Effects of Habitat on Mallard Duckling Survival in the Great Lakes Region

Research Article Effects of Habitat on Mallard Duckling Survival in the Great Lakes Region JOHN W. SIMPSON, 1,2 Ducks Unlimited, Inc., 331 Metty Drive, Suite 4, Ann Arbor, MI 48103, USA TINA YERKES, Ducks Unlimited, Inc., 331 Metty Drive, Suite 4, Ann Arbor, MI 48103, USA THOMAS D. NUDDS, Integrative Biology, University of Guelph, Guelph, ON N1G 2W1, Canada BARRY D. SMITH, Canadian Wildlife Service, Environment Canada, Pacific Wildlife Research Centre, 5421 Robertson Road, Delta, BC V4K 3N2, Canada ABSTRACT Habitat provides food and shelter resources for prefledgling waterfowl and thus plays a critical role in their growth, development, and survival. However, few studies have examined whether and how particular elements of habitat affect duckling survival. We investigated relationships of duckling survival rates with distance of overland travel, wetland vegetation composition, water permanency, and surrounding upland vegetation for 116 mallard (Anas platyrhynchos) broods in the Great Lakes region from 2001 to 2003. We found that the probability, on hatch day, that a mallard duckling will survive to 55 days was positively related to the proportion of wetland area that was vegetated and negatively related to the proportion of forest cover within 500 m of duckling locations. We found little support for relationships between duckling survival rates and the proportions of grasslands or seasonal wetlands or to distances traveled overland by broods. Our results suggest that conservation groups and wildlife managers in the Great Lakes region can improve mallard duckling survival rates by managing for, creating, and protecting vegetated wetlands and focusing efforts within lightly-forested areas. (JOURNAL OF WILDLIFE MANAGE- MENT 71(6):1885 1891; 2007) KEY WORDS Anas platyrhynchos, duckling, Great Lakes, habitat, mallard, survival, wetland. DOI: 10.2193/2006-204 Habitat, by definition, provides elements critical to the growth and development of waterfowl ducklings. However, little is known about the role-specific elements of broodrearing habitat play in determining duckling survival rates, particularly outside of the traditional prairie breeding areas. Survival rates of mallard (Anas platyrhynchos) ducklings vary considerably across the Great Lakes region of the United States and Canada, and this variation might be due to differences in habitat utilized by ducklings (Hoekman et al. 2004, Simpson et al. 2005). Our goal was to quantify which elements of habitat, specifically water permanency, wetland vegetation, and surrounding upland vegetation, are most critical to the survival of mallard ducklings within the Great Lakes region. Duckling and brood survival rates in the prairie region of North America are positively related to increasingly wetter conditions, particularly water conditions within seasonal wetlands (Krapu et al. 2000, Pietz et al. 2003). However, seasonal wetlands are proportionally less abundant in the Great Lakes region than in the prairies, and they generally receive relatively little use by breeding waterfowl (Merindino et al. 1995). In addition, the temperate climate of the Great Lakes region does not produce the strong annual wet and dry cycles within seasonal wetlands that are characteristic of prairie wetlands. It is unknown if these relationships exist between seasonal wetland conditions and duckling survival rates within in the Great Lakes region, and if they do not, it is unknown which alternative habitat elements are of importance, if any. 1 E-mail: jsimpson@asti-env.com 2 Present address: ASTI Environmental, 10448 Citation Drive, Brighton, MI 48116, USA Mallard ducklings prefer to feed in stands of emergent vegetation, and the presence of emergent vegetation within wetlands positively affects wetland use by broods (Lokemoen 1973, Mack and Flake 1980, Monda and Ratti 1998). It is not clear, however, whether these preferences confer a survival advantage. Emergent vegetation provides a variety of essential functions for growing ducklings, including foods, shelter from severe weather, and escape cover from predators (Sedinger 1992, Cox et al. 1998). Because of the importance of these functions to prefledgling waterfowl survival, it