Land subsidence is the downward movement of the ground surface caused by deformation, displacement, or loss of material beneath it. It ranges from gradual lowering across an entire sedimentary basin to abrupt collapse above an underground cavity. Natural causes include sediment compaction, dissolution of soluble rock, and tectonic movement; human causes include the withdrawal of groundwater, extraction of oil and gas, mining, drainage of organic soils, and construction-related loading or dewatering. The term generally distinguishes downward ground movement from surface lowering caused directly by erosion, although the processes can interact.
Subsidence matters because even small changes in elevation can alter drainage, damage buildings and transport networks, reduce groundwater-storage capacity, and increase flood exposure. In coastal regions it combines with sea-level rise to raise water levels relative to the land. Its investigation brings together hydrology, engineering geology, soil mechanics, and geodesy. Understanding a subsiding area requires identifying the depth and mechanism of deformation, separating reversible movement from permanent damage, and measuring how rates vary across space and time.
Forms, scales, and geological setting
Land subsidence is a description of movement rather than a single geological process. Its geometry depends on the material that deforms and the way stresses or underground openings are distributed. Withdrawal of groundwater from a sedimentary basin commonly produces a broad, shallow depression extending across many square kilometres. Underground mining can generate a more sharply defined subsidence trough. Collapse above a dissolving limestone cavity may instead create a small, steep-sided depression within minutes. These forms demand different methods of investigation even when their visible result is the same: a lower ground surface.
A useful distinction separates continuous subsidence from discontinuous deformation. In continuous subsidence, displacement changes gradually across the landscape, although differences between neighbouring locations may still damage structures. Discontinuous deformation involves abrupt offsets, open fractures, or collapse. A sinkhole is a localized closed depression associated with subsurface material removal or collapse; not all subsidence produces sinkholes, and not all depressions indicate active subsidence. Similarly, settlement commonly refers to the response of soil or a foundation to loading, whereas subsidence often denotes a broader land-surface process. Their usage overlaps in engineering practice.
Geological setting strongly influences susceptibility. Thick sequences of unconsolidated alluvium, lake deposits, and marine sediment can contain compressible clay, silt, and organic matter. Such deposits are widespread beneath coastal plains, intermontane basins, and river deltas. In these environments, coarse sand and gravel may transmit groundwater readily while clay-rich layers deform slowly as their pore pressure changes. Bedrock areas generally resist this form of compaction more strongly, but soluble rocks, underground workings, and highly fractured formations introduce other pathways to subsidence.
Rates span several orders of magnitude. Geological lowering may proceed at fractions of a millimetre to a few millimetres per year, while heavily pumped basins can experience centimetres or, locally, decimetres per year. Collapse events cannot be represented adequately by an annual average: a surface that appears stable for decades may fail suddenly. Cumulative displacement and differential displacement are therefore as important as velocity. A broad area that settles almost uniformly may experience limited structural distortion, whereas a much smaller vertical difference across a building can be damaging.
Subsidence must also be distinguished from movement of the reference system used to measure it. A survey records displacement relative to a benchmark, satellite orbit, or geodetic datum. If the benchmark itself moves, apparent stability or apparent acceleration can be misleading. Describing a subsidence rate meaningfully requires a location, measurement interval, reference frame, and indication of uncertainty.
Effective stress and sediment compaction
The principal mechanism of groundwater-related subsidence is the transfer of load from pore water to the sediment skeleton. Saturated sediment supports overlying material partly through contact forces between grains and partly through fluid pressure. In the simplest formulation, the vertical effective stress is
where is vertical effective stress, is total vertical stress, and is pore-water pressure. If groundwater extraction reduces while the overburden remains approximately unchanged, the grain framework must carry more load. Grains rearrange, clay aggregates deform, and pore space contracts.
