Nine municipal indicators used in the PM2.5 benchmarking study: 289 Chinese cities, 2010–2023

This dataset contains the frozen nine-indicator municipal analytical panel used in the study “Benchmarking machine-learning screens against concentration history for prioritizing PM2.5 verification in monitored Chinese cities”, together with archived component statistics, quality-control flags, a variable dictionary, source-verification records and deterministic preprocessing reconstruction code. The panel covers 289 Chinese cities during 2010–2023 (4,046 city-years). The 1,346 missing indicator cells are retained without additional imputation. The reconstruction script verifies 35,068 non-missing indicator values against the archived model input. Indicators originate from public official statistical releases, statistical yearbooks, communiqués and departmental releases. Edition-level provenance was not recorded for the complete compilation; known locators and existing verification discrepancies are documented. The archive does not redistribute full yearbooks, pollution or meteorological products, map source files or fitted model binaries. Data compilation and documentation: CC BY 4.0. Source code: MIT. Reconstruction verifies analytical-input consistency, not complete official-source authentication or model refitting.