Multivariate analysis of experimental marine ecosystems

Abstract

Twelve replicate 150-l laboratory microcosms were developed using whole water samples and natural benthic communities from Narragansett Bay, Rhode Island (USA). The microcosms were scaled to the bay in terms of salinity, temperature, light input, volume pelagic community to area of benthic community, density of macrofauna, turbulent mixing, and flushing time. The microcosms were self maintaining during a six month study period with over 35 species of phytoplankton and 30 species of macro-invertebrates. Some 25 species of meroplankton entered the microcosms and sucessfully colonized the benthic communities. Zooplankton were present in all life stages. After an initial study of replication among the 12 tanks, the microcosms were perturbed with 3 levels of treated urban sewage for a three month period. Three microcosms were maintained at each level with 3 tanks remaining as controls. At the end of 3 months, the sewage input was terminated and the response of the systems was followed for an additional 2 months. Both time series data and multivariate statistical analysis of over 10 different parameters indicated that the replication of the microcosms was adequate to show the effects of experimental treatments. Control microcosms were generally within the range of variation expected in Narragansett Bay. Moreover, the results suggested that the microcosms responded to the gradient of sewage input in a manner similar to that of the bay. During the two month period after the sewage was discontinued, all of the microcosms became increasingly similar, though the tanks that had been subjected to higher levels of sewage remained distinct. It was apparent throughout the study that comparisons of microcosms and natural systems must account for the large variation characteristic of each. For this, and other reasons, multivariate statistical techniques appear to provide a powerful tool for experimental ecosystem analysis.