This principle underlies soil consolidation, formalized in the twentieth century through the work of Karl Terzaghi and others. In a simple one-dimensional approximation, the compaction of a sediment layer can be expressed as
where is initial layer thickness and is a coefficient of volume compressibility. The relationship is an approximation over a specified stress range rather than a universal linear law. Compressibility changes with sediment type, stress history, fabric, and the magnitude of loading. A thick sequence of moderately compressible sediment can contribute more total subsidence than a thin, highly compressible layer.
The distinction between elastic and inelastic compaction is especially important. At stresses below a sediment’s previous maximum effective stress, deformation is commonly relatively small and largely recoverable. Water-level recovery may then produce measurable uplift. Once effective stress exceeds the preconsolidation stress, additional grain rearrangement produces much larger, predominantly irreversible deformation. The corresponding groundwater level is sometimes described as a critical head, but it is not necessarily uniform across a basin or even through a single borehole.
Fine-grained layers are particularly important because their pore water drains slowly. An aquifer may contain interbedded or adjacent aquitards whose low hydraulic conductivity delays pressure equilibration. Pumping can lower pressure rapidly in permeable sand while water continues draining from clay layers for years or decades. The ground can consequently keep subsiding after pumping has declined or measured water levels in the principal aquifer have begun to recover.
The characteristic timescale of consolidation increases approximately with the square of the drainage distance. A thick clay bed draining only through one boundary responds much more slowly than a thin bed draining through both boundaries. This explains why similar groundwater declines can produce very different surface responses in neighbouring districts. Laboratory consolidation tests provide useful parameters, but field-scale behaviour also depends on buried channels, faults, discontinuous clay lenses, and previous cycles of groundwater withdrawal.
Groundwater withdrawal and delayed basin response
Groundwater pumping changes hydraulic head, the measure of water’s potential energy expressed as an equivalent elevation. Around an operating well, head generally declines toward the intake, producing a cone of depression. Where numerous wells operate, their effects overlap and can create regional pressure declines. Subsidence is especially likely when pumping affects confined or semiconfined systems containing compressible fine-grained layers. In such settings, pressure can fall substantially without an equivalent fall in a visible shallow water table.
Groundwater withdrawal is not identical to groundwater depletion, and neither alone specifies a subsidence rate. Pumping can be partly balanced by groundwater recharge, increased leakage, or reduced groundwater discharge to rivers. Subsidence depends on the resulting pressure history and mechanical properties of the subsurface. A fractured-rock aquifer may sustain considerable head decline with relatively little compaction, while a clay-rich basin may subside substantially under a smaller decline. Water-level observations therefore need to be interpreted alongside geological information.
Seasonal pumping often produces seasonal ground movement. During a dry irrigation season, declining pressure can cause contraction; subsequent recharge may generate partial rebound. A long-term downward trend can be superimposed on these cycles if successive droughts or expanding demand drive heads below previous minima. Drought can intensify the process in two ways: recharge may decrease while farms and cities substitute groundwater for diminished surface-water supplies. The resulting acceleration need not cease immediately when rainfall returns because delayed aquitard drainage continues.
Agricultural demand is an important driver in many sedimentary basins because irrigation can require large volumes over extensive areas. Urban and industrial pumping can instead create concentrated pressure declines around well fields. Changes in municipal supply can therefore shift the location of subsidence rather than simply reduce its basin-wide total. The depth of extraction also matters: neighbouring wells screened in different layers may induce different deformation patterns, even when their surface locations are close together.
Compaction changes the groundwater system itself. Inelastic deformation reduces pore volume and can permanently reduce storage capacity; it may also alter permeability and the exchange of water between layers. The water released by compacting sediment contributes temporarily to pumped supply, but this represents the loss of subsurface storage rather than a renewable addition. Recovering groundwater levels does not normally recreate the original pore structure.
Consequently, groundwater-level recovery and land-surface recovery are different outcomes. A basin may show rising heads while retaining metres of historical elevation loss. Evaluations of pumping reductions must distinguish reduced subsidence rates, cessation of inelastic compaction, and actual rebound. These are related but not interchangeable measures of change.
