1. Compute the distribution of mean number of sightings from a sample of size , with mean number per sighter , and then use that to determine the required sample size for the sort of mean number of sightings you are expecting (I suspect, but would have to do some research to confirm this, that with observers the distribution of the total number of sightings is Poisson with mean , where is the mean number of sightings from a single observer)
2. Generate approxinate distributions for the mean using a bootstrap technique.
3. Using the Chebyshev or Vysochanski´-Petunin inequalities to give conservative estimates of the required sample size.