The impact of 3D geometry complexity on the accuracy of simulating radiative, convective and conductive fluxes in an urban canyon was explored. The research involved the collection of meteorological data in an urban canyon in the city of Strasbourg, France, for input into a 3D model called LASER/F, which simulates the aforementioned fluxes. The key input into LASER/F is a 3D geometric model representing the scene (urban canyon). It was hypothesized that more geometric detail in the model would lead to improved accuracy in the simulation but increased computing time. To test this, seven geometry scenarios were prepared with different complexities and test run in LASER/F. The results were validated with thermal images of two facades collected during the field campaign in Strasbourg. The results show that LASER/F systematically underestimates facade surface temperatures possibly due to various model assumptions and input geometry. One of those is the oversimplified 1D parameterization of the vertical wind profile and was demonstrated by comparing it with a vertical wind profile extracted from a CFD model. It was concluded that the most complex 3D geometry does not necessarily improve simulation accuracy, especially during warming periods of the day, but does affect simulation time. The inclusion of balconies on the facades is influential and should be used for future simulation experiments when it is a significant feature of a facade. This implies that 3D simulations for the management of urban heat at canyon scale will need to consider the particularities of facades on a case-by-case basis when determining level of detail required for the input geometry. Further research is required to better understand canyon geometry effects such as canyon orientation, aspect ratio and the volumetric influence on internal heat storage in buildings. The investigation was carried out in association with ICube Laboratory (UMR 7357 CNRS-Université de Strasbourg), for the 2012 Geomatics Synthesis Project.