Application of multivariate statistical analysis methods to the dam hydrologic impact studies

Introduced in the early 1990s in aquatic ecology as a new paradigm for restoration of rivers modified by anthropic activities, the “natural flow regime” concept has allowed the formulation of new problems in dam hydrologic impact studies. Within the context of this study, we posed two problems: (1) selection of the hydrologic variables most modified by dams and (2) identification of the factors that influence the extent of the hydrologic changes. Resolving these two problems necessitated the application of multivariate statistical analysis methods, in this instance, principal component analysis and canonical correlation analysis. These two methods thus are being applied for the first time to analysis of the hydrologic impacts of dams. The first method allowed selection of the hydrologic variables most modified downstream from dams in Québec. These are the variables best correlated to the first two principal components. As for the factors influencing the magnitude of these changes, canonical correlation analysis exposed the influence of the type of regulated regime (dam management mode), and to a lesser degree, the watershed size.