1. The Marais Poitevin, one of the principal wintering and passage sites for ducks and waders in the East Atlantic flyway, underwent agricultural intensification in the 1980s. The numbers of ducks declined in the period 1979–96 in the main roost, Baie de l’Aiguillon; this was a site-specific phenomenon for which local factors were responsible. We describe here long-term changes in some key characteristics of the wetlands, paying particular attention to the principal factors affecting the quality of waterbird habitat: the area of semi-natural grasslands, water management and wildfowling.
2. The area of grasslands declined by 50% between 1970 and 1995, principally as a result of conversion to arable farmland. The pattern was observed over the whole area and was slightly more marked close to the day-roost of the ducks. The rate of change was greatest in the 1980s. Water levels, and therefore the flooded area, were reduced over the study period during the months of May to November. The numbers of wildfowlers declined from the 1970s, but the number of artificial wildfowling ponds increased, especially after 1980. The number of birds killed did not decline and may have increased. The quality of the ducks’ habitat therefore declined by all three measures.
3. Four species, mallard, shoveler, teal and gadwall, showed an increase in other sites in north and west France either over the whole period or over the last part (1984–96), while declining (shoveler and mallard) or remaining stable in the Baie de l’Aiguillon. Pintail remained stable in the other sites, while showing a long-term decline (over 30 years) in the Baie. Wigeon did not use the Baie in large numbers in the 1996 cold spell as they did until the 1980s.
4. The uncoupling of the trends for these species in the Baie from their trends in the other sites may reflect the loss of grassland feeding habitat in the Marais Poitevin. However, for mallard, pintail and shoveler the declines in the Baie were much stronger than the 50% loss of grasslands so other factors, such as the reduction in autumn flooding and/or shooting, are likely to have played a part.
5. Since the three measures of long-term trends in duck habitat co-varied it was not possible to separate their effects statistically. The hypotheses arising from this study need to be tested: this should be done using comparative and experimental approaches. Suggestions are made for the management of water levels and shooting in the feeding habitat close to the Baie de l’Aiguillon.