Using MARK for juvenile survival where p<1

[fionaewingburns]

Hi to you all,

I am studying a plover species where the young leave the nest after a few hours and I am trying to calculate survival from hatching to fledging. However, detectability is less than 1 so I thought that MARK would be a good way to do this.

I visited each brood weekly, but not all on the same day as I had several sites. So, although I could code the encounter history as visit 1, visit 2 ... my sampling period would not be instantaneous, which is one of the assumptions of MARK. Additonally, although you would get estimates for survival and detectability for broods of different ages (assumming you know the age at first capture), you would be ignoring the possibility of changes in survival (and detectability) over time.

Alternatively, the nest survival mode of MARK would be ideal, if there was some way of estimating detectability as well as survival.

I would like to know if anyone has tried to do this previously, or if anyone has any comments or suggestions for me.

Thank you for your help,
Fiona Burns

[abreton]

You say that you had several 'sites'. Assuming you have multiple nests at each site, and that all nests at a site were visited on the same day (or two perhaps), then it seems reasonable to estimate survival over the 5-6 day interval between occasions and estimate detection probabilities over the 1-2 day sampling occasions for each site. As you alluded to, it may not be reasonable to assume that detection probabilities were the same at each site when they weren't sampled on the same day (or two; but you can test this by comparing the appropriate models).

To try this approach, import your data into MARK as sites=groups (Chapter 6, gentle intro to MARK) or use a site-level individual covariate (Chapter 12). Either of these approaches will allow you to account/test for differences in detection probabilities (and apparent survival) among sites. However, the group approach may be the simplest and most intuitive.