Advances in digital soil mapping based on machine learning

• Published in: Journal of agricultural resources and environment Vol. 41; no. 4; p. 744
• Main Authors: Shuai, Mei, Tong, Tong, Ying Chunyang, Wang, Tiantian, Zhang, Mei, Tang Mengmeng, Cai Tianpei, Ma, Youhua, Wang, Qiang
• Format: Journal Article
• Language: Chinese
• Published: Tianjin Journal of Agricultural Resources and Environment (JARE) 01.01.2024
• Subjects: Digital computers; Digital mapping; Geostatistics; Knowledge representation; Learning algorithms; Machine learning; Mapping; Real time; Soil mapping; Soil maps; Soils; Spatial distribution
• ISSN: 2095-6819; 2095-6819
• DOI: 10.13254/j.jare.2023.0347

通过数字土壤制图可以高效、精准地获取土壤信息。近年来，随着计算机学科的快速发展和土壤-景观模型被广泛认同，采用机器学习方法进行数字土壤建模已成为数字土壤制图的主流思路，这为土壤空间分布的定量解释提供了一种不同于地统计学、专家知识和个体代表性等传统制图技术的新模式。本文综述了国内外数字土壤制图领域的研究成果，从利用机器学习技术进行土壤制图的基本理论、制图方法、展望三个方面完整系统地阐述了数字土壤制图领域的主要进展，其中数字土壤制图方法包括特征信息的选择、制图模型的选择和土壤图的精度验证。研究结果为全面、实时和精确获取土壤信息空间分布提供参考。Digital soil mapping can facilitate acquiring soil information efficiently and precisely. In recent years, owing to the rapid development of computer disciplines and widespread recognition of soil-landscape models, digital soil modeling using machine learning has become a mainstream idea to provide new models for soil spatial distribution interpretation. These models differ from traditional mapping techniques such as geostatistics, expert knowledge, and individual representation. This study reviews the recent findings in the field of digital soil mapping nationally and internationally, and provides a complete and systematic description of digital soil mapping from three perspectives：basic theory, mapping method and outlook of soil mapping us