How Science Works | Why Land Rises After Heavy Ice Disappears

A column of grounded ice exerts a load because it has weight. A first estimate of the pressure added by a uniform thickness H is p = ρ_i gH, where ρ_i is ice density and g is gravitational acceleration.

For an original example, assign an average ice density of 900 kilograms per cubic metre and a thickness of 2000 metres. With g = 9.81 metres per second squared, the added pressure is 17.658 million pascals, or about 17.7 megapascals.

This estimate ignores detailed geometry, basal water pressure and stress transmission within the ice. It is a scale calculation showing why a continent-sized load cannot be treated as mechanically negligible.

The lithosphere, Earth’s relatively stiff outer shell, supports and spreads part of the load through bending. Below it, the mantle responds over a range of timescales. The resulting deformation is distributed over a region rather than confined to a perfectly vertical column beneath each patch of ice.

A useful cross-section shows three stages: an unloaded reference surface, a depressed region beneath growing ice and a recovering region after ice loss. Include the surrounding land, because its motion can have the opposite sign from the former ice centre.

The source of the unloading also needs precision. Melting, calving and changing ice flow can alter grounded ice thickness in different places. The Cryosphere Science guide explains those ice-system processes. This article follows the solid Earth’s response to their changing load.

Contents · Next chapter

An elastic material changes shape under a force and recovers when the force is removed, within its elastic range. The Earth therefore has a relatively rapid deformation response when its surface load changes. “Rapid” here means short compared with the long mantle-adjustment timescale, not that a whole continent moves instantaneously without wave propagation.

Over much longer periods, mantle rock can accumulate deformation through processes within and between mineral grains. It remains predominantly solid while behaving as a highly viscous material on geological timescales. A liquid-water analogy can illustrate redistribution, but it should not be mistaken for the mantle’s actual physical state.

A viscoelastic model combines those behaviours. An elastic spring and a viscous element are a useful conceptual representation, although a real planet has layered structure, gravity, curved geometry and many interacting deformation modes.

NASA’s Ice-sheet and Sea-level System Model documentation describes one simplified GIA formulation using an elastic lithosphere over a viscoelastic mantle. It explicitly includes the history of surface loading in the displacement calculation. That history dependence is the key difference between a static weight calculation and a prediction of motion through time. Read the model’s physical basis.

The material’s viscosity influences how readily slow deformation occurs. Temperature, composition, water content and deformation mechanism can all matter. It is therefore unsafe to assign one universal rebound speed to every formerly glaciated region.

Nor is a material relaxation timescale automatically the same as the observed decay time of a regional uplift pattern. Load size, lithospheric stiffness, mantle layering and self-gravity enter the regional response too.

Contents · Previous chapter · Next chapter

A deliberately simple local-compensation model helps estimate the possible size of depression. Ignore lithospheric bending and suppose an ice load is balanced by displacement of mantle material of density ρ_m. Equating the added ice mass per unit area with displaced mantle mass gives ρ_i H ≈ ρ_m w.

Using our assigned 2000-metre ice thickness, ice density 900 kilograms per cubic metre and mantle density 3300 kilograms per cubic metre gives w ≈ 545 metres.

This is a long-term, idealized compensation scale. It is not a prediction that every place under two kilometres of ice sinks exactly 545 metres, or rises by that amount immediately after melting. A real load may not persist long enough for full adjustment, and lithospheric bending distributes the response laterally.

The model also omits changes in ocean loading, erosion, sediment deposition, density structure and the gravitational field. Its value is to distinguish two questions: how large an equilibrium displacement might be, and how much of it has actually developed at a particular time.

A region loaded only recently can still be moving downward toward its loaded state. Another region with the same current ice thickness may be closer to equilibrium because the load has lasted longer. Present thickness alone cannot establish the deformation history.

Likewise, removing equal amounts of ice from two regions need not produce equal present uplift rates. The mantle beneath them may differ, and the timing and spatial pattern of unloading may differ. A static balance supplies only one part of the explanation.

Contents · Previous chapter · Next chapter

Now use a separate hypothetical site, with parameters chosen for arithmetic rather than fitted to observations. Suppose sudden complete unloading produces a rapid upward displacement of 30 metres and a further eventual delayed recovery of 270 metres.

Represent the delayed part with a single exponential timescale τ = 3000 years. For elapsed time t after unloading, the total recovery is u(t) = 30 + 270[1 − exp(−t/3000)] metres. The value at t = 0 means just after the idealized rapid response.

