Early Warning for Water-Retaining Infrastructure: Internal Erosion Detection in Dams and Levees with MCGPR-HDR

Internal erosion — the migration of soil particles through an embankment dam or levee — is responsible for more than a third of all dam failures worldwide, and the process is well advanced before it becomes visible at the downstream face. This is not a statistical anomaly in dam failure data; it is the defining characteristic of the failure mode. The movement of soil particles through an embankment occurs within the body of the structure, along seepage pathways that are invisible from the surface and undetectable by conventional monitoring until the particle migration has progressed to a scale where it affects measurable flow rates or pore pressures. By that stage, the window between detection and failure may be measured in hours.

Events documented in dam safety literature reveal a consistent pattern: failure events attributed to overtopping or operational causes were preceded by internal erosion that went undetected because the investigation methods employed were not capable of identifying it. Water found the path of least resistance because the path had already been prepared by years of particle migration. USGS^[1] water resources science underpins the understanding of seepage and piping mechanisms that govern internal erosion development. Conventional embankment monitoring — piezometers measuring pore pressure, seepage measurement at the toe, inclinometers tracking deformation — detects the consequences of internal erosion at a stage when the erosion process is already established. None of these methods detect particle migration itself.

MCGPR-HDR characterizes embankment condition in terms that are directly relevant to internal erosion risk: moisture content variation, density anomalies indicating zones of material loss, and developing flow pathways within the embankment body. Applied as part of a proactive special technologies inspection program, it provides early warning data before the event that conventional monitoring is waiting to detect.

Internal Erosion: How Embankment Failure Begins

Internal erosion in embankment dams and levees initiates through one of four primary mechanisms. Backward erosion piping begins at the downstream face or toe, where seepage exits the embankment, and progresses upstream as successive soil particles are entrained in the exit flow. The erosion works against the flow direction — backward through the embankment — developing a pipe or channel that grows progressively toward the upstream face. When the pipe connects upstream reservoir water to the downstream exit, the hydraulic gradient drives rapid erosion and failure can follow within hours.

Concentrated leakage through internal cracks is a second mechanism. Cracks through the embankment body — caused by differential settlement, desiccation of clay cores during drawdown, or hydraulic fracturing under reservoir pressure — provide pre-formed pathways along which concentrated flow develops. These are particularly common in clay-core earth-fill dams where the clay core has experienced shrinkage during low-reservoir periods and re-saturation during impoundment. The crack may seal partially when wetted, but the seepage pathway it represents can initiate erosion progressively.

Contact erosion at the boundary between embankment materials of different grain size — typically the contact between a clay core and a coarser shell material — occurs when hydraulic gradients drive fine particles through the coarse matrix. This mechanism is often associated with poorly designed or degraded filter zones whose function is to prevent cross-boundary particle movement. A degraded filter zone between clay core and rockfill shell is an internal erosion risk that neither piezometer data nor visual inspection can identify. Only a subsurface investigation method capable of characterizing the contact zone condition can provide early warning of this failure mode. Water infrastructure assessment programs that incorporate subsurface geophysical investigation address this specifically.

The fourth mechanism — suffosion, or internal instability — involves the selective removal of fine particles from a broadly graded soil matrix, leaving a coarser skeleton that progressively loses fines until it collapses. This mechanism is insidious because it can operate for extended periods without producing changes in measurable parameters. It is the type of internal erosion least likely to be detected by conventional monitoring, and the type most likely to produce sudden, rapid failure without adequate warning. Understanding these mechanisms against the water system context of a specific structure is fundamental to designing an effective investigation program.

What Conventional Monitoring Misses

Piezometers measure pore water pressure at the point of installation. They detect changes in hydraulic gradient within the embankment body, which can indicate developing seepage pathways — but only when those pathways have grown to a scale where they affect measurable pressure distribution. A developing piping channel that has not yet produced a measurable pore pressure change is invisible to the piezometer network. Moreover, piezometer networks installed at design stage typically reflect the assumed internal geometry of the embankment; where internal erosion develops along an unanticipated pathway, it may travel between piezometer locations without intercepting any instrument.

Seepage measurement at the drainage toe quantifies the total flow exiting the embankment. It cannot distinguish between seepage distributed across the full embankment cross-section — which is the design condition — and concentrated seepage exiting through a developing pipe. Both produce measurable flow at the toe. A change in seepage measurement is a confirmation that something has already changed within the embankment; it is not advance warning of change. In the context of dam safety, this distinction is the difference between preventive intervention and emergency response.

Visual inspection of the downstream face and toe identifies surface expressions of internal erosion — boils, wet spots, sinkholes, and slope failures. These are late-stage indicators that appear after the internal erosion process has progressed from initiation through continuation to the stage where it is beginning to affect the embankment surface. As with crack patterns in tunnel linings, these visible indicators are downstream evidence of a process that started elsewhere and earlier. They are valuable for confirming a problem; they are not capable of detecting it at the stage where intervention is straightforward.

