Tag / Battery Life - Removals When Programmed To 'DIE'

[Halfyard]

Hello all,

I am looking for help with designing a CJS project that deals with tag / battery life in acoustic tags. I’ve been using MARK for a few years, but only with fairly basic CJS models.

I am looking to design a long-term (i.e. 10 year) acoustic telemetry project where striped bass will be tagged every year and likely encountered every spring and fall (near their over-wintering habitat). As such, I will have 20 sampling occasions. I plan to mark fish only in the first 6 years, so there will be 6 ‘cohorts’. The tags will have a battery life of 5 years (programmed to have a ‘drop dead date’). As such, tags released in year one will be available for the first 10 sampling occasions, but will be unavailable thereafter. I’ll also need to include several covariates.

My question is; what model should I consider for such a scenario? Can I adapt a CJS or do I need to branch out?

Standard coding for removals in CJS models (i.e. 2.2 in “A Gentle Intro.”) requires that individuals are encountered at the time of removal. (e.g. 10100 -1 = released occasion 1, encountered on occasion 3 and removed at that point). Similarly, I have considered a joint live encounter – dead recovery model, but I think it suffers from the same problem (i.e. tag is not recovered but would be removed). It would not be known if the animal was alive at that point, BUT, we know it has “died” IF it was alive.

Any nudge in the right direction would be a help.

Thanks

Note: This was originally placed in the "general questions" section, but I think it may be more appropriate here.

[Halfyard]

As so often happens, the combination of writing out my problem (and a good sleep) might have helped.

I am now considering using a Time-Since-Marking (TSM) model (Ch. 7.4 in “A Gentle Intro.”). The idea of permanent emigration makes sense, however in my case, I’d like to have the individual ‘emigrate’ (i.e. tag stop transmitting) during the 10th interval.

I think I may set it up as follows: 1) allow for a ‘tagging effect’ by coding for a separate survival during the first period post-tagging, 2) allow full time-dependence in all subsequent years, 3) accommodate the tag life issue by coding all survival post-tag expiration as the same. This will be ‘thrown out’.



[Note: the image is cut-off on the right hand side]

Does this appear to be a reasonable approach? Does anyone know of any worked examples similar to this with covariates?