Open Collections

Abstract

Anthropogenic climate change is accelerating glacier mass loss, with projections indicating that 70–90% of current glacier volume could disappear by 2100 in many mountain regions, including western Canada. However, these projections carry large uncertainties at the scale of individual glaciers and watersheds, where freshwater impacts are critical. A key uncertainty stems from the widespread use of empirically calibrated temperature-index melt models, which rely solely on temperature and lack process-based understanding and transferability. This thesis investigates the potential of physics-based surface energy balance (SEB) models as alternatives, addressing three key components for regional-scale glacier modeling in western Canada: (1) the effectiveness of dynamical downscaling for climate input to the SEB model, (2) the ability of a glacier evolution model to reproduce observed glacier changes since 1980, and (3) the impact of different downscaling methods on uncertainties in future climate projections derived from a global climate model (GCM). First, SEB input data were evaluated using ERA5 reanalysis and dynamically downscaled Weather Research and Forecasting (WRF) model outputs at multiple glaciers. Both datasets aligned strongly with multi-season meteorological and glaciological measurements, supporting the use of ERA5 for historical melt reconstructions and WRF for downscaling GCM outputs for future projections. Second, an SEB model, enhanced with neural network-based glacier albedo simulations, was integrated into a glacier evolution model driven by ERA5 for 1980–2022. Simulated glacier thinning rates across the region closely matched in-situ observations and remote sensing estimates, revealing that declining summer albedo and increasing longwave radiation and turbulent heat fluxes were primary contributors to mass loss. Finally, a hybrid downscaling approach integrating WRF with statistical downscaling, ranging from simple bias correction to machine learning, was used to downscale climate fields for 2025–2100 and project glacier mass changes. Uncertainty partitioning quantified contributions from downscaling method, training dataset (i.e., whether future projections were included), and emission scenarios. By integrating dynamical downscaling, physics-based SEB models, and neural networks, this thesis advances large-scale glacier mass balance modeling, offering refined insights into glacier-climate interactions and the impacts of climate change on glacier-fed water resources. These findings inform mitigation and adaptation strategies in glacierized regions.