||Though the geothermal reservoir at Soultz-sous-Forêts is targeted in a granite horst, exploration and production results tend to show that the Buntsandstein Formation, above the granitic basement, may also have geothermal potential. Indeed, temperature gradient anomalies observed in the Bunstsandstein, which is comprised of inter-bedded sandstones and argillite in the Rhine Graben, indicate that the Bunstsandstein could be exploited to produce heat or even electricity. Petrophysical properties measured on the Buntsandstein sandstone cores of borehole EPS1 (Soultz-sous-Forêts, France), and comparison between theoretical and measured temperature gradient logs, suggest that fluid flows exist and are strongly localized in the reservoir (presented in an associated paper of S. Haffen et al.). It has also been established that the fluid flow network consists in two types of conductive structural and geological features: sub-vertical faults and permeable sedimentary layers (the playa-lake and fluvio-aeolian marginal Erg facies). In order to identify flow boundary conditions that would explain the observed temperature gradient anomalies, a phenomenological study has been carried out using TOUGH2 to simulate fluid flow and heat transfer for various boundary conditions. Whether the fluid flows are mainly and independently concentrated in faults, or connect faults through permeable layers, different temperatures and temperature gradient logs are obtained that must be fitted to those measured in borehole EPS1. A simplified 2D vertical conceptual model has been used that integrates the different horizontal lithostratigraphic layers, associated with petrophysical properties measured on cores, and is bounded laterally with two vertical faults. Two flow and heat transport simulations were conducted, based on different flow boundary conditions, with the same temperature boundary conditions applying to both (prescribed temperatures at the top and bottom). In the first simulation, the same hydraulic potentials were applied to the two faults, resulting in non-interfering upward flows along the two faults. In the second simulation, the hydraulic potential was increased at the top of one fault, thus leading to transverse flows from one fault to the other through the sedimentary permeable layers, in addition to the vertical upward flows in the faults. For the first TOUGH2 model, the simulated temperature gradient log shows an opposite global trend compared with the one measured in the borehole. For the second TOUGH2 model, simulated and measured temperature and temperature gradient logs fit quite well at a distance of 40 m from the ‘outlet’ fault where the flow rates are higher. These results suggest that the temperature gradient anomalies observed in borehole EPS1 can only be explained by lateral flows connecting the two faults through the three lithostratigraphic levels previously identified as possible fluid circulation zones from petrophysical analyses. The differential of hydraulic potential gradient from one fault to the other required to allow fluid flow inside the rock formation can be explained by a structural tilt about 5°, as already noticed at Soultz-sous-Forêts. The tilt is the result of the extensive tectonic history of the Upper Rhine Graben, which is formed by a set of blocks acting as horst and graben. This model points out the interactions between microstructural parameters, macroscopic structures with tectonic and sedimentary origins, and a regional network.