Limitations of removal design

[zwallace]

Hi All,

I am fitting dynamic occupancy models for a removal sampling design with <= 3 surveys per season. I understand survey specific detection probabilities are not identifiable without an equality constraint between at least 2 occasions. Accordingly, I am only considering models with (1) constant p within seasons, (2) seasonal or site-level individual covariates for p among seasons, and/or (3) equality constraint for 2nd and 3rd occasions (reflecting use of aircraft for occasion 1 and ground surveys for occasions 2 and 3). I have coded the removals as missing observations. Do I need to customize the model in any other way to reflect the removal design? Or, will the estimates, variances, and SEs in the output be correct for a removal model?

Thanks!
Zach

[jhines]

Hi Zach,

You should be able to estimate survey-specific p's in the multi-season model, as long as you have at least 2 surveys in each season. So, as long as you have enough data, you don't need to constrain p constant within season, or constant over the last two surveys in a season. You can test this using GENPRES to generate data with 2 survyes/season and different p's for each survey.

If removal means you stopped visiting a site, then using missing values for the surveys after you stopped going is OK.

Jim