Long-distance bird migration consists of several flight episodes interrupted by a series of resting and refuelling periods on stopover sites. We assessed the role of food availability as the determinant of staging decisions focusing on the following three aspects of food availability: intake rates, stochasticity in intake rates and onset of spring. Using stochastic dynamic modelling, we investigated their impact on staging times and expected fitness. Subsequently, we compared relations in the use of the stopover sites as predicted by the model with empirical data of the Svalbard-breeding population of Pink-footed Goose Anser brachyrhynchus collected in the period 1990–2002. Our results indicate that, for the case of Pink-footed Geese, spring phenology determines a major part of the migration schedule. In contrast to our expectations, intake rates were generally only of minor importance; however, when approaching the breeding grounds their significance increased. Expected fitness at arrival on the breeding grounds showed that the geese can compensate for changes in a broad range of food availability and also cope with varying degrees of stochasticity. However, declining intake rates at the last stopover site or very late onsets of spring clearly decreased fitness. As predicted by the model, the use of stopover sites was interdependent – from empirical data we derived negative relationships between the staging durations of subsequent sites. These results lend credit to an integrated spatially explicit approach focussing on multiple stopover site characteristics when attempting to improve our understanding of bird migration.