Accounting for variability when designing wildlife surveys

Occupancy-based monitoring programs rely on survey data to infer presence or absence of the target species. However, species may occupy a site and go undetected, leading to the erroneous inference of absence (‘false absence’). If detectability is influenced by the time of year or weather conditions, survey protocols can be adjusted to minimise the chance of false absences.

In this study, detection probabilities for three frog species from SE Australia were modelled using a Bayesian approach. For surveys based on frog calls, the researchers compared basic models (which only included effects of survey date, duration and time of day on detection) to models including additional effects of weather. Model selection using deviance information criterion (DIC) suggested that the basic model was the best for Crinia signifera, while models including relative humidity and water temperature were most supported for Limnodynastes dumerilii and L. Tasmaniensis respectively.

When predictive performance was assessed by cross validation, DIC results were largely matched for C. signifera and L. dumerilii, while models of detection for L. tasmaniensis were indistinguishable, AUC scores suggesting inadequate performance.

We show how results such as these can be used to design surveys, developing protocols for individual surveys and estimating the number of surveys required under those protocols to achieve a threshold cumulative probability of detection. Conservation managers can use these models to maximize the efficiency of
surveys. This will improve the accuracy of occupancy data, and reduce the risk of misdirected conservation actions resulting from false absences.


Canessa S, GW Heard, KM Parris & MA McCarthy (2012). Integrating variability in detection probabilities when designing wildlife surveys: a case study of amphibians from south-eastern Australia. Biodiversity Conservation 21:729–744.

Leave a Reply