Zooplankton proportion estimates from non-uniform sample volumes

The relative abundance of organisms from different taxa provides information about ecosystem health and diversity. When the numbers of sampled organisms are modelled as Poisson counts, and the sample volumes are not uniform, variance for the proportion attributable to each taxon is difficult to compute. We present a method for computing approximate variances for this situation. The point estimates and their standard errors reduce to the standard multinomial maximum likelihood results when sample volumes are uniform. Further, given initial estimates of population densities for the taxa of interest, optimal sample volumes can be computed. The methods are illustrated for zooplankton counts from Andrus Lake, Michigan.