In this paper we assess the capacity of satellite images to explain and predict bird community patterns in farm-wood landscapes in southwestern France. Our goal is to examine the effect of the images’ acquisition date and spatial resolution on the models’ performance. We also seek to assess whether unclassified images provide results comparable with classified data (i.e. land-cover map). To do that we constructed species richness models (generalized additive models) based on a sample of 573 counting points and on non-classified images made up of NDVI data and digital height model (DHM), making it possible to quantify the spatial and vertical heterogeneity of habitats. To assess the acquisition date effect, we compared the performance of NDVI data acquired on four different dates (February 4th, June 24th, August 19th and October 18th, 2009) by the same sensor (SPOT-5). To assess the spatial resolution effect, we compared five NDVI images acquired over an identical period (September 2010) but by different sensors (WorldView-2, SPOT-5, SPOT-4, Landsat, MODIS) as well as two DHMs obtained from LiDAR (1 m) and radar (5 m) data. Our results show that for a constant spatial resolution (10 m) it is the NDVI data acquired at the beginning of the autumn that provide the best performance. These data better reveal the landscape requirements of birds during the breeding period. For a given period (September 2010), the higher resolution spatial data (2 m) are the highest performing. However, in view of the cost of WorldView images, we suggest that 10 m data (SPOT-5) provide a good trade-off for studying the distribution of bird communities. For the height data (DHM), the effect of the spatial resolution is not significant. The differences of performance between the spatial resolutions of NDVI data are not as great as those between the data acquisition periods. The performance of unclassified data (NDVI or DHM) is also comparable with that of land-cover maps. This suggests on the one hand that the choice of the NDVI image date is more important than that of the spatial resolution and on the other hand that the NDVI or DHM data are good alternatives to classified data when constructing a bird-habitat predictive model.