A Bibliometric Review of Machine Learning Applications in Soil Science Using Digital Spectroscopy (2008–2024)

Abstract

Machine learning (ML) has rapidly advanced soil science, particularly in the last decade. This bibliometric study examines the scholarly production on machine learning in soil science from 2008 to 2024, analysing research impact, leading groups, and trends. The study reveals a significant increase in research output, with China emerging as the most prolific country after 2020. Collaboration networks highlight significant partnerships, with Zhejiang University and the Czech University of Life Sciences acting as central hubs. The analysis underscores the transition of ML in soil science from a nascent field to a well-established area characterized by increased research activity and interdisciplinary collaboration. Soil property prediction using ML techniques has been a major focus, with approximately 60% of studies dedicated to predicting organic carbon, nitrogen, moisture, texture, pH, and heavy metals. Digital soil mapping, soil profile analysis, and soil classification using ML account for about 20% of the research. Soil health assessment, including predicting soil health indicators, optimizing nutrient management, and developing remediation strategies, accounts for around 15% of the studies. Emerging trends include the rise of deep learning, the increased focus on soil health assessment, and the growing integration of ML with remote sensing techniques. While Chinese institutions lead in publication volume, organizations from Greece, Brazil, and Australia demonstrate higher research impact. The study underscores the importance of international collaboration in advancing this field and suggests a need to strengthen international networks and explore emerging interdisciplinary connections.