Statistical methods for detecting ichthyoplankton density patterns that influence entrainment mortality

Samples of drifting American shad eggs were collected with 0.5 mm mesh nets at two transects in the Savannah River near industrial water intakes. At each transect the river was divided horizontally and vertically into four sectors that were sampled at two hour intervals for 24 hour periods to assess horizontal and vertical patterns of egg distribution that could affect vulnerability to entrainment. Egg densities varied by an average factor of 20 over a 24 hour cycle and were consistently highest between 2300 and 0300 and lowest between 1300 and 2000. We used two methods to separate spatial from temporal variance, analysis of covariance (ANCOVA) with a polynomial model for time of day as the covariate and a detrending technique in which diel variation was subtracted from each data point. These procedures resulted in substantially more sensitive tests of sector differences as indicated by minimum detectable ratios of sector means that were only half as great as for unadjusted data. Spatial patterns differed between transects but were generally consistent within transects where three to four-fold differences among sector means were common. The actual risk of entrainment was approximately 35–50% lower than if the shad eggs were uniformly distributed, and the risk of entrainment was lower at one intake than at the other. Our results indicate the importance of site specific assessments of ichthyoplankton distribution near existing or proposed water intakes using statistical designs that permit sensitive resolution of spatial patterns.