Non-standard, non-rotating panel multi-season design

[KWC]

I am interested in fitting multi-season models with site and survey covariates to a suite of species. Data span 225 sampling points over a period of 4 years, with 4 surveys in any given year. However, due to logistics, no sampling location was ever sampled in 2 consecutive years because sites were sampled either every-other-year, or every 4 years. There is a good mix of sites with 2 years of data (with no data in between) and sites with 1 year of data (with no data before or after) within this period of 4 years. It is my understanding that this design is not a standard nor a rotating panel design, rather a hybrid of the sort.

If colonization and extinction are calculated based upon the occupancy state in the previous time step, how can these dynamic processes be modeled in Program PRESENCE?

Currently, constant and time varying models produce estimates for colonization and extinction, however, I am suspicious of them due to their confidence intervals being very wide, and large standard errors. Bailey et al. 2007 provides some guidance concerning a rotating panel design assuming at least a subset of the sites were surveyed each consecutive year, which is not the case here.

Any thoughts and advice is appreciated, thank you.
Kevin

[darryl]

Hi Kevin,
I'd be very suspicious of any estimates from models where colonization and extinction have been specifed as year specific. Without any surveys in successive years, there's no direct information about the year specific rates.

The constant models will get the information from the surveys that are 2 years apart, although the estimates have been annualised. For example, if a species is present in year 1 and year 3 then the species may have been there the whole time, or went locally extinct then recolonized the site., ie, (1-epsilon)*(1-epsilon) +epsilon*gamma.

Ideally, if you're interested in these processes then you should be resurveying the units at the timescale which you want to interpret the resulting estimates (eg., annually, 2-yearly, 3.5 months, etc).

Cheers
Darryl