Datasets and codes used in "Mechanisms Governing Boulder Distribution and Size Dynamics in the Lunar Apollo Basin: Insights from Thermal Stress, Impacts, and Geological Units"

This catalog presents all lunar boulder datasets and processed inputs within the XGBoost-SHAP analytical framework for the research study "Mechanisms Governing Boulder Distribution and Size Dynamics in the Lunar Apollo Basin: Insights from Thermal Stress, Impacts, and Geological Units". Primary Boulder Dataset The primary boulder dataset is derived from Bickel (2021; https://edmond.mpg.de/dataset.xhtml?persistentId=doi:10.17617/3.OG927P). The analysis utilizes the 60°-capped catalog with confidence threshold (CT) values > 0.8, from which boulders within the Apollo basin were extracted. Key focus areas include boulder location coordinates and diameter estimations. For detailed methodologies behind dataset generation and boulder diameter calculations, refer to: Bickel et al. (2020; https://www.nature.com/articles/s41467-020-16653-3) Bickel et al. (2018; https://ieeexplore.ieee.org/document/8587120) Analytical Framework and Data Structure XGBoost-SHAP Codebase:The computational framework is provided in lunar_boulder_xgboost_shap.rar (Python implementation). The input dataset (raster_df.csv) is structured as follows: The study area is gridded into 1°×1° cells, with each row representing a grid and each column an attribute. Column raster0 denotes the kernel density of boulders in the grid. Subsequent columns contain control variables (e.g., thermal stress, impact parameters, geological units). Data Normalization:Boulder density values and continuous control variables were normalized prior to modeling. Categorical numerical variables (geological unit types) were excluded from normalization.