Open-population SCR & sex-specific recruitment identifiability

[Zoe Woodgate]

Hi all,

I'm working on fitting an open-population spatially explicit capture-recapture model to six years of camera trapping data (following Rostro-Garcia et al. [2023] pretty closely). I'm working in NIMBLE, and it's on a large carnivore.

I'm appearing to have an issue with my female per-capita recruitment rate (gamma_f). More specifically, I have set the prior to be:

gamma_f ~ dunif(0, 0.5) 

This is the same as for males. This is a bit above a 'realistic'  value, but I didn't want to constrain the prior too much. However, the posterior mean is 0.39 and 51% of the posterior is above 0.4. Gamma_m, in contrast, performs as expected (posterior mean of 0.27, with only 5% above 0.4). This runaway gamma_f is also impacting estimated abundance- Nfemale in year one is 115, and Nfemale in year 7 is 167, implying a 45% increase, which is biologically implausible for this population. The Nmale remains pretty stable over time. Survival (phi) for both males and females is similar, hovering around 0.75. Annual density also seems to be much higher than what a multisession ML secr model would suggest. For some additional context, 63 unique females were detected across all six sessions

I was initially worried that sigma and detection were impacting gamma. Consequently, I now specify that sex-specific sigma and baseline detection probability (p0) priors are informed by a pre-fitted ML secr model (constant density across all years, h2 sex). I'm currently running a model with a tighter prior- dunif(0, 0.30)- but not sure if this is appropriate? I'm also considering dropping sex-specific gamma, but given that this model seems to be fine for other sites with more data (~10 years), I'm reluctant to do this.

My main questions are thus:
1. Is female recruitment simply not identifiable from this data structure (6 surveys, moderate individual recapture rates), and if so, what would be your recommended approach? 
2. What additional methods should I use to help distinguish "your data can't identify gamma_f" from "gamma_f is genuinely is high"?

Thanks in advance for any advice! Happy to share more information/data privately; sadly, I can't share much on the public forum due to data sensitivities.

Cheers,
Zoe

Some key refs:
Rostro-García, S., Kamler, J.F., Sollmann, R., Balme, G., Augustine, B.C., Kéry, M., Crouthers, R., Gray, T.N.E., Groenenberg, M., Prum, S. and Macdonald, D.W., 2023. Population dynamics of the last leopard population of eastern Indochina in the context of improved law enforcement. Biological Conservation

Chandler, R.B. and Clark, J.D., 2014. Spatially explicit integrated population models. Methods in Ecology and Evolution

Augustine, B.C., Kéry, M., Olano Marin, J., Mollet, P., Pasinelli, G. and Sutherland, C., 2020. Sex‐specific population dynamics and demography of capercaillie (Tetrao urogallus L.) in a patchy environment. Population Ecology