In this work we used several complementary techniques (TEM, TPR, CO chemisorption, EXAFS and FTIR spectroscopy) to understand the effects of the activation temperature and activation atmosphere (air or H2) on the particle size distribution, the fraction, and the type of exposed surface sites of Pd nanoparticles supported on a high surface area SiO2−Al2O3 (SA) support. Pd particle distribution has been carefully determined by a high statistic TEM study, from which the cuboctahedral-like shape of the metal particles is demonstrated. Assuming a model of perfect cuboctahedral particles, from the TEM particle size distribution we estimated the expected average Pd first shell coordination number. This value is slightly larger than that directly found by EXAFS owing to the fraction of very small Pd particles (d < 6−8 Å) that basically escape TEM detection. The same geometrical model allows prediction, from TEM particle size distribution, of the metal dispersion observed by CO chemisorption (S/VChemi). The S/VChemi value drops significantly upon increasing the H2-reduction temperature. According to TEM, the sintering process can account only for a very small fraction of the S/VChemi decrease, suggesting an important poisoning of the potentially available Pd surface. This hypothesis is supported by a parallel experiment of thermal decomposition at the same temperature (in absence of H2), showing a S/VChemi value almost unchanged. FTIR spectroscopy of adsorbed CO, probing the nature of the Pd surface available for adsorption, confirms the hypothesis.