Data and Code for Stage-Specific Shoreline Evolution under Offshore Artificial Island Construction and Removal

Published: 9 November 2026| Version 1 | DOI: 10.17632/hphwdpydg2.1
Contributor:
少勤 李

Description

This dataset supports the article “Stage-Specific Shoreline Evolution under Offshore Artificial Island Construction and Removal: An Interpretable Machine Learning Assessment of Riyuewan Bay, China.” It contains 35,586 transect-date-pair records, date-level model data, processed mainland shorelines covering 562 unique acquisition dates from 2000 to 2025, 292 fixed transects, feature definitions, engineering-stage chronology, image quality-control and registration records, exact date-grouped, chronological, and spatial-block split assignments, model outputs, and Python scripts for model training, validation, sensitivity analysis, TreeSHAP attribution, and figure generation. Raw Landsat, Sentinel-2, ERA5, and EOT20 products are not redistributed; their sources and relevant identifiers are documented in the archive. The released material supports reproduction of the article-level conditional diagnostic analyses. TreeSHAP results represent model attribution rather than causal effects, and the model is not intended as a deterministic forecast of future shoreline positions.

Files

Steps to reproduce

1. Download and extract the supplementary archive while preserving its folder structure. 2. Create a Python environment and install numpy, pandas, scikit-learn, xgboost, and matplotlib. 3. Run the principal model-family comparison: python code/run_reviewer_model_family_comparison.py 4. Generate the final XGBoost TreeSHAP attribution outputs: python code/run_final_xgboost_shap_attribution.py 5. Run the remaining validation, sensitivity, association, and diagnostic analyses: python code/run_reviewer_remaining_analyses.py 6. Reproduce the associated diagnostic figures: python code/plot_reviewer_remaining_figures.py The model-ready input tables are provided in the data directory. Exact date-grouped, purged chronological, and spatial-block assignments are provided in the splits directory. Model outputs are written to the results directory. Random seeds used in the reported analyses are 11, 23, 42, 77, and 101.

Institutions

Categories

Computer Science, Engineering, Environmental Science

Funders

  • Science & Technology Fundamental Resources Investigation Program
    Grant ID: 2022FY202404
  • National Natural Science Foundation of China
    Beijing, Beijing
    Grant ID: 41906184
  • Fund of Key Laboratory of Marine Ecological Conservation and Restoration, Ministry of Natural Resources/Fujian Provincial Key Laboratory of Marine Ecological Conservation and Restoration
    Grant ID: EPR2024003
  • Fujian Provincial Science and Technology Innovation Project of Natural Resources
    Grant ID: KY-080000-04-2025-014

Licence