EPA PM2.5 parameter
SCIENTIFIC PACK 07.1
From an official EPA snapshot to a verified forecasting report.
This workflow acquires real PM2.5 observations from the US EPA Air Quality System, freezes an immutable snapshot, selects one monitoring station by a declared temporal-continuity rule and runs the paper-ready HeatAQ Nexus benchmark.
OFFICIAL DATA IDENTITY
US EPA AQS · Alameda County · PM2.5 · 2025
California · Alameda County
Strict contiguous segment
Snapshot and experiment integrity
PIPELINE
One declared path from acquisition to evidence
Secret-free plan
The committed plan records required environment-variable names and sanitized request parameters without credential values.
Immutable snapshot
Normalized observations, quality reports, source metadata, Dataset Card and checksums are frozen together.
Station selection
Stations are ranked by longest contiguous hourly run, total observations and deterministic station-ID tie breaking.
Complete benchmark
Persistence, seasonal, moving-average, linear, regularized and tree models remain visible in the complete results.
Uncertainty reporting
Coverage, interval width, event metrics, rolling-origin results and explicit limitations accompany point errors.
Candidate release
The run produces a self-contained archive without automatically minting or claiming a DOI.
Why NOAA is not silently merged into this hourly run
NOAA GHCN Daily weather values have a different temporal resolution. A later experiment will define a defensible daily protocol or a separately validated alignment method rather than manufacturing hourly precision.
SCIENTIFIC BOUNDARY
Station-level forecasting, not exposure or warning authority
A successful run demonstrates reproducible use of an official data identity and a declared evaluation protocol. It does not establish population exposure, countywide representativeness, causal effects, medical risk or official warning authority.