Glaciers are often described as rivers of ice, yet the comparison is misleading in one important respect. A river responds to rainfall within hours, whereas a glacier may take decades to register a change in climate. This delay, which researchers call the response time, is what makes glaciers such valuable long-term recorders of environmental change. A team led by the Norwegian glaciologist Ingrid Halvorsen has argued that, for mid-latitude ice bodies of moderate size, the full adjustment to a shift in winter snowfall can require between forty and eighty years.

Movement begins at the surface, where fresh snow accumulates and is gradually compressed into a granular material known as firn. Under the weight of further snowfall, the trapped air is expelled and the firn becomes dense glacier ice. The timescale for this transformation varies with temperature: in the high Andes it may take several centuries, while in the wet maritime climate of southern Alaska it can be completed in less than fifteen years. Once formed, the ice flows downslope under its own weight through two distinct mechanisms. The first, known as internal deformation, involves the slow rearrangement of ice crystals within the body of the glacier. The second, basal sliding, occurs when a thin film of meltwater lubricates the contact between ice and bedrock.

The relative importance of these two mechanisms explains why some glaciers advance steadily while others are prone to dramatic surges. Cold-based glaciers, whose beds remain frozen to the underlying rock, move almost entirely by internal deformation and rarely exceed a few metres per year. Warm-based glaciers, by contrast, can slide at speeds of several kilometres per year during a surge event. The Bruarjokull glacier in Iceland, for example, advanced almost ten kilometres between the autumn of 1963 and the spring of 1964, then entered a quiescent phase that has lasted for more than fifty years. Halvorsen's group has proposed that such surges are triggered when meltwater, prevented from draining by an insulating layer of softer sediment, accumulates at the base of the ice until it reaches a critical pressure.

The internal structure of a moving glacier provides further evidence of these processes. Crevasses open where the ice is stretched too quickly to deform plastically, and their orientation reveals the local direction of flow. Where the ice passes over a rock step it may break into a chaotic jumble known as an icefall, in which blocks the size of houses rotate and collide as they descend. Below the icefall, the broken surface heals and the ice resumes a more orderly motion. At the lower margin, or terminus, the rate of melting eventually balances the rate of supply, and the glacier appears stationary even while its interior continues to flow.

Measuring that flow is a central task of modern glaciology. Early surveyors drove wooden stakes into the surface and returned each summer to record their displacement. The method was laborious but surprisingly accurate: figures collected on the Mer de Glace in France between 1895 and 1914 compare well with satellite measurements taken a century later. Today, glaciologists rely on radar interferometry from polar-orbiting spacecraft, which can detect movements of a few millimetres a day across entire ice sheets. The new data have revealed that several outlet glaciers in Greenland accelerated sharply between 2000 and 2010 before stabilising at higher velocities, a pattern that had been predicted by theoretical models but never directly observed.

The causes of glacier retreat are more subtle than is often assumed. An increase in summer air temperature raises the melt rate at the surface, but the effect can be partially offset by greater winter snowfall in a warmer, wetter atmosphere. In many coastal regions, the dominant control on glacier length is not air temperature at all but the temperature of the seawater in contact with the floating tongue. Warm currents that penetrate beneath the ice can thin it from below, allowing the grounding line to retreat into deeper water where further thinning occurs. Halvorsen has warned that this process, once begun, is difficult to reverse, because the bed of many polar fjords slopes inland and the ice becomes increasingly unstable as it withdraws.

For communities living downstream of glaciated valleys, the behaviour of the ice has practical as well as scientific importance. Sudden releases of water from glacier-dammed lakes have destroyed settlements in the Himalayas and the Peruvian Andes. In Iceland, the routine monitoring of subglacial geothermal activity has reduced but not eliminated the risk of the meltwater floods known locally as jokulhlaups. As the pace of retreat accelerates, glaciologists face the twin task of refining their theoretical understanding and translating it into warnings that ordinary people can act upon.