A coupled ocean-wave model was used to investigate residence times of the water masses on the Gulf of Lion shelf and their export routes during autumn 2010 and winter 2010-2011. Particular attention is paid to the Cap de Creus region and submarine canyon, a key site for the export of water from the Gulf of Lion shelf. First, model results were compared to numerous observations taken during the same period. The timing of strong current pulses on the shelf and at 300 m depth within the Cap de Creus Canyon, linked to easterly winds during winter, were well reproduced by the model. Lagrangian particle trajectories were used to calculate residence times of water masses on the Gulf of Lion shelf. Those of waters located near the Rhone River ranged from 10 to 40 days for autumn 2010, a period which was dominated by frequent strong winds, to 40-60 days for winter 2010-2011, which has been linked to less frequent strong winds and a slope current flowing farther away from the shelf. In the Cap de Creus region the volumes of water exported were estimated at 747 km(3) in autumn and 1513 km(3) in winter. Results show that, in autumn, only 4% of the water was exported at depths below 200 m while, in winter, this percentage was 25% because it was related to coastal dense water cascades. Yet, this export remains low compared to other winters. It is likely that the low heat losses that characterized the second part of the winter were responsible for the shallow export depth through the Cap de Creus Canyon. These conditions favoured an export of water from the Gulf of Lion to the Spanish coastal zone that would represent 70% of the total exported volume. Interannual variability of the distribution of this export was investigated over the longer period of 2010-2017. Heat losses in February and March appear to be an indicator of dense shelf water cascading, suggesting about 3 or 4 out of 8 winters being affected by deep cascading (reaching 1000 m depth). Understanding the variability of exports of continental inputs to this region is essential to anticipate their potential impacts on ecosystems and human activities in environments as contrasting as the coastal zone and deep canyons.