Classification of land cover in urban areas can play an important role in urban planning decisions and in characterizing urban materials properties such as reflectance. Taking into account the large offer of new and future remote sensing sensors with different spectral and spatial characteristics, it is important to compare their classification performances in urban area. To this aim, this work simulates from airborne data the at sensor images acquired bythree space borne instruments (Pléiades, SENTINEL-2 and HYPXIM) in the Visible Near Infrared (0.4 µm-1.0 µm) and Shortwave Infrared (1.0 µm-2.5µm) spectral ranges. Five classification maps with 8 land cover classes over the city of Toulouse (France) are generated with a Support Vector Machine rule. Correct values of accuracy are obtained in all cases (kappa coefficient higher than 0.65 and overall accuracy better than 70 %). Nevertheless, coarser spatial resolutions do not allow mapping urban details and SWIR data was necessary to discriminate between classes.