Estimating open population site occupancy from presence-absence data lacking the robust design

Area of Science:

• Ecology
• Wildlife population monitoring
• Statistical modeling

Background:

• Estimating animal site occupancy is crucial for conservation and management.
• Traditional models struggle with imperfect detection probabilities (p < 1) and lack of robust sampling designs.
• Existing multiseason models perform poorly without repeated sampling within seasons.

Purpose of the Study:

• To develop an alternative statistical model for estimating seasonal site occupancy (Ψt) and detection probability (p).
• To improve occupancy estimation in the absence of a robust sampling design.
• To provide more accurate estimates of wildlife distribution and abundance.

Main Methods:

• Constructed a marginal likelihood by conditioning on and summing out latent occupied sites per season.
• Developed a novel statistical approach to estimate detection probability and site occupancy.
• Compared the proposed model with MacKenzie et al.'s multiseason model using simulations and real-world data.

Main Results:

• The proposed model yields less biased estimates of detection probability (p) compared to MacKenzie et al.'s model.
• Improved estimates of seasonal site occupancy (Ψt) were achieved in the absence of a robust design.
• The new model's occupancy estimates closely aligned with those from a point count model for American robins.

Conclusions:

• The proposed likelihood model offers a significant advancement for estimating wildlife occupancy and detection probabilities.
• This method enhances ecological study accuracy when robust sampling designs are not feasible.
• The findings have implications for wildlife management and conservation strategies requiring precise population distribution data.