How much clean-air progress was retained under joint PM2.5-ozone attainment in China, 2015 to 2019
Year: 2026
Type: dataset
Abstract: Data, code and quality record supporting the manuscript How much clean-air progress was retained under joint PM2.5-ozone attainment in China, 2015 to 2019. The deposit reconstructs a city-day ledger of joint PM2.5 and ozone attainment for 328 Chinese prefectures over 2015–2019 (533,786 city-days) and classifies each city-day into one of four exposure states: jointly clean, PM2.5-only exceedance, ozone-only exceedance, or both. From that ledger it computes what fraction of the apparent clean-day gain measured on PM2.5 alone survives when ozone is scored at the same time. Headline result, reproduced from this bundle: 48.43% of the apparent population-weighted clean-day gain is retained under the joint test (95% CI 39.22 to 57.41); 54.09% unweighted. The paired weighting difference is +5.66 percentage points (CI +0.39 to +11.42, p = 0.038). Contents 03_Results_Tables/ — the classified city-day ledger, every derived table, and CANONICAL_NUMBERS.csv, which records 983 quantities with the source file and computation behind each one 04_Data_Inputs/ — city-daily PM2.5 and ozone reconstructions, population weights, boundary file, input checksums 05_Code/ — reconstruction, population aggregation, ledger, table and figure builders 06_Validation/ — the preregistered blind sample, the official-record comparison, run logs 08_Bootstrap_20260821/ — cluster bootstrap (prefecture clusters, 2,000 replicates, seed 20260821), balanced-panel and annual-rule tests What runs, and what does not Every downstream script runs from this bundle with no configuration: unzip anywhere and run, since each script locates the bundle from its own file position. The four upstream reconstruction scripts consume hourly station netCDF archives that cannot be redistributed here; they are included so the method is inspectable, and their outputs are deposited, so everything downstream of them is fully reproducible. README.md states this file by file. Verified 25 August 2026 from a cold shell with no arguments and no environment variables: the cluster bootstrap reproduced all three archived outputs byte-identically, and the ledger builder reproduced the published state shares to machine precision. The bootstrap enforces its own gate, aborting if the ledger it rebuilds departs from the published shares by more than 1e-6 percentage points; it did not abort. Related work The city-day reconstruction described above is shared with a companion manuscript by the same authors, under consideration elsewhere, which applies a different accounting over a different window and reports no result used here. Stated limitations The validation compares the reconstruction against daily files from the same archive family as the hourly inputs. It tests the aggregation; it is not an independent external check against a source a reader can consult. The deposited panel carries per-city validity flags as constants; per-city-day station counts are not included. Released under CC BY 4.0. Software in 05_Code/ and 08_Bootstrap_20260821/ is released under the same terms.
Language: English
Subfield: Health, Toxicology and Mutagenesis
Field: Environmental Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7204258454