Subsidence from mining, dissolution, drainage, and other processes
Mining removes solid material and redistributes the load carried by surrounding rock. In longwall coal mining, roof strata are commonly allowed to collapse behind the advancing extraction face, producing a subsidence trough at the surface. Its width, depth, and timing depend on seam thickness, extraction geometry, mining depth, and rock properties. Room-and-pillar workings may initially remain stable because unmined pillars support the roof, but pillar deterioration or failure can cause delayed subsidence long after production ends. Flooding of abandoned workings can further alter stresses and material strength.
Withdrawal of petroleum and natural gas can cause reservoir compaction through pressure decline, broadly analogous to groundwater-induced compaction. The magnitude depends on reservoir thickness, rock compressibility, depletion history, and the mechanical coupling between the reservoir and overburden. The Wilmington oil field near Los Angeles is a major historical example: parts of the Long Beach area subsided by about 9 metres during the twentieth century before pressure-maintenance measures greatly reduced the problem. Fluid injection can limit compaction in some reservoirs, but its effects depend on local geology and operating conditions.
In karst terrain, groundwater dissolves limestone, dolomite, gypsum, or other soluble materials, creating openings that may enlarge beneath a relatively intact surface. Subsidence can occur gradually as loose sediment migrates into these openings or abruptly when a cavity roof or overlying soil arch fails. Pumping, drainage changes, excavation, and leaking water infrastructure can modify groundwater flow and trigger instability. Solution mining of salt introduces a related human mechanism by deliberately creating underground cavities whose long-term stability requires assessment.
Drainage of peat produces subsidence through both physical and biological processes. Lowering the water table removes buoyant support, promotes shrinkage and consolidation, and exposes previously waterlogged organic matter to oxygen. Microbial decomposition then converts part of the soil’s carbon into carbon dioxide, causing actual loss of solid material. Continued drainage can maintain this process for decades. It is important in agricultural lowlands of the Netherlands and in drained tropical peatlands, where elevation loss can progressively increase dependence on pumping and flood defences.
Other mechanisms operate without resource extraction. Thawing of ice-rich permafrost causes settlement when excess ground ice melts and the remaining soil cannot retain its original volume. Wetting of collapsible loess can destroy the weak bonds supporting an open soil structure. Earthquake shaking may cause loose saturated sands to densify through soil liquefaction. At larger scales, plate tectonics, sediment loading, and glacial isostatic adjustment can lower entire regions. These mechanisms may overlap, so identifying a single cause from surface displacement alone is often impossible.
Effects on infrastructure, flooding, and ecosystems
Damage depends strongly on differential movement. Uniform lowering does not necessarily distort a structure, but variations in settlement create bending, rotation, and tension. Buildings may develop cracks, doors and windows may become misaligned, and buried pipes may separate at joints. Roads and railways can lose their designed profiles. Where foundations extend into deeper, more stable strata while surrounding soil settles, relative movement can impose additional downward forces on piles and disrupt connections to utilities at ground level.
Water infrastructure is particularly sensitive because it often depends on small elevation gradients. Subsidence can reduce the capacity of gravity sewers, reverse local drainage directions, and create depressions where rainfall accumulates. In irrigation canals, uneven lowering can reduce freeboard—the vertical distance between water level and the canal bank—or create reaches that cannot carry their intended flow. Repairs may require raising embankments, altering control structures, or adding pumping, even where the canal itself has not visibly fractured.
Coastal subsidence increases relative sea level, which is the height of the sea with respect to the adjoining land. If upward movement is taken as positive, a simplified relationship is
where is relative sea-level change, is sea-surface elevation change, and is vertical land motion. Negative therefore increases relative sea-level rise. Local ocean dynamics and reference-frame choices complicate detailed calculations, but the physical implication is direct: sinking land can experience rising water levels even without a local increase in ocean volume.
