This week’s “AI in Agriculture” 🌱
📖 Harnessing AI to decode the rhizosphere microbiome: https://www.sciencedirect.com/science/article/pii/S2662173825002127
📊 AI models adapted from human microbiome analytics handled dimensionality, compositionality, and sparsity through feature selection and normalisation.
⚙️ Future directions include federated learning for multi-site data integration, and both inside-out (hologenome) & outside-in (SynCom) strategies for microbiome-based improvement.
This week’s “AI in Agriculture” 🌱
📖 Exploring crop health and its associations with fungal soil microbiome composition using ML and remote sensing (Communications Earth & Environment, 2025: https://doi.org/10.1038/s43247-025-02330-0)
📊 Random forest models corrected NDVI for abiotic drivers, then associated residuals with fungal microbiome clusters.
✅ Combining remote sensing + microbiome data can be a promising route to early-warning systems.
This week’s “AI in Agriculture” 🌱
📖 A study of Universal ODE approaches to predicting soil organic carbon (arXiv, 2025: https://arxiv.org/abs/2509.24306)
📊 Hybrid Universal ODE model achieved R² ≈ 0.9999 in clean datasets, with moisture, temperature and microbial turnover as dominant SOC drivers.
⚠️ Performance drops with noisy inputs — a key challenge before field use.
✅ Hybrid physics–AI offers interpretable, process-based SOC forecasting for sustainable soil management.
Soil Organic Carbon (SOC) is a foundation of soil health and global climate resilience, yet its prediction remains difficult because of intricate physical, chemical, and biological processes. In this study, we explore a Scientific Machine Learning (SciML) framework built on Universal Differential Equations (UDEs) to forecast SOC dynamics across soil depth and time. UDEs blend mechanistic physics, such as advection diffusion transport, with neural networks that learn nonlinear microbial production and respiration. Using synthetic datasets, we systematically evaluated six experimental cases, progressing from clean, noise free benchmarks to stress tests with high (35%) multiplicative, spatially correlated noise. Our results highlight both the potential and limitations of the approach. In noise free and moderate noise settings, the UDE accurately reconstructed SOC dynamics. In clean terminal profile at 50 years (Case 4) achieved near perfect fidelity, with MSE = 1.6e-5, and R2 = 0.9999. Case 5, with 7% noise, remained robust (MSE = 3.4e-6, R2 = 0.99998), capturing depth wise SOC trends while tolerating realistic measurement uncertainty. In contrast, Case 3 (35% noise at t = 0) showed clear evidence of overfitting: the model reproduced noisy inputs with high accuracy but lost generalization against the clean truth (R2 = 0.94). Case 6 (35% noise at t = 50) collapsed toward overly smooth mean profiles, failing to capture depth wise variability and yielding negative R2, underscoring the limits of standard training under severe uncertainty. These findings suggest that UDEs are well suited for scalable, noise tolerant SOC forecasting, though advancing toward field deployment will require noise aware loss functions, probabilistic modelling, and tighter integration of microbial dynamics.
This week’s “AI in Agriculture” 🌱
📖 AgroLLM: Domain-specific LLM built with RAG over curated agri texts. ChatGPT-4o Mini reached 93% accuracy on 108 agri questions, outperforming Gemini & Mistral.
⚠️ Real-world deployment will need broader data and safeguards.
✅ Shows how specialised LLMs + retrieval can evolve into reliable digital advisors for farmers and researchers.
AgroLLM is an AI-powered chatbot designed to enhance knowledge-sharing and education in agriculture using Large Language Models (LLMs) and a Retrieval-Augmented Generation (RAG) framework. By using a comprehensive open-source agricultural database, AgroLLM provides accurate, contextually relevant responses while reducing incorrect information retrieval. The system utilizes the FAISS vector database for efficient similarity searches, ensuring rapid access to agricultural knowledge. A comparative study of three advanced models: Gemini 1.5 Flash, ChatGPT-4o Mini, and Mistral-7B-Instruct-v0.2 was conducted to evaluate performance across four key agricultural domains: Agriculture and Life Sciences, Agricultural Management, Agriculture and Forestry, and Agriculture Business. Key evaluation metrics included embedding quality, search efficiency, and response relevance. Results indicated that ChatGPT-4o Mini with RAG achieved the highest accuracy at 93%. Continuous feedback mechanisms enhance response quality, making AgroLLM a benchmark AI-driven educational tool for farmers, researchers, and professionals, promoting informed decision-making and improved agricultural practices.
This week’s “AI in Agriculture” 🌱
📖 Microbiome-driven ML for predicting soil suppressiveness against Rhizoctonia solani (Crop Protection, 2025: https://www.sciencedirect.com/science/article/abs/pii/S0929139325005475
📊 Random forest classifiers on microbial community data predicted soil suppressiveness vs Rhizoctonia. Certain taxa were consistently linked with suppression.
✅ Encouraging step: microbiome data can underpin quantitative models of soil disease resilience, moving closer to diagnostic soil health tools.
This week’s “AI/ML in Agriculture” hot take: Blueprints for sustainable plant production via crop wild relatives + microbiomes (Nat Commun, July 2025: https://www.nature.com/articles/s41467-025-61779-x
💡 CWRs carry microbes that boost nutrient use, stress resilience, disease suppression. ML helps untangle which wild relatives + microbial partners deliver the best traits.
✅ Fits into our HE COUSIN project, where aim to unravel CWR microbiome-mediated trait transfers to modern crops.
Authors discuss the potential of conserving crop wild relatives, together with the associated communities of microorganisms, to unlock strategies for improving crop resilience and achieving food security.
This week's "AI/ML in Agriculture"
📖 Exploring crop health and its associations with fungal soil microbiome composition (Communications Earth & Environment: https://www.nature.com/articles/s43247-025-02330-0
💡 Insight: Accounting for weather and soil, crop health patterns associate with fungal community clusters.
⚠️ Limitation: 115 samples → association study, not yet farm-ready.
✅ Takeaway: Satellites can screen, fungi tell the story. Use both to prioritise plots for bio-based interventions.
Crop health is improved by abundance of beneficial soil fungal genera and reduced by abundance of soil fungal pathogens, according to analysis of satellite-derived normalized difference vegetation index data and fungal soil microbiome taxonomic data.
Pierre Hohmann 🇺🇦