is possible that vegetation within wetlands might play a role in determining survival rates of mallard ducklings. Additionally, the presence of forest cover nearby wetlands in the prairies might negatively affect the survival of ducklings reared in those wetlands because trees provide hunting perches for avian predators (Murphy 1997, Stafford et al. 2002) and cover for mammalian predators of ducklings (Talent et al. 1983, Arnold and Fritzell 1987, Pietz et al. 2003). This relationship between forest cover and predators might have strong implications for duckling survival in the Great Lakes region where forest cover often dominates wetland and surrounding upland habitats. Similarly, upland vegetation might also affect duckling survival rates during periods of travel overland. Ducklings can travel great distances from the nest to the first brooding wetland, and these overland movements can negatively affect duckling survival rates (Ball et al. 1975, Duncan 1986, Rotella and Ratti 1992). Nest success in mallards is positively related to area of grassland cover, presumably because predator foraging efficiency is reduced in large, intact, and dense grasslands (Reynolds et al. 2001, Phillips et al. 2003). During overland travel, waterfowl broods can be Simpson et al. Mallard Duckling Survival and Habitat 1885

subjected to predation threats similar to nests; thus, dense grasslands might provide similar benefits to broods traveling across uplands. Our objective was to test the hypotheses that mallard duckling survival rates in the Great Lakes region are affected by 1) seasonal wetlands as a proportion of the surrounding area and total wetland area, 2) vegetated wetlands as a proportion of the surrounding area and total wetland area, 3) forest and grassland cover as a proportion of the surrounding area, and 4) distances traveled overland by broods of ducklings. We use observational data from 116 mallard broods to estimate duckling survival rates relative to these habitat elements. STUDY AREA We selected 9 study sites that had a variety of landscapes representative of the Great Lakes region of the United States. In 2001, study sites were located nearby Port Clinton, Ohio (41830 0 N, 84859 0 W); Riverdale, Michigan (43823 0 N, 84850 0 W); and Ripon, Wisconsin (43850 0 N, 88850 0 W). In 2002, study sites were located nearby Stueben County, Indiana (41838 0 N, 84859 0 W); Battle Creek, Michigan (42827 0 N, 85813 0 W); and Shiocton, Wisconsin (44826 0 N, 88834 0 W). In 2003, study sites were located nearby Akron, Ohio (41804 0 N, 81813 0 W); Big Rapids, Michigan (43837 0 N, 85813 0 W); and New Richmond, Wisconsin (45811 0 N, 92823 0 W; Fig. 1). METHODS Data Collection We trapped 60 female mallards and radiomarked them with abdominal implant transmitters at each site prior to nesting. We radiotracked females 1 6 times a day using truckmounted null-array systems to locate nests and determine nest fates. We used hand-held tracking when we could not obtain locations with null-array systems, and we conducted aerial telemetry flights as needed to locate missing hens. We attempted to count ducklings within 1 day of hatch and every 7 days thereafter until fledging (55 d; Bellrose 1976). We made duckling counts from concealed locations to avoid disturbing the hen and her brood. We tracked radiomarked hens with surviving broods daily and estimated their Universal Transverse Mercator locations via triangulation using LOCATE software (Pacer, Truro, NS, Canada). In cases where brood hens were killed (,5% of the sample), we assumed total brood loss. For additional information on radiotracking, telemetry, and duckling count methods see Simpson et al. (2005). Habitat Classification At each study site, we ground-truthed all wetland and upland habitat within 0.8 km of all brood locations to carefully map habitat types onto digital aerial photographs. We classified upland habitat into 7 general categories: grassland, pasture, hayland, cropland, woodland, scrubland, and residential. We numbered individual wetland basins for identification and assigned a system, sub-system, class, subclass, and subordinate class according to the methods of Cowardin et al. (1979). We also assigned individually numbered wetlands a permanency class, cover type, dominant, and