Detection Models

[joseph.nett87]

Hi all,

I am new to this software, but I am an Ecology M.S. student interested in learning more about these models. Most of what is found within these models centers around occupancy, but I am simply interested in Probability of Detection and associated models. However, due to the nature of my project, occupancy models cannot be conducted. My project centers around the creation of a sampling protocol for an invasive fish with the use of three gear types. The main reason that occupancy models could not be performed on the data is that all sites had a known abundance (ranging from sparse to dense) in accordance with the grant protocol, so all sites had a known population.

So my question is how to start up and utilize the detection figures provided. I went through the software and conducted single-season, single species tests, and within that I compared the use of both 'constant P' and 'survey-specific P'. Those models then gave me probability of detection for each gear type for each survey number. Does this sound like a correct assumption that I can use these Pdetection figures that are given to me? Is there something that I am completely overlooking? Are there any equations that I can plug my information into that would examine the number of surveys at each site necessary to determine whether or not the fish is present? I appreciate any help and would be happy to possibly clarify the situation further if necessary.

Thanks,
Joe

[joseph.nett87]

After I run the models, I have a problem with one of the findings. As stated previously, all sites had known populations of the organism of interest. When I ran the analysis for all the gear types, the most ineffective one popped up as having the greatest Pdetection (p = 0.75, as compared to p = 0.69 and 0.46 for the other two gear types which had a higher occurrence of detection at sites). I tried fixing the PSI parameter at 1, but that didn't help. Could this be because when the gear type was successful in detecting, it was commonly detecting the organism in all three trials? Because this is obviously not the correct assumption to be made if it only detects the fish in half of the sites when all of the sites have a known abundance.

Again, thank you for any feedback.

[jhines]

Joe,

From your first message, yes, you can still use occupancy analysis when you're only interested in detection probabilities and you know all sites are occupied. Your question in the 2nd paragraph is confusing, though (where you ask how to determine the number of surveys necessary to assess occupancy of a particular site). Did you mean to ask how many surveys would be necessary to detect the fish, given occupancy=1.0? If that's the case, then you can graph the probability of detecting at least one fish for a given number of surveys and decide how many is enough to get the probability close to 1.0.

The results you mentioned in your 2nd post seem to go against intuition. Since occupancy is 1.0, overall detection probability should be the proportion of detections. I'd be happy to look over your output if you'd like to send it to me offline (jhines_at_usgs.gov).

Cheers,

Jim