After 3000 years, the total recovery is about 200.7 metres. After 9000 years it is about 286.6 metres, still short of the eventual 300 metres. The remaining delayed recovery at that later time is about 13.4 metres.

The uplift rate is the derivative: du/dt = 0.090 exp(−t/3000) metres per year. It starts at 90 millimetres per year in this assigned model, falls to about 33.1 after 3000 years and about 4.48 millimetres per year after 9000 years.

A slowly rising landscape can therefore be the late stage of a much larger adjustment. Small present motion does not imply that the total past displacement was small.

The exponential is a teaching approximation for one relaxation mode. Actual Earth models can contain many timescales and continuing load changes. The numbers should not be transferred to a named coast, used as a forecast or interpreted as a universal maximum uplift rate.

Contents · Previous chapter · Next chapter

Compare two unloading histories using the same linear response model. In history A, the full assigned load disappears at year 0. In history B, half disappears at year 0 and the other half at year 3000.

Assume each half-load contributes half of the rapid and delayed response, so we can add their effects. This superposition is an explicit linear-model assumption, not a guarantee for every real ice–Earth system.

At year 6000, history A has recovered about 263.5 metres and is rising at about 12.2 millimetres per year. For history B, the first half has had 6000 years to respond, but the second half only 3000 years.

Adding the two half-responses gives about 232.1 metres of total recovery for history B and a current rate of about 22.6 millimetres per year. Both sites are now completely unloaded and have the same eventual recovery, yet their present heights and velocities differ.

The more recently unloaded example is still farther from its final state and is moving faster. A measurement of today’s uplift cannot uniquely determine mantle viscosity unless the load history is constrained as well.

This is an inverse problem. A model with a different viscosity and a different ice history may reproduce a similar present rate. Geological evidence for ice extent and retreat timing, multiple measurement sites and older relative-sea-level records help reduce that ambiguity.

The lesson is practical even without solving a full Earth model: ask whether a reported rate is being attributed to present ice loss, past unloading or a combination of both.

Contents · Previous chapter · Next chapter

During loading, deformation can include a depressed central region and a peripheral raised region, often called a forebulge. The pattern reflects lithospheric bending and redistribution within the underlying Earth.

As the ice load is removed, the centre can rebound upward while the former forebulge relaxes downward. A place that was outside the thickest ice can therefore experience subsidence related to the same glacial history that drives uplift farther inland.

This is why “the ice melted, so all nearby land rises” is an unreliable rule. Location relative to the old load and the shape of the deformation response matter. Horizontal motion and changes in gravity are also part of the adjustment.

Draw the unloaded reference as a horizontal line. In the loaded panel, put a broad depression beneath the ice and a smaller raised shoulder beyond its edge. In the recovery panel, use an upward arrow in the depression and a downward arrow on the shoulder. Label the drawing as schematic: it does not supply real distances, amplitudes or boundaries.

The position of the forebulge and the region of present subsidence depend on the evolving ice geometry and Earth structure. They should not be inferred merely by drawing a fixed ring around a modern glacier.

Other processes can also raise or lower land. Tectonic deformation, volcanic activity, groundwater changes and sediment compaction can overlap the glacial signal. The task is to identify their spatial and temporal fingerprints rather than assign every vertical movement to one cause.

Contents · Previous chapter · Next chapter

A tide gauge measures sea level relative to the land supporting it. Let s be sea-surface height in a chosen common reference frame and u be land height in that frame. Relative sea level is R = s − u, so its change is dR/dt = ds/dt − du/dt.

For an original local example, assign an ocean-surface rise of 4 millimetres per year and land uplift of 7 millimetres per year. Relative sea level falls by 3 millimetres per year. At a second site with the same ocean-surface trend but land subsidence of 2 millimetres per year, relative sea level rises by 6 millimetres per year.

Those opposite shoreline experiences do not contradict one another. They combine water motion with different land motion. The Geodesy guide explains the reference frames needed to compare them consistently.

Actual ice loss also redistributes ocean water and changes gravity and the geoid. Losing a large ice mass reduces its attraction on surrounding ocean water, while deforming the ocean floor changes basin geometry. Local sea-surface change therefore need not equal a uniform global average.

Pan and colleagues’ 2021 modelling showed that uplift of marine-based West Antarctic sectors after ice retreat can reduce the space available for water and drive additional outflow into the wider ocean. This illustrates why bedrock rising locally need not reduce global mean sea-level rise. It is a conditional model result about a coupled system, not a claim that all rebound everywhere has the same effect. Read the original rebound-and-sea-level study.