subdominant emergent vegetation according to the methods of Stewart and Kantrud (1971). We then manually digitized the ground-truthed and classified photographs in ArcView 3.2 to produce a digital habitat layer for each study site. Habitat Covariates We plotted brood locations onto the digitized habitat layer using ArcMap 8.3. We used an ArcObjects macro to buffer radiolocations by a radius of 500 m and to dissolve those buffers by brood to create a single irregularly shaped buffer. We then intersected these buffers with the digitized habitat layer to obtain area measurements of upland and wetland vegetation within the buffered region. To further reduce candidate variables, we combined structurally similar woodland and scrubland when calculating forest cover and grassland and hayland when calculating grassland cover. We determined proportional measurements of forest cover (PAFOR) and grassland cover (PAGRS) surrounding each brood by dividing the total area of PAFOR or PAGRS within the 500-m buffered region by the total area of all classified habitat within the buffer. Similarly, we calculated the proportion of seasonal wetland (PASEAS) by dividing the total area of permanency class 3 (seasonal) wetlands (Stewart and Kantrud 1971) by the total area of all classified habitat within the buffer. We excluded permanency class 1 (ephemeral) and 2 (temporary) wetlands (Stewart and Kantrud 1971) from all wetland analyses because they were dry and unavailable to broods during the brood rearing period in all years and at all sites (J. Simpson, Ducks Unlimited, Inc., personal observation). We calculated the proportion of vegetated wetland (PAWEM) by dividing the area of all vegetated wetland types by the total area of classified habitat within the buffer. We considered wetlands a vegetated wetland type if the wetland was field classified with a class of emergent, forested, scrub shrub, or aquatic bed according to the methods of Cowardin et al. (1979). We also measured seasonal and vegetated wetlands as a proportion of the total wetland area within the buffered region. We calculated the proportion of wetland area that was seasonal (PWPM3) by dividing the total area of wetlands classified as permanency class 3 by the total area of all permanency class 3, 4 (semipermanent), and 5 (permanent) wetland types within the buffered region (Stewart and Kantrud 1971). We determined the proportion of wetland area that was vegetated (PWVEG) by dividing the total area of wetlands classified as emergent, forested, scrub shrub, and aquatic bed (Cowardin et al. 1979) by the total area of all wetland types within the buffer, excluding permanency class 1 and 2 wetlands. We calculated minimum distance of overland travel (OTOT) as the shortest possible straight-line measurement from a successful nest to the first wetland in which we radiolocated the brood. We created and measured straight lines connecting the nest location and the nearest point of 1886 The Journal of Wildlife Management 71(6)

Figure 1. Locations of the 9 study sites within the Great Lakes region, USA, where we studied mallard duckling survival, 2001 2003. (OH01 ¼ Port Clinton, OH; MI01 ¼ Riverdale, MI; WI01 ¼ Ripon, WI; OH02 ¼ Stueben County, IN; MI02 ¼ Battle Creek, MI; WI02 ¼ Shiocton, WI; OH03 ¼ Akron, OH; MI03 ¼ Big Rapids, MI; WI03 ¼ New Richmond, WI). the first wetland using the Nearest Features Extension v3.6 (www.jennessent.com) in ArcView 3.2. Survival Estimation We used the Clutch and Brood Survivorship Model (V.1.1.0, Smith et al. 2005) to obtain maximum-likelihood estimates of the probability of duckling survival to 55 days post hatching as a function of our putative habitat covariates (Simpson et al. 2005). Our data met the assumptions of this model in that 1) our counts of the number of juveniles in a brood were considered accurate, 2) all data records for each brood exhibited a steady or declining number of individuals over time, 3) hatching was essentially synchronous within each clutch, and 4) all juveniles were of known age. We tested goodness-of-fit for survival models using randomized Pearson deviates (p) and randomized deviance using the likelihood parameter estimates (Smith et al. 2005) and more conservatively by an a posteriori estimate of overdispersion (^c, White and Burnham 1999). We considered models to fit the data well if 0.025, p, 0.975, where extremely small values of p suggest an under-fitted model and extremely large values suggest an over-fitted model, and ^c was judged close to unity. Once we obtained maximumlikelihood estimates of the model parameters and satisfactory goodness-of-fit statistics (Burnham and Anderson 2002), we used parametric bootstrapping to obtain duckling survival estimates with standard errors. Simpson et al. Mallard Duckling Survival and Habitat 1887

