New research led by Budiman Minasny and Alex McBratney, professors at the University of Sydney Institute of Agriculture, explores how artificial intelligence (AI) tools can facilitate the adaptation of soils and the systems they support in response to climate change. The study, published in Frontiers in Science, delineates how AI can enhance soil science by accelerating preliminary research, refining predictions for land use, carbon management, and climate adaptation, efficiently processing complex data, and allowing scientists to concentrate on more nuanced questions that require expert judgment.
McBratney noted, “In partnership with experts, AI could help us better match the complexity and ever-changing nature of soil ecosystems. Unlike existing machine learning tools that focus on singular tasks, these systems can simulate scientific collaboration to a highly sophisticated degree—integrating reasoning, planning, and interdisciplinary insights to assist researchers and drive significant progress. As awareness of the crucial role of soil in planetary functioning grows, soil science will continue to evolve and thrive with the guidance of scientist-led AI.”
Currently, soil science employs machine learning techniques such as digital soil mapping and spectroscopy. AI systems could expand on these capabilities by generating digital soil twins from sensor data, improving soil microbiome monitoring, and testing climate adaptation strategies in simulations before applying them in the field to yield faster results.
To demonstrate this AI capability, the research team engaged a multi-agent AI system to review pertinent scientific literature and formulate ideas regarding carbon storage in soils and the factors that govern its limits. The study indicated that the AI agents successfully produced five hypotheses, encompassing climate influence, saturation thresholds, biological and chemical controls, interdisciplinary feedback, and management strategies.
These hypotheses were subsequently assessed through expert evaluation and simulated peer review, with the AI system effectively replicating essential aspects of the scientific process, yielding outputs that align closely with expert research.
Minasny stated, “Our findings highlight the potential for AI to accelerate soil research, which can have beneficial implications for our food and climate systems. A deeper understanding of soils could enable more sustainable agricultural practices, improved soil management, and enhanced climate adaptation, aiding land managers in early detection of nutrient loss, water stress, compaction, and erosion.”
The study also evaluated the system’s capacity for perceptual processing, strategic planning, and scientific reasoning, illustrating the promise that multi-agent AI systems present, with significant global ramifications for soil—a crucial, yet often undervalued resource.
Co-author Mercedes Román Dobarco from the Basque Institute for Agricultural Research and Development in Spain cautioned, “While the use cases are clearly compelling, and although AI can emulate certain aspects of expert reasoning, it cannot replace the contextual judgment, creativity, and critical interpretation that scientists contribute to research. AI agents also raise concerns regarding data quality, interpretability, creativity, and dataset bias—particularly in the absence of human oversight and domain expertise. Consequently, we should regard AI as an augmentative tool that enhances, rather than replaces, human scientific endeavors.”
The study was published on May 21, 2026.
