Encounter History question...

[ChelonianMan]

Greetings all,

I am currently working on a mark-recapture study using pond turtles. Upon starting to run my data through MARK, I found that one aspect of my experimental design may pose somewhat of a problem. In my project, I check my turtle traps every other day for three weeks, which gives me 11 encounter occasions. When a turtle is first captured, it is returned to the lab to be marked and measured. The turtle is then released the next time the traps are checked. This, however, makes the turtle unavailable to be trapped in that second trapping occasion. If an already processed turtle is captured again, it is simply noted where it was captured, and then immediately released.

My question is, is there a way I can indicate in my encounter history, or another aspect of the program, that the turtle is still alive, but not available for capture? By leaving the encounter histories the way they are, the data shows that detectability increases over time, which isn't always the case. Any help or advice you can offer will be greatly appreciated.

[ganghis]

I assume that since you're talking about a turtle still being alive that you are running a CJS model to estimate survival (if you're looking at abundance it's a bit more complex). I'll simplify your situation by assuming that you have only 4 capture periods. Assuming a time specific model, the PIM for survival could be

1 2 3
1 3
1


and the PIM for capture probability could be

4 5 6
4 6
4

then, when you go to run the model, fix parameter 1 = 1 and parameter 4=0 (if you give the encounter histories a 0 for the period of release). In effect, you're saying that animals always survive the first period following intitial release and that they are never recaptured in the period following intial release. As usual, parameters 3 and 6 (the last p and phi) cannot be estimated separtely in the fully time dependent model.

Cheers, Paul

[abreton]

An alternative would be to code the encounter history to reflect when the animal was actually released. So rather than (e.g.) 101, you'd code it 011 even though it was actually captured on occasion one. This approach assumes a live encounter CMR model, such as CJS, which condition on first release rather than first capture - stated differently, these models do not attempt to model initial capture probabilites.