WP2_Chengdu_NC_zenodo_deposit.zip

This deposit contains all derived data products and analysis scripts supporting a twelve-fire-season (2015–2026), data-driven assessment of air-quality resilience for Chengdu, Sichuan Basin, China. The framework fits meteorologically conditioned counterfactual baselines (four specifications), detects pollution anomalies with a leave-one-window-out scan requiring no prior event definitions, decomposes the largest events into pre-/event-/post-shock phases (resilience integral), and attributes event sources through station-level concentration fingerprints and hourly speciated supersite measurements. Contents: (1) daily panel with counterfactual predictions (690 season-days); (2) ten-event catalogue with LOWO-corrected resilience integrals and scan-null p-values; (3) event-cycle phase decomposition with bootstrap intervals; (4) station and hourly fingerprints for all catalogue events plus the 2022 Chengdu–Chongqing comparison; (5) supersite composition indicators; (6) difference-in-differences estimates for the September 2022 Chengdu stay-home order. Analysis scripts (Python) are included in execution order; see README.md inside the zip. Raw monitoring records and restricted supersite data are not redistributed (provenance documented in README). Data files are released under CC BY 4.0; scripts under the MIT License. Related manuscript submitted to Nature Communications; citation to be updated upon publication.