Table 1. Pearson s correlation coefficients for variables included in competing models explaining mallard duckling survival to 55 days in the Great Lakes region, USA, 2001 2003. Covariate a OTOT PAGRS PAFOR PAWEM PWPM3 PWVEG PASEAS OTOT 1.000 0.027 0.012 0.053 0.046 0.105 0.081 PAGRS 1.000 0.074 0.269 0.233 0.133 0.074 PAFOR 1.000 0.249 0.256 0.151 0.149 PAWEM 1.000 0.303 0.450 0.158 PWPM3 1.000 0.104 0.617 PWVEG 1.000 0.078 a OTOT ¼ distance (m) from the nest to first wetland; PAGRS ¼ proportion of grassland within buffered area; PAFOR ¼ proportion of forest within buffered area; PAWEM ¼ proportion of vegetated wetland within buffered area; PASEAS ¼ proportion of seasonal wetland within buffered area; PWPM3 ¼ proportion of total wetland within buffered area that is seasonal; PWVEG ¼ proportion of total wetland within buffered area that is vegetated. Model Building and Selection Similar to Simpson et al. (2005), we developed models by adding covariates to a base model containing variables that treated duckling survival as a mixture of age-dependent random and correlated mortality (hereafter the AGE model; see Smith et al. 2005). Because we observed variation among sites in random mortality processes (Simpson et al. 2005), we allowed covariates to influence random mortality processes only in this analysis (Smith et al. 2005). We used Pearson s correlation coefficients to ensure that covariates included in competing models were sufficiently uncorrelated (Zar 1996). We constructed a set of 27 candidate models based upon a priori hypotheses, rather than simply considering all possible combinations of putative covariates. These models included combinations of wetland and upland covariates, overland travel, hatching date (DATE), and study site (SITE). We included hatching date in paired models to account for seasonal variation in habitat conditions, particularly within seasonal wetlands. We included the best competing model (AGE þ SITE) from Simpson et al. (2005) as a measure of how well habitat models performed relative to the integrated effects of random mortality processes across study sites. We also included in the candidate set the base model containing only age effects as a relative comparison of how well habitat covariates improved the fit of that model. Models investigating site-specific habitat relationships and complex interactions were too over-parameterized to be considered with this data set. We ranked candidate models using Akaike s Information Criterion corrected for small sample sizes (AIC c ; Anderson et al. 2000, Burnham and Anderson 2002). For each model, we calculated the AIC c score, the difference in AIC c between each model and the model with the minimum AIC c, and the Akaike weight (Burnham and Anderson 2002). RESULTS We were able to obtain brood observations (x observations/ brood ¼ 5.1, range ¼ 2 19), radio-telemetry locations (x locations/brood ¼ 34.0, range ¼ 1 90) and measurements of habitat covariates for 116 mallard broods across the 9 study sites. Pearson s correlation coefficients between all habitat variables considered for analysis were judged to be sufficiently close to zero (Table 1) and certainly well below the jrj 0.95 upper limit recommended by Burnham and Anderson (2002). The most parsimonious model contained the parameters PWVEG and PAFOR in addition to the AGE parameters and was ranked 1.82 AIC c units ahead of its closest competitor (Table 2). This model identified a positive relationship between the probability, on hatch day, that a duckling will survive to 55 days of age (P[Fledge (0,55)]) and the proportion of wetland area classified as vegetated wetland (Fig. 2). It also identified a negative relationship between (P[Fledge (0,55)]) and the proportion