Machine learning frontiers in rhizosphere engineering: A comprehensive review of AI-driven nutrient bioavailability and soil carbon sequestration modelling

Abstract

The rhizosphere, a narrow zone of intense tripartite interactions between plants, soil, and the microbiome, is a critical biogeochemical hotspot for global food security and climate mitigation. Despite its significance, the "hidden half" of the plant remains difficult to manage due to the non-linear complexity of nutrient cycles and microbial stabilization pathways. This review evaluates the emerging frontiers of Machine Learning (ML) and Deep Learning (DL) in transitioning from traditional empirical soil models to predictive Rhizosphere Engineering. We focus on the application of advanced computational frameworks, including Physics-Informed Neural Networks (PINNs) and Graph Neural Networks (GNNs), to decode the mechanisms of nutrient bioavailability specifically the role of Potassium Solubilizing Bacteria (KSB) in mineral weathering. Furthermore, we analyse the integration of Explainable AI (XAI) and Digital Twins in modelling Soil Organic Carbon (SOC) stocks and Carbon Saturation Deficits, bridging the gap between micro-scale microbial necromass stabilization and macro-scale regional carbon sequestration. The synthesis highlights how AI-driven Variable Rate Application (VRA) of bio-fertilizers can optimize resource use efficiency while maximizing soil carbon sinks. However, challenges regarding data scale mismatch, "black box" interpretability, and the digital divide in small-holder farming systems remain. We conclude that the convergence of AI and soil biology marks a paradigm shift toward a carbon-negative, information-intensive agricultural future where the rhizosphere is no longer a misunderstood frontier but an engineered asset for planetary health.