclusionary applies.
A station assessment that fails one or more requirements is recorded as not assessed. It is not treated as having passed the screen.
Operational trigger
For each eligible PM2.5 station assessment, the current 24-hour mean is compared with a short station-specific recent-history window using a robust standardised score. A raw candidate requires both a 24-hour PM2.5 mean of at least 15 µg/m³ and a robust score of at least 3.
The 15 µg/m³ value is an operational screening floor. Crossing it does not by itself establish a legal exceedance, health finding or confirmed pollution event.
Multiple-testing adjustment
The pipeline derives an operational one-sided standard-normal tail score from each eligible robust score and applies the Benjamini–Hochberg procedure across the defined family using an operational decision parameter of 0.05.
Formal false-discovery-rate control has not been established. The tail scores are not calibrated inferential p-values, no empirical null has yet been established, measurements are dependent across time and space, and repeated-run error control over days or weeks is not established.
How to read zero
A zero result means that no eligible station assessment crossed all current operational screening criteria during that run. It does not mean that pollution was absent, concentrations were harmless, every station was assessed or all pollutants were screened.
Methodology
SCC Air Quality harvests primary observations first, applies quality controls and target-neutral screening, then executes transparent cross-source joins. Facilities are introduced only after discovery as possible context. Legacy Google Drive material is excluded unless the original source, date, authority, SHA-256, lineage, parser and scientific quality are validated.
How SCC Air Quality works
Source-attribution caution
The automated site must not collapse these evidence classes into a single causal claim.
Evidence checks
- Source acquired
- Provenance recorded
- Parser validated
- Time alignment checked
- Monitor role checked
- Meteorology checked
- Comparator checked
- Reviewer challenge recorded
Matching discipline
Public records are linked to a post-discovery context location only by location-specific phrases or official identifiers. Generic source-class terms are not enough on their own.
Public-safety rule
The public build excludes protected bundles, service-account keys, credentials, ZIP evidence packages and unredacted registers. The deployment guard fails if these appear in the public folder.
Forensic scientific evidence review
Inspect evidence coverage, freshness, source maturity, executed cross-references, uncertainty and explicit non-conclusions.
AI and model assurance
Review current deterministic execution, prohibited automated uses, evaluation gates and model provenance.
Evidence-assurance checks
Four controls govern every public release
evidence assurance is the internal technical shorthand. Public outputs use the full assurance language and never imply external approval or endorsement.
Requirements, source contracts, calculations, models, accessibility and publication rules are tested before release.
The live website and published social assets must match the exact reviewed evidence package and release receipt.
Secret leakage, unsupported claims, duplicate publication, unreviewed routes and evidence substitution are refused.
Sources, requests, transformations, model versions, artifacts, deployment and platform publication remain traceable.
Latest reviewed package timestamp: 2026-09-17T08:35:17+00:00. See sources, readiness and external review material.
Scientific evidence and software assurance are different
Scientific evidence
Measurements, station and regional coverage, descriptive calculations, source records and stated limitations are the evidence available for review.
Engineering assurance
Automated checks confirm that the pipeline followed its declared rules. They do not constitute peer review, accreditation, regulatory approval or institutional endorsement.
From a facility-context prototype to an England-wide, data-led observatory
The project began with a narrower facility-context purpose. The present pipeline discovers and evaluates official monitoring evidence first. Supplementary locations do not seed the monitoring screen, and unvalidated inherited material has no scientific influence.
This is an independent open-source project, not an official, accredited or peer-reviewed authority. Reviewers are invited to challenge the sources, station selection, units, timestamps, missing-data treatment, calculations, coverage rules and public interpretation.
Operational PM2.5 screen: threshold and multiplicity boundary
The current screen uses operational screening parameters including a 24-hour PM2.5 mean floor of 15 µg/m³ and a robust station-history score of 3 or greater. These are project screening parameters, not UK legal limits, WHO guideline values, health thresholds, or evidence of a pollution source.
Calibration status. Until an empirical null, sensitivity analysis and dependence-aware calibration are published, the project does not claim that the nominal Benjamini–Hochberg alpha establishes a validated false-discovery rate. The scores are screening statistics rather than calibrated p-values.
WHO context is not UK legal compliance
For NO₂, the WHO 2021 annual air-quality guideline is 10 µg/m³, while the UK statutory annual-mean limit value/objective is 40 µg/m³. WHO-context pollutant-day counts must not be interpreted as UK legal exceedance counts.