The measurements of surface temperature are prone to important directional anisotropy related to the structure of the canopy and the radiative and energy exchanges inside of it. Directional effects must be taken into account for a number of practical applications such as the correction of large swath satellite data, the assimilation of thermal infrared (TIR) measurements in surface models, the design of future spatial missions… For urban canopies, experimental measurements of TIR directional anisotropy previously performed during summer days over Marseille in the framework of the ESCOMPTE campaign (2001) revealed significant angular surface temperature variations with noticeable hot spot effects whose intensity was related to the canopy structure. The CAPITOUL project (http://medias.cnrs.fr/capitoul/) provided the opportunity to extend these results to other seasons and to nighttime conditions. The experimental setup is based on the use of 2 airborne TIR cameras with different lenses, inclination and resolution, and installed aboard a small aircraft. The flight protocol allowed the retrieval of directional anisotropy in all azimutal directions and in a range of zenith viewing angles between nadir and 62°. Measurements were performed during several intensive operation periods (IOP) in summer (2004 july), autumn (2004 September and October) and winter (2005 February). Only the first results of the 2004 autumn and 2005 winter IOPs are presented in this paper. The results obtained in daytime conditions confirm the systematic hot spot effects observed in previous experiments over cities. The variations found seem to be particularly important in winter when sun elevation is low: for instance they range between −4 and 10 K between oblique and nadir viewing in February. During nighttime conditions, angular variations are much lower (always less than 2 K between nadir and 60° zenithal viewing angle), whichever the azimutal viewing direction.