Contents · Previous chapter · Next chapter

Repeated GNSS measurements on stable bedrock estimate motion relative to a terrestrial reference frame. A time series may contain a long-term trend, seasonal loading, equipment offsets, earthquake-related motion and correlated measurement noise. A straight line fitted without those checks can be misleading.

Raised shorelines and dated sea-level indicators provide a longer record, but they measure past relative sea level rather than land motion alone. Their original environmental height, age uncertainty and later disturbance must be considered.

Gravity measurements add another constraint because mantle redistribution changes Earth’s gravity field. Satellite gravity used to infer ice loss therefore needs a solid-Earth correction; otherwise part of the signal can be assigned to the wrong mass change.

Barletta and colleagues’ 2018 study used six GPS stations in the Amundsen Sea Embayment and reported uplift rates reaching 41 millimetres per year. Their viscoelastic modelling supported unusually low shallow-mantle viscosity beneath the studied region. The inference depended on exploring uncertainty in ice history and Earth structure, rather than treating the measured rate as a direct viscosity reading. Read the original Amundsen Sea study.

This does not mean all Antarctica rises at that rate or that every glacial response takes millennia. Regional material properties can permit much faster adjustment than a simple global average would suggest.

The strongest test combines different observations. A proposed model should explain not only one vertical rate but also the broader pattern and, where available, independent shoreline, gravity and ice-history evidence within their uncertainties.

Contents · Previous chapter · Next chapter

For ice grounded below sea level, rising bedrock can reduce water depth near the grounding line, where grounded ice begins to float. Changes in bed slope, water depth and nearby sea surface can influence how the ice retreats.

This creates feedback: ice loss changes the Earth and ocean, and those changes alter subsequent ice behaviour. Under suitable conditions, the feedback can slow retreat compared with a calculation that holds the bed and gravitational field fixed.

The outcome depends on the speed of Earth response relative to ice loss, the bed geometry and the climate and ocean forcing. It cannot be reduced to “the ground rises, therefore the ice sheet is safe.” Nor is the possible stabilizing feedback inconsistent with the separate water-displacement effect described in Chapter 7.

In September 2026, Adhikari and colleagues presented an efficient method for incorporating leading solid-Earth effects into ice-sheet models. Their approach uses precomputed response functions for radially symmetric Earth structures and combines them with changing ice loads to calculate evolving bedrock and geoid fields. The stated scope matters: an efficient approximation is not a complete representation of every three-dimensional mantle structure. Read the original coupled-modelling study.

A response function is, conceptually, the answer to a small standardized loading event. Combining many such responses reconstructs a complex history when linearity is appropriate. Our two-step unloading exercise is a very simple illustration of that logic.

Scientific progress here involves testing useful approximations against more complete models and observations, while keeping the uncertainty visible. A colourful uplift map is an output of that process, not an unqualified forecast by itself.

Contents · Previous chapter · Next chapter

First calculate the pressure of the assigned two-kilometre ice column. Using 900 kilograms per cubic metre and g = 9.81 gives about 17.7 megapascals. Then explain why the 545-metre local-compensation estimate does not specify when that displacement occurs.

Next sketch the recovery curve u(t) = 30 + 270[1 − exp(−t/3000)]. Mark the initial rapid response, the eventual 300-metre limit and the roughly 286.6-metre recovery at year 9000. The slope becomes smaller as the curve approaches its limit.

At year 9000, simply extending the current 4.48-millimetre annual rate for another thousand years predicts about 4.48 metres more uplift. The actual assigned exponential gives about 3.81 metres, because the rate continues to decline. This is a demonstration of model behaviour, not a coastal forecast.

Compare the two unloading histories at year 6000. The later two-step history has less completed recovery but faster current uplift. State why observing current ice extent alone would not distinguish the histories’ effects on today’s motion.

Finally, apply R = s − u to the two coast examples. One has a relative fall of 3 millimetres per year and the other a relative rise of 6. Identify whether each number describes the ocean surface, the land or their difference before interpreting it.

The central idea is mechanical memory: the solid Earth responds to a changing load through time, and its present movement contains information about both the past ice and the material beneath it. Continue through Geophysics or the How Science Works master guide to connect this example with other inverse problems.

Contents · Previous chapter · Continue to Geophysics