Help with fish survival study

[Snook]

Hello all,

I am trying to determine if the MARK analysis I have performed for my current telemetry project is appropriate. Any feedback would be greatly appreciated, and I will try to keep my description as brief and simple as possible.

I mark juvenile fish with PIT tags beginning in November (I catch fish I mark with a seine net). I have two years of data, and I keep each year separate. Marking is done somewhat haphazardly. One year I marked for a week in November, a week in January, and a week in February. I marked more often in my first year, and also marked in December and April. My first year I marked 850 fish, my second year 590.

I have PIT tag antenna in my study creeks, recapturing marked individuals on a continuous basis. I have nearly 400,000 recaptures over 2 years, and an 80% recapture for my first year, and a 89% recapture for fish marked in the second year. I recaptured the majority of fish I tagged in more than one month.

I would like to calculate survival beginning in November, and going through August for each year. As I have 400,000 data points, I collapse my recaptures into monthly sampling periods to handle them more efficiently. Using recapture histories (which include the marking events and a rare physical recapture), I calculate monthly apparent survival and capture probability using the Recaptures Only CJS model (I often just do (t) and (.), but also try different seasonal approaches).

I am curious if this approach is ok, given that since I am collapsing all recaptures in a month into a single 0 or 1, there is no time-lapse between sampling periods and therefore may not have instantaneous sampling periods.

Basically, since I am recapturing fish continuously with no time gap between sampling events, is using the Recaptures Only model OK, or should I be looking at another model such as Barker?


This is just my initial analysis. Once I complete this, I would like to use a mutli-state model, as the tagging I am conducting (and antennae I have placed) occurs over 4 separate creeks.

[dhewitt]

You need the Barker type model, and there is an example in Transactions of the American Fisheries Society in 2008 (v. 137:1709–1722). That said, I don't think that all the issues with these continuous remote PIT tag recorders have been resolved, and you have a variety of issues that will complicate things.