Seasonal occupancy eps(1-gam) and detection parameterization

[ReneeLorica]

Hi,

I was trying out the other parameterizations in the hopes of getting less astronomical SEs and better CIs but the 'Seasonal occupancy and colonization, detection', and the 'Seasonal occupancy and local extinction, detection' are giving me worse SEs so I tried the 'Seasonal occupancy eps(1-gam) and detection' parameterization instead and have been getting much better SEs and CIs.

In the Donovan and Hines 2007 manual on Exercises on Occupancy estimation and modeling, the authors made a comment (Ch. 13, p. 43) that they don't know what the significance of this model is, and that the "Levins model balance extinction with recolonization".

Does this mean that 'Seasonal occupancy eps(1-gam) and detection' parameterization is invalid?

Thank you for your time.

[darryl]

If you're balancing extinction with colonization that implies some of equilibrium for the system (note this is in terms of overall number of extinctions and colonization, not necessarily equality of the 2 probabilities). The constraint in the 4th parameterization is different; it's saying that regardless of whether a site was occupied or not in the previous year, all sites have the same probability of occupancy in the current year, ie changes in the occupancy status are random rather than Markovian. If you have access to the MacKenzie et al. (2006) book then there's more detail from around page 206 onwards.

This 4th parameterization is also essentially equivalent to fitting multiple single-season models to each year of data, with the advantage that you can apply constraints across the different years unlike if you analyzed each year separately.

There's nothing 'invalid' about this parametrization.

Darryl

[ReneeLorica]

Darryl,

Many thanks for the clarification. I wasn't sure what to make of the authors' comments though I hope I didn't offend anyone with my question.

[ReneeLorica]

P.S.

Does this mean that the Init occupancy, local colonization, extinction, detection parameterization are Markovian?

[darryl]

Yes. Occupancy state in the previous year influences probability of occupancy in the current year. occupied previously -> 1-exinction, unoccupied previously -> colonization

2nd and 3rd parametrizations are also Markovian, just some of the parameters are derived rather than calculated directly.

[ReneeLorica]

Thanks Darryl!