The case for AI-integrated monitoring in soft fruit production: development of the SPADE system

Soft fruit production is highly sensitive to environmental fluctuations and inconsistent human monitoring, which often leads to significant data gaps and suboptimal yields. The paper presents the SPADE system, an independent AI-integrated monitoring plat form designed for longitudinal field data collection in soft fruit production. Through a 103-day field trial, we demonstrate that manual monitoring by volunteers resulted in a 45.9% data loss rate, whereas the SPADE system maintained a consistent monitoring schedule regardless of operator availability. Technical evaluation shows that our edge-deployed quantized MobileNet V2 model achieved a plant identification accuracy of 90% and a precision of 1.000, ensuring reliable resource allocation without cloud dependency. Critically, plants under AI-integrated precision monitoring reached the f lowering stage approximately one week earlier than those subjected to manual feeding and watering schedules. While current yields were limited by first-season bare-root growth cycles, these results establish that automated precision monitoring provides a distinct developmental advantage. Future work will transition the SPADE system from a passive monitoring tool to an active control framework, integrating federated learning to optimize nutrient delivery across decentralized growing networks, thereby reducing the environmental footprint of soft fruit production.