Non random Resight Probability

[AWalpole]

Hello all,

I am looking for some advice about appropriate models for some PIT tag data. I am new to Mark and I have begun working my way through the “Gentle Guide”. So far I am kind of stumped. Our objectives are to estimate survival and abundance of an arboreal rodent. We began live trapping in the study area and implanting individuals with PIT tags.
This trapping has been almost continuous throughout the last three years and we can treat the entire population as marked. Following our initial captures we began monitoring individuals at communal nesting sites (cavities) with our PIT-tag recorders. To find these cavities we followed radio collared individuals until the radio collars died or were removed. It is this PIT-tag data that we wish to use as our resighting data

My problem is with how to treat the non-random resight probabilities per individual (initially we were interested in the social behaviour of the species) as at different times different groups of squirrels were followed. Each individual does not have an equal probability of being resighted. The Poisson-log normal mark-resight model seemed appropriate assuming I separate the data into primary and secondary periods. However, the non-random resight probabilities complicate things. Any thoughts or suggestions about how I could go about estimating survival and abundance given non-random resight probability would be much appreciated? Or account for it in modeling?

Thanks,
Aaron

[bmcclintock]

Hi Aaron,

It can be tough to retrofit data into a model it wasn't designed for. Although survival estimation shouldn't be a problem based on what you describe, it sounds like abundance may be a bit of a pickle. The problem is not so much non-random sighting probabilities, but the (potentially severe) individual heterogeneity induced by the non-random timing and placement of PIT-tag recorders relative to the target population. In other words, if only a segment of the population within the study area is available to be sighted during a given closed primary period, then only the population size for the observable population (i.e., those individuals visiting cavities with recorders) can be estimated. Depending on the specifics, this could be substantially smaller than the entire population size in the study area. You probably need to think really hard on a few issues:

1) What is the exact population of interest during each closed period?
2) Was the entire population of interest observable, given the timing and placement of the recorders? If not, what segment of the population was observable?
3) Do you actually believe the entire population is marked? In the absence of unmarked individual sightings, both the mark-recapture and the mark-resight estimators will only estimate the marked population size.
4) Given the non-random placement of recorders, temporal and individual covariates addressing individual heterogeneity (e.g., perhaps using information on live trapping locations relative to recorders?) should be utilized if you're using the robust design Poisson log-normal model. Spatially explicit capture-recapture (SECR) may also be worth investigating for estimating abundance or density. Of course, any approach you choose assumes points 1-3 can be adequately fleshed out.

Cheers,
Brett