Crop-water assessment in Citrus (Citrus sinensis L.) based on continuous measurements of leaf-turgor pressure using machine learning and IoT

• Published in: Expert systems with applications 2022-12, Vol.209, p.118255, Article 118255
• Main Authors: Barriga, Jose A., Blanco-Cipollone, Fernando, Trigo-Córdoba, Emiliano, García-Tejero, Iván, Clemente, Pedro J.
• Format: Article
• Language: English
• Subjects: Expert system; IoT; Irrigation scheduling; Machine learning; Stem water potential; Turgor pressure
• ISSN: 0957-4174; 1873-6793
• DOI: 10.1016/j.eswa.2022.118255

Water is the most limiting natural resource in many semi-arid areas. This, together with the current climate change scenario, is fostering a context of uncertainty and major challenges concerning the sustainability and viability of existing agroecosystems. Crop water status based on three pre-established values (severe, mild, and no stress) is the essential datum needed to implement optimised irrigation scheduling based on deficit irrigation. Currently however, its calculation is a repetitive, tedious, and technical process carried out by hand. This communication presents a novel system based on continuous measurements of leaf turgor pressure to assess the crop water status when deficit irrigation strategies are being applied and/or to optimise irrigation scheduling in water scarcity scenarios. To this end, a novel expert system based on machine learning, together with an IoT infrastructure based on continuous measurements of leaf turgor pressure, is able to predict the citrus crop ψstem with a 99% F1 score. Thus, crop irrigation strategies involving irrigation-restriction cycles can be applied based on stem water potential. •An expert system to determine crop ψstem has been defined based on the IoT and machine learning.•A novel leaf-turgor pressure sensor has been presented.•A large experiment has been carried out to validate the expert system.•A crop irrigation-restriction strategy can be implemented from the ψstem predicted.•The machine learning experiments carried out to obtain the suitable model are described.