Every earthen embankment passes water, and its design assumes so. What is not assumed is soil moving with that water. The distinction between flow and transport separates a structure doing its job from one quietly dismantling itself. Internal erosion is not one mechanism but a family of them, each with a different initiating condition, a different progression rate, and different implications for what the design should have contained. Sorting out which one was operating is where a dam or levee investigation usually starts, and it governs much of what follows.
Seepage is expected; transported soil is not
Design accounts for seepage with drains, filters and a phreatic surface kept inside the downstream shell. A toe drain running clear is working. The condition that matters is discharge carrying fines: cloudy water, a sediment cone at an exit point, or seepage rising without a corresponding rise in reservoir level.
This is why weir and piezometer records are read as trends rather than snapshots. A rising seepage quantity at constant head, or a piezometric response that decouples from the reservoir, describes a flow path that is changing shape.
Backward erosion piping
Backward erosion begins at a seepage exit, a sand boil on the landside of a levee or a spring at a downstream toe, where the gradient is high enough to lift and carry individual grains. Erosion works upstream from that exit, forming a shallow pipe beneath a cohesive layer that roofs it.
The mechanism needs a specific combination: a pervious cohesionless stratum, an unfiltered exit, and sustained gradient. It is characteristically a foundation phenomenon in levees, and it can advance for a long time before reaching the water source.
Concentrated leak erosion
Concentrated leak erosion requires an opening that holds its shape: a transverse crack from differential settlement, desiccation cracking near the crest, hydraulic fracture along a steep abutment, or a gap left beneath a conduit haunch. Flow through the opening strips material from its walls and enlarges it.
It can move quickly because the flow is concentrated from the outset rather than diffuse. It also depends on the soil having enough cohesion to hold the crack open, which is why well-compacted fine-grained cores are not immune, and why dispersive clays are a recognised concern.
Suffusion and internal instability
Suffusion is subtler. In a broadly graded or gap-graded soil, the fine fraction washes out through the pore network of the coarse fraction without the soil losing its skeleton. Volume barely changes, permeability rises and strength falls.
Because the change is internal rather than visible, suffusion is usually identified afterwards from gradation testing rather than field observation. A material whose grain-size curve shows a gap, or whose fines pass freely through the voids of its own coarse fraction, is internally unstable by geometry, independent of how well it was placed.
Contact erosion and the conduit interface
Where a coarse layer sits against a fine one, flow parallel to the contact carries fines into the voids of the coarse material. The same geometry is created artificially at every penetration through an embankment: outlet works, spillway conduits and utility crossings each produce a soil-to-structure contact with a potential seepage path along it.
Conduits also generate their own defects through joint separation, corrosion perforation or crushing, turning a controlled water passage into a seepage source inside the fill. FEMA's dam safety guidance treats conduits through embankment dams as an internal erosion subject for that reason.
What a filter is supposed to do
A properly designed filter is the defence common to all four mechanisms. Its function is not to block water but to arrest particles: pores fine enough to retain the base soil while remaining pervious enough to relieve pressure without clogging. Filter criteria in USACE seepage and levee engineering manuals, and in FEMA's guidance on filters for embankment dams, are expressed as ratios between filter and base-soil gradations.
The forensic question is rarely whether a filter appears on the drawings. It is whether the material placed met the criteria, whether it was continuous, and whether segregation during construction left windows in it. Many older embankments predate modern filter practice entirely, which is an era finding rather than a defect finding.
Gradation and permeability testing
Grain-size analysis under ASTM D6913, with hydrometer methods for the fine fraction, establishes whether embankment, filter and foundation materials are filter-compatible and whether a soil is internally unstable. Hydraulic conductivity under ASTM D5084 constrains how much flow a zone could plausibly have carried.
Samples have to be tied to location and elevation to be worth anything. Material recovered from a breach face without that control describes soil in general rather than soil at the seepage path.
Initiation, continuation, progression, breach
Dam safety guidance frames internal erosion in stages: initiation of particle movement, continuation where an unfiltered exit permits transport, progression as a pipe enlarges and holds its roof, and breach. The framework matters because it separates questions otherwise argued as one.
Whether a condition initiated is a materials and gradient question. Whether it continued is a filter question. Whether it progressed is a question about the embankment's capacity to arrest it, and about whether anyone intervened. Deficiencies at different stages implicate different decisions.
Where these opinions are challenged
Predictably: that the mechanism was inferred from post-breach geometry rather than established from material and instrumentation evidence; that gradation results came from samples of uncertain provenance; that the filter criteria applied were current-day rather than those governing at construction; and that a seepage trend was drawn from records too sparse to show one.
Analysis that names the mechanism, identifies what evidence each stage rests on, and states plainly which stages remain inference will survive that. A conclusion of piping offered without those separations will not.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.