IoT and Machine Learning for the Forecasting of Physiological Parameters of Crop Leaves

Abstract

In the current context of accelerating climate change, agriculture faces a pressing need for proactive solutions that can help maintain crop health and improve resource use efficiency. Physiological parameters of crop leaves, such as Leaf-Turgor Pressure ( P leaf ) and Leaf Temperature ( Tleaf ), serve as valuable indicators of crop water stress and health, providing essential information for orchard management. However, although Internet of Things (IoT) systems equipped with sensors attached to crop leaves make it possible to monitor these parameters, they only provide real-time measurements, which are often insufficient to anticipate and mitigate adverse future conditions. Therefore, this study proposes a novel integration of IoT and machine learning technologies to enable the forecasting of physiological parameters of crop leaves. For this purpose, an experimental plot of Japanese plum trees has been monitored for five months using P leaf and T leaf sensors. Additionally, weather conditions were recorded. Using the data gathered, machine learning algorithms have been applied to train learning models for Pleaf and Tleaf forecasting. Results report Support Vector Regression as the best algorithm with R-squared values of 0.96 and 0.99 in the forecasting of Pleaf and Tleaf , respectively (one-week forecast horizon). In addition, a comprehensive digital twin software system integrating the forecasting models has been proposed. Thus, this study represents a significant breakthrough in proactive crop management, laying the groundwork for more sustainable and resilient farming practices.