Using simulation to improve wildlife surveys: Wintering mallards in Mississippi, USA

Wildlife conservation plans generally require reliable data about population abundance and density. Aerial surveys often can provide these data; however, associated costs necessitate designing and conducting surveys efficiently. We developed methods to simulate population distributions of mallards (Anas platyrhynchos) wintering in western Mississippi, USA, by combining bird observations from three previous strip-transect surveys and habitat data from three sets of satellite images representing conditions when surveys were conducted. For each simulated population distribution, we compared 12 primary survey designs and two secondary design options by using coefficients of variation (CV) of population indices as the primary criterion for assessing survey performance. In all, 3 of the 12 primary designs provided the best precision (CV???11.7%) and performed equally well (WR08082E1d.gif _diff???0.6%). Features of the designs that provided the largest gains in precision were optimal allocation of sample effort among strata and configuring the study area into five rather than four strata, to more precisely estimate mallard indices in areas of consistently high density. Of the two secondary design options, we found including a second observer to double the size of strip transects increased precision or decreased costs, whereas ratio estimation using auxiliary habitat data from satellite images did not increase precision appreciably. We recommend future surveys of mallard populations in our study area use the strata we developed, optimally allocate samples among strata, employ PPS or EPS sampling, and include two observers when qualified staff are available. More generally, the methods we developed to simulate population distributions from prior survey data provide a cost-effective method to assess performance of alternative wildlife surveys critical to informing management decisions, and could be extended to account for effects of detectability on estimates of true abundance. ?? 2009 CSIRO.