Northern Prairie Wildlife Research Center
Canvasback Mortality from Illegal Hunting on the Upper Mississippi River
Methods
Canvasback population
Surveys from aircraft and boats were used to estimate the number of canvasbacks
that used Navigation Pool 7 during the 1991 and 1992 waterfowl hunting seasons.
We counted canvasbacks separately in areas open and areas closed to waterfowl
hunting.
Number of hunters
The number of hunters in our study area were estimated from vehicle counts and
boat observations. Vehicles were counted at boat landings and at other points
around Lake Onalaska where hunters could access the study area. Vehicles at
boat landings near the Lake Onalaska study area were counted on 6 weekend-days
and on 5 weekdays during the 17-day hunting period in 1991. All observations
were terminated during a severe winter storm on 1 November. In 1992, vehicles
were counted on 7 weekdays and 7-weekend days.
An elevated observation blind was used to count boats that entered the northeast
portion of Lake Onalaska each day throughout each season. Observers occupied
the blind from approximately 45 minutes after sunrise to approximately 20
minutes before sunset and logged information about each boat with hunters
that entered or exited the study area. They recorded the time of observation,
number of hunters, boat description and registration number, and direction
of travel.
Observers compared their records of boats entering and exiting the study
area by boat description and registration number. Boats that exited the study
area before 1500 hours without a recorded entry time were assumed to have
begun hunting at the legal opening time (sunrise in 1991, 0.5 hour before
sunrise in 1992). Boats that entered the study area after 1200 hours without
recorded return times were assumed to have completed hunting at sunset. For
calculations of hunting durations, we assumed hunting began 15 minutes after
a boat entered and ended 15 minutes before a boat exited the study area.
Information about hunters
We contacted hunters in the field and at boat landings throughout the 1991-1992
hunting seasons. We collected information on the number of hunters/party, number
of vehicles/party, number of shots fired, number and species of harvested or
crippled waterfowl, and hunters' knowledge of the local concerns over canvasback
food resources on the river and canvasback mortality from hunting.
Observations by law-enforcement personnel
Actions of waterfowlers in the field were monitored by concealed law-enforcement
personnel throughout the 1991-1992 hunting seasons. For each hunting party observed,
law-enforcement officers recorded each encounter of the hunting party with a
duck, goose, American coot (Fulica americana), or common snipe (Gallinago
gallinago), time of each encounter, species encountered, flock size, number
of shots fired, and the results. An encounter was defined as 1 or more birds
that flew within shooting distance of a hunting party. Shooting distance varied
and was based on the action of the individual party. For some parties, shooting
distance may have been ≥90 m; for others the maximum distance was ≤
45 m. Results were tallied as passed, missed, downed and retrieved, downed and
not retrieved, or hit but kept flying.
Attempts by hunters to shoot canvasbacks
A ratio estimator (Cochran 1977) was used to compute the total number of attempts
to shoot canvasbacks. In 1991, observations were divided into 4 strata based
on day of the hunting season and time of day. The number of attempts to shoot
canvasbacks in a given stratum I (YRi ) was estimated
by:
where yi = total attempts to shoot canvasbacks by parties
under observation in stratum i, xi = total hours
parties were observed hunting in stratum i, and Hi
= total hours all parties hunted in stratum i.
The variance of the ratio estimator (YR ) was estimated
by:
where yj = total attempts to shoot canvasbacks by party
j, xj = total hours party j hunted, H
= total hours all parties hunted in strata, n = total parties under
observation within strata, and z = the average length of observation
for each hunting party. This variance estimate accounts for the portion of
the hunting trip observed by the law-enforcement officers.
 |
| Law-enforcement personnel monitored the actions of waterfowl
hunters in the field. Photo by J.M. Nissen, 1990. |
Assumptions for this model include: the total number of hunting parties
and the hours that all parties hunted in stratum i are known; sampling
(hunter observation) was conducted at random throughout strata; and observed
hunting parties were selected at random. We estimated the total number of
hunting parties and hours hunted within a given stratum and used conservative
values in our analysis. With these assumptions and preliminary data from 1990
in mind, we developed the following guidelines for scheduling law-enforcement
observations in 1991 and 1992:
- Conduct 60 hours of law-enforcement observations/stratum (2 season portions
× 2 daily periods = 4 strata; 240 hours).
- Distribute observations throughout the hunting season (i.e., do not clump
observations during a few days).
- Obtain 40% of observations on weekends and 60% on weekdays.
- Select hunting parties at random from all parties in the study area.
Hunting effort (total number of hunting parties and hours hunted by all parties)
was determined from boat observations and vehicle counts. The average hunt length
used to calculate the total number of hours hunted from vehicle count data was
based on hunter survey data.
To reduce variability in the model, 1991 data were divided and analyzed
by 2 daily time periods (morning and afternoon) and by 2 segments of the hunting
season (16 - 20 October; 21 October - 1 Nov), yielding 4 strata. The season
division was based largely on canvasback encounter rates because few canvasbacks
were encountered during the first season segment (although 10,000 canvasbacks
were present on 15 Oct). A noticeable increase in the canvasback encounter
rate occurred on 21 October. In 1992, we observed consistent canvasback encounter
rates, and no clear seasonal division existed. As a result, 1992 data were
analyzed by only 2 daily periods (2 strata).
T-tests using the Satterwaite approximation for degrees of freedom
(SAS, PROC TTEST, SAS Inst., Inc. 1989) were used to compare canvasback flock
size and encounter-, attempt-, and downed-rates between years. We rejected
statistical hypotheses with P≤0.05.
Previous Section -- Study Area
Return to Contents
Next Section -- Results
U.S. Department of the Interior |
U.S. Geological Survey