cmr with highly abundant stream fish

[royworth]

Dear forum,

My colleagues and I are planning a cmr study that will involve closed, removal sampling (multiple removal passes via backpack electrofishing) of a potentially very abundant stream fish. We would like to be able to estimate seasonal survival, lambda, the components of lambda (via Nichols et al. 2000), and (perhaps) stage/size transition probabilities among a number of sites draining different land use types.

However, I am assuming that we will not be able to individually mark all the animals in any single closed, seasonal sample (potentially > 2000 individuals). Therefore, I was hoping that someone might be able to provide a quick rundown of the feasibility of estimating the parameters we would like given this potential limitation. The two marking options, I believe, would be batch/group/cohort marking of all fish per site (no individual identifiability) or individual marks (likely VIE) of a subsample (~200-300 individuals might be possible).

Thanks in advance, for any advice/comments.

High Regards,

Roy

[dhewitt]

If you can individually tag 200-300 fish per event and select them in a random, representative way, do that rather than batch marking all of them. The options for modeling are much greater with that approach. And you're going to need to account for heterogeneity in detection with removal sampling... it can be bad and bugger things up. Plenty of literature on that issue. Is there a reason sampling has to be done by removal sampling?

- Dave

[royworth]

Thanks a bunch, Dave.

Actually, no there really isn't any reason that we'd HAVE to use 3-pass removal sampling, other than it is the method I've gone with traditionally for estimating p-hat and N-hat. I also have wrestled first hand with the heterogeneity problem, now that you mention it. I would be very grateful for any other suggestions you might have that would still fit the robust design framework (Lincoln-Peterson?).

Thanks again,

Roy

[dhewitt]

I guess I am confused as to why you need L-P or removal sampling when abundance is not in your list of parameters of interest (good for you!). Why not just thorough random sampling at intervals through time -- say 3-4 sampling events, one per day or every few days (secondary periods; keep the block nets up), repeated each month, or every 6 months or whatever seems reasonable given the life history of the fish (I'm thinking this must be some rather short-lived minnow/darter/sculpin, or maybe trout [sigh]).

- Dave

[royworth]

Good points. Abundance is not necessarily of primary concern, although an estimate of p* in the context of the (multi-state) robust design is. Our primary concern is getting estimates of the components of lambda by size class (i.e., survival, local recruitment, immigration) among the site groupings (i.e., land use categorization).

Also, your guess was pretty good: we are going to try to work with the central stoneroller. Although, that could change to something like a darter.

We have 12 sites that we are planning to sample seasonally and we have about a 4 to 6 week time frame to complete a primary session. In the past, I had generally stuck to the removal design (e.g., Peterson, Fausch, et al. 2004) given that I figured we could cover fewer sites in a shorter time frame than with a design that required multi-day sampling. Heterogeneity has been an issue in some of those cases in the past, but not too overwhelmingly due to the fact that we were working with fishes with pretty high capture efficiencies (> 0.6-0.7) in quite small streams (e.g., trout). Nevertheless, I appreciate the likelihood that capture efficiency could be much more variable for these small fishes in somewhat larger streams, so I would definitely consider the other approach.

Since I am somewhat ignorant of the non-removal-based literature, I suppose my next question to you would be how many secondary sessions would you expect to need to get reliable estimates of p*? Would one day of (subsample?) marking and one day of follow-up sampling (and perhaps more marking) per site (so 2 secondary sessions per primary) suffice in your experience?

Thanks again Dave,

Roy

[dhewitt]

Abundance (with some nuanced definition) is available from the robust design of course. I was just working backwards from your design to try and decide what your primary objectives were.

Two secondary sampling events per session would not get you much (anything?), especially in terms of heterogeneity in detection probability. Which is why you see general recommendations for 3+ removal passes. I'd probably want 4 for the first go-round and then maybe drop to 3 if things looked "good". But this mostly applies to abundance, which is based on the histories in the closed period (secondary sampling events), although p* would also be based on those secondary histories. Hmmm...

I guess the only difference between the removal strategy and the non-removal strategy is that without removals (and provided you allow fish to get back to "normal" by waiting a day or two) you are not putting any restrictions on detection probability. Right off I can't picture how a removal design would affect inferences about p*, esp. since I am not sure why that is of primary interest. For abundance, correct me if I'm wrong but my recollection is that the removal strategy can be doomed for abundance when the depletion pattern doesn't behave, esp. with only 2-3 passes (e.g., 200 on pass 1, 500 on pass 2, 5 on pass 3). The Peterson et al. (2004) method works great when you can remove 90% of fish in two passes(!), but that is far from general for fish sampling. I'd bet an old ball cap that you won't get that with a darter species.

Overall, I'm not sure of the overall effects of these two alternatives on the various parameters as a function of the design elements. Sorry I can't be of more help on that, but I think you would want to use the MARK simulation routines to get firmer answers anyway (Appendix A in the GIM book). And since Gary was involved in the Peterson et al. analysis he can chime in if I'm off base.

- Dave

FACT: Darters are cooler than stonerollers.