Accounting for imperfect detection in estimates of yearly site occupancy

Abstract

Estimating site occupancy is an essential component of any monitoring program because changes in occupancy over time can serve as a metric of population status (MacKenzie et al. 2003). Detection is often imperfect during occupancy surveys, however, and estimating the proportion of occupied sites without accounting for detection error can lead to an underestimation of occupancy (Kéry and Schmidt 2008). Hierarchical occupancy models have become a popular tool to address such issues because they provide a framework for estimating parameters that drive occupancy while accounting for detection probability that is <1 (Marsh and Trenham 2008). In this chapter we use a simulated data set that includes survey observations for 30 Gyrfalcon breeding sites across 20 years. The goal of this analysis is to gain a better understanding of the observed patterns of yearly occupancy in this simulated sample population, and to derive parameters of occupancy that tell us about population status using the multi-season occupancy model outlined in MacKenzie et al. (2003). 10.2 Occupancy model The multi-season occupancy model is said to be hierarchical because it is composed of two submodels: 1) a model that describes the occupancy state, and 2) a model that describes the detection process (i.e., our observations in the field).

Woody plant encroachment drives population declines in 20% of common open ecosystem bird species

“…(1) Woody cover coefficients for initial occupancy were classed as either positive, negative or if 90% confidence intervals (CIs) overlapped with zero—neutral. (2) Average rate of occupancy change ($$$ \lambda $$$) was calculated using the predicted annual occupancy probabilities from the below equation:$$$$ {\lambda}_t=\frac{{\hat{\varPsi}}_{t+1}}{{\hat{\varPsi}}_t} $$$$where $$$ {\lambda}_t $$$ is the change in occupancy at time $$$ t $$$, $$$ {\hat{\varPsi}}_t $$$ is the occupancy estimate at time $$$ t $$$, and $$$ {\hat{\varPsi}}_{t+1} $$$ is the occupancy estimate at time $$$ t+1 $$$ (Hedlin & Franke, 2017; MacKenzie et al., 2003). Mean occupancy trends ($$$ \lambda $$$) with 90% CIs greater than one were classed as increasing, less than one as decreasing and if 90% CIs overlapped with one—stable.…”

“…These models allow for the inclusion of several covariates that in- where t is the change in occupancy at time t, Ψt is the occupancy estimate at time t, and Ψt+1 is the occupancy estimate at time t + 1 (Hedlin & Franke, 2017;MacKenzie et al, 2003). Mean occupancy trends ( ) with 90% CIs greater than one were classed as increasing, less than one as decreasing and if 90% CIs overlapped with one-stable.…”

Variation in nest‐site occupancy and breeding outcome in a threatened Subantarctic raptor

“…Each nest-site was visited at least twice each season and a discrepancy in occupancy status was never observed (i.e. unoccupied on the first visit and occupied on the second or vice versa), which would have allowed us to model the detection probability of an occupied site (Hedlin & Franke 2017). Therefore, we had to assume that the occupancy status of the sites was assigned without error.…”