Search for: All records

Although vegetables are important for healthy diets, there are concerns about the sustainability of food systems that provide them. For example, half of fresh-market vegetables sold in the United States (US) are produced in California, leading to negative impacts associated with transportation. In Iowa, the focus of this study, 90% of food is imported from outside the state. Previous life cycle assessment (LCA) studies indicate that food consumption patterns affect global warming potential (GWP), with animal products having more negative impacts than vegetables. However, studies focused on how GWP, energy, and water use vary between food systems and vegetable types are less common. The purpose of this study was to examine these environmental impacts to inform decisions to buy locally or grow vegetables in the Midwest. We used a life cycle approach to examine three food systems (large-, mid-, and small-scale) and 18 vegetables commonly grown in/near Des Moines, Iowa. We found differences in GWP, energy, and water use (p ≤ 0.001 for each) for the three food systems with the large-scale scenario producing more emissions. There were also differences among vegetables, with the highest GWP for romaine lettuce (1.92 CO2eq/kg vegetable) approximately three times that of leaf lettuce (0.65 CO2eq/kg vegetable) at the large scale. Hotspots and tradeoffs between GWP, energy, and water use were also identified and could inform vegetable production/consumption based on carbon and water use footprints for the US Midwest. 

Most people in the world live in urban areas, and their high population densities, heavy reliance on external sources of food, energy, and water, and disproportionately large waste production result in severe and cumulative negative environmental effects. Integrated study of urban areas requires a system-of-systems analytical framework that includes modeling with social and biophysical data. We describe preliminary work toward an integrated urban food-energy-water systems (FEWS) analysis using co-simulation for assessment of current and future conditions, with an emphasis on local (urban and urban-adjacent) food production. We create a framework to enable simultaneous analyses of climate dynamics, changes in land cover, built forms, energy use, and environmental outcomes associated with a set of drivers of system change related to policy, crop management, technology, social interaction, and market forces affecting food production. The ultimate goal of our research program is to enhance understanding of the urban FEWS nexus so as to improve system function and management, increase resilience, and enhance sustainability. Our approach involves data-driven co-simulation to enable coupling of disparate food, energy and water simulation models across a range of spatial and temporal scales. When complete, these models will quantify energy use and water quality outcomes for current systems, and determine if undesirable environmental effects are decreased and local food supply is increased with different configurations of socioeconomic and biophysical factors in urban and urban-adjacent areas. The effort emphasizes use of open-source simulation models and expert knowledge to guide modeling for individual and combined systems in the urban FEWS nexus. 

Abstract Insect pests cause significant damage to food production, so early detection and efficient mitigation strategies are crucial. There is a continual shift toward machine learning (ML)‐based approaches for automating agricultural pest detection. Although supervised learning has achieved remarkable progress in this regard, it is impeded by the need for significant expert involvement in labeling the data used for model training. This makes real‐world applications tedious and oftentimes infeasible. Recently, self‐supervised learning (SSL) approaches have provided a viable alternative to training ML models with minimal annotations. Here, we present an SSL approach to classify 22 insect pests. The framework was assessed on raw and segmented field‐captured images using three different SSL methods, Nearest Neighbor Contrastive Learning of Visual Representations (NNCLR), Bootstrap Your Own Latent, and Barlow Twins. SSL pre‐training was done on ResNet‐18 and ResNet‐50 models using all three SSL methods on the original RGB images and foreground segmented images. The performance of SSL pre‐training methods was evaluated using linear probing of SSL representations and end‐to‐end fine‐tuning approaches. The SSL‐pre‐trained convolutional neural network models were able to perform annotation‐efficient classification. NNCLR was the best performing SSL method for both linear and full model fine‐tuning. With just 5% annotated images, transfer learning with ImageNet initialization obtained 74% accuracy, whereas NNCLR achieved an improved classification accuracy of 79% for end‐to‐end fine‐tuning. Models created using SSL pre‐training consistently performed better, especially under very low annotation, and were robust to object class imbalances. These approaches help overcome annotation bottlenecks and are resource efficient.