of forest within the brood-rearing area (Fig. 3). Goodness-of-fit tests of model adequacy indicated that the model fit the data well (p ¼ 0.74 6 0.04) and bootstrapped estimates of ^c did not differ significantly from unity (^c ¼ 0.89 6 0.15). The second-ranked model contained only the covariate PWVEG in addition to the AGE parameters. The third-ranked model was 1.90 AIC c units away from the highest-ranked model and contained the proportion of grassland, PAGRS, in addition to the parameters contained in the highestranked model. However, examination of this model s direct competitors revealed that its ranking was determined by the presence of the parameters PWVEG and PAFOR and not by the presence of the parameter PAGRS. All remaining models that contained the parameter PWVEG and additional habitat covariates were ranked above both the basic AGE model and the site effects model. With the exception of the proportion of wetland that was vegetated, PWVEG, no single covariate was able to improve the fit of the basic AGE model or able to rank higher than the site effects model, AGE þ SITE, revealing that those covariates alone captured little of the differences in survival among sites. However, the proportion of forest cover, PAFOR, when added to its paired competitor, AGE þ PWVEG, was able to improve its AIC c ranking by 1.82 AIC c units. Additional models containing combinations of wetland permanency, forest cover, grassland cover, overland travel, and hatching date were unable to improve AIC c scores for the basic AGE model and received minimal support relative to the other models ranked in this set of competing models, with AIC c scores.13 units lower than the highest-ranked model. 1888 The Journal of Wildlife Management 71(6)

Table 2. Competing models explaining mallard duckling survival at 9 sites in the Great Lakes region, 2001 2003. Table includes the model description, the number of parameters (K), the Akaike s Information Criterion corrected for small sample sizes score (AIC c ), the difference from the highest-ranked model (DAIC c ), and the relative AIC c weight (w i ; Burnham and Anderson 2002). Model a K AIC c DAIC c w i AGE þ PWVEG þ PAFOR 8 1101.13 0.00 0.28 AGE þ PWVEG 7 1102.95 1.82 0.11 AGE þ PWVEG þ PAFOR þ PAGRS 9 1103.03 1.90 0.11 AGE þ PWVEG þ PASEAS 8 1103.33 2.20 0.09 AGE þ PWPM3 þ DATE 8 1103.63 2.50 0.08 AGE þ PWVEG þ DATE 8 1103.78 2.65 0.08 AGE þ PWVEG þ PAWEM 8 1104.25 3.12 0.06 AGE þ PWVEG þ PWPM3 8 1104.31 3.18 0.06 AGE þ PWVEG þ PAGRS 8 1104.67 3.54 0.05 AGE þ PWVEG þ OTOT 9 1104.98 3.85 0.04 AGE þ SITE 14 1105.62 4.49 0.03 AGE 6 1114.75 13.61 0.00 AGE þ PASEAS þ PWPM3 8 1114.81 13.68 0.00 AGE þ PASEAS 7 1115.23 14.10 0.00 AGE þ PAGRS 7 1115.82 14.69 0.00 AGE þ PAFOR 7 1115.82 14.69 0.00 AGE þ PAWEM 7 1116.29 15.15 0.00 AGE þ PWPM3 7 1116.64 15.50 0.00 AGE þ OTOT 8 1116.99 15.86 0.00 AGE þ PASEAS þ PAWEM 8 1117.05 15.92 0.00 AGE þ PASEAS þ DATE 8 1117.07 15.94 0.00 AGE þ PAFOR þ PAGRS 8 1117.11 15.98 0.00 AGE þ PAWEM þ DATE 8 1117.99 16.86 0.00 AGE þ PAFOR þ OTOT 9 1118.14 17.01 0.00 AGE þ PAGRS þ OTOT 9 1118.28 17.15 0.00 AGE þ PWPM3 þ PAWEM 8 1118.28 17.15 0.00 AGE þ OTOT þ DATE 9 1118.63 17.50 0.00 a AGE parameters treat duckling mortality as a mixture of age-dependent random and correlated processes without the effect of putative covariates; OTOT ¼ distance (m) from the nest to first wetland; PAGRS ¼ proportion of grassland within buffered area; PAFOR ¼ proportion of forest within buffered area; PAWEM ¼ proportion of vegetated wetland within buffered area; PASEAS ¼ proportion of seasonal wetland within buffered area; PWPM3¼ proportion of total wetland within buffered area that is seasonal; PWVEG ¼ proportion of total wetland within buffered area that is vegetated; DATE ¼ hatching date of the brood; SITE ¼ categorical variable representing the 9 study sites. Figure 2. The probability, on hatch day, that a mallard duckling will survive to 55 days of age, P[Fledge (0,55)], as a function of the proportion of wetland area that is vegetated (PWVEG). Individual points represent estimated survival probabilities (6SE) for each of the 116 studied mallard broods as predicted by the highest-ranked model, AGE þ PWVEG þ PAFOR (PAFOR ¼ the proportion of forested area). We collected data in the Great