The purpose of this study is twofold -first, to introduce the application of high-order discontinuous Galerkin methods to buoyancy-driven cargo hold fire simulations; second, to explore statistical variation in the fluid dynamics of a cargo hold fire given parameterized uncertainty in the fire source location and temperature. Cargo hold fires represent a class of problems that require highly-accurate computational methods to simulate faithfully. Hence, we use an in-house discontinuous Galerkin code to treat these flows. Cargo hold fires also exhibit a large amount of uncertainty with respect to the boundary conditions. Thus, the second aim of this paper is to use tools from the uncertainty quantification community to ensure that our efforts require a minimal number of simulations. We expect that the results of this study will provide statistical insight into the effects of fire location and temperature on cargo fires, and also assist in the optimization of fire detection system placement.