Abstract. 1. The effects of habitat restoration are usually studied using crosssectional comparisons of species assemblages among sites of various ages or disturbance levels. Longitudinal studies, however, are necessary for detecting longterm responses to habitat restoration and for understanding annual demographic variation.2. To investigate the time course of bee community restoration in sites previously made uninhabitable by anthropogenic disturbance, we studied a former landfill site for 10 years from initial revegetation in 2003 until 2013, comparing two restored sites with three nearby, undisturbed control sites. We used permutational multivariate analysis of variance and generalised additive mixed models to investigate how bee abundance and species richness varied over time (years), between seasons and between restoration levels.3. Landfill restoration and the creation of foraging and nesting habitat resulted in rapid and persistent occupation by bees, suggesting that efforts to restore bee communities can be successful when a source of colonists exists nearby.4. Based on earlier studies, we predicted that in restored sites there would be an initial rapid increase in bee abundance and species richness, followed by a decline to a stable intermediate level. This prediction was supported as bee abundance and species richness in restored sites increased until 2006/2007 and subsequently declined.5. In control sites, there were significant declines in abundance and species richness over time, despite a lack of anthropogenic disturbance. Possible contributors are changing weather patterns, especially severe droughts; plant community succession resulting in loss of bare ground for nesting sites; and increasing suburbanisation of the surrounding landscape.