It is critically important to understand and improve the robustness and adaptability of our Food Supply Chain, making it more resilient to disturbances in food supply and demand networks. There is inherently a risk of network fracturing and delayed recovery during extreme weather events, wildfires, floods, and other natural hazards. In the face of uncertain natural hazards of increasing frequency and severity, it is vital that the implications of these disruptions are evaluated for the source nodes of our supply chains, such that resiliency in the whole supply chain can be promoted. Ensuring this requires bridging across topics and domains, but data are siloed, distributed, variously described, and have regional variations that create different needs and solutions. To solve this challenge, we are partnering with the Food Industry Association (FMI), which has identified food safety and food quality issues rising from environmental disasters or disturbances as a high-priority industry concern. Together we make use of the graph and enrichment services to link across data silos from multiple domains with rich public data to address the following pilot use cases:
Our initial work has focused on leafy greens tainted in the field by ash falling from wildfires. FMI members want to understand better when and where a tainted product is and how to supplement similar products (i.e., from alternate sources) in the supply chain. Example graph queries include: (1) What percentage of product from a specific smoke-affected area is potentially tainted? (2) Can we identify and notify key stakeholders in the supply chain about a product that may be (or may become) tainted? (3) How can we best anticipate and react to shortages of a product that the current disaster may taint? (4) Are the current events comparable to previous situations both in the short term and in relation to long-term trends or are they outliers? FMI members currently carry out these analyses internally, with limited data inputs, dependent on available internal knowledge, and often only in retrospect.
Through this partnership, we have delivered graph-based tools that substantially enhance current capabilities for assessment and strategic planning during real-time events/issues by providing online analysis, forecasting and alerts, enriched with location and context-specific intelligence, ensuring key stakeholders throughout the supply chain are ready with backup strategies to keep products moving. Such a system also has large implications for the farmers/growers. By providing informative analytics and predictive and informative capabilities, it allows them to identify how they can be better prepared to mitigate and/or build resilience in the face of such events. Interestingly, one of the key challenges from a graph perspective is that data are widely available but difficult to interlink, e.g., data layers for air quality, smoke plume simulations and observations, and wildfires and their burn intensity exist but cannot be easily combined to answer questions about the impacts of one specific fire. Within our graph, we perform parts of the analysis and processing and then serve the pre-integrated data together with topological information about the affected areas. While each of these datasets come from public sources, such as the National Oceanic and Atmospheric Administration, information about individual fields and farmers is not. Public information is linked to the graph to enable our use case and provide further services such as link prediction. In addition, the KnowWhereGraph supports authenticated access to privileged data that is not accessible as part of the publicly available graph.