Regulatory provenance and event-selection sensitivity in reported startup NOx emissions
Description
This dataset contains reproducibility materials for the analysis of reported startup NOx events in regulatory monitoring data. The frozen 2023 Rhode Island analysis comprises five facilities, ten reporting units, 87,600 unit-hours, 1,257 operational startup events and 3,709 event-hours. Original event definitions, NOXR classifier assignments, policy totals and scientific numerical references are preserved. The original Version 1 files retain the historical production materials and figure/table sources for that version. The additional archive supplies subsequent explanatory diagnostics and evidence supporting the manuscript Auditable selection of regulatory emissions events using parameter provenance. These additions include direct bulk-versus-parameter membership comparisons, a decomposition of complete-event exclusions, the fixed 63-event C review list and eligibility-exception ledgers, twelve synthetic mechanism conditions, a primary-document claim/evidence matrix, exact computational task references and reconstructions, and the supplied Blackstone ORIS 1594 monitoring-plan export with its acquisition record and content audit. Figure 2 source values and optional plotting code are included. Historical table and figure numbering in Version 1 should not be interpreted as numbering in the current manuscript. Policy C uses the frozen NOXR-level classifier: MODC 01/14 are retained, 11 denotes the code-identified historical substitution pathway, and 12 remains unclassified by the study dictionary despite having an official EPA meaning. Policy D requires every observed hour of an event to satisfy C. An hourly subtotal and an eligible whole event are different analytical outputs. The additional diagnostics are post-hoc explanatory analyses. Computational checks demonstrate answer preservation and traceability; they are not an external human expert review or a human analyst-effectiveness experiment. The supplied Blackstone plan supports reported dated source linkage; it does not independently establish hourly compliance or physical measurement accuracy. The materials characterize reported-data handling, not true or corrected emissions. Original hourly EPA source bytes must be obtained separately; the supplied monitoring-plan export is included as expressly identified evidence.
Files
Steps to reproduce
Download both ZIP files from this version of the dataset and extract them into separate folders. To reproduce the original Rhode Island 2023 reference results, open JEM_Startup_NOx_Reproducibility_Package_v4.zip and follow its README.md, which documents python code/run_all.py and the optional reconstruction from original EPA source files. For the additional ESD diagnostics, audit tasks and Figure 2, open ProvScreen_ESD_Additional_Evidence.zip and follow its top-level README.txt and the instructions in the relevant subfolders.
Institutions
- Kocaeli ÜniversitesiKocaeli, İzmit