SECR when lots of re-captures but little/no spatial recatpures

[Tim Henderson]

Hi all,

I have a trap array that is sufficiently spaced for my target species. However, I am also getting lots of captures of another species, and would be interested in obtaining a density for them as well without changing the design. The problem is, this species has a smaller home range and I get very little or no spatial recaptures due to the larger spacing of traps. Using the capture-recapture data from my current design, what alternatives are there for estimating density of this non-target species without sufficient spatial information? Could I use other movement data of this species as a proxy for sigma? I.e. if I obtained movement data from GPS tracking, or from a separate secr study with more closely spaced traps, can I incorporate this data into the secr modeling of the current design?

For extra context. Each trap site has 5 traps, surveyed for 4 nights, 150 sites total, spaced ~500m apart. The preference is to run one trapping survey to cover multiple species (as this is an ongoing study).

Thanks for your help!

[Murray Efford]

I think you understand the issues, and that there is unlikely to be a 100% convincing answer, short of compromising on the design e.g. injecting some closer spacings. As you seem to be thinking ahead, is that not still possible? If not then I can only offer some opinions. Applying sigma estimates derived from another study or from telemetry is quite feasible (i.e. the software allows you to fix sigma to any value and see the result). However, you carry the burden of justifying this biologically, and the estimates will always be qualified ('assuming that sigma applies'). Results from other trapping studies seem to me preferable to telemetry data. Although we think of sigma (and g0/lambda0) as independent of trap layout, there are some suggestions this is not always true (e.g. salamander example of Fleming et al. (2021) https://doi.org/10.1002/eap.2419 cited in the SECR book).

Others will have different opinions - I hope we hear them!

[Tim Henderson]

Thanks for you response Murray. The current design has been in place for several years, hence I'm keen to keep it unchanged (and mainly due to logistical constraints of adding more trap sites). It sounds like (with some limitations), I could potentially replicate a subset of the trapping array with more closely spaced traps, as a one-off survey to obtain a sigma value for this species, and then use this value for the current array going forward (the survey is conducted annually). The main limitation of this would be that I would have to assume that sigma also remains the same for subsequent surveys. Would this be correct?

Cheers  

[Murray Efford]

Yes. There is also the issue that uncertainty in the estimate of sigma is not automatically propagated to the density estimates, but I think if the one-off estimates are good that's a lesser concern. You might also want to keep an eye on known drivers of change in home range size - big changes in density over time are expected to drive inverse changes in sigma.

[James Russell]

I don’t have much to add here, and agree that (at least in my experience with invasive rodents) sigma can be less variable and so transfers better among sessions than g0 (which can be very session specific), unless as Murray has noted substantial density variation comes in to play.

 

I have been known to run a single live-trapping grid to estimate sigma and then additional kill-trap grids (less labour intensive but which by design can’t estimate sigma) to get density estimates across different sites. If you run these as separate sessions in one analysis and use a null model for sigma the sigma estimate is used to calculate density for all grids even though in practice its only estimated from one grid. You could do the same here (use a supplementary grid to estimate sigma once and then include it in analyses of all data).

 

If you wanted to be real rough and ready you could just guesstimate a lower and upper value for sigma for this species and plug them in as fixed values to get a ball-park range for density from existing data, but I’d only advise that if someone else has estimated sigma for the species as converting other movement estimates to sigma is fraught.

 

James

--
You received this message because you are subscribed to the Google Groups "secr" group.
To unsubscribe from this group and stop receiving emails from it, send an email to secrgroup+...@googlegroups.com.
To view this discussion, visit https://groups.google.com/d/msgid/secrgroup/717db00e-db98-425b-91da-d6dabfd5e9d7n%40googlegroups.com.