Detection and Characterization of Glacier Melt Signatures in Western Canadian Streamflow

Glacier-fed rivers are reliable sources of freshwater used for municipal water supply, agricultural irrigation, and hydropower generation; critically, they are characterized by relatively high total summer streamflow and low summer streamflow variability as compared to non-glacier-fed rivers. Studies of glacier contribution to streamflow tend to be either highly detailed for a single glacier or basin (with melt models that rely on dense observational data) or are generalized at a larger scale (with melt models that rely on uncertain parameterizations). Regional studies that explore complex spatio-temoporal patterns of streamflow are less common, and rarely place glacier fed rivers in the context of regional flow regimes. Regional studies in Canada tend to focus on British Columbia over Alberta, as Alberta's glaciers are comparatively fewer, smaller, and concentrated near the British Columbia border. We take a data analysis approach to determine how glaciers in Alberta influence streamflow regime, with the goal of identifying regions of water vulnerability. We use long term (>40 years) daily streamflow measurements from the Canadian HYDAT database to detect features that are attributable to glacier melt, using self-organizing maps (SOMs) and principal component analysis (PCA) to extract patterns within the large spatio-temporal dataset. We use SOMs, an unsupervised machine learning algorithm, to detect characteristic features in average seasonal hydrographs, and a coupled PCA and SOM method on daily summer streamflow to characterize a river's ability to sustain summer streamflow. This analysis highlights Alberta's dependence on rivers that demonstrate a strong glacial signal, and we provide a metric that can be used to track the relative importance of glacier melt in a particular basin without having to model the glacier melt itself.