A Field Test for Differences in Condition among Trapped and Shot Mallards

Abstract

We tested predictions from the condition bias hypothesis (Weatherhead and Greenwood 1981) regarding the effects of sampling methods on body weights of mallards (Anas platyrhynchos) at White River National Wildlife Refuge (WRNWR), Arkansas, during 24 November-8 December 1985. Body weights of 84 mallards caught with unbaited rocket nets in a natural wetland were used as experimental controls and compared to the body weights of 70 mallards captured with baited rocket nets, 86 mallards captured with baited swim-in traps, and 130 mallards killed by hunters. We found no differences (P > 0.27) in body weight among sampling methods, but'body condition (wt/wing length) of the birds killed by hunters was less (P 0.75 for differences >50 g. The condition bias hypothesis probably applies to ducks killed by hunters but not to trapping operations when substantial (>20 at 1 time) numbers of birds are captured. J. WILDL. MANAGE. 52(2):227-232 Waterfowl researchers recognize the need to obtain representative samples when estimating survival, recruitment, and other population parameters (Brownie et al. 1985:6, Cowardin et al. 1985:29). Unfortunately, random sampling of waterfowl populations is impractical and detecting nonrepresentative samples requires knowledge of population parameters (Burnham and Nichols 1985). An alternative is to look for interpretable patterns of variation among sampling methods (Greenwood et al. 1986). Weatherhead and Greenwood (1981) recently reported that red-winged blackbirds (Agelaius phoeniceus) caught with mist nets differed in body weight and social structure from those captured in decoy traps. They interpreted the data as evidence that the blackbirds captured in decoy traps were in poor condition and not representative of the population. Considerable discussion regarding the implications of nonrandom sampling for waterfowl banding programs and population management ensued (Weatherhead and Ankney 1984, 1985; Burnham and Nichols 1985), but few new data were generated (Burnham and Nichols 1985, Greenwood et al. 1986). Our objectives were to design and conduct an experiment that would sample a population of mallards in Arkansas during early winter using 1 recovery and 3 capture methods, and test predictions from the condition bias hypothesis (Weatherhead and Greenwood 1981) regarding differences in body weight, condition, and social structure of mallards in relation to sampling methods. We are grateful to D. E. Delnicki, W. R. Noble, and A. M. Strong for help in trapping and examining mallards, and to J. E. Fraize for access to birds taken by hunters. C. M. Bunck and B. K. Williams assisted with data analysis and experimental design, and R. M. Erwin, G. M. Haramis, R. M. Kaminski, J. R. Longcore, and J. D. Nichols provided helpful comments on various drafts of the manuscript. CONCEPTUAL FRAMEWORK Heterogeneity of physical condition within populations can result in 2 types of bias. Statistical bias occurs when representative but heterogeneous samples cause estimators to be biased. This problem was addressed by Nichols et al. (1982). Sampling bias occurs when the initial capture or subsequent recovery processes (or both) are not representative with respect to a particular variable (e.g., condition), and that variable is correlated with a parameter of interest (e.g., survival). Our study was concerned with capture and recovery sampling biases. Weatherhead and Greenwood (1981) suggested that the biological basis for the differential condition between blackbirds captured in mist nets and decoy traps was the birds' foraging behavior. Blackbirds often feed by local enhancement wherein birds seeking food join oth-