We present a network model of the United States (U.S.) interstate food transfers to analyze the trade dependency with respect to participating regions and embodied irrigation impacts from a food–energy–water (FEW) nexus perspective. To this end, we utilize systems analysis methods including the pointwise mutual information (PMI) measure to provide an indication of interdependencies by estimating probability of trade between states. PMI compares observed trade with a benchmark of what is statistically expected given the structure and flow in the network. This helps assess whether dependencies arising from empirically observed trade occur due to chance or preferential attachment. The implications of PMI values are demonstrated by using Texas as an example, the largest importer in the U.S. grain transfer network. We find that strong dependencies exist not only just with states (Kansas, Oklahoma, Nebraska) providing high volume of transfer to Texas but also with states that have comparatively lower trade (New Mexico). This is due to New Mexico’s reliance on Texas as an important revenue source compared to its other connections. For Texas, import interdependencies arise from geographical proximity to trade. As these states primarily rely on the commonly shared High Plains aquifer for irrigation, overreliance poses a risk for water shortage for food supply in Texas. PMI values also indicate the capacity to trade more (the states are less reliant on each other than expected), and therefore provide an indication of where the trade could be shifted to avoid groundwater scarcity. However, some of the identified states rely on GHG emission intensive fossil fuels such as diesel and gasoline for irrigation, highlighting a potential tradeoff between crop water footprint and switching to lower emissions pumping fuels.

The feasibility of a hydrogeological modeling approach to simulate several thousand shallow groundwater-fed lakes and wetlands without explicitly considering their connection with groundwater is investigated at the regional scale (~40,000 km²) through an application in the semi-arid Nebraska Sand Hills (NSH), USA. Hydraulic heads are compared to local land-surface elevations from a digital elevation model (DEM) within a geographic information system to assess locations of lakes and wetlands. The water bodies are inferred where hydraulic heads exceed, or are above a certain depth below, the land surface. Numbers of lakes and/or wetlands are determined via image cluster analysis applied to the same 30-m grid as the DEM after interpolating both simulated and estimated heads. The regional water-table map was used for groundwater model calibration, considering MODIS-based net groundwater recharge data. Resulting values of simulated total baseflow to interior streams are within 1% of observed values. Locations, areas, and numbers of simulated lakes and wetlands are compared with Landsat 2005 survey data and with areas of lakes from a 1979–1980 Landsat survey and the National Hydrography Dataset. This simplified process-based modeling approach avoids the need for field-based morphology or water-budget data from individual lakes or wetlands, or determination of lake-groundwater exchanges, yet it reproduces observed lake-wetland characteristics at regional groundwater management scales. A better understanding of the NSH hydrogeology is attained, and the approach shows promise for use in simulations of groundwater-fed lake and wetland characteristics in other large groundwater systems.