Overdeepenings under the Greenland Ice Sheet and their control on ice dynamics.

Ice loss from the Greenland Ice Sheet (GrIS) is currently the most significant single global contributor to barystatic sea level rise. The discharge of ice directly into the ocean from marine terminating glaciers is the cause of approximately 40% of this sea level rise. Understanding the processes that control how ice slides over the bed is fundamental to improving predictions of future GrIS mass loss. This thesis aims to investigate the possible significance of bed topography for sliding rates and mechanisms by: first, improving understanding of the location and morphology of overdeepenings under the GrIS, which are key landscape features that dictate the locations of riegels and adverse slopes; and second, investigating the dynamics of ice flow through the major outlet glaciers that dominate discharge of ice to the ocean, which involves flow through complexly overdeepened glacially eroded troughs. The thesis begins by compiling the first high resolution, GrIS-wide dataset of overdeepenings for Greenland, and analysing velocity patterns of ice flowing through these overdeepenings: overdeepenings are found to be ubiquitous features under the GrIS, and a small but statistically significant increase in ice velocity is observed on the adverse slopes of overdeepenings. Next, the control exerted on ice dynamics by overdeepenings is explored in detail for two separate glaciers with overdeepened beds near the terminus (Helheim and Upernavik Isstrøm II, where high-quality bed topography was available) were observed to exert a ‘marine-isolating’ effect on the flow of inland ice, with ice dynamics dominated by marine processes downstream of the riegel and by melt processes inland of the riegel. Further, intriguing patterns of seasonal velocity variation were observed within the overdeepening under high melt conditions at Upernavik Isstrøm II that supported the possibility that adverse slopes of overdeepenings suppress the development of efficient channelised subglacial drainage, which is a key mediator of rates of sliding. A machine learning based automated method utilising k-means clustering for identifying distinct seasonal velocity typologies characteristic of those at Upernavik Isstrøm II was then developed and used to classify all areas of fast-flowing ice of the GrIS. This enabled an ice sheet-wide statistical analysis of whether such the patterns, presumed caused by overdeepenings, are widespread. Limited evidence was found to support systematic and consistent differences in seasonal velocity typology as a result of the presence of overdeepenings; however, examples of modulation of ice dynamics by overdeepenings in single and adjacent groups of catchments were easily identified and these examples were widespread across the GrIS, suggesting more work on this topic is needed. Finally, a wider analysis of seasonal ice dynamics using the typology dataset for the whole of the GrIS makes three key findings: 1) a statistically significant relationship between melt availability and seasonal velocity typology is demonstrated for the first time; 2) seasonal velocity typologies driven by marine process are found to be relatively rare suggesting that marine processes are not a dominant control on seasonal ice dynamics for the GrIS; 3) the spatial patterns of seasonal velocity typologies generated by automated classification that were interpreted to be representative of channelised subglacial hydrology aligned very closely with published field and modelling data estimating the inland channelisation limit. The study therefore shows a possible importance of overdeepenings in controlling ice dynamics by modulating subglacial hydrology and demonstrates the value of investigating flow typology beyond the immediate vicinity of outlet glacier termini.