JCU ePrints

Abstract

The probability of detecting an animal during a survey consists of 2 components: (1) the probability of an animal being available for detection (availability) which can be highly variable in heterogeneous environments, and (2) the probability of an animal being detected conditional on its being available for detection (perception). Many surveys only estimate the latter probability, because modeling the availability process requires additional information. We illustrate estimation of both probabilities in an application to aerial surveys of dugongs (Dugong dugon) in Northern Australian coastal waters where water clarity varies greatly over relatively small spatial scales. Using artificial dugong models and timed depth recorders deployed on 15 wild dugongs to obtain dive profile, we carried out experiments to determine zones of detectability for dugongs at the range of depths, turbidities and sea states which spanned the environmental heterogeneity encountered on dugong surveys. Resulting probability estimates were heterogeneous and dependent on the measured conditions. To model the perception process we used a tandem team of 2 observers on either side of the aircraft. This permitted the fitting of generalized Lincoln-Petersen models with Program MARK. We then used the generalized Horvitz-Thompson estimator based on the overall detection probability of each individual group to generate population estimates. We also developed a new simulation-based method for estimating standard errors and coThe probability of detecting an animal during a survey consists of 2 components: (1) the probability of an animal being available for detection (availability) which can be highly variable in heterogeneous environments, and (2) the probability of an animal being detected conditional on its being available for detection (perception). Many surveys only estimate the latter probability, because modeling the availability process requires additional information. We illustrate estimation of both probabilities in an application to aerial surveys of dugongs (Dugong dugon) in Northern Australian coastal waters where water clarity varies greatly over relatively small spatial scales. Using artificial dugong models and timed depth recorders deployed on 15 wild dugongs to obtain dive profile, we carried out experiments to determine zones of detectability for dugongs at the range of depths, turbidities and sea states which spanned the environmental heterogeneity encountered on dugong surveys. Resulting probability estimates were heterogeneous and dependent on the measured conditions. To model the perception process we used a tandem team of 2 observers on either side of the aircraft. This permitted the fitting of generalized Lincoln-Petersen models with Program MARK. We then used the generalized Horvitz-Thompson estimator based on the overall detection probability of each individual group to generate population estimates. We also developed a new simulation-based method for estimating standard errors and confidence intervals. We contrast absolute abundance estimates of dugongs in the Torres Strait and Northern Great Barrier Reef regions using both the new and original approaches (Marsh and Sinclair 1989a). Both estimates were similar in value with standard errors similar for Torres Strait but the new method had a substantially larger standard error for the Northern Great Barrier Reef survey.nfidence intervals. We contrast absolute abundance estimates of dugongs in the Torres Strait and Northern Great Barrier Reef regions using both the new and original approaches (Marsh and Sinclair 1989a). Both estimates were similar in value with standard errors similar for Torres Strait but the new method had a substantially larger standard error for the Northern Great Barrier Reef survey.