Lakes region, 2001 2003. However, this increased survival could also be the result of an increase in availability of flooded vegetation and not seasonal hydrology per se, although it is likely that the 2 factors are correlated. We found weak support for models containing proportions of seasonal wetland area, supporting our hypothesis that wetland vegetation, and not necessarily seasonal wetland availability, is a better predictor of mallard duckling survival in the Great Lakes region. DISCUSSION Our results provide insight into relationships between duckling survival and specific elements of habitat within the Great Lakes region, in particular wetland vegetation and forest cover elements. These results might also highlight differences in habitat-survival relationships between Great Lakes and prairie mallard ducklings. Several studies, including this one, report that the habitat composition of wetland vegetation is positively associated with duckling survival (Sayler and Willms 1997, Stafford et al. 2002). Sayler and Willms (1997) also found that duckling survival rates were higher in years where high water levels increased accessibility to flooded emergent vegetation, compared to years with low water levels and little access to emergent vegetation. Similarly, in wet years when wetlands were fully inundated, both mallard and gadwall ducklings survived better due to a greater availability of seasonal wetlands (Krapu et al. 2000, Pietz et al. 2003). Figure 3. The probability, on hatch day, that a mallard duckling will survive to 55 days of age, P[Fledge (0,55)], as a function of the proportion of forested area (PAFOR). Individual points represent estimated survival probabilities (6SE) for each of the 116 studied mallard broods as predicted by the highest-ranked model, AGE þ PWVEG þ PAFOR (PWVEG ¼ the proportion of wetland area that is vegetated). We collected data in the Great Lakes region, 2001 2003. Simpson et al. Mallard Duckling Survival and Habitat 1889

The importance of wetland vegetation to mallard duckling survival might be magnified in the Great Lakes region by patterns of wetland use and availability. Breeding mallards in the Great Lakes region use a wide variety of semipermanent and permanent wetlands; however, use of seasonal wetlands is not high (Merindino et al. 1995). This is unlike the prairie pothole region, where waterfowl make extensive use of seasonal wetlands for brood rearing (Rotella and Ratti 1992, Krapu et al. 2000, Pietz et al. 2003). Permanent wetland basins produce fewer invertebrate food resources than do seasonal wetlands, and invertebrate abundance in more permanent wetlands is often determined by the amount and structure of emergent vegetation (Neckles et al. 1990, Murkin and Caldwell 2000, Murkin and Ross 2000). Thus, the amount of wetland vegetation might be especially important in determining survival rates of mallard ducklings in the Great Lakes region. In addition to providing invertebrate and vegetative food sources for waterfowl ducklings, wetland vegetation provides escape cover from predators. Permanent wetland basins attract and support populations of mink and this relationship negatively influences duckling survival rates in the prairie pothole region (Arnold and Fritzell 1990, Krapu et al. 2004). Although we did not quantify the individual causes of duckling loss, predation contributes to duckling mortality in other regions (Talent et al. 1983, Pietz et al. 2003) and presumably in the Great Lakes region as well. Because wetland vegetation is important in mediating predation by providing escape cover, it might once again magnify the importance of wetland vegetation to mallard duckling survival in the Great Lakes region. Our hypothesis that forest cover acts as a predator source was supported by a decline in duckling survival rates as the proportion of forest cover increased in brood rearing areas. This relationship has been hypothesized before; Murphy (1997) found that waterfowl comprised a large proportion of the diet of prairie-dwelling great horned owls, and Stafford et al. (2002) found that many radiomarked ducklings were lost to predation by raptors. Both authors suggested that large trees nearby wetlands served as raptor roosts and attracted these predatory birds. Other