Mallard Study

Why research mallards in the Great Lakes?

How was the Great Lakes Mallard Study accomplished?

What were the results of the Mallard Study?

How these results are used in landscape level planning

Why reseach mallards in the Great Lakes?

How was the Great Lakes Mallard Study accomplished?

Nine study sites were chosen, during a 3-year period from 2001 to 2003, based on the existing variation in land cover in the Great Lakes states. At each site 60 mallard hens were decoy trapped in early spring and surgically implanted with transmitters. The hen was followed via radio-telemetry throughout the spring and summer. Tracking the mallards with the telemetry allowed biologists to gather data on breeding, nesting, and brooding, as well as data about the habitat that the hen uses for each of these activities. This data was then analyzed and used to make important decisions about future habitat conservation efforts in the Great Lakes region. To view illustrations and step-by-step diagrams describing how the Great Lakes Mallard Study field methods are conducted click here.

What were the results of the Mallard Study?

Results of the Mallard Study

Over three years (2000-2002), 536 female mallards were radioed and followed from trapping through brood rearing. A total of 602 nests were initiated and 104 hatched.  Initiation dates ranged from 15 March to 1 July, but most initiations occurred between the second week of April to the third week in June.  Reproductive parameter definitions (Table 1) and estimates (Table 2) are reported below and were tested for differences between SY (second year) and ASY (after second year) age classes.  Individual analyses by 3 graduate students focused on the factors that effected nest success, duckling survival and home range.  Nest success was a function of the proportion of row crops:  the more acres of row crops, the lower nest success.  Duckling survival was positively influenced by wetland area and negatively influenced by forest cover.  Home range was reduced in areas that contained a lot of grass and wetland acres.

Table 1.  Parameters for a demographic model of female Great Lakes mallards.

Table 2.  Population parameters and associated standard deviations (SD) used to model demographics of Great Lakes female mallards.

^a Nest success varied by age class.  Means for ASY and SY females were 0.2281 and 0.0960, respectively.

^b Breeding incidence varied by age class.  Means for ASY and SY females were 0.8950 and 0.7935, respectively.

^c Nonbreeding survival = annual survival ÷ breeding survival.

^d Annual survival was not a model parameter but was used to in conjunction with breeding survival to estimate nonbreeding survival.

Population Model and Sensitivity Analyses

Using the estimates above, a population model was developed using the graphical programming language STELLA (2000).  The flow diagram below (Fig. 1) shows the model structure and the functional relationships among parameters within each time step.  This model was used to perform sensitivity analyses which determines how much each parameter influences population growth.

The majority of the explained variation in lambda (population growth) occurs on the breeding grounds (61%, Fig. 2), a finding consistent with our understanding of many waterfowl species.  However, duckling survival was more influential than nesting success, which has been identified as the most important limiting factor for prairie ducks:  28% of the variation in population is accounted for by ducking survival whereas only 16% of that variation is accounted for by nest success.  This finding on the prairies lead to the development of habitat programs designed to address nest success, specifically grass-based programs such as Grasslands for Tomorrow.  In the Great Lakes states, duckling survival, which is linked to wetlands, is the most influential.  This finding focuses our attention on the development of programs that address wetland losses specific to the Great Lakes states. 

Fig. 1.  Flow diagram showing functional relationships among parameters for a demographic model of Great Lakes female mallards.

 Figure 2.  Proportion of variation in Lambda explained by variation in each parameter.

How these results are used in landscape level planning:  linking habitat goals to population goals for breeding mallards

The end-point goal of this research was to be able to link habitat objectives to population objectives for breeding mallards under the North American Waterfowl Management Plan and the Upper Mississippi/Great Lakes Joint Venture.  In order to accomplish this, the research data were incorporated into the planning process in such a way as to determine the changes in the current landscape that must occur to support 1.3 breeding pairs of mallards (can be established at any level) at ‘historic distributions’ in the Great Lakes states. 

Data do not exist to establish historic distributions, so potential distributions were modeled and predicted from survey data and digital wetland data (Yerkes et al. in prep).  See figure to left.

Another assumption tested was the delivery of 4:1 ratio of grass acres to wetlands acres that is currently used in the prairies.  It has always been assumed that this ratio would also apply to Great Lakes habitat delivery.  Through modeling of limiting factors and population responses, it was determined that a 1:1 ratio on average was adequate to support breeding mallards in the Great Lakes states.  There will always be variation in space, such that some areas may need greater or smaller ratios to support local populations.  In fact, the suggested delivery scheme (see below) indicates a ratio 1:3 in areas of low lambda, low mallard distribution and 1:1 in areas of low lambda, high mallard distribution.  These recommendations can be viewed by geographic location by using HEN.

