A multi-variate methodology for analyzing pre-existing lake water quality data

Environmental agencies are given the task of monitoring water quality in rivers, lakes, and other bodies of water, for the purpose of comparing the results with regulatory standards. Monitoring follows requirements set by regulations, and data are collected in a systematic way for the intended purpose. Monitoring enables agencies to determine whether water bodies are polluted. Much effort is spent per monitoring event, resulting in hundreds of data points typically used solely for comparison with regulatory standards and then stored for little further use. This paper devises a data analysis methodology that can make use of the pre-existing datasets to extract more useful information on water quality trends, without new sample collection and analysis. In this paper, measured lake water quality data are subjected to statistical analyses including Principal Component Analysis (PCA) to deduce changes in water quality spatially and temporally over several years. It was found that the lake as a whole changed temporally by season, rather than spatially. Storm events caused the greatest shifts in water quality, though the shifts were fairly consistent across sampling stations. This methodology can be applied to similar datasets, especially with the recent emphasis by the U.S. EPA on protection of lakes as water sources. Water quality managers using these techniques may be able to lower their monitoring costs by eliminating redundant water quality parameters found in this analysis.