Proximal Soil Sensing through AI-Driven Edge-Cloud Workflows in Agriculture Fields

In the face of global soil degradation and increasing demands for sustainable land management, advancing in-situ soil monitoring iscritical. This abstract presents the outcomes of the EMPYREAN Horizon Europe project, specifically focusing on the use case “Proximal Sensing in Agricultural Fields”. This use case leverages edge computing, AI-based data analytics, and robust IoT infrastructure toenable high-resolution, real-time soil data acquisition and decision-making support.The architecture integrates soil spectrometry systems, deployed on mobile platforms, with edge-enabled devices that handle initial datapreprocessing and inference using locally trained AI models. These models are optimized to extract soil organic carbon (SOC) indicatorsfrom hyperspectral and Vis-NIR reflectance data. Workflows developed within the EMPYREAN platform, enable dynamic orchestrationacross devices, edge nodes, and cloud backends. This architecture ensures efficient handling of large datasets, continuous systemadaptation, and seamless communication within heterogeneous and sometimes intermittently connected environments.Central to this approach is the novel use of AI-driven orchestration mechanisms for adaptive workload placement, vertical autoscaling ofcontainerized inference tasks, and integration with EMPYREAN’s distributed Decision and Telemetry Engines. These capabilities allowfor intelligent routing of soil sensing tasks to optimal nodes based on network conditions, computational load, and energy efficiencyconsiderations.Preliminary integration of the system components has demonstrated the feasibility and robustness of the proposed architecture underrealistic agricultural conditions. Initial deployments show promising performance in handling heterogeneous sensor inputs, executingAI inference pipelines efficiently at the edge, and dynamically orchestrating workloads across resource-constrained environments. Thesystem exhibits a high degree of modularity and adaptability, supporting low-latency data processing and scalable model deploymentacross varying field configurations. These developments pave the way for comprehensive evaluation in diverse agricultural scenarios,aiming to validate system performance in real-time SOC monitoring, data-driven decision support, and resource optimization.This contribution reflects a step forward in embedding digital innovation into soil science and monitoring. By integrating AI, edge-cloudorchestration, and proximal sensing technologies, EMPYREAN’s approach addresses key soil preservation challenges and supportsthe broader EUROSOIL mission to foster actionable, technology-driven strategies for sustainable soil stewardship.