European Journal of Sustainable Agroecosystems

Background: Background: The increasing water shortage, unpredictable precipitation, and heightened evaporation requirements due to climate change are combining to pose a serious danger to rainfed and irrigated cropping systems. Coupled soil-plant-atmosphere models, which describe the persistent transfer of energy and water in the soil-root-canopy-atmospheric chain, are becoming the main means of converting this machine process into a practical irrigation guide.
Aim of the review: In this review, we summarize the existing peer-reviewed articles on coupled soil-plant-atmosphere modeling applied to irrigation purposes. We discuss the general structure of the mechanistic models, various types of sensing technologies and machine-learning irrigation techniques, as well as their impact on the decision-making process in smarter agriculture. 
Literature search process: A systematic review of international literature was conducted through Scopus, Web of Science, PubMed, Google Scholar, and Food and Agriculture Organization of the United Nations (FAO) reports published during the period of 2015–2026 and pre-2015 landmark studies. Out of 842 preliminary references, 96 were selected for thematic synthesis after checking the inclusion criteria. 
Important findings of the research: Mechanistic SPAC models, such as CropSPAC, GEOSPACE, and CFD-coupled canopy models, have been found to provide satisfying results consistent with field observations of soil water retention and crop growth, but are too demanding in terms of computation and data. Remote sensing and proximal sensor technologies have achieved sub-field soil moisture and evapotranspiration information at the required resolutions for operational scheduling, while machine learning and deep learning techniques, especially hybrid and reinforcement methodologies, yielded water conservation results between 30 and 50% compared to traditional scheduling techniques with accuracy of prediction at least 90%. The introduction of digital twin IoT technologies has resulted in the integration of above-mentioned techniques into the process of decision making, however many of them are still being tested. 
Research gaps: Robust gaps still persist concerning a lack of validation across scales, limited inter-operability between mechanistic and data-driven models, unavailability of standardized benchmark datasets, limited heterogeneity of microclimates, and the deployment gap for smallholders in contexts with a lack of resources. 
Conclusion: Combining physical SPAC models and decision support boosted with sensor technology and AI has the highest potential for developing climate-resilient and efficient irrigation technologies, although success relies on the availability of benchmark open datasets, models of understood architecture, and cost-effective solutions.