The response of peatlands to changes in the climatic water budget is crucial to predicting potential feedbacks on the global carbon (C) cycle. To gain insight on the patterns and mechanisms of response, we linked a model of peat accumulation to a model of peatland hydrology, then applied these models to empirical data spanning the past 5000 years for the large mire Store Mosse in southern Sweden. We estimated parameters for C sequestration and height growth by fitting the peat accumulation model to two age profiles. Then, we used independent reconstruction of climate wetness and model reconstruction of bog height to examine changes in peatland hydrology. Reconstructions of C sequestration showed two distinct patterns of behaviour: abrupt increases associated with major transitions in vegetation and dominant Sphagnum species (fuscum, rubellum–fuscum and magellanicum stages), and gradual decreases associated with increasing humification of newly formed peat. Carbon sequestration rate ranged from a minimum of 14 to a maximum of 72 g m−2 yr−1, with the most rapid changes occurring in the past 1000 years. Vegetation transitions were associated with periods of increasing climate wetness during which the hydrological requirement for increased seepage loss was met by rise of the water table closer to the peatland surface, with the indirect result of enhancing peat formation. Gradual decline in C sequestration within each vegetation stage resulted from enhanced litter decay losses from the near-surface layer. In the first two vegetation stages, peatland development (i.e., increasing surface gradient) and decreasing climate wetness drove a gradual increase in thickness of the unsaturated, near-surface layer, reducing seepage water loss and peat formation. In the most recent vegetation stage, the surface diverged into a mosaic of wet and dry microsites. Despite a steady increase in climate wetness, C sequestration declined rapidly. The complexity of response to climate change cautions against use of past rates to estimate current or to predict future rates of peatland C sequestration. Understanding interactions among hydrology, surface structure and peat formation are essential to predicting potential feedback on the global C cycle.