HOW THE DATA WERE BUILT

A representative sequence, with traceable origins.

A concise guide to the adopted manuscript workflow. The web tool reads the results; it does not rebuild TAPY.

TAPY construction

Follow the stages from observations to a complete representative year.

Data sources & preprocessing

EPA AQS observations underpin both products. Site-level PM₂.₅ uses FRM observations first, with non-FRM records filling missing site-hours; the multi-pollutant product uses county-hour observations.

Quality control & reconstruction

Coverage screening precedes reconstruction. Accepted short gaps are interpolated; remaining gaps use the adopted models and fallback rules. The framework-specific steps below distinguish site eligibility from the retained county roster.

PM₂.₅-only · monitoring sites

  1. Start with FRM-primary AQS observations; non-FRM fills missing site-hours.
  2. Import the adopted site roster and pre-filling monthly eligibility mask.
  3. Remove negative observations and interpolate internal 1–2-hour gaps.
  4. Reconstruct remaining gaps with the selected two-track Ridge models and documented fallback.
  5. Rank the 7–10 eligible years for each month by hourly F–S; keep the top three.
  6. Apply the daily-percentile persistence safeguard and select each typical month.

Multi-pollutant · counties

  1. Prepare nonnegative county-hour observations and interpolate accepted short gaps.
  2. Use the 60-county roster retained after initial coverage screening.
  3. Fill remaining gaps using pollutant-specific models and fallback rules.
  4. Rank all ten completed years using the adopted five-pollutant DALY weights.
  5. Apply the corresponding weighted persistence safeguard within the top three candidates.

Empirical CDFs & F–S ranking

Compare candidate-month empirical concentration distributions with the corresponding long-term calendar-month distribution using the Finkelstein–Schafer statistic. Lower scores indicate closer distributional similarity; the multi-pollutant framework combines pollutant-specific scores using the adopted DALY weights.

Candidate ranking & persistence screening

Retain the top three ranked candidates and apply the adopted daily-percentile persistence safeguard. Select the first passing candidate in F–S rank order; retain rank 1 if all three are flagged.

One complete annual product

Assemble the 12 selected months, remove February 29, then smooth only the first two hours at each of the 11 internal month boundaries. Verify a complete, nonnegative 8,760-hour product and preserve the selected source years.

Reading the comparisons

Seasonal profiles: local monthly means divided by the same location’s full-reference annual mean. The local view has no across-location IQR band.

Normalized Wasserstein distance: W1 divided by the local reference IQR. Lower indicates closer distributional agreement; zero IQR makes this metric undefined.

Profile correlation describes seasonal shape. NRMSE divides monthly-profile RMSE by the mean reference monthly profile. These use the existing manuscript calculations.

The 2015–2024 reference includes cleaned and filled records. Comparisons are in-sample representativeness, not independent tests of gap-filling accuracy.

Initial county screening is author-documented; its original generating code is not fully traced in the project. The final code imports the roster. This application does not reinterpret or repeat that screening.

Paper and citation

Paper DOI and full citation: to be provided by the project authors.

Background: air quality and representative years

What is air pollution?

Air contains gases and tiny particles. Some can affect health and the environment. This atlas follows fine and coarse particles, ozone, nitrogen dioxide and sulfur dioxide.

Why does it change?

Emissions, weather and atmospheric chemistry change through the day and across seasons. A daily average can hide a short peak, while an annual average can hide a seasonal pattern.

What is a typical year?

A Typical Meteorological Year assembles representative months of weather. TAPY uses a related idea for air pollution: select representative historical months and join them into a complete annual sequence.

Why not choose a single year?

One historical year may include unusually high or low pollution. TAPY is designed to reflect a longer record. It is not a forecast, a worst-case year, or a guarantee that every peak is reproduced.

More ways to learn, in a future phase

Build Your Own TAPY, F–S activities, persistence screening exercises, high-school exercises and undergraduate laboratory assignments are planned. They are not part of this research prototype.

Workflow figure

Adopted site-level and county-level TAPY construction workflows