Thomas Barnes

Mapping and interpreting subglacial landforms is important for understanding the processes in the interior of Quaternary glaciations and present-day ice sheets such as the Greenlandic and Antarctic ice sheets. While most studies have focused on marginal landforms such as terminal moraines, which have clearly defined morphology and little debate as to their formational processes, there have only been few previous studies into the subglacial environment. While these works provide a strong basis of knowledge surrounding bedforms subject to detailed study, at a large scale they were limited by historic data quality, which has since been improved. This means that today, there are limitations on how accurately we can model and understand Quaternary glaciations, such as the last glacial with the Fennoscandian ice sheet. Recent attempts to fill this research gap have primarily focused on Finland and Sweden, with a lesser focus on Norway. Thus, this thesis attempts to use former research and newly available high-quality and high-resolution data related to the interior of the Fennoscandian ice sheet to map and interpret subglacial landforms in Norway, with a focus on several study areas.

Recent studies which have focused on murtoos, hummocks and ribbed moraines in Finland and Sweden have established a basis for this research. This thesis, specifically, focuses on examining ribbed moraines in greater detail, with a secondary focus on related bedforms and subglacial hydrology. Further, the research area of this thesis begins with inventorying ribbed moraines, but is aimed towards working with numerical modelling of geomorphology, rather than reconstructive mapping and cataloguing as an end goal. Ribbed moraines are a subglacial landform which occur as sinuous, elongate ridges which form transverse or oblique to former glacial flow orientation, and are typically found near transitional zones in the basal thermal regime of ice sheets. Despite the formational processes still being a topic of debate, recent works discuss concepts of a “landform continuum” and “equimorphology”, which provide a basis for considering ribbed moraines as a diverse indicator of subglacial conditions, from their formation to their preservation. It is, therefore, possible that ribbed moraines may be valuable in the geomorphological reconstruction of interior regions of former ice sheets.

Updated, high-resolution spatial data provides a solid foundation for this thesis’ works, with regularly updated high-resolution elevation models and detailed superficial geological maps for the entire mainland of Norway. Further, data from the numerical modelling of the Fennoscandian ice sheet, combined with this high-resolution data enable this thesis to map ribbed moraines at a 10-metre scale across all of Norway, and then to infer glacial, hydrological and morphological conditions of ribbed moraines. Furthermore, using the inferred conditions, I produce a detailed and geomorphologically-grounded reconstruction of local ice dynamics in the interior of the Fennoscandian ice sheet. Furthermore, this thesis develops understanding into the distribution of ribbed moraines, providing observations which are then used for the detailed reconstruction of two study regions near the ice divide of the former Fennoscandian ice sheet. For this, I used the time period between the last glacial Maximum to the complete deglaciation of the former ice sheet (20 – 25 ka BP – 9 ka BP). This was done by using a combination of K-means clustering (unsupervised machine learning algorithm) on elevation model data, spatial statistical analysis of the results, and geomorphological observations. With this, I address the overall goals by first adding new understanding of how ribbed moraines are formed relating to their related glacial dynamics, hydrology and morphology; and then using this to provide insight into the processes and conditions in the interior of the Fennoscandian ice sheet.

I conducted this research in several steps, producing three articles. The first addresses the problem of mapping and cataloguing ribbed moraines, the second addresses their formational conditions via spatial statistical analysis, and the third covers a detailed reconstruction of the ice dynamics and hydrology of the interior of the Fennoscandian ice sheet. Furthermore, the third paper also ties much of this research together, by forming connections between research outputs as a means of producing this reconstruction.

Based on the results from paper 1, I find that machine learning methods are of great value to the automated mapping of glacial bedforms. While the focus was primarily on ribbed moraine, I found this methodology to be transferrable as long as characteristic morphologies were identifiable for the target landform. Furthermore, scalability was possible within this methodology, allowing for the machine learning methodology to be scaled up to larger areas and useable across a diverse set of general landscape morphologies. Paper 2 offers insights that support previous hypotheses of ribbed moraine formational conditions and shows that they can be formed and preserved under a general set of conditions. Specifically, these refer to where ice is relatively slow flowing, there is presence of water under high pressures but with low hydraulic gradient, and there is a general low slope to the landscape. These findings, whilst not groundbreaking in and of themselves, are useful in that they support previous research and understanding of how ribbed moraines form. As such, these findings supported and enabled the detailed reconstruction in paper 3.

In combining this thesis’ body of work, it is clear that the reconstruction as presented and interpreted here offers strong implications towards the conditions of the interior of former ice sheets. Further, my findings open up new directions and opportunities for future research, First, in addition to the hydrological relationships in this work, ribbed moraines display a spatial relationship with landforms produced by subglacial drainage such as eskers, drainage tracts, canyons and murtoos. Further, the quantity of water required for formation appears to be variable, with ribbed moraines being found in areas with differing footprints of hydrological activity in the geomorphological record, thus suggesting an equimorphological explanation for their formation. Furthermore, the spatial relationship that ribbed moraines display with glacially produced bedforms shows that ribbed moraines are transient, and develop continuously until they are preserved by exposure or cold-based ice. In connection with this, this thesis’ findings suggest that the hydrological network under the Fennoscandian ice sheet may have been more destructive and dynamic than previously assumed, when considering the concept of “hydraulic drainage”, as several possible subglacial drainage canyons are investigated through each study area. Finally, in coupling all the data used in Paper 3, there are implications of subglacial lakes being present in the interior of the former Fennoscandian Ice Sheet, indicating areas requiring further research. In all, the data and theoretical understanding offered up through my work in this thesis shows the potential of machine learning methods, high-resolution spatial data, and geomorphological reconstructions in developing new insights into former ice sheets.