WATERSHED CLASSIFICATION BY DISCRIMINANT ANALYSES OF LAKEWATER-CHEMISTRY AND TERRESTRIAL CHARACTERISTICS

We used canonical discriminant analyses to improve an existing classif ication scheme by identifying those lakewater-chemistry and terrestria l variables that can jointly differentiate among a set of lakes in the Adirondack region of the northeastern USA. We then used predictive di scriminant analysis to examine the ability of the models constructed i n predicting class membership for lakes with unknown hydrogeological s ettings into known classes. We used two sets of data collected during summers of 1984-1987 and 1994. For both periods, two main underlying d imensions were identified reflecting watershed hydrogeology and lake d issolved organic carbon. The hydrogeological construct was defined by lakewater concentrations of Ca, Mg, Na, and Si. This construct provide s a continuous index from several previously assigned class variables and can help in data reduction and elimination of multicolinearity in further multivariate studies of ecosystem processes in the study lakes . The importance of Ca in defining the hydrogeological construct decre ased from 1984-1987 to 1994, due to differential decrease in Ca concen tration in the lake classes. The periodic decrease in Ca concentration of many drainage lakes suggests that they continue to lose buffering capacity to acidic deposition and that the current decreased emission and deposition of acidic S compounds may not be sufficient for chemica l recovery of these lakes. The construct reflecting lakewater concentr ation of dissolved organic carbon was defined by lakewater Fe concentr ation and the lake-volume-weighted watershed areas covered by wetlands and forest. Our methodology and results should provide useful informa tion for studies conducted in the Adirondack watersheds and may be app licable to other regions with properties and environmental issues comp arable to those in the Adirondack region.