This research combines Geographic Information Systems (GIS) and Mixed Integer Programming (MIP) model to determine feedstock supply, preprocessing facility, and biorefinery locations at a highly resolved spatial scale with the objective of meeting a large-scale annual biofuel production and demand goal for Tennessee, USA. Simultaneous determination of this supply chain network using MIP is computationally expensive. Memory limitation of typical personal computers constrain the network problem. High Performance Computing environments may require days to solve, and without any assurance of a feasible solution. This paper proposes a two-stage optimization procedure to overcome these computational challenges while maintaining theoretical consistency with conventional firm location theory paradigms. The two-stage procedure assumes that firms first identify biorefinery locations with comparative advantage in terms of supplying biomass. Step two entails determining if costs can be lowered with the addition of preprocessing facilities, and their optimal location relative to the biorefinery sites. In contrast, a single step optimization procedure siting multiple biorefineries and concomitantly preprocessing units and biomass supply areas assumes a different strategy for a single firm. Results suggest that the two-stage optimization approach is able to identify a feasible solution of multiple feedstock areas, feedstock preprocessing and fuel production locations using high-resolution spatial layers over a relatively large geographic region in a fraction of the computer resources required to solve a single, simultaneous location problem. The optimal supply chain network solutions generated from the two-stage and single simultaneous approaches are different. The absolute difference in the smaller scaled models' objective values is less than 1%.