Reproducibility data for “Overall rank association can mask weak upper-tail concordance between TROPOMI and official urban NO₂ networks”
Description
This dataset contains the reproducibility materials supporting the manuscript “Overall rank association can mask weak upper-tail concordance between TROPOMI and official urban NO₂ networks.” Version 3 preserves the frozen provenance, observation-support gates, event definitions and primary analytical outputs deposited in the earlier repository versions, and adds a self-contained analytical extension for cross-platform upper-tail concordance between Sentinel-5P/TROPOMI observations and official surface-monitoring networks in Tokyo, Warsaw and Mexico City for 2023–2024. The scientific V2 analysis extension included in repository Version 3 contains frozen city-day and station-day inputs; code and complete outputs for distribution-wide Spearman and Kendall rank association; full 2 × 2 upper-tail contingency metrics; direction-specific concordance; TSS, ETS, CSI, F1, MCC and kappa; 20,000 month-stratified circular-shift permutations per city-year; threshold sensitivity from q = 0.80 to 0.95; lag sensitivity from −2 to +2 days; network-coherence diagnostics; 4,000 equal-size seven-station network draws; reconstruction checks; audit tables; and publication figures. The q = 0.90 reconstruction check exactly reproduces the frozen primary event classifications and counts. Neither platform is treated as ground truth. The reported quantities describe observation-conditioned concordance among TROPOMI-supported days and should not be interpreted as direct validation or failure of regulatory monitoring networks. Provider-native monitoring files and original Copernicus products are not redistributed; their provenance and access routes are documented in the archive. No confidential, personal or respondent-level data are included.
Files
Steps to reproduce
1. Download and extract 07_NO2_Upper_Tail_Concordance_Reproducibility_Package_V3.zip while preserving its directory structure. 2. Review the top-level README.md. The 01_frozen_primary_pipeline/ directory preserves the Version 1 audit trail; the reproducible revised analysis is in 02_v2_upper_tail_analysis/. 3. Use Python 3.11 or later and install the required packages with pip install -r requirements.txt. 4. From the extracted archive, run python 02_v2_upper_tail_analysis/analysis/run_v2_analysis.py. 5. The script uses the fixed random seed 20261021, 20,000 month-stratified circular-shift permutations per city-year and 4,000 equal-size network draws. It recreates the CSV/JSON outputs in 02_v2_upper_tail_analysis/analysis/outputs/ and the PNG/TIFF figures in 02_v2_upper_tail_analysis/analysis/figures/. 6. Confirm exact recovery of the frozen q = 0.90 primary classifications with v2_q90_reconstruction_check.csv and verify archive integrity with SHA256SUMS.txt. 7. Provider-native monitoring files and Copernicus source products are not redistributed. The revised analysis is reproducible from the included frozen derived inputs; upstream acquisition and provenance are documented in the frozen primary-pipeline directory.
Institutions
- Kocaeli ÜniversitesiKocaeli, İzmit