Page 1 of 1

within and between season survival

PostPosted: Mon Mar 07, 2022 3:20 pm
by Chance
Hi, is there an analysis method that incorporates within and between season captures/recaptures. I am interested in winter survival for a migratory bird. Similar to the paper in the link below.

In this paper the authors have several trapping sessions per winter, but combine all captures/recaptures within a winter into a single column for the capture history and then use several years to estimate winter survival. It seems like recapture probability (and probably survival too) would be more accurate if both within and between season trapping sessions were incorporated.

This may be addressed elsewhere but I have not been able to find it. Thank you for any help provided.


Re: within and between season survival

PostPosted: Mon Mar 07, 2022 3:53 pm
by cooch
Dn't have time to look at the paper you linked, but if I'm interpreting you correctly, what you're describing is Pollock's 'robust design'. See Bill Kendall's chapter in the MARK book: ... chap15.pdf

Re: within and between season survival

PostPosted: Mon Mar 07, 2022 4:58 pm
by jlaake
Evan may have misunderstood your post because the Pollock robust design assumes closure between secondary sessions within a primary period. I have not looked at the article you referred to but if p is small for each session or capture effort does not cover entire area each session then it is can be best approach to pool. But given you want to estimate survival within season then you can simply after different time intervals between each capture occasion and have a different survival rate within season vs between in a CJS analysis. Alternatively you can use an open robust design if you have losses within season. It has a lot more capabilities but requires more data with many more types of parameters. All depends on your situation which you have not explained. That said, it sounds like you need to do a lot more reading including the chapter Evan suggested.

Re: within and between season survival

PostPosted: Wed Mar 09, 2022 12:25 pm
by Chance
A little more on the project: I am capturing birds during three winter capture sessions for three years. Sessions are approximately 1 month apart. I have been capturing about 100 birds per winter at three sites 1-3 km from each other. I have only one bird that has moved from 1 site to another. 30-50% of my captures are already marked. I can age a portion of the birds but not 100%.

I dont think Pollocks robust design will work because I have mortality between secondary sampling periods.

I think I understand Jeff's suggestion but would like confirmation. Essentially birds captured in different time periods would have a covariate that would be the number of days since the last trapping session. I would also need some sort of covariate that would indicate whether the time passed since the previous capture session was during winter or not winter which may require an interaction between the two.

So something like this:

0 30 60 365 395 425 790 820 850 - days since last capture period
011 011 011 - 0 means not winter before last capture period, 1 means was winter in last capture period

I can see how this would work in a cjs model for the initial cohort of birds moving through three years but still not sure how that incorporates birds hatched between winters. I could run two analyses, one for the initial cohort moving through three years to get between season survival and another combining all seasons into a single season (so only three column capture history) with a covariate for season id and random effect to account for individuals captured during more than 1 season to get single season survival.

Thank you for all of the help yall are providing. I have used Mark and marked (prefer) with simple cjs models in the past. I have been looking at the Multivariate Multistate CJS models in marked and maybe I could manipulate one of those to suit my needs?


Re: within and between season survival

PostPosted: Wed Mar 09, 2022 3:25 pm
by jlaake
Time intervals would not be covariates. They are used in calculation of survival between capture intervals. If S is annual survival and you had 3 occasions each a month apart and then a gap to next winter, the time intervals are 1/12,1/12,10/12 and survival would be S^1/12*S^1/12*S^10/12 and the values for S could differ for each interval.

CJS models use data after first entry (first marking) so when they enter does not matter.

You need to do a lot more reading. MVMS models in marked aren't really necessary for your data. I would continue to work with MARK and if you to work in R, go read the documentation for RMark.