This interaction increases exposure to tides, storm surges, and river flooding. It can also reduce the effectiveness of levees and seawalls as their crest elevations fall relative to water. In deltas, subsidence interacts with sediment supply: floodplain deposition can compensate for some elevation loss, but embankments and upstream sediment retention may interrupt that contribution. Coastal wetlands persist only where mineral deposition, organic-soil production, and other elevation-building processes keep pace with relative water-level rise.
Subsidence can alter groundwater gradients and contribute to conditions favouring saltwater intrusion, although pumping-induced pressure decline is often a direct driver of saline movement. Open ground fissures can become preferential pathways for runoff and contaminants, while uneven subsidence may damage well casings. These effects connect land deformation with water quality and groundwater management.
The distribution of harm is also shaped by urbanization and infrastructure design. Informal settlements, shallow drainage networks, and districts lacking reliable piped water may face overlapping exposure to flooding and continued private pumping. Economic losses include not only repair costs but also interrupted transport, increased drainage expenditure, reduced agricultural productivity, and the gradual loss of usable land.
Measuring surface motion and subsurface deformation
Traditional precise levelling measures elevation differences between benchmarks along a survey route. Repeated surveys can reveal millimetre-scale changes under suitable conditions and have established many of the longest subsidence records. Their main limitations are the cost of repeated fieldwork, uneven spatial coverage, and the possibility that supposedly stable benchmarks are themselves moving. A levelling network becomes more informative when connected to independent reference stations outside the deforming basin.
Global navigation satellite system measurements provide three-dimensional positions relative to a defined terrestrial reference frame. Continuously operating stations can resolve long-term vertical trends and seasonal signals, although vertical positions are generally noisier than horizontal positions. Monument stability, antenna changes, atmospheric effects, and processing choices all matter. A station founded on deep material may not represent movement of nearby shallow soils, so the physical construction of the station is part of the measurement’s interpretation.
Interferometric synthetic aperture radar, commonly called InSAR, maps displacement by comparing the phase of radar signals acquired from orbit at different times. It has transformed subsidence investigation by revealing spatial patterns between ground instruments. Multitemporal methods identify persistent radar targets or combine many image pairs to reduce noise. Buildings often provide strong, stable reflections, making the technique particularly useful in urban areas; dense vegetation, changing agricultural surfaces, water, and snow can reduce measurement coherence.
InSAR measures movement along the satellite’s line of sight rather than vertical displacement directly. Converting that measurement to vertical motion usually requires an assumption about horizontal movement or additional observations from different viewing geometries. Atmospheric water vapour, orbit errors, topographic inaccuracies, and phase-unwrapping mistakes can produce misleading patterns. Very rapid movement or sharp displacement gradients may also be difficult to recover. Reliable interpretation commonly combines radar data with levelling, satellite-navigation stations, and field observations rather than treating a coloured velocity map as a complete diagnosis.
Borehole extensometers measure changes in the thickness of a subsurface interval, usually relative to a deep anchor. They can distinguish aquifer-system compaction from deeper geological movement, but their depth determines which layers they capture. Nested instruments, pressure sensors, and wells screened at different depths help identify where pressure decline and deformation occur. Laboratory tests on sediment cores provide complementary measurements of compressibility and consolidation behaviour.
A digital elevation model records surface elevation, not automatically subsidence. Comparing models from different dates requires careful alignment, compatible vertical datums, and separation of ground movement from construction, excavation, or vegetation change. Integrating these observations within a geographic information system allows deformation to be compared with wells, geology, infrastructure, and flood exposure while retaining the different resolutions and uncertainties of each dataset.
Modelling, attribution, and uncertainty
A subsidence model translates assumptions about geology, water flow, and mechanical behaviour into predicted displacement. The simplest empirical models relate observed subsidence to groundwater-level change or pumping. Such relationships can be useful within the conditions under which they were measured, but they may fail when pumping shifts to another depth, previously uncompressed sediments exceed their stress threshold, or thick aquitards continue draining after aquifer heads stabilize.