Within the predicted distribution, several scenarios were modeled to determine the best option for setting habitat goals to reach population goals.  These alternatives are outlined below and ALL approaches assume no further habitat loss.  Unfortunately, this is a bad assumption given recent analyses to update NWI data (which detected approximately 50,000 acres of loss between the 1978 NWI data and 1998 photos in the lower half of lower MI alone).

Scenarios:

1.      Do nothing.  If no further loss occurs, mallards would reach the JV goal within 27 years, but many areas would not continue to support mallard pairs (subpopulations would become extinct). Those areas include all of the low lambda-low mallard and low lambda-high mallard sub-populations (see figures below)

2.      Do the minimal.  Under this scenario, a minimal amount of work would occur to reduce the time needed to reach populations goals, however, large variation in lambda would still occur and many subpopulations would disappear (all low lambda areas).  Under this scenario, goals could be reached in 16 years by restoring 338,000 grass and 350,000 wetland acres (loss of 688,000 acres of row crops).

3.      Do everything. Under this scenario, the goal was to eliminate variation in lambda across the distribution area.  This ensures no loss of breeding pairs and growth of the population.  Under this scenario, one would need to protect 8.4 million acres of grass and 1.1 million acres of wetlands, and restore 2.7 million acres of grass and 2.2 million acres of wetlands.  Unfortunately, this scenario is unrealistic because 4.9 million acres of row crop cannot be taken out of production – it is not sustainable.

4.      Do what is most efficient and effective. Under this scenario, several options were examined which included combinations of lambda, mallard numbers, and reductions in row crop agriculture of 25% to 40%. This scenario is discussed below under ‘Recommended Approach’.

Explanation of sub-populations

Subpopulations were defined by variation in lambda and predicted pair density, and further subdivided by the impact on agricultural lands, specifically row crops.  Under an optimal scenario, the elimination in excess of 50% of the row crop area is recommended to maximize mallard density and pair numbers.  This is an unrealistic scenario and not sustainable on the landscape due to socio-economic factors.  Therefore, we examined scenarios in which 25%, 30% and 40% of the row crop acreage was converted to wetlands and/or grass.  We modeled various combinations of the above to determine the most efficient and effective approach to reaching breeding pair population goals.

Below are the delineated subpopulations by lambda and pair density.  51 % of the total area falls under the high lambda-low mallard delineation, whereas 43% is low lambda-low mallard, 1.5% is low lambda-high mallard, and 5% if high lambda-high mallard.

Recommended approach:The most efficient and effective approach involves affecting a small portion of the high lambda, low distribution area to act as a source population for the low lambda, low distribution area until the habitat programs in that area can sustain high lambda. Under this scenario, all work is restricted to affecting only 25% of the agricultural base and the majority of restoration work is concentrated in the low lambda, low distribution area whereas the majority of protection work is concentrated in the high lambda, low distribution and high lambda, high distribution area with a minimal amount of restoration.  This approach assumes delivery of 1:3 grass to wetlands (approximately 119,000 grass, 391,000 wetlands, within 20,942 square mile blocks) in low lambda, low mallard distribution over 17 years, in addition to delivery of 1:1 ratio of grass to wetlands (approximately 5200 acres grass, 5200 wetlands acres within 8,420 square mile blocks) in high lambda, low mallard distribution over 17 years to result in a 1.3 million breeding pair prediction by year 24.  See map to the right for areas recommended by this analysis.  Note:  this is not the only way to achieve goal.  Many areas outside this distribution can positively affect lambda via larger reductions in row crop acres (>25%); these data are presented in HEN.

Under this scenario, the current habitat base in the high lambda, low distribution and high lambda, high distribution areas support significant production.  If that base is lost or significantly reduced, this area may no longer serve as a source.  Therefore, the protection aspect of this analysis is enormous in which no further loss of wetlands can occur; this equates to protecting approximately 1,088,000 acres of wetlands currently on the landscape (see high lambda, low distribution and high lambda, high distribution maps above).

As stated before, this does not take into account background wetland losses that we know are occurring. Once we determine the average rate of wetland loss per state per year, we will adjust delivery estimates upwards to account for that loss.

HEN: Habitat Evaluation Network

Great Lakes Habitat Evaluation Network (HEN) is a basin wide conservation tool that uses GIS/Internet technology to plan and target conservation activities within the Great Lake states.  HEN is a Decision Support System that will allow both DU and its’ partners to enhance and refine wetland and upland habitat programs based on the most current information and technology available.  HEN incorporates data from the mallard study as well as existing landscape features with current habitat information to guide conservation activities to high priority areas.  The overriding objective of HEN is to allow DU biologists and our partners to target restoration ‘hot spots’ in the most effective and efficient manner for the benefit of waterfowl and other wetland species.  Ultimately, this system will allow federal, state, and local land use planners along with conservation organizations to use limited conservation resources in the most efficient and effective manner.  HEN will also allow on-the-ground delivery biologists to evaluate conservation activities on a site-by-site basis.  This will allow optimization of conservation activities across the landscape.