A Methodology for Clustering Lakes in Alberta on the basis of Water Quality Parameters

In this study, a methodology for clustering 18 lakes in Alberta, Canada using the data of 19 water quality parameters for a period of 11 years (1988-2002) is presented. The methods consist of (i) principal component analysis (PCA) to determine the dominant water quality parameters, (ii) cluster analysis techniques to develop the characteristics of the clusters, and (iii) pattern-match lakes to determine the appropriate cluster for each of the lakes. The PCA revealed that three principal components (PCs) were able to explain similar to 88% of the variability and the dominant water quality parameters were total dissolved solids, total phosphorus, and chlorophyll-a. We obtained five clusters for the period 1994-1997 by using the dominant parameters with water quality deteriorating as the cluster number increased from 1 to 5. Upon matching cluster patterns with the entire dataset, it was observed that some of the lakes belonged to the same cluster all the time (e. g., cluster 1 for lakes Elkwater, Gregg, and Jarvis; cluster 3 for Sturgeon; cluster 4 for Moonshine; and cluster 5 for Saskatoon), while others changed with time. This methodology could be applied in other regions of the world to identify the most suitable source waters and prioritize their management. It could be helpful to analyze the natural controlling processes, pollution types, impact of seasonal changes and overall quality of source waters. This methodology could be used for monitoring water bodies in a cost effective and efficient way by sampling only less number of dominant parameters instead of using a large set of parameters.