Effectively feeding a burgeoning world population is one of the main goals of sustainable agricultural practices. Digital technology, such as edge artificial intelligence (AI), has the potential to introduce substantial benefits to agriculture by enhancing farming practices that can improve agricultural production efficiency, yield, quality and safety. However, the adoption of edge AI faces several challenges, including the need for innovative and efficient edge AI solutions and greater investment in infrastructure and training, all compounded by various environmental, social and economic constraints. Here we provide a roadmap for leveraging edge AI at the intersection of food production and sustainability.

Reliable and readily available data on actual evapotranspiration (ET_a) over large-scale areas throughout the crop growing season are critical for improved agricultural irrigation and water resource management. On-site data collection is costly, labour-intensive, and very challenging in resource-limited conditions. Thus, open-source satellite-based approaches might be adopted as cost-effective alternatives. In this study, the performance of a cost-effective and open source satellite-based approach for estimating ET_a over a large-scale (1200 ha) irrigation system, the Kou Valley Irrigation Scheme (KVIS), in Burkina Faso was assessed. ET_a values over the critical irrigation period during the 2014 dry season (January-April) were estimated using the Python module for Surface Energy Balance Algorithm for Land model (PySEBAL). Then, they were compared against the Water Productivity Open-access (FAO-WaPOR), and United States Geological Survey-Famine Early Warning Systems Network Operational Simplified Surface Energy Balance (USGS-FEWS NET's SSEBop) ET_a over the same period at different temporal scales. Overall, ET_a values were satisfactorily estimated throughout the crop growth season across the Kou Valley irrigation scheme using PySEBAL. They spatially varied depending on the soil type and crop, with daily values ranging from 4.09 mm day⁻¹ to 7.7 mm day⁻¹, for a seasonal average of 619 mm. The finer spatial resolution (30 m) of PySEBAL outputs allowed better estimations compared to the FAO-WaPOR and SSEBop-based approaches. Our findings help ascertain the use of the PySEBAL model in semi-arid environment in Burkina Faso, and could serve as a basis for developing strategies for improved irrigation water management in countries experiencing similar conditions such as Burkina Faso.

Traditional methods based on field campaigns are generally used to assess the performance of irrigation schemes in Burkina Faso, resulting in labor-intensive, time-consuming, and costly processes. Despite their extensive application for such performance assessment, remote sensing (RS)-based approaches remain very much underutilized in Burkina Faso. Using multi-temporal Landsat images within the Python module for the Surface Energy Balance Algorithm for Land model, we investigated the spatiotemporal performance patterns of the Kou Valley irrigation scheme (KVIS) during two consecutive cropping seasons. Four performance indicators (depleted fraction, relative evapotranspiration, uniformity of water consumption, and crop water productivity) for rice, maize, and sweet potato were calculated and compared against standard values. Overall, the performance of the KVIS varied depending on year, crop, and the crop’s geographical position in the irrigation scheme. A gradient of spatially varied relative evapotranspiration was observed across the scheme, with the uniformity of water consumption being fair to good. Although rice was the most cultivated, a shift to more sweet potato farming could be adopted to benefit more from irrigation, given the relatively good performance achieved by this crop. Our findings ascertain the potential of such RS-based cost-effective methodologies to serve as basis for improved irrigation water management in decision support tools.