waterfowl predators such as mink and raccoon also associate with wooded habitats adjacent to wetland areas (Fritzell 1978, Allen 1986). In the Great Lakes region the amount of forest cover can be highly variable, and often, much of the remaining forest cover is surrounding wetlands and other areas inaccessible to agriculture ( J. Simpson, personal observation). These patterns could have important implications for mallard breeding in this region, because patterns of forest cover and fragmentation might concentrate predators around mallard brood-rearing areas. MANAGEMENT IMPLICATIONS Waterfowl conservation planners in the Great Lakes region should focus efforts on conserving and creating wellvegetated wetland areas, whereas wetland managers should focus on providing densely vegetated wetlands available to mallards during the brood-rearing season. Similar positive relationships have been demonstrated in other regions of the country, and our study reveals that this relationship might be particularly strong in the Great Lakes region. Additionally, given the negative relationship between forest cover and duckling survival rates, conservation plans and management efforts might be most effective when delivered within lessforested areas of the Great Lakes landscape. ACKNOWLEDGMENTS Funding and support for this research was provided by the Institute for Wetlands and Waterfowl Research of Ducks Unlimited Canada; Ohio Department of Natural Resources (DNR), Division of Wildlife; Indiana DNR; Michigan DNR; Wisconsin DNR; United States Fish and Wildlife Service (USFWS) Upper Mississippi River Great Lakes Joint Venture; United States Environmental Protection Agency Great Lakes National Program Office; USFWS Coastal Program; Saginaw Bay Water Initiative Network; The Bruning Foundation; The Christel DeHaan Family Foundation; Winous Point Marsh Conservancy; West Rosendale Hunt Club; Michigan State University Kellogg Bird Sanctuary; Herbert H. and Grace A. Dow Foundation; and several anonymous donors. Earlier versions were improved by comments from J. Hupp, M. Losito, and an anonymous reviewer. LITERATURE CITED Allen, A. W. 1986. Habitat suitability index models: mink, revised. U.S. Fish and Wildlife Service, Biological Report 82 (10.127), Fort Collins, Colorado, USA. Anderson, D. R., K. P. Burnham, and W. L. Thompson. 2000. Null hypothesis testing: problems, prevalence and an alternative. Journal of Wildlife Management 64:912 923. Arnold, T. W., and E. K. Fritzell. 1987. Food habits of prairie mink during the breeding season. Canadian Journal of Zoology 65:2322 2324. Arnold, T. W., and E. K. Fritzell. 1990. Habitat use by male mink in relation to wetland characteristics and avian prey abundances. Canadian Journal of Zoology 68:2205 2208. Ball, I. J., D. S. Gilmer, L. M. Cowardin, and J. H. Riechmann. 1975. Survival of wood duck and mallard broods in north-central Minnesota. Journal of Wildlife Management 39:776 780. Bellrose, F. C. 1976. Ducks, geese, and swans of North America. Stackpole, Harrisburg, Pennsylvania, USA. Burnham, K. P., and D. R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. Springer Verlag, New York, New York, USA. Cowardin, L. M., V. Carter, F. C. Golet, and E. T. Laroe. 1979. Classification of wetlands and deepwater habitats in the United States. U.S. Department of the Interior, Fish and Wildlife Service, Washington, D.C., USA. Cox, R. R., Jr., M. A. Hanson, C. C. Roy, N. H. Euliss, Jr., D. H. Johnson, and M. G. Butler. 1998. Mallard duckling growth and survival in relation to aquatic invertebrates. Journal of Wildlife Management 62:124 133. Duncan, D. C. 1986. Survival of dabbling duck broods on prairie impoundments in southeastern Alberta. Canadian Field Naturalist 100: 110 113. Fritzell, E. K. 1978. Habitat use by prairie raccoons during the waterfowl breeding season. Journal of Wildlife Management 42:118 127. Hoekman, S. T., T. S. Gabor, H. R. Murkin, R. H. Maher, and L. M. Armstrong. 2004. Factors affecting survival of mallard ducklings in southern Ontario. Condor 106:485 495. Krapu, G. L., P. J. Pietz, D. A. Brandt, and R. R. Cox, Jr. 2000. Factors 1890 The Journal of Wildlife Management 71(6)

