Bats: Study Design and Model Considerations

[Bryan Hamilton]

I'm setting up a capture-recapture study with bats. Bats will be captured with mist nets at cave entrances, PIT tagged, and released. The cave and mine entrances will be armed with antennas and PIT tag readers, so when a bat enters and exits the cave, the tag number, date and time will be recorded. Each PIT tag reader array costs around $6,000, so I'll only have three of these to deploy simultaneously, at three different sites. The main goals of the study are to quantify bat survival by species, gender, and age and capture seasonal use patterns of the caves.

I'm trying to figure out how to model these data. The set-up I'm proposing is very similar to single release models used in fisheries. One issue that has come up is that detection of bats entering and exiting the caves is not 1.0. Its going to vary as a function of bat velocity and distance to the antenna.

Beyond survival and seasonal use patterns, is there anything else I could do with these data? Does anyone have any experience with studies like this they could share? I’m thinking that a robust design framework may be appropriate. But I don’t really get any data on unmarked individuals, except when we mist net. The PIT tag readers can only record tagged bats.

I appreciate any thoughts or discussion.

[simone77]

Hi Bryan,
For sure there are several ways of modeling your data, this is just the way I would probably do. It seems to me that the issue of the detection probability <1 is crucial because otherwise it doesn’t make too much sense using capture-recapture modeling. First, you should consider how to define the time length of each session (from which pooling you observations) and the intervals between sessions. These decisions should be based on statistical criteria [1,2] and the knowledge of the species biology and the likely dynamics in the populations. If you think that the detection probability varies among individuals in a completely random way you can use a CJS model without further complications. You will find that the detection probability is somehow uncertain because of that but I don’t believe this has serious consequences on the estimate of survival that I understand to be your real parameter of interest.
However, in case you think that some individuals that you are not able to group together by some known feature (sex, age, etc.), are more (or less) probable to be detected than others, you could use a capture heterogeneity model. These models consider and handle the existence of a certain number of groups with different detection probabilities. For instance, I remember a nice example made by J.D. Lebreton at a workshop I assisted some years ago where a crowded colony of birds was being monitored and detection of marked individuals recorded: those individuals in the periphery of the colony were much more likely to be detected than those nesting in the core of it. This kind of models is easy to be suited in the multievent framework [3] because it represents a special case of state uncertainty. Some practical examples should exist in the E-SURGE help or in the manual [4] (but see also [5,6]). However, recently it has been shown by means of simulated data sets that these models are convenient to be fitted only if a large proportion of individuals have a very low detection probability [7]. Probably this will not be your case but it is a tool you might think of when modeling your data once you have them available. I don’t know if this kind of models may be fitted in other software, for example in MARK.
Just an idea.
Good luck!

1. O’Brien S, Robert B, Tiandry H. Consequences of violating the recapture duration assumption of mark-recapture models: a test using simulated and empirical data from an endangered tortoise population. J Appl Ecol. 2005;42: 1096–1104. doi:10.1111/j.1365-2664.2005.01084.x
2. Hargrove JW, Borland CH. Pooled population parameter estimates from mark-recapture data. Biometrics. 1994;50: 1129–1141.
3. Pradel R. Multievent: an extension of multistate capture-recapture models to uncertain states. Biometrics. Wiley Online Library; 2005;61: 442–447. doi:10.1111/j.1541-0420.2005.00318.x
4. Choquet R, Nogue E. E-SURGE 1.8 user’s manual. Population (English Edition). Montpellier, UMR 5175, France: CEFE; 2011.
5. Péron G, Crochet P, Choquet R, Pradel R, Lebreton J, Gimenez O. Capture – recapture models with heterogeneity to study survival senescence in the wild. OIKOS. 2010;119: 524–532. doi:10.1111/j.
6. Pledger S, Pollock KH, Norris JL. Open Capture-Recapture Models with Heterogeneity : I . Cormack-Jolly-Seber Model. Biometrics. 2003;59: 786–794.
7. Abadi F, Botha A, Altwegg R. Revisiting the Effect of Capture Heterogeneity on Survival Estimates in Capture-Mark-Recapture Studies: Does It Matter? PLoS One. 2013;8: 20–22. doi:10.1371/journal.pone.0062636