Bulk density datasets using pedotransfer functions in the Qinghai-Tibetan Plateau

• Published in: Scientific data Vol. 12; no. 1; pp. 649 - 10
• Main Authors: Gu, Jun, Ye, Ming-Liang, Song, Xiao-Dong, Yang, Fei, Zhang, Gan-Lin
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 17.04.2025; Nature Publishing Group; Nature Portfolio
• Subjects: 704/158/1144; 704/47/4113; Accuracy; Carbon; Climate change; Data Descriptor; Datasets; Ecology; Environmental protection; Environmental restoration; Humanities and Social Sciences; Machine learning; Mapping; multidisciplinary; Neural networks; Particle size; Regions; Regression analysis; Science; Science (multidisciplinary); Soil compaction; Soil properties; Soil sciences; Spatial heterogeneity; Sustainable development; Tibetan Plateau
• ISSN: 2052-4463; 2052-4463
• DOI: 10.1038/s41597-025-04983-0

Bulk Density (BD) represents a critical soil property essential for understanding soil compaction, element stocks, and ecological processes. However, obtaining accurate BD data in large and diverse regions, especially in complex terrain such as the Qinghai-Tibetan Plateau (QTP), remains challenging owing to the labor-intensive measurement procedures. In this study, we integrated soil pedotransfer functions and machine learning techniques to predict and generate a 90 m resolution BD grid dataset for the QTP. This dataset effectively captures the spatial heterogeneity of BD across the region. We rigorously validated the accuracy of the maps and assessed the reliability of uncertainty quantification, achieving RMSE values ranging from 0.144 to 0.190. The high spatial and depth resolution renders the data suitable for various practical applications, including element storage estimation and soil water transport modeling. These maps provide a robust scientific foundation for informing regional ecological restoration efforts, environmental protection initiatives, and the formulation of sustainable development strategies.