Data and code for comparing methods for assessing annual urban water supply conditions in four Yangtze River cities, 2018–2024
Description
This dataset supports the manuscript “Comparing Methods for Assessing Annual Urban Water Supply Conditions: A Case Study of Four Yangtze River Cities”. The study covers Wuhan, Yichang, Yueyang and Jiujiang during 2018–2024, comprising 28 city–year observations and 17 indicators organised within a Driving force–Pressure–State–Response (DPSR) framework. The analysis input comprises 476 indicator values compiled or derived from official statistical yearbooks, statistical communiqués, water resources bulletins and the China Urban Construction Statistical Yearbook. The files include the final numerical input, an indicator dictionary, source and derivation documentation, portable Python code, and computed results. They support comparisons of five evaluation methods, a supplementary combined-weight linear reference, 15 weight–aggregation configurations, and sensitivity analyses of indicator deletion, direction and grade-function parameters. FCE monotonicity probes are also included. The data describe retrospective annual composite water supply conditions and should not be interpreted as direct measurements of operational water-supply safety. The source documentation preserves historical verification statuses and remaining provenance limitations. Computational reproduction checks agreement with archived analysis outputs; it does not certify every primary-source value. Original publications and third-party GIS assets are not redistributed.
Files
Steps to reproduce
1. Download Water_Supply_Data_Code.zip and extract it. Read README.md and documentation/SOURCES_AND_DERIVATIONS.md before using the data. 2. Use Python 3.12 with NumPy 2.3.5 (the tested environment used Python 3.12.14 on Windows). From the extracted Water_Supply_Data_Code folder, install the dependency with: python -m pip install -r requirements.txt 3. Optionally verify file integrity with: python -B code/verify_integrity.py 4. Reproduce the numerical analyses with: python -B code/reproduce.py --output reproduced_results The output directory must be new. Input data and supplied results are not overwritten. 5. Inspect reproduced_results/validation.json and summary.json. A successful run reports passed=true and checks 168 baseline scores and 92,116 numerical fields against archived final outputs. The outputs include full-panel normalisation, three weight vectors, six baseline scores, 15 weight–aggregation configurations, 17 single-indicator deletions, planned direction/deletion scenarios, parameter sensitivity, transition diagnostics, contribution decompositions and FCE monotonicity probes. The canonical input is data/input_data.csv. Preserve its numerical precision. See documentation/OUTPUT_GUIDE.md for file definitions and documentation/SOURCES_AND_DERIVATIONS.md for source limitations. Original source PDFs and third-party cartographic assets are outside this package; the numerical scores underlying Figure 2 are supplied, but pixel-identical figure recreation is not included.
Institutions
- Changjiang River Scientific Research InstituteHubei, Wuhan