Dataset for: “From Biological to Structural Controls: Explainable Machine Learning Reveals a Fundamental Shift in Soil Carbon Drivers Across Depth in Subtropical Plantations”

This dataset supports the findings of the manuscript entitled “From Biological to Structural Controls: Explainable Machine Learning Reveals a Fundamental Shift in Soil Carbon Drivers Across Depth in Subtropical Plantations,” submitted to the journal “Catena”.
Study Overview:
The study aimed to quantitatively uncover the depth-dependent drivers of soil organic carbon (SOC) and microbial biomass (MBC, MBN) in subtropical plantations using explainable machine learning (Random Forest and XGBoost coupled with SHAP analysis). Data was collected from 13-year-old monoculture and intercropping plantations of three species (Ligustrum lucidum, Osmanthus fragrans, and Cinnamomum camphora) in the Shihe River Basin, Henan Province, China.
Data Content:
The dataset comprises two main parts:
1. Stand Inventory Data: Tree species, planting pattern (monoculture/intercropping), mean diameter at breast height (DBH, cm), mean tree height (m), canopy density (%), and stand density (trees/ha).
2. Soil Property Data: Samples were collected from two depth intervals (0-10 cm and 30-50 cm). Analyzed properties include: Soil organic carbon (SOC, g/kg), Particulate organic carbon (POC, g/kg), Easily oxidized organic carbon (EOC, g/kg), Total nitrogen (TN, mg/kg), Ammonium nitrogen (NH₄⁺-N, mg/kg), Microbial biomass carbon (MBC, mg/kg), Microbial biomass nitrogen (MBN, mg/kg).
Data Availability and Usage:
This dataset is embargoed until December 31, 2026 to allow for the publication of the associated manuscript. Upon expiration of the embargo, the data will be fully accessible under a CC-BY 4.0 license. Researchers are encouraged to cite this dataset if used in their work.