FoodGravity: understand food flows using classic Gravity model and explainable Artificial Intelligence techniques

In the past few years, unexpected disruptions such as the Ukraine war and the global pandemic reveal the fact that our food systems are highly connected and the physical flow of food is vulnerable to shocks. This calls for insights into how food is moved in space, which is essential to build resilient food systems that are connected by the flows. However, only national-level food flows (e.g., FAO trade data) are publicly available and most studies are conducted at country level. Data on inter-district food flows are typically not collected and thus not available to enhance resilience for disasters or analyze regional-specific deficits in food production and supply. Our knowledge of why and to where food moves in space are impeded by the empirical and model gaps. Yet, inter-district food flows (i.e., the amount of food that moves from production districts to consumption districts) are extremely useful to identify the weak links and disasters of the flows at a high-precision level and map the regions and districts that may need humanitarian support.