The roughness of the bed beneath ice is an important control on ice stream location and dynamics. Deglaciated terrain provides the opportunity to explore bed roughness in greater detail and over larger areas compared to glaciated terrain. This thesis examines three differ- ent aspects of palaeo-ice streams bed roughness. Firstly, this thesis explores methods used to measure bed roughness in glaciology. The choices made by researchers on transect orien- tation, window size, detrending and roughness methods have an impact on results. The Fast Fourier Transform analysis and Standard Deviation methods are both useful for calculating bed roughness in glaciology.

Secondly, this thesis directly compares the roughness of contemporary and palaeo-ice stream beds. The bed roughness of Minch Palaeo-Ice Stream (MPIS) is compared to the Institute and M ̀ˆoller Ice Streams (IMIS). The MPIS has a rough bed along major flow paths in the onshore onset zones. The results from the MPIS demonstrate that the presence of sediment does not necessarily correspond with fast flowing ice. The spacing of Radio Echo Sounding (RES) transects (10 x 30 km) used to measure bed roughness under contemporary- ice streams was too wide to capture bed roughness of MPIS glacial landforms.

Thirdly, this thesis investigates whether glacial landforms have unique bed roughness signatures. The results show that groups of glacial landforms have unique bed roughness signatures when anisotropy is taken into account. Bed roughness signatures of glacial land- forms have the potential to be compared with known and unknown areas of glacial landforms at the bed of contemporary-ice streams.

Future studies should acquire RES transects where a rough bed or glacial landforms are inferred. 250 m transect spacing would be desirable but 1 km transect spacing is likely to be more practical because it would allow orthogonal and parallel transects to be acquired.