www.phidot.org

Hi All,

I have a dataset for a laboratory PIT tagging experiment with (YY) Trojan male brook trout, and would like some feedback on the best mark-recapture model to analyze the results.

Experimental groups were tagged with two different sizes PIT tags in three different body locations, and there are both single- and double-tagged treatment groups. After tagging, fish were placed in circular tanks and resampled ~monthly (on 7 occasions) to confirm tag retention. Tanks were checked daily for rejected tags, and dead fish. During monthly sampling, fish that lost all their tags were removed (censored) and fish that ingested shed tags were removed (censored). Because some fish ingested shed tags before the tags could be physically recovered in daily checks, the date of shedding for ingested tags could only be ascribed to the interval between the monthly sampling.

The parameters of interest are tag retention and survival. The fate of individual fish is known and the fate of all tags is known (with a single exception), so we are assuming probability of detection is 1. Individual covariates can be included (e.g., size at tagging).

There were few mortalities so overall survival was high (96-100% per treatment group). Retention by tag size and location appears to be independent of whether a fish was single- or double-tagged but that question awaits analysis.

Jointly estimating tag retention and survival (i.e., in the same model) would be preferred, but given the high survival (and potential boundary issues with estimation) I could also envision analyzing tag retention separately and handling mortalities/removals by censoring at the time of death/removal. Including the single- and double-tagged fish in the same analysis would be preferred, so that I could evaluate whether tag retention is independent, but I'm not sure if that is possible.

It seems like a number of models could potentially be used: known-fate model (S for tag retention), nest survival model (S for tag retention), multi-state model (S for survival, psi for tag retention), occupancy/two species occupancy model (psi for tag retention), etc.

I’d be interested in any suggestions on which would be preferred, given the information above.

Thanks! Doug

Hmm. Interesting situation where you know the fate of each critter. Unless you are concerned about tag affect on survival, I can't see a good reason to estimate survival. Your analysis will be easier if you just censor those individuals at the time of death and focus on estimation of tag retention. Whatever you do, I suggest that you consider models with dependence in tag loss. Do you know which individuals lost both tags or just the number of fish that lost both tags between occasions? If you decide to use the multistrata approach you may want to look at

Laake, J. L., Johnson, D. S., Diefenbach, D. R., & Ternent, M. A. (2014). Hidden Markov Model for Dependent Mark Loss and Survival Estimation. Journal of Agricultural, Biological, and Environmental Statistics, 19(4), 522–538. https://doi.org/10.1007/s13253-014-0190-1

In particular the supplemental material for the article will be the most useful as it shows how to fit the models with MARK via RMark. Paper and supplement are available on my ResearchGate site.

Hi Jeffrey,

Thank you for the reply.

jlaake wrote:Unless you are concerned about tag affect on survival, I can't see a good reason to estimate survival.

From a design standpoint, we were interested in the effect of tagging on survival and a few fish did die shortly after tagging, but, overall, survival was very high. Statistically, I suppose we could demonstrate that survival of 97% and 100% are different, but perhaps not that isn't that meaningful.

jlaake wrote:Do you know which individuals lost both tags or just the number of fish that lost both tags between occasions?

There were no individuals that lost both tags. Some single-tagged fish lost their tag, and some double-tagged fish lost one of their two tags.

I've seen double-tagging studies where the tagging was used to help estimate retention (based on the proportion of recaptures that retain one or both tags). In our case, we were/are specificially interested in whether double tagging might help increase the probability of a fish being individually identifiable over multiple years. The most common PIT tag implantation site for salmonid fishes is the body cavity, but the tags can be expelled during spawning, particuarly for females, so having a second tag somewhere in the body is helpful.