The energy and moisture exchange between the surface and the atmosphere in urban areas is characterised by the net radiation, the sensible and latent heat flux, the storage of heat in the construction materials as well the anthropogenic heat flux. The anthropogenic heat flux is mainly due to heating and cooling of buildings, industrial activities, electricity consumption for household equipment and lighting as well as human metabolism. The energy demand for heating and cooling of buildings depends on the meteorological conditions. For this reason, it has become usual to couple urban canopy parametrisations (UCP) for atmospheric models like the Town Energy Balance (TEB) with a simple Building Energy Model (BEM) to interactively simulate the heating and cooling energy demand as a function of the prevailing meteorological conditions. With this approach, one part of the anthropogenic heat flux is no longer prescribed (e.g. via an inventory), but interactively simulated and thus becomes a prognostic variable. Various parameters characterising the urban morpohology (e.g. building surface cover fraction, building height), the construction materials and the building use (e.g. design temperature for heating and cooling) need to be known in order to perform a simulation with an UCP-BEM model. These parameters cannot be known perfectly and therefore the approach of interactively simulating the heating and cooling energy demand introduces a new uncertainty in the simulation of the urban surface energy budget. For this reason it is necessary to critically evaluate the spatio-temporal variability of the anthropogenic heat flux simulated using the UCM-BEM approach. In this study, we use TEB-BEM to simulate the spatio-temporal variability of the meteorology-dependent anthropogenic heat flux for the agglomeration of Toulouse in southern France. Data describing the urban morphology is taken from the French 3D vector description of the territory (BD TOPO). Information on building construction materials and the presence of insulation materials as a function of building type, construction period and building use is retrieved from a database describing building archetypes in France. The most relevant human-behaviour-related parameters are the design temperatures for heating and cooling, the internal heat release and the usage of shading and ventilation by the inhabitants. These parameters are estimated via a statistical modelling approach based on surveys and socio-economic data available from French census data. In a first step, TEB-BEM is forced with meteorological parameters observed at a meteorological mast at ~30 m above roof level during the CAPITOUL campaign for March 2004 to March 2005 („offline“ approach). In a second step, TEB-BEM is coupled to the mesoscale atmospheric model Méso-NH. For both, the „offline“ and „coupled“ approach, the spatio-temporal variability of the simulated anthropogenic heat release is evaluated against observed electricity and gas consumption data available for the CAPITOUL period at daily temporal resolution and a spatial resolution of 100x100 m².