Data and code for: When can a pumping record answer what-if questions about stopping or raising pumping? An identifiability number tested in twin aquifers

Version 1.2.0 keeps the two archives of version 1.0.0 (twin-aquifer generator, candidate-family search, Pastas and TimesFM runs, statistics and earlier figures) and replaces the version 1.1.0 addition with pumping-record-answerability-v1.2.0-addition.zip for the revised manuscript. The addition contains: scripts that compute the linearized identifiability number Z for 4,434 records (1,950 designed records and 2,484 active operating-calendar records at 200 m and 20 m); a refit of the full-model minimum for every record with the same search engines, with the tolerance and zero-gain decision recomputed and the spread W re-searched with the production budget for the 3,372 records whose tolerance changed; the per-record tables, refitted minima, inclusion check and summaries from which every number of the revised manuscript is taken; the W search records of the re-search; and the scripts and PDF files of Figures 1–5. The folders mirror the version 1.0.0 project tree; the README and run_all.sh give the run order. All outputs are included; a rerun additionally needs the record inputs and truth arrays, which, as in version 1.0.0, are regenerated from the KMA station rainfall and not distributed. TimesFM 3.0 weights are not included. Changes from version 1.1.0: extras.py and table3_extra.json add the intervention result split by whether the 100 m³/d control week supplies the queried rate; make_figures.py changes two axis and legend labels, and Figures 1 and 5 are redrawn with them. No other file changed.