Date of Award

1994

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

First Advisor

J. W. Day, Jr.

Second Advisor

J. P. Geaghan

Abstract

The utility of species composition and abundance to detect environmental impacts in environmental baseline studies, was examined. It was shown that a high proportion of studies will falsely report significant changes (type I statistical errors) in species composition from putative environmental impacts using a chi square metric due to the way species abundance is sampled. A high rate of false complacency (type II statistical errors) results from before-after-control-impact (BACI) studies of individual species under the same, common, sampling schemes. Spatial aggregation and abundance parameters were measured for a nekton data set and provided a realistic range of values for the subsequent analysis. Correspondence analysis was used as a data exploratory technique to investigate the effect of spatial aggregation, abundance, number of samples and number of species on the sample variability. Aggregation alone and in combination with the other factors was observed to increase sample variability. Type I statistical error rates were shown to be significantly and dramatically affected by realistic levels of spatial aggregation as well as by the number of species and the mean abundance in the presence of ecologically realistic levels of aggregation. Four types of natural environmental disturbances: hurricanes, freezes, low dissolved oxygen and low offshore salinities were investigated as models of pulse perturbations on nekton species composition. Species composition was found to change simultaneously at stations close together more than at stations far apart, but none of the natural disturbances was associated with changes in species composition that occurred at two stations simultaneously. An estimator for the non-centrality parameter of the F distribution was developed for mixed statistical models. This estimator can be used to determine the statistical power to detect environmental impacts in studies with controls that have data from before and after an environmental impact. Parameter estimates from a nekton database indicated that, for the given level of replication, a simple BACI statistical model would not provide sufficient power to detect a doubling or halving of the geometric mean for most of the nekton species.

Pages

134

Available for download on Wednesday, December 01, 9999

Share

COinS