Species interaction study design question

[sixtystrat]

We hope to use occupancy data to determine if there is a negative relationship between river otter and muskrats on streams in Kentucky. The hypothesis is that otter reintroduction negatively affects muskrat populations.

Our idea is to look for otter and muskrat sign upstream from randomly selected bridge crossings. We would repeatedly sample the sites to estimate occurrence and detection probabilities, using covariates to account for habitat differences, factors affecting detectability, etc. The model and test criteria to detect multi-species interactions seem pretty straightforward. My question is about study design.

Would it be legitimate to sample several sections of stream at each bridge crossing, counting each section (say 100m) as a sampling event? This would be more cost effective (less gasoline) than repeatedly visiting the same sites. Spatially, our assumption of course is that otters and muskrats would be equally present/not present for all of the stream sections at a bridge crossing. Would this approach be reasonable? Thanks!

[darryl]

Biggest question would be whether the sections are independent from one another, and that you select and survey sections randomly; with replacement (so same section may get survey multiple times) if there's a limited number of sections and each crossing that you could survey.

[sixtystrat]

Darryl:
The protocol would be to randomly select, say 20 bridge crossings in a county, go to each one and walk upstream for, say 500 meters. The survey would consist of 5 100-m sections. So, we would have 5 presence/absence data points for that site (bridge crossing). We would not return to that bridge to sample again. The actual length of the section would depend on how much otter/muskrat sign we find. We would determine the actuallength of the sections after we do some pilot work so that our detection probabilities are reasonable.
Obviously, the sections are not independent from one another. But I was not aware that that was an issue, since revisiting sites over and over amounts to repeated measures and is not independent either. The species would be assumed to be present or absent for all of the sections at a bridge crossing. Am I missing something?
Hope this helps to clarify things. Thanks!

[sixtystrat]

In thinking about it a little more, it might be that time-specific weather factors could introduce heterogeneity into the detection probabilities if each bridge crossing were sampled as multiple sections of stream at one time. For example, a recent rain could reduce detection probabilities and that would be the same for all the sections of a stream at a particular bridge crossing. Another bridge crossing might be surveyed a week later when detection probabilities have improved, resulting in heterogeneity. We could use rainfall as a covariate but, alternatively, surveying 1 section of a bridge crossing on multiple occasions might reduce the weather-related biases. Does that make sense?

[darryl]

You could use a covariate to account for this sort of thing, but then you're reliant on the covariate to explain the heterogeneity. There may also be other similar issues that arise whenever you try and do multiple surveys (either spatial or temporal) in a single visit. We discuss some of this issues in Chapter 6 of our book.

Cheers
Darryl