Search

ABSTRACT Gobbling call‐count surveys have been frequently used as a means for monitoring distributions and population trends of eastern wild turkeys (Meleagris gallopavo silvestris). However, gobbling activity can be highly variable and potentially leads to biased estimates if detection probability is left unaccounted for. Occupancy modeling based on detection–nondetection data offers an effective way of exploring changes in species distribution and provides logistical flexibility with sampling design. Recent advances in occupancy modeling have enabled detection probability and inferences about occupancy dynamics to be estimated from transect sampling designs with spatially replicated survey stations, similar to those of gobbling call‐count surveys. We analyzed data collected from 1,815 gobbling call‐count surveys in northern Wisconsin, USA, 2014–2017, to compare multiseason correlated‐replicate occupancy models to standard multiseason models. Model selection results strongly supported multiseason correlated‐replicate models (Σwi = 1.00) over standard multiseason occupancy models. We observed evidence for spatial autocorrelation in local occupancy—probability of local availability was lower when turkeys were unavailable at the previous station (θ¯ˆ = 0.18 ± 0.07 [SE]) compared with when turkeys were available (θ¯ˆ' = 0.79 ± 0.13). Probability of detection was imperfect and varied by time of day (β = −0.503) and wind speed (β = −0.362). Turkeys occurred throughout a variety of habitat conditions, but occupancy was greatest (ψˆ = 0.990 ± 0.03) along survey routes with 40–60% forest cover. We fit simulated data to standard occupancy models and correlated replicate models and found occupancy and detection estimates were underestimated by as much as 10% with standard occupancy models, but bias was negligible under the correlated replicate modeling approach. Our findings indicate that transect sampling designs that account for spatial correlation among adjacent survey stations can yield improved estimates of detection and occupancy, and such sampling designs could lead to greater ecological interest in factors that influence species occurrence and patterns of habitat use. Gobbling call‐count surveys in conjunction with occupancy‐based modeling may be useful for assessing reintroduced or natural expanding wild turkey populations at the edge of their range. Additionally, we suggest the combination of gobbling surveys and occupancy modeling is useful for evaluating habitat–occurrence relationships across large‐scale landscapes or state‐based wild turkey management units. © 2019 The Wildlife Society.

Occupancy surveys collecting data from adjacent (sometimes correlated) spatial replicates have become relatively popular for logistical reasons. Hines et al. () presented one approach to modelling such data for single-season occupancy surveys. Here, we present a multiseason analogue of this model (with corresponding software) for inferences about occupancy dynamics. We include a new parameter to deal with the uncertainty associated with the first spatial replicate for both single-season and multiseason models. We use a case study, based on the brown-headed nuthatch, to assess the need for these models when analysing data from the North American Breeding Bird Survey ( BBS), and we test various hypotheses about occupancy dynamics for this species in the south-eastern United States., The new model permits inference about local probabilities of extinction, colonization and occupancy for sampling conducted over multiple seasons. The model performs adequately, based on a small simulation study and on results of the case study analysis., The new model incorporating correlated replicates was strongly favoured by model selection for the BBS data for brown-headed nuthatch ( Sitta pusilla). Latitude was found to be an important source of variation in local colonization and occupancy probabilities for brown-headed nuthatch, with both probabilities being higher near the centre of the species range, as opposed to more northern and southern areas., We recommend this new occupancy model for detection-nondetection studies that use potentially correlated replicates. [ABSTRACT FROM AUTHOR]