Physically based groundwater models represent recharge, pumping, leakage, and hydraulic connections among layers. Compaction calculations then use simulated pressures together with elastic and inelastic storage properties. More complete poroelastic models couple fluid pressure and deformation in both directions, allowing changes in stress and pore volume to influence flow. The appropriate complexity depends on the question: a regional management model need not reproduce every foundation-scale deformation, whereas predicting a ground fissure requires more detailed treatment of material contrasts and horizontal strain.
Geological heterogeneity is a major source of uncertainty. A borehole samples only a narrow column, while clay beds and buried channels may change markedly between wells. Faults can impede groundwater flow, concentrate pressure differences, or separate blocks with different compressibilities. Bedrock highs can produce sharp changes in sediment thickness and therefore in cumulative settlement. These contrasts help explain why a smooth regional pressure decline can generate an uneven surface-deformation pattern.
Attribution requires separating overlapping signals. Coastal observations may combine groundwater-related compaction, shallow organic-soil loss, tectonic motion, sediment loading, and construction settlement. Seasonal surface-water loads can produce elastic movement unrelated to local pumping. A statistical correlation between urban growth and subsidence does not by itself show that building weight is the principal cause; both can coincide with increasing groundwater demand. Stronger explanations combine deformation patterns with pressure records, extraction histories, sediment properties, and independent geological evidence.
Forecasts are conditional on future water use, recharge, climate, and management. A model may compare continued pumping, reduced withdrawals, alternative supply, or recharge scenarios, but none is a prediction independent of those assumptions. Calibration to past deformation is necessary but does not guarantee that the model represents the correct mechanism. Multiple parameter combinations can reproduce the same surface record, especially when subsurface observations are sparse.
Modern machine learning methods can map susceptibility from geological, hydrological, and land-use variables. Their outputs require a distinction between susceptibility, present deformation rate, and risk. Susceptibility describes the potential for subsidence; deformation describes observed or modelled motion; risk additionally depends on exposed people and assets and their vulnerability. A high-susceptibility map is not evidence that rapid subsidence is currently occurring at every mapped location.
Regional histories and comparative evidence
California’s San Joaquin Valley is a classic example of groundwater-induced subsidence in an agricultural basin. Large-scale pumping expanded during the twentieth century, lowering groundwater pressure in thick sedimentary deposits containing compressible fine-grained layers. At one extensively documented site southwest of Mendota, the land surface fell by about 8.5 metres between 1925 and 1977. Imported surface water subsequently reduced pumping in some districts, but drought and renewed groundwater dependence produced further episodes of accelerated subsidence. Canals and other water-delivery infrastructure have been affected by the resulting changes in elevation and gradient.
Mexico City illustrates the importance of highly compressible lake sediments. Much of the city occupies deposits of the former lake system in the Basin of Mexico, where groundwater extraction has contributed to cumulative lowering of many metres. Nabor Carrillo’s work in the 1940s helped establish the mechanical connection between pressure decline and consolidation. The city’s deformation is highly uneven because sediment thickness, foundations, and pumping histories vary. Historical buildings, drainage systems, and transport infrastructure have therefore experienced different combinations of settlement and structural distortion.
In Japan, industrial groundwater use caused severe twentieth-century subsidence in districts of Tokyo and Osaka. Groundwater controls, industrial-water legislation, and substitution of other supplies reduced pumping and greatly lowered subsidence rates. The Industrial Water Law of 1956 and legislation regulating groundwater use for buildings enacted in 1962 were important parts of this response. These cases show that controlling the pressure decline can substantially change the trajectory, while also demonstrating that previously lost elevation remains a long-term flood-management concern.
