Short-distance detectability in camera trap surveys: implications for population assessment.

Area of Science:

• Wildlife ecology
• Conservation biology
• Statistical modeling

Background:

• Camera trapping is crucial for monitoring unmarked wildlife populations.
• Existing statistical methods often assume perfect detection near camera traps, potentially biasing estimates.
• The Random Encounter Model (REM) is a common method for density estimation.

Purpose of the Study:

• To test the assumption of perfect detectability near camera traps for wild boar (Sus scrofa).
• To evaluate the impact of imperfect near-camera detection on population density estimates.
• To compare the performance of different camera trap models.

Main Methods:

• Conducted experiments to measure wild boar detection probability at varying distances (0-6m) from camera traps.
• Compared population density estimates using a derived detection function versus standard REM functions.
• Assessed three camera trap models (Bolyguard SG2060-K, Uovision UV595-HD, Browning Spec Ops HP5) for performance.

Main Results:

• Detection probability was found to be less than 1 even at close proximity to camera traps.
• The Random Encounter Model underestimated wild boar population density by 17% when not accounting for imperfect detection.
• No significant performance differences were observed among the tested camera trap models for wild boar.

Conclusions:

• The assumption of perfect detectability near camera traps is invalid for wild boar.
• Camera trap population density estimates can be biased if imperfect detection is not incorporated.
• Accurate wildlife population monitoring requires models that explicitly address detection probabilities below 1.