Aerial surveys of waterfowl were conducted in subarctic Quebec in 1989
and 1990 on randomly selected 100 km(2) plots. We used logistic regre
ssion for modelling relationships between the presence of waterfowl br
oods and habitat characteristics. For each species, models of habitat
use were generated with one data set and tested with two others. We al
so compared the frequency distributions of broods of each species in d
ifferent habitat types. Correct-classification rates of models varied
between 0.53 and 0.77. Sensitivity of models generally increased when
applied to validation data sets. Models showed that green-winged teals
(Anas crecca) avoided lakes > 501 ha and used vegetated lakes, ponds
and fens or bogs more than predicted by their availability. Black scot
ers (Melanitta nigra) and surf scoters (M. perspicillata) were found m
ostly on small lakes (< 10 ha) and medium lakes (10-100 ha). Black sco
ters were associated with the presence of ponds and lakes with sedge a
nd grass, whereas surf scoters were not. The presence of scaup (Aythia
spp.) broods was associated with the presence of vegetation. Green-wi
nged teals were observed more often on bogs than were either scoters o
r scaups. Black scoters were observed more often on ponds and less oft
en on medium-sized lakes than surf scoters and scaups. The latter diff
ered in their use of lakes < 10 ha: surf scoters, like black scoters;
mostly used unvegetated lakes, while scaups and green-winged teals mos
tly used vegetated lakes. The accuracy and precision of logistic model
s may be enhanced by additional habitat variables, careful selection o
f sample-cell size and further investigation of the breeding biology o
f the surveyed species.