Characterizing and modeling subnational virtual water networks of US agricultural and industrial commodity flows

Virtual water flows have been extensively analyzed at the national level using complex network approaches. However, less is known about the regional structure of subnational virtual water flows, even though virtual water flows can vary greatly within a country. Additionally, subnational-level studies are needed since water policy and decision-making tend to be local or regional in scope. Here, we characterize and model virtual water flows for aggregated agricultural and industrial commodities at the US subnational level. Using subnational trade data for the US, we build and analyze unweighted and weighted, directed virtual water trade networks (VWTNs). To model and explore the drivers of subnational virtual water flows, we build and implement a gravity-type spatial interaction model. Using different network metrics, we find that the subnational VWTNs differ from previous well-studied networks, including national-level VWTNs. The network metrics also show a high connectivity for the unweighted VWTNs with no community structure, while the weighted VWTNs reveal spatially coherent communities of intense trade activity. The gravity model shows that the subnational weighted VWTNs are mainly controlled by distance, agricultural land, gross domestic product, and population. Despite the high connectivity of the VWTNs, the presence of community structure indicates that large volumes of virtual water are traded regionally. This suggests the possibility of having hydroeconomic boundaries that differ from known physical boundaries, e.g., watersheds and aquifers. Such boundaries could have implications for the design of consumption-based strategies for water sustainability.