SPATIAL CHARACTERIZATION OF WATER-QUALITY IN FLORIDA BAY AND WHITEWATER BAY BY MULTIVARIATE ANALYSES - ZONES OF SIMILAR INFLUENCE

We apply an objective statistical analysis to a 6-yr, multiparameter d ataset in an effort to describe the spatial dependence and inherent va riation of water quality patterns in the Florida Bay-Whitewater Bay ar ea. Principal component analysis of 16 water quality parameters collec ted monthly over a 6-yr period resulted in five principal components ( PC) that explained 71.8% of the variance of the original variables. Th e ''organic'' component (PCI) was composed of TN, TON, APA, and TOC; t he ''inorganic N'' component (PCII) contained NO2-, NO3-, and NH4+, th e ''phytoplankton'' component (PCIII) was made up of turbidity, TP, an d Chl a; DO and temperature were inversely related (PCIV); and salinit y was the only parameter included in PCV. A cluster analysis of mean a nd SD of PC scores resulted in the spatial aggregation of 50 fixed mon itoring stations in Florida Bay and Whitewater Bay into six zones of s imilar influence (ZSI) defined as Eastern Florida Bay. Core Florida Ba y, Western Florida Bay, Coot Bay, the Inner Mangrove Fringe, and the O uter Mangrove Fringe. Marked differences in physical, chemical, and bi ological characteristics among ZSI were illustrated by this technique. Comparison of medians and variability of parameter values among ZSI a llowed large-scale generalizations as to underlying differences in wat er quality in these regions. For example, Eastern Florida Bay had lowe r salinity, TON, TOC, TP, and Chl a than the Core Bay as a function of differences in freshwater inputs and water residence time. Comparison of medians and variability within ZSI resulted in new hypotheses as t o the processes generating these internal patterns. For example, the C ore Bay had very high TON, TOC, and NH4+ concentrations but very low N O3-, leading us to postulate the inhibition of nitrification via CO pr oduction by TOC photolysis. We believe that this simple, objective app roach to spatial analysis of fixed-station monitoring datasets will ai d scientists and managers in the interpretation of factors underlying the observed parameter distribution patterns. We also expect that this approach will be useful in focussing attention on specific spatial ar eas of concern and in generating new ideas for hypothesis testing.