RIVERINE SYSTEMS AS CONDUITS OF MATERIAL FLOW: THE USE OF R-MODE FACTOR ANALYSIS AND CLUSTER ANALYSIS TO EXAMINE THE INFLUENCE OF LAND USE ON RIVER CHEMISTRY

Examining biogeochemical processes that influence the chemistry of riverine systems is an important step to understanding terrestrial inputs to lacustrine and marine systems. While recent emphasis on watershed scale investigations has illuminated the need for reliable methods of sample and data collection (e.g., a multi-element, base-flow, synoptic sampling strategy using clean techniques), systematic analysis of the data is also of concern. This is in part because the resulting multivariable water-quality datasets are complex and reveal little about the dominant biogeochemical processes without extensive statistical manipulation. Similarly, statistical regression of these datasets does not adequately resolve the relationships between terrestrial variables such as land use and river chemistry.

Here we examine the use of exploratory factor analysis as a method of data reduction and for investigating and quantifying the relationships between land use variables and river chemistry. The combination of R-mode factor analysis and cluster analysis allows us to reduce the data set into a more manageable number of factors, facilitating investigations into the processes controlling the effects of land use distribution on river chemistry. R-mode factor analysis, which examines the relationships among the variables, is performed with and without land use as a variable. Cluster analysis is used to identify groups (or clusters) of sites that exhibit similar biogeochemical behavior. The nature of these groups is examined through plots of the log ratio of the cluster median to the overall population median for each chemical parameter. Through this analysis, the chemical signatures of riverine samples can be related to land use. We have found that this method reveals similar relationships between land use and water quality in two large watersheds in Michigan (Muskegon River, Grand Traverse Bay). GIS technology is used to develop sourcesheds for each sampling site and land use/land cover percentages are calculated. Thus, land use may have a consistent terrestrial or biogeochemical fingerprint.