REPRODUCIBILITY PACKAGE: SYSTEMS SPECIAL ISSUE MANUSCRIPT

Title:
Digital Twin Ready Predictive Risk Management of Subsurface Injection Nodes Using Mechanics Informed Cross Fault Machine Learning

Run order:
1. python digital_twin_risk_analysis.py
2. python make_figures.py
3. python make_figures_9_10.py

Analytical input:
- ml_prepared.npz: processed 1 Hz event arrays used by the systems level analysis.
- event_catalog.csv: frozen 18 event catalogue and identifiers.

Primary outputs:
- policy_metrics.csv
- domain_support.csv
- domain_threshold_sensitivity.csv
- event_level_policy_results.csv
- risk_state_traces.csv
- uncertainty_summary.csv
- analysis_summary.json

The analysis reproduces the compact seven variable digital state, source calibrated risk policies, source only probability correction, cross fitted domain support gate, support eligible coverage metrics, event level outcomes, Wilson intervals, event block bootstrap intervals, and the data driven manuscript figures. It intentionally does not reproduce the companion geomechanics manuscript's broad 71 feature benchmarking, GRU, CORAL, permutation control, or SHAP result set.

The original experimental workbooks are publicly archived in Mendeley Data:
Wang, L. (2023), DOI: 10.17632/xcyvspjk3c.1.
