discussion / Camera Traps  / 16 March 2022

Advice Needed: Camera Trap Occupancy Modeling for Small Data Set

Hello All! 

I'm currently trying to analyze predator-prey spatial overlap using multi-species occupancy modeling. I've created 2-way (between two species) multi-species occupancy models using unmarked in R with a fairly small data set (30-80 observations per species across 30 sites over 2 months). However, some of my models are not converging. I use 0-3 covariates per model, and there does not seem to be any pattern as to which covariates are causing the lack of convergence (for example, a model with cov1+co3 or cov2-only both work, but not an intercept-only ~1 model). Some models produce NaNs for estimates and/or SEs, and others have both NaNs and no convergence. Nealry all the models have crazy SEs (which may be expected with the small sample size). 

I've checked and do have detections from all species for each combination of covariates. I've also tried using single species models with detection frequency as a covariate, and am still not able to get convergence. I've checked all my input files (detection history, site-level covariates, etc.), and everything is correct. 

Does anyone have any suggestions about how to move forward from here? Whether that be to continue trying to use multi-species occupancy modeling or another way to analyze spatial overlap between two species?

Any and all guidance is highly appreciated! 




One thought, if you haven't tried it yet and depending on how you're creating detection histories, would be to increase the occasion length such that there are fewer 0s in the data. I've found convergence issues if most of the occasions in the histories are 0s, with a few 1s here and there.

 

Also, there are a couple of new "continuous time" occupancy models out there that might help address your convergence issues. I have no idea if this is the case, but given that they don't treat data the same way in terms of discrete occasions it might be worth a try. And, one is explicitly mult-species, so that might be good for you as well! Here are the papers:

Rota CT, Ferreira MAR, Kays RW, Forrester TD, Kalies EL, McShea WJ, Parsons AW, Millspaugh JJ. 2016. A multispecies occupancy model for two or more interacting species. Methods in Ecology and Evolution 7:1164–1173.   https://besjournals.onlinelibrary.wiley.com/doi/10.1111/2041-210X.12587 

 

Emmet RL, Long RA, Gardner B. 2021. Modeling multi‐scale occupancy for monitoring rare and highly mobile species. Ecosphere 12.  https://esajournals.onlinelibrary.wiley.com/doi/full/10.1002/ecs2.3637  and attached

 

 

 

 

emmet_et_al_2021_modeling_multi-scale_occupancy_for_monitoring_rare_and_highly_mobile_species.pdf

Couple other papers looking at predator-prey interactions with camera trap data & multi-species occupancy models (though you may/probably know about some of them already!) -

Van der Weyde, L. K., Mbisana, C., & Klein, R. (2018). Multi-species occupancy modelling of a carnivore guild in wildlife management areas in the Kalahari. Biological Conservation, 220, 21-28. https://www.sciencedirect.com/science/article/abs/pii/S0006320717316464. 

Murphy, A., Kelly, M. J., Karpanty, S. M., Andrianjakarivelo, V., & Farris, Z. J. (2019). Using camera traps to investigate spatial co‐occurrence between exotic predators and native prey species: a case study from northeastern Madagascar. Journal of Zoology, 307(4), 264-273. https://zslpublications.onlinelibrary.wiley.com/doi/abs/10.1111/jzo.12645. 

Kleiven, E. F., Barraquand, F., Gimenez, O., Henden, J. A., Ims, R. A., Soininen, E. M., & Yoccoz, N. G. (2021). A dynamic occupancy model for interacting species with two spatial scales. bioRxiv, 2020-12. https://www.biorxiv.org/content/10.1101/2020.12.16.423067v2.abstract.