How MCGPR-HDR Characterizes Embankment Condition

Ground penetrating radar characterizes embankment condition through its sensitivity to the electrical properties of the fill materials — properties that change measurably when moisture content increases, when density decreases due to particle migration, or when a concentrated seepage pathway develops. The EPA geophysical methods reference^[2] provides technical context for how electromagnetic soil properties reflect moisture and density variation. In embankment surveys, these variations are the primary target rather than discrete buried objects.

Moisture content variation is the most direct indicator of seepage pathway development. A zone of elevated moisture within the embankment body — detectable as a zone of reduced signal velocity and increased attenuation in the GPR profile — may represent a seepage concentration that does not yet produce a measurable change in toe seepage or piezometer readings. Soil moisture GPR-HDR integration methodology specifically addresses the calibration of GPR velocity data for moisture content estimation, providing quantitative rather than qualitative characterization of moisture distribution within the embankment.

Density anomalies indicative of material loss — zones where particle migration has reduced the density of the fill matrix — produce characteristic changes in GPR reflection amplitude and geometry. These anomalies are subtler than the strong void reflections that characterize air-filled karst features, but they are distinguishable from the background reflection character of competent, undisturbed fill in properly calibrated MCGPR-HDR data. The AI-assisted subsurface data analysis applied to MCGPR-HDR embankment surveys improves the detection sensitivity for low-contrast density anomalies by processing the full multi-channel dataset simultaneously rather than analyzing individual profiles in isolation.

Survey Methodology for Dams and Levees

The primary survey configuration for earthfill dams and levees is a vehicle-mounted MCGPR-HDR array traversing the dam crest longitudinally. For levees of significant length — 10 kilometers or more — this configuration allows the full crest to be surveyed in a single mobilization at traverse intervals of 0.5 to 1 meter. Multiple longitudinal passes at different offsets across the crest width provide a three-dimensional picture of embankment condition within the crest zone. For rocky terrain dams — rockfill dams with clay or concrete cores — the crest survey provides the primary dataset, with the core zone characterization being the key deliverable.

Where access to the downstream face permits vehicle-mounted or pedestrian survey, additional traverse lines across the embankment slope provide data on the lower downstream zones that are not covered by the crest survey. Downstream face surveys are particularly valuable for identifying developing seepage concentration in the lower embankment, where pore pressure effects are most pronounced. For large dams with wide crest roads and accessible downstream berms, a combined crest-plus-berm survey provides coverage of the full embankment cross-section.

The multi-frequency approach used in MCGPR-HDR surveys is important in embankment applications because different zones of the embankment have different investigation requirements. Clay core characterization requires sufficient penetration through the overlying fill to reach core depth — typically 3 to 8 meters from crest level — which favors lower-frequency antenna configurations. Near-surface condition assessment of the crest zone and drainage layer uses higher-frequency configurations for resolution. A multi-frequency survey in a single mobilization provides both, optimizing the data return from the investigation. The geothermal borefield subsurface moisture mapping methodology, where multi-depth characterization of moisture and thermal properties is required, uses analogous multi-frequency approaches to MCGPR-HDR embankment surveys.

Early Warning as a Dam Safety Strategy

The strategic argument for proactive embankment GPR surveys is not primarily technical — it is about when along the failure progression curve the dam owner chooses to intervene. A passive monitoring strategy that relies on piezometers and seepage measurement detects internal erosion at a late stage: when it has already developed a pathway, is moving significant quantities of material, and is producing measurable downstream effects. An active investigation strategy that includes periodic MCGPR-HDR surveys detects the precursors of that stage: moisture and density anomalies that indicate where the embankment is most vulnerable before a pathway has fully developed. The cost differential between intervention at the precursor stage — targeted grouting, filter improvement, drainage rehabilitation — and intervention at the late stage — emergency stabilization, partial reconstruction — is substantial. Non-destructive utility mapping principles applied to embankment inspection embody this preventive philosophy.

Regulatory requirements for dam safety inspection vary by jurisdiction, but the trend across major dam safety programs is toward risk-informed inspection frequency and scope that goes beyond visual inspection of accessible surfaces. Geophysical investigation is increasingly referenced in dam safety guidelines as an appropriate method for characterizing embankment condition, particularly for dams classified as high hazard potential. The underground infrastructure of a dam — its internal drainage layers, filter zones, core geometry, and service penetrations — is the component that governs its long-term safety and the component that routine visual inspection cannot reach.

For levee systems, where inspection resources are often distributed across large network lengths, the efficiency of MCGPR-HDR survey makes systematic condition assessment of the full levee network a practical program rather than an aspirational one. Prioritizing remediation investment on the basis of condition data — identifying the specific reaches with elevated anomaly density and directing repair resources there — is a more defensible allocation of dam safety budgets than distributing maintenance resources uniformly across inspected and uninspected reaches alike. Pipe rehabilitation programs within levee systems — addressing the drainage infrastructure embedded in the embankment — are most effectively scoped when GPR data identifies the specific drain locations and conditions that require intervention.