Quantification of Error and Potential Avenues for Improvement of Modeled Ice-Thickness Distributions for Individual Glaciers

Abstract:

Models of glacier dynamics are becoming increasingly complex and widely applied, requiring data that are often difficult to acquire or produce. With the growing availability of global datasets, such as ITS_LIVE for surface velocity fields or multi-model estimates of ice-thickness fields, usage of these datasets for individual glaciers is becoming more common. Independent site-specific data allow us to assess the mismatch between measurements and derived/modeled products to evaluate the strengths and limitations of using the latter for studies of individual glaciers. Here we evaluate the discrepancy between modeled ice-thickness distributions and ice-penetrating radar (IPR) data for five glaciers in western Canada with highly variable IPR coverage. We attempt to better the modeled ice-thickness estimates using different statistical methods. For some of the glaciers, the discrepancy between modeled and measured values is distinctly non-linear as a function of ice thickness. Simple linear methods (e.g. scaling), however, have the potential to improve the modeled values depending on the IPR data density and coverage for a given glacier. Attempts to improve the modeled values using more sophisticated methods can introduce additional error in the absence of sufficient IPR data. Our results highlight some of the conditions under which modeled ice thicknesses can be used or improved for site-specific studies, and when different approaches may be advised.