We use an ice sheet model (GLIMMER) with an erosion component to examine the evolution of landscapes under ice sheets over long time scales in hypothetical situations. GLIMMER is a fully coupled thermomechanical 3-dimensional ice sheet model with a sliding component. The model thus has the ability to predict ice thickness, the distribution of areas at pressure melting point and basal velocities. The erosion rate is assumed to be a simple linear function of the basal velocity and therefore erosion only occurs where the bed is not frozen and basal sliding is possible. In addition, sliding velocities are assumed to be related to basal water production, thus giving a mechanism for smooth temporal and spatial transitions between eroding and non-eroding portions of the bed. The eroding glacial landscape system is modelled for long-term persistent glaciation assuming a variety of initial landscapes. Results indicate that without the inclusion of basal water production in the calculation of sliding velocities, any thermal feedbacks are exaggerated because of steep horizontal basal temperature gradients. Additionally, bed morphology exerts a greater influence under low basal slip conditions because ice cannot respond readily to thermal instabilities. We suggest that erosion and valley overdeepening may aid stabilisation of the thermal regime of an ice sheet and that it is possible to generate glacial signals in previously fluvial systems after only 100 kyr of glaciation at conservative erosion rates.