Assessing bias

[lefa]

I am surveying a variety of habitats in the Northeast for two relatively rare species, lynx and bobcat. As a primary method for detecting the occurrence of these species, I have been using dogs that are trained to locate scat. I'm planning on using occupancy modeling framework in MARK to assess detection and site occupancy probability for both species. The problem is that some of the habitats used by these species are so densly overgrown that they are not workable with the survey method unless I use existing snow-mobile trails, hiking paths, or old logging roads to conduct the survey. I suspect that these trails enhance detection probability because 1) I can view the dogs better and determine when they have a located a scat, and 2) the scent of scat may travel down these lanes. Last year, I conducted surveys on trails and was successful at finding a large number of scats. However, I am well aware of the potential bias associated with conducted surveys only on trails. I am wondering if anyone has some ideas of how I can assess the potential bias associated with surveying trails.

[Summer Burdick]

You could assess the difference in detection probability by entering your data as two groups. One group for surveys done the normal survey method and one group for surveys conducted in dense forest. Then compare models where you make the detection probablies equal or unequal between groups.

[Fish_Boy]

I would agree with Summer Burdick. It is always fun when you plan the perfect sampling protocol and then nature lets you do it in some places, but not others. If the research is for a thesis you have a whole chapter on comparing field sampling methods. Professors love those, it shows them initiative and creativity which are the two halves of science... the other halves are luck and funding.