Recent experimental and theoretical studies highlighted the importance of the growing surface structure on the final morphology of GaN. Actually, optimum morphology is achieved by growth in presence of a Ga bilayer adsorbed on the GaN surface. The threshold fluxes limiting the region of the Ga bilayer adsorption have been measured as a function of the GaN substrate temperature, giving rise to a Ga adsorption phase diagram. The Ga flux limiting the regions in the adsorption phase diagram exhibit a linear behavior in an Arrhenius plot. However, both energy activation (about 5 eV) and prefactor (in the 1025 range) are surprinsingly high. These questions were adressed by studying the adsorption/desorption of Ga adatoms and small islands (consisting of 2 and 3 Ga adatoms) on the Ga bilayer surface employing first principle density functional theory calculations. We find a desorption barrier of 2.1 eV and a binding energy between two Ga atoms of approximately 0.3 eV. Using these numbers we derived a simple growth model (based on rate equations). An analysis of the experimental data with the model revealed the origin of the large difference in the activation energies and the unusually large prefactor. We find that the nucleation of the droplets cannot be described by a simple Arrhenius behavior (as commonly assumed to fit experimental data) but that the nucleation energy is temperature dependent.