Despite a strong current interest in ecological networks, the bulk of studies are static descriptions of network structure, and very few analyze their temporal dynamics. Yet, understanding network dynamics is important in order to link network patterns to ecological processes. We studied the day-to-day dynamics of an Arctic pollination interaction network over two consecutive seasons. First, we found that new species entering the network tend to link to already well-connected species, although there are deviations from this trend due, for example, to morphological mismatching between plant and pollinator traits and non-overlapping phenophases of plant and pollinator species. Thus, temporal dynamics provides a mechanistic explanation for previously reported network patterns such as the heterogeneous distribution of number of links across species. Second, we looked for the ecological properties most likely mediating this dynamical process, and found that both abundance and phenophase length were important determinants of the number of links per species.