A necessary condition for a glacier to surge by thermal instability is that the glacier be cold with basal ice at or near the melting point. Deep temperature measurements show that two small surge-type glaciers in the Yukon Territory meet this requirement, but shallow measurements in three other surge-type glaciers suggest a temperate regime. If the latter observations are accepted, not all surges are thermally controlled: if a single mechanism accounts for all surges it cannot be thermal instability.
In this paper it is argued that thermal instability remains a viable mechanism for explaining surges of many cold glaciers, and that regardless of the underlying mechanism, thermal processes must at least have a major influence on the surge behavior of cold glaciers.
Two numerical modelling experiments are described. The first involves a one-dimensional model which shows that thermal control can account for the remarkably constant surge cycle found in some glaciers. The second, a two-dimensional model of the time-dependent temperature structure of a surge-type glacier, shows that the relative amounts of temperate and cold basal ice can change considerably as the surge cycle progresses. This variation alone may be sufficient to explain surges, but even if this is not the case, thermal processes must affect the timing of surges in many cold glaciers.
A compelling feature of the thermal instability mechanism is that it offers an explanation of the factors controlling the non-random geographical distribution of surge-type glaciers. For a glacier to have a cold surface and near-temperate bed, the ice thickness, temperature, and geothermal flux must be fortuitously related.