Traps as sites

[Wijid]

Hi there

I hope this is OK to ask - I have done some searching through the forums and MacKenzie book but seem to still be confused.

I have capture data over grids (>100m apart) of 56 traps (10m apart). Currently I have analysed my data at the grid level as biologically the grids are truely independent (same individual not seen moving between grids). I have analysed to estimate the number of grids required to detect the species at 95% level. My question is whether this can be done at the trap level?

My original concern was independence - individuals move around 30m a day. However I can not capture the same individual within a survey at more than one trap (as described in MacKenzie book p108 about this) as all animals are kept until the grid has been checked and then released at the end.

Any advice would be greatly appreciated.

Wendy

[dhewitt]

I think we need more information about the structure of the model you are fitting. There would be bias in p and Psi if you are treating each trap as a survey (spatial instead of temporal replicates for occupancy), but I don't think that's what you're doing (just a guess).

A recent paper by Nichols et al. addresses hierarchically structured occupancy models and might be of interest. I think it was in Journal of Applied Ecology.

[darryl]

Are individuals highly territorial with non-overlapping home ranges? How many grids do you have? What's biologically interesting about looking at within grid occupancy?

[Wijid]

Hi, Sorry if I haven't given enough info, was trying to be concise

The main question being asked is how much trapping do we need to do to detect the species. A lot of building work etc is going on around the area and this is an endangered species for which contractors are asked to look for as part of their environmental report. My grids were originally designed for mark recapture of individuals (hence grid layout to increase ability to recapture the same individuals). The traps were checked 20 times over 6.5 weeks each year from 2006 - 2009. I analysed each year separately as I am not so interested in extinction or colonisation of grids (or traps) but survey effort.

I have 8 grids (4 at each of two locations) for the first three years and then we increased it to 6 at each of these two location and 4 at one additional location in 2009 (16 total).

They have overlapping homeranges but do seem to show site fidelity to traps (note the traps are burrow mimics and so animals are not really caught). So there may be territoriality at the trap level but not at the grid level.

Thanks for the suggestion of the paper - I will look for it and see if it is at all applicable.

Wendy

[murray.efford]

Wendy
This is a spatially explicit capture-recapture problem, not an occupancy problem. Use your capture-recapture data to fit a detection function and then evaluate the probability of detecting at least one individual given the detection function and a particular trap layout and population density (either the density you estimate by SECR or a range of hypothetical densities). This evaluation is most easily done by simulation (it may even be the only way). The tools you need are in the Density program or the R package 'secr' - see the website www.otago.ac.nz/density.
Murray

[Wijid]

Hi Murray

I had a look at this method, we designed the study as a robust design only to discover our population is not closed over the six weeks of trapping. I understand that is an assumption for the density model? I was a little confused about this from the 2004 paper with mistnetting as an example, due to all the pooling done, if your definition of closed is the same as in mark recapture for survival analysis.

Reducing the number of weeks over which we use data from to try and make it closed doesn't work either largely as we reduce the recaptures down too low (recapture rates were pretty low).

Cheers for the reply I appreciate hearing any opinions!
Wendy

[murray.efford]

For SECR we assume a population is closed in the usual sense, except that animals may move around in their (stationary) home ranges.
What species are we talking about and how do you know the population is not closed over 6 weeks? I would not take any of the existing tests at face value when the data are from trapping grids.

You may be able to fit a more finely subdivided model by fitting detection parameters common to multiple 'sessions', thus pooling information. I admit that can be fiddly, but it's highly effective. The pooling method applied in Efford Dawson & Robbins 2004 was a clunky precursor of this, appropriate only to the simulation-based methods then used. It assumed within-breeding-season closure.

Murray

[Wijid]

I am talking about Tympanocryptis pinguicolla (Grassland Earless Dragon), an endangered lizard in the ACT and NSW. We focussed on a couple of ACT populations. We discovered large declines over my trapping years.

I tested for closure using CLOSE TEST (Stanley and Burnham, 1999) and Otis's test.
Season Otis et al 1978 Stanley and Burnham, 1999
z-test P x2 df P
Summer 2006 -999 2 4.573 7 0.71189
Spring 2006 -3.76547 <0.001 83.67182 28 <0.001
Summer 2007 -3.17360 <0.001 87.11066 20 <0.001

We sampled in two seasons each year summer is during the hatching season so it is likely we have dispersal on and off grids of hatchlings (as well as possible births on the grids). This is less likely in the spring sample (just prior to hatchling emergence) but I either had insuffiecient data to complete the tests or they indicated open populations.

Wendy

[murray.efford]

Thanks for the background. As I indicated before - I would be very surprised if the Stanley & Burnham closure test is effective in the presence of the extreme individual heterogeneity that arises naturally on trapping grids, but I cannot offer definitive evidence on this.
Murray

[egc]

Since this topic is not 'software specific', I moved it over to the general discussion forum.