Data-Driven Optimisation of Irrigation Dose Using Machine-Learning Ensembles for Sustainable European Agriculture

This study focuses on predicting irrigation doses using digital technologies and statistical modelling to enhance water resource management in agriculture. Conducted as part of the CODECS project in the semi-arid Nitra region of Slovakia, this study aimed to evaluate the effectiveness of various irrigation systems and to develop predictive models for optimal irrigation doses. The methodology integrates environmental sensor data, agronomic models, and machine learning techniques, utilizing IoT sensors alongside Valley and Irriga control software. A significant challenge was the incompatibility of heterogeneous data from different sources, leading to the creation of a unified method-ology for data collection, validation, and analysis. Analytical tools, such as ex-ploratory data analysis, correlation techniques, and regression models, were employed to identify key factors affecting irrigation efficiency, including precipitation, temperature, soil moisture, and energy consumption. The findings aim to inform sustainable irrigation strategies that reduce water usage, enhance crop productivity, and safeguard soil resources under changing climatic con-ditions.