Regionalization of low flows based on Canonical Correlation Analysis

Regional analysis of low-flow statistics is a critical step in solving water resources management problems related to the requirements of the Water Framework Directive. Important element in this analysis is the determination of homogeneous sub-regions based on physiographic characteristics of the corresponding basin. The purpose of this paper is to investigate the use of the canonical correlation method for partitioning the set of drainage basins of a region into a number of homogeneous sub-regions and determining the relations between the physiographic and low-flow statistics of the basins of each sub-region. The method is also proposed to be used for classifying an ungauged basin in a sub-region of gauged basins.

► We use canonical correlation and multiple regression for low-flow regionalization.
► We classify a set of drainage basins of a region into homogeneous sub-regions.
► We determine the relations between the physiographic and low-flow statistics.
► We predict low-flow statistics of ungauged basins after classification.
► Regional estimates of low-flow statistics were better than the sub-region-based ones.