limiting mallard brood survival in prairie pothole landscapes. Journal of Wildlife Management 64:553 561. Krapu, G. L., P. J. Pietz, D. A. Brandt, and R. R. Cox, Jr. 2004. Does presence of permanent fresh water affect recruitment in prairie-nesting dabbling ducks? Journal of Wildlife Management 68:332 341. Lokemoen, J. T. 1973. Waterfowl production on stock-watering ponds in the northern plains. Journal of Range Management 26:179 184. Mack, G. D., and L. D. Flake. 1980. Habitat relationships of waterfowl broods on South Dakota stock ponds. Journal of Wildlife Management 44:695 700. Merindino, M. T., G. B. McCullough, and N. R. North. 1995. Wetland availability and use by breeding waterfowl in Southern Ontario. Journal of Wildlife Management 59:527 532. Monda, M. J., and J. T. Ratti. 1998. Niche overlap and habitat use by sympatric duck broods in Eastern Washington. Journal of Wildlife Management 52:95 103. Murkin, H. R., and P. J. Caldwell. 2000. Avian use of prairie wetlands. Pages 249 286 in H. R. Murkin, A. G. van der Valk, and W. R. Clark, editors. Prairie wetland ecology: the contribution of the Marsh Ecology Research Program. University of Iowa Press, Ames, USA. Murkin, H. R., and L. M. Ross. 2000. Invertebrates in prairie wetlands. Pages 201 248 in H. R. Murkin, A. G. van der Valk, and W. R. Clark, editors. Prairie wetland ecology: the contribution of the Marsh Ecology Research Program. University of Iowa Press, Ames, USA. Murphy, R. K. 1997. Importance of prairie wetlands and avian prey to breeding great horned owls (Bubo virginianus) in northwestern North Dakota. Pages 286 293 in J. R. Duncan, D. H. Johnson, and T. H. Nicholls, editors. Biology and conservation of owls of the Northern Hemisphere. U.S. Department of Agriculture Forest Service General Technical Report NC-190, Winnipeg, Manitoba, Canada. Neckles, H. A., H. R. Murkin, and J. A. Cooper. 1990. Influences of seasonal flooding on macro-invertebrate abundance in wetland habitats. Freshwater Biology 23:311 322. Phillips, M. L., W. R. Clark, M. A. Sovada, D. J. Horn, R. R. Koford, and R. J. Greenwood. 2003. Predator selection of prairie landscape features and its relation to duck nest success. Journal of Wildlife Management 67: 104 114. Pietz, P. J., G. L. Krapu, D. A. Brandt, and R. R. Cox, Jr. 2003. Factors affecting gadwall brood and duckling survival in prairie pothole landscapes. Journal of Wildlife Management 67:564 575. Reynolds, R. E., T. L. Shaffer, R. W. Renner, W. E. Newton, and B. D. J. Batt. 2001. Impact of the Conservation Reserve Program on duck recruitment in the U.S. Prairie Pothole Region. Journal of Wildlife Management 65:765 780. Rotella, J. J., and J. T. Ratti. 1992. Mallard brood movements and wetland selection in south-western Manitoba. Journal of Wildlife Management 56:508 515. Sayler, R. D., and M. A. Willms. 1997. Brood ecology of mallards and gadwalls nesting on islands in large reservoirs. Journal of Wildlife Management 61:808 815. Sedinger, J. S. 1992. Ecology of prefledging waterfowl. Pages 109 127 in B. D. J. Batt, A. D. Afton, M. G. Anderson, C. D. Ankney, D. H. Johnson, J. A. Kadlec, and G. L. Krapu, editors. Ecology and management of breeding waterfowl. University of Minnesota Press, Minneapolis, USA. Simpson, J. S., T. J. Yerkes, B. D. Smith, and T. D. Nudds. 2005. Mallard duckling survival in the Great Lakes Region. The Condor 107:898 909. Smith, B. D., W. S. Boyd, and M. R. Evans. 2005. A clutch and brood survival model that discriminates random and correlated mortality. Ecological Applications 15:281 293. Stafford, J. D., L. D. Flake, and P. W. Mammenga. 2002. Survival of mallard broods and ducklings departing overwater nesting structures in eastern South Dakota. Wildlife Society Bulletin 30:327 336. Stewart, R. E., and H. A. Kantrud. 1971. Classification of natural ponds and lakes in the glaciated prairie region. U.S. Fish and Wildlife Service Resource Publication 92, Washington, D.C., USA. Talent, L. G., R. L. Jarvis, and G. L. Krapu. 1983. Survival of mallard broods in south-central North Dakota. Condor 85:74 78. White, G. C., and K. P. Burnham. 1999. Program MARK: survival estimation from populations of marked animals. Bird Study 46 (supplement):120 138. Zar, J. H. 1996. Biostatistical analysis. Prentice-Hall, Upper Saddle River, New Jersey, USA. Associate Editor: Hupp. Simpson et al. Mallard Duckling Survival and Habitat 1891