www.phidot.org

Hello,
I am estimating the population abundance of the federally endangered Atlantic Sturgeon using long term capture-recapture data. The sampling area is a one kilometer stretch of river that is sampled 30-40 times a year over a five-month period. These fish are very mobile and move in and out of the sampling area frequently. For estimates I am using the open population model POPAN and my parameter of interest is the superpopulation size each year. I know that one of the assumptions of this model is that emigration is permanent which is violated in our analysis because my fish move in and out of the sampling area so much. My question is how strong of an effect is this violation likely to have on my population estimates. Would there be a strong bias and if so is there anything I can do to avoid it?

I cannot use a robust design for several reasons. I cannot assume closure because my sampling area is so small and my sampling period is so long. Also, there are years in the study that have a lot of capture data and many years with little to no capture data. Thank you for the help.

The model might be estimating what you want it to. If your study reach is sampling the fish with probability a*p, where a is the probability a sturgeon is in your reach when sampling occurs and p is the probability you catch it given it is there, and your interest is in the population of sturgeon that could use that reach at that time, then your estimate of detection reflects that product and therefore your estimate of N(t) reflects the population of interest. So, completely random emigration should produce an unbiased estimate of those that use the reach, which is greater than the number present in the reach when sampling occurs. If availability is something other than completely random, like Markovian, then some bias would be induced.

Bill Kendall,
Thank you so much for responding so quickly. Your help is greatly appreciated.