Glacier Snowline Altitude Mapping in the Canadian High Arctic Using Feature-Oriented Principle Component Analysis

Abstract:

The snowline on temperate glaciers are typically easy to identify in the mid-latitudes from optical imagery, by examining differences in albedo between snow, firn and bare ice and applying a threshold value of 0.6 for the Normalized Difference Snow Index (NDSI). Many studies of glacier mass balance use the snowline at the end of the ablation season as an indicator for the equilibrium line altitude (ELA) and as a proxy of glacier health. The position of equilibrium line on the glacier is controlled by the climatic conditions and net mass budget of the glacier in that hydrological year. Therefore, the variations in ELA can be used in remote regions where no available direct/field measurements exist as a proxy for regional climate change. Since most glaciers in the Canadian High Arctic are snow/firn-covered year-round, this widely used NDSI thresholding method to estimate snowline as a substitute for ELA cannot be used. In the Canadian Arctic, the ELA is instead estimated by evaluating how the various spectral signatures of snow and firn are distributed over the glacier surface.

In this study, we used feature-oriented principal component analysis (FPCA), a multivariant statistical technique that can be applied to multispectral satellite images to extract desired features based on their spectral signatures. FPCA reduces data dimensionality by implementing a linear orthogonal coordinate transformation, which results in an uncorrelated set of principal components. Relevant features such as fresh snowcover, old snow, and firn can be extracted from these principal components. The snowline is delineated from the principle component representing the snow cover in the summer season. We applied FPCA to a time series of Landsat imagery from 2000 and 2020. To validate the relationship between satellite-derived snowline and the ELA, we focused on White Glacier, a benchmark glacier located on Axel Heiberg Island, Nunavut in the Canadian Arctic for which direct field measurements of mass balance and ELA are available since 1960. The application of FPCA we present here is a first step towards extending the estimation of ELA for all glaciers in the Canadian Arctic and to understanding their responses to ongoing rapid regional climate change.