AI-Driven Soil Moisture Forecasting for Enhanced Precision Agriculture

This paper presents the development of a Machine Learning model aimed at accurately forecasting soil moisture levels, a crucial aspect of precision agriculture. The model leverages data from tensiometers, devices that measure soil water potential, installed in two distinct agricultural areas in Trentino, Italy. Utilizing a Long Short-Term Memory neural network, the model effectively captures and predicts the temporal dynamics of soil moisture. The data, sourced from multiple tensiometers with varying temporal and spatial frequencies, undergoes preprocessing to align sampling times and integrate environmental factors such as air humidity, temperature and irrigation data. Our results demonstrate the strong trend-following capabilities of the model and an inherent ability to predict when soil moisture values cross critical agronomic thresholds, which is essential for optimizing irrigation schedules. These preliminary findings suggest that the proposed model could be a valuable tool in advancing precision irrigation practices, contributing to more sustainable and efficient agricultural production.