Digitalización del sector agroalimentario: de la investigación a la aplicación práctica mediante herramientas de monitorización y predicción

The global agri-food sector is currently facing multiple challenges arising from the pressure to increase production while using fewer resources, in a context shaped by population growth, climate change and market instability. In this scenario, digitalization and data analytics become essential tools to drive a significant transformation in agronomic management, optimizing resource use and supporting decision-making. However, advanced technologies still have a limited uptake in the sector, creating a gap between scientific progress and its practical application on farms, particularly in agriculturally important regions that face their specific challenges such as Andalusia. This PhD thesis, developed within the framework of an Industrial PhD program between University of Córdoba and the company Hispatec, addresses the issue with a dual objective: first, to identify the real needs and barriers of digitalization in the Andalusian agri-food sector; second, to develop and validate advanced analytical tools that, primarily using public or accessible data, respond to the identified demands and are directly applicable in an industrial environment. The research is structured into three main studies. First, an analysis is conducted on the state and perception of digitalization in the agri-food sector of Andalusia and its comparison with global research trends, combining bibliometric studies and surveys of sector stakeholders. The results reveal that, beyond complex technologies, farmers demand simple, interoperable solutions with a clear return on investment, oriented toward optimizing agronomic management. Second, phenological and yield prediction models for soybean cultivation are developed and validated in a real industrial context, showing reasonable accuracy (RMSE of 5-7 days for harvest date and MAPE of 8-18% for yield). Finally, traditional thermal models are compared with machine learning algorithms for the phenological monitoring of olive groves in southern Spain (Andalusia). The models yielded promising results for key phenological stages in olives (RMSE of 8 days for flowering and 12 days for veraison onset), thus demonstrating and analyzing their potential not only for farm-level crop management but also for regional monitoring of crop seasons by sectoral organizations and public administrations. The results reported herein provide predictive tools that are not only robust and cost-effective but also directly applicable, helping to bridge the gap between research and agricultural practice. The findings highlight the need for close collaboration between the public and private sectors to promote knowledge transfer, relying on methodologies that emphasize the use of open data and analytical techniques across diverse agricultural scenarios. Increasing the scalability and interoperability of solutions while balancing robustness and simplicity, this thesis contributes to the digitalization process of the agri-food sector and also lays the foundation for future research within the digital agricultural ecosystem.