A novel algorithmic framework for identifying changing streamflow regimes: Application to Canadian natural streams (1966–2010)

Abstract. Climate change significantly affects natural streamflow regime. To assess alterations in streamflow regime, typically few streamflow characteristics are considered and their significant variations in time and space are taken as a notion of change. Although, this approach is informative, intuitively appealing and widely-implemented, (1) it cannot see simultaneous changes in multiple streamflow characteristics; (2) it does not utilize all the available information contained in a streamflow hydrograph; and (3) it cannot describe how and to what extent one streamflow regime evolves into other regime types. To address these gaps, we conceptualize streamflow regimes as intersecting spectrums that are formed by multiple streamflow characteristics. Accordingly, we recognize that changes in streamflow regime should be diagnosed through gradual, yet continuous changes in an ensemble of streamflow characteristics. To incorporate these key considerations, we propose a fuzzy clustering-based approach to classify the natural streamflow into a finite set of intersecting regime types. Accordingly, by analyzing how the degrees of membership to regime types change, we quantify monotonic shifts between regime types in time and space. Our proposed algorithm eliminates the subjectivity in quantifying shift between flow regimes, and can extract valuable knowledge stored in the shape and variability of annual streamflow hydrographs. We apply this approach to the natural streamflow data, obtained from 106 Canadian gauges, during the period of 1966 to 2010. We show that natural streamflow in Canada can be categorized into six regime types, with clear physical and geographical distinctions. Analyses of trends in membership values during the study period show that alterations in natural streamflow regime are vibrant and can be different within and between major Canadian drainage basins. We show that gradual changes in natural streamflow regimes in Canada can be attributed to simultaneous changes in a large number of streamflow characteristics, some of which have been previously unknown or not well-attended. Our study introduces a generic algorithmic framework for identifying changing streamflow regime at regional and global scales, and provides a fresh look at streamflow alterations in Canada, which can be seen as another line of evidence for the complex and multifaceted impacts of climate change on streamflow regime, particularly in cold regions.