New technologies in the mix: Assessing N-mixture models for abundance estimation using automated detection data from...

Area of Science:

• Wildlife ecology
• Conservation technology
• Statistical modeling

Background:

• Effective management of threatened species relies on accurate abundance estimates.
• Integrating data from advanced technologies like remotely piloted aircraft systems (RPAS) into population models presents challenges due to unique error sources.

Purpose of the Study:

• To evaluate two distinct modeling approaches for estimating koala abundance from automated detections in RPAS imagery.
• To compare the performance of a generalized N-mixture model against a modified Horvitz-Thompson (H-T) estimator.

Main Methods:

• Two abundance estimation methods were assessed: a generalized N-mixture model and a modified H-T estimator.
• The H-T estimator incorporated generalized linear and additive models to address detection, false detection, and duplicate detection probabilities.
• Model estimates were validated against true koala counts obtained via telemetry-assisted ground surveys.

Main Results:

• The modified H-T estimator demonstrated superior performance, accurately capturing true koala counts within 95% confidence intervals across all four surveys.
• This approach successfully estimated abundance for 138 detected objects in the testing dataset.
• N-mixture models, in their current form, were found less suitable for wildlife abundance estimation using automated RPAS detections.

Conclusions:

• The modified H-T estimator offers a robust method for estimating wildlife abundance from RPAS data with automated detection.
• Accounting for spurious detections is crucial for accurate population estimates in this context.
• This study highlights the potential of advanced statistical methods to overcome limitations of current models in wildlife monitoring technology.