Closure assumption and occupancy estimation

[AWalpole]

Hello all,

I am building an occupancy model from replicated surveys but I am concerned about the assumption of closure at my sites. Do to the wide ranging nature of my study species, it is likely that my sample units do not have constant occupancy over the course of my survey period. I am concerned that this violation of the closure assumption will bias my occupancy estimates. Are there any suggestions about how I could address this issue in order to produce reliable occupancy estimates?

Thanks,
Aaron

[murray.efford]

Aaron
I think it might help if we knew your study species, the layout of your sites, and the reason you expect non-closure. I'm guessing that you might be applying occupancy methods beyond their limits.
Murray

[AWalpole]

Thanks Murray.

I'm studying the Canada lynx which has a fairly large home range. I snow-shoed transects in the winter in search of lynx tracks to build the occupancy model. Each transect was visited three times to estimate detection probability. The transects are positioned along a gradient of cover types representing the landscape. Unfortunately, the transects cover an area smaller than the average home range size of the lynx. I can't be sure of constant occupancy throughout the entire survey period (during all 3 visits). The transects cover an area of about 1km-squared while a lynx home range is about 50 km-squared. However, because snow tracks stick around for a while and because I visited each transect three times times I'm fairly confident that if a lynx were using the habitat covered by the transect I would detect it. None the less, I'm not quite sure how to handle this situation. How robust are occupancy estimates to small violations of the closure assumption?

Hope this clarifies things a bit. Thanks,

Aaron

[robertlong]

Hi Aaron,

I experienced a similar situation during my PhD work when using scat detection dogs for an occupancy analysis of black bears, fishers, and bobcats in Vermont. My situation was quite similar in that I was able to detect the target species after they may have left the survey sites--which like in your situation were in most cases much smaller than the species' home range. After much discussion with various occupancy aficionados, I ended up included the following statement in my dissertation:

"Scat-based detection methods increase dramatically the time frame during which the detection of a carnivore species is possible, making it feasible to detect individuals that are not currently using a surveyed portion of their home range. In cases such as this, where the species likely ranges beyond the survey site, the system cannot be considered closed. Assuming changes in true occupancy of the site over time are random, the occupancy estimates are still unbiased (MacKenzie et al. 2005), but should be interpreted in terms of sites “used” (i.e., sites where the species is sometimes present during the survey “season”)."

Cited: MacKenzie, D. I., J. D. Nichols, J. A. Royle, K. H. Pollock, L. L. Bailey, and J. E. Hines 2005. Occupancy estimation and modeling. Academic Press, Burlington, Massachusetts.

Hope this helps,

Robert

Robert Long
Road Ecology Program
Western Transportation Institute (WTI)
Montana State University
robert.long@coe.montana.edu

[AWalpole]

Robert,

That helps a lot and I’ll definitely alter my interpretation of the results to use rather than occupancy.

Thanks,
Aaron