Pradel models and heterogeneous capture probabilities

[mmcleod]

I am working on an 11-year mark-recapture dataset for willow flycatchers and am trying to get at one of the big questions -- namely, what is the population at these study areas doing, going up, going down, or remaining steady. Calculating lambda via adult and juvenile survivorship and fecundity (we know # fledges via nest monitoring) clearly doesn't get at the full question for a couple of reasons: 1) lambda estimates suggest the population is plummeting, but there is no evidence of this just based on # of resident adults detected each year, and 2) in some study areas, we detect far more (as in, sometimes nearly an order of magnitude) unbanded adults each year than can be accounted for from the known production of unbanded fledges at the study area in previous years. Thus, immigration is clearly happening and needs to be accounted for. So, I want to use Pradel models. I know that one of the assumptions of Pradel models is violated, in that capture probability for unmarked individuals is lower than "capture" (i.e., identified to individual via resighting) probability for banded individuals. Each year we do intensive territory & nest monitoring, which entails resighting all the resident adults, and attempting to band any adults that are not already banded. Thus, each season we know how many resident adults we had at each site, how many remained unbanded throughout the season, how many were banded but could not be identified to individual via resighting, etc. Overall, at the end of each season, the ratio of newly-banded:still-unbanded adults is lower than the ratio of adults-IDed-via-resighting:banded-adults-not-positively-IDed. I.e., we're more likely to be able to identify a previously banded adult than we are to capture an unbanded adult, thus violating the equal capture probability assumption. Given this unequal capture probability, it seems that the Pradel models will give me an estimate of recruitment (or lambda) that is biased low? Is there any way to correct for this, given that we know the numbers of banded vs. unbanded and resighted vs. unidentified adults? I'm assuming that the probability of being detected at all is equal for banded and unbanded adults.

Thanks,
Mary Anne