Shanghai experienced approximately 2.6 metres of subsidence in its central urban area between 1921 and 1965, associated principally with intensive groundwater withdrawal from its multilayer aquifer system. Restrictions on extraction, changes in supply, and artificial recharge subsequently reduced rates. In the wider Yangtze River Delta, however, geological variability, construction dewatering, and differing local pumping histories mean that one city-wide average cannot describe every district. Monitoring has consequently combined surface surveys with measurements of groundwater pressure and deformation at depth.
Jakarta and Bangkok demonstrate the interaction between urban demand, compressible deltaic sediments, and coastal exposure. Both have experienced substantial subsidence associated with groundwater extraction, although rates vary considerably by district and measurement period. Bangkok’s management has included groundwater regulation, charges, and expansion of alternative water supplies. In Jakarta, groundwater dependence intersects with uneven access to piped water, drainage limitations, and coastal flooding. These differences make direct comparisons based on a single headline rate unreliable.
Around Houston and Galveston Bay, groundwater withdrawals produced extensive twentieth-century subsidence. The establishment of the Harris–Galveston Subsidence District in 1975 helped institutionalize pumping regulation and conversion to surface-water supplies. Subsequent patterns have varied as water demand and supply sources shifted geographically. Across these regional histories, pressure management has often reduced subsidence, but the timing of improvement reflects delayed drainage as well as the effectiveness and spatial coverage of controls.
Prevention, mitigation, and long-term management
Measures to limit subsidence must target its mechanism. In groundwater basins, the principal objective is generally to prevent damaging pressure declines, especially declines below levels that trigger additional inelastic compaction. This differs from simply balancing annual pumping and recharge. A basin-wide water balance may conceal severe local drawdown, while seasonal pumping can cross a compaction threshold even if water levels recover later. Management therefore often combines extraction volumes with limits or triggers based on groundwater heads and observed deformation.
Reducing withdrawals can involve changes in municipal supply, industrial reuse, irrigation demand, or the spatial distribution of pumping. Each option has wider consequences. Imported surface water may transfer pressure to another basin; pumping relocation can shift subsidence to a new district; and irrigation-efficiency measures do not always reduce basin-wide consumption if saved water supports additional cropping. Effective assessment distinguishes water withdrawn from water consumed and considers the hydraulic connections among aquifers, rivers, and ecosystems.
Managed aquifer recharge can help maintain groundwater pressures where geology, water availability, and water quality permit. Methods include spreading water in infiltration basins and injecting it through wells. Recharge into a shallow layer does not necessarily restore pressure quickly in a deep confined aquifer separated by thick aquitards. It can also encounter clogging, geochemical reactions, or limited infiltration capacity. Even successful recharge generally cannot reverse the permanent pore-volume loss caused by past inelastic compaction.
Engineering measures address consequences rather than the regional cause. Structures may use foundations designed for anticipated differential movement; pipelines may incorporate flexible connections; canals and levees may be raised or regraded. In mining areas, backfilling, pillar assessment, and control of underground water can reduce particular hazards. Karst investigations may use boreholes and geophysical methods to locate unstable ground before construction. Such interventions are site-specific and cannot substitute for identifying the deformation mechanism.
In coastal lowlands, subsidence management must be coordinated with flood protection and urban planning. Drainage pumps, embankments, and tide barriers are designed against relative water levels, which include land motion as well as ocean change. Restoring sediment delivery or rewetting organic soils can address particular elevation-loss processes, but suitability depends on land use and ecological conditions. Where continuing elevation loss makes protection increasingly difficult, land-use restrictions or managed retreat may form part of formally adopted adaptation policies.
Long-term programmes depend on sustained observation and institutions capable of responding to it. Useful records connect well construction, extraction volumes, groundwater heads, benchmark histories, satellite measurements, and infrastructure damage. Reporting both cumulative displacement and recent rates helps distinguish inherited elevation loss from active deformation. Monitoring must continue after apparent stabilization, because renewed pumping, delayed compaction, or changes in extraction depth can reactivate subsidence without reproducing the exact pattern observed in the past.