Internal erosion of dams is the process of washout of fine-grained particles from a dam’s filling material by seepage. Internal erosion initiates generally by concentrated leak erosion, backward erosion, and/or suffusion. In the continuation phase it typically manifests itself by the formation of sinkholes, and sudden increased muddy leakage. For internal erosion to occur it generally requires: i) sufficiently large seepage; ii) a fill unable to retain its fines (fine-graded particles); and iii) an interfacing material unable to filter its base. Provided that the core-foundation contact is adequately treated (Ripley, 1988), a dam’s downstream filter is generally considered the most important protection against internal erosion.

Despite the critical importance of filters, during the 1960s and 1970s – the peak period of modern dam construction – there were still uncertainties in criteria in terms of filters protecting cores of broadly graded soils. In addition, it appears there was an over-reliance on broadly graded soils’ ability to “self-heal” (i.e. the core itself being able to self-filter and stop movement of its fine-grained particles), which in turn seems to have induced a trend of using a single filter downstream of the core instead of multiple transition layers. Dams in Scandinavia, parts of North America and Russia (i.e. dams erected in areas once glaciated) are typically the type of dam composed of broadly graded glacial materials (see Figure 1).

Although there has been significant development in the field of internal erosion assessment, internal erosion in dams composed of broadly graded glacial material remains fairly difficult to analyze (Rönnqvist, 2006). In order to improve the understanding of internal erosion in this type of dam, this paper studies the long-term behaviour of 30 dams afflicted with internal erosion. The objective of this review is to investigate: i) typical signs and observations of internal erosion; ii) the timing of internal erosion incidents; iii) the possible early-warning signs; iv) the possible warning-time; and v) the location of the internal erosion incident on the dam body. A better understanding of such factors may provide possible tell-tale signs of dams vulnerable to internal erosion and thereby aid in judgment and tools for early-warning of internal erosion in existing dams composed of broadly graded glacial soils.

In Foster et al. (2000) embankment dam incidents are closely studied, and they conclude that dams with glacial cores are more likely to experience piping accidents than dams constructed of other materials, but that piping is less likely to lead to failure. This paper elaborates further into dams comprising broadly graded glacial soils and their long-term behaviour in terms of internal erosion.

Notation and terminology used in this paper is as follows: DF15 – particle size in filter for which 15% by weight is finer; dB95 – particle size in the core base soil for which 95% by weight is finer; fines – percent finer than a particle size of 0.075mm.

Filter characteristics

Gradings are plotted in Figure 2 of the coarse limits of the filters of dams that have had incidents related to internal erosion (e.g. sinkhole formations). Also in Figure 2, the Sherard & Dunnigan (1989) filter criterion DF15 ≤ 0.7mm is plotted, which by testing (Foster & Fell, 2001) has been found to be the no-erosion-boundary for base soils with fines content 35-85% (soil group 2A and glacial soils). The continuing erosion boundary, i.e. DF15 > 9xdB95 (as proposed by Foster & Fell, 2001), is the limit beyond which the filter is too coarse to be sealed by the fines eroded from the core, and thereby the majority of the impervious core may erode through the filter. Based on the dams reviewed in this paper, the continuing erosion boundary is approximately DF15 = 10.0 mm (if the core is regraded on the 4.75mm sieve), and as shown in Figure 2, only a few of the filters of dams with internal erosion incidents (i.e. sinkholes) are excessively coarse (coarser than the continuing erosion boundary). The majority of the filters’ DF15 falls within the range between the no-erosion boundary and the continuing erosion boundary.

In Rönnqvist (2008) the following set of indicators, typical of dams that have experienced internal erosion, are proposed: a coarsely graded filter (generally DF15 ≥ 1.4mm, as plotted in Figure 2); grading instability of the core and filter; a filter susceptible to segregation; and insufficient filter-transition components to the shell (in terms of rockfill dams). Such unfavorable filter characteristics may adversely affect the filter’s as-placed properties (such as the ‘effective’ DF15-value), which may explain why filters perform poorly, although without excessively coarse gradings. Unstable gradings of soils may cause redistribution of fine-grained particles within the soil (given high enough seepage). Furthermore, depending on the filtering ability of the connecting downstream zone to the filter, or rather the lack thereof, there is also a risk for piping of filter fines into its downstream zone.

The investigated piping accidents involving glacial core materials in Foster et al. (2000) were generally attributed to the use of coarse or segregated filters. The guidelines for assessing potential for segregation of soils are less clear, but Foster & Fell (2001) found that soils with maximum particle size coarser than 75mm, and low on sand (less than 40% finer than 4.75mm) are highly susceptible for segregation. The filters in Figure 2 may have segregated during handling, since the maximum particle exceeds 75mm in many of the filters, and all of them are low in sand content. In order to limit the risk for filter segregation, Ripley (1986) suggests an upper limit of particle size in filters of 18mm and also a high content of sand fraction (at least 60% to be finer than 4.75mm).

Timing of internal erosion related incident

The timing of the first internal erosion incident refers to the elapsed period of time from the dam being put into service to the first clear sign of continuing internal erosion. Internal erosion in the continuation phase usually manifests itself in a sinkhole on the crest of a dam, and is usually formed by way of a large, progressive concentrated leak erosion, backward erosion and/or suffusion.

In close agreement with Foster et al (2000); the timing of the internal erosion incident of dams analyzed in this paper indicates that the majority of incidents occur on first filling or during the first year of operation (Figure 3). Out of the 30 dams analysed, 13 of those were found to have an internal erosion incident that occurred in the early life of the dam, with rockfill dams accounting for the majority of these cases (10 dams). This may be influenced by the fact that most of these rockfill dams are dams with i) coarsely graded potentially unstable filters, ii) filters potentially susceptible to segregate, and iii) no filter-transition component to the rockfill shell (which are characteristics suggested in Rönnqvist (2008) as possible indicators of internal erosion prone dams).

However, there are still a notable number of dams (more than 25%) where internal erosion surfaced after 10 years of service or more (see Figure 3). Such examples point out that an older dam can also experience the first signs of internal erosion.

Possible early-warning signs

Figure 4 shows the observed signs prior to an incident of internal erosion. There may be several surfacing signs leading up to an internal erosion incident. Based on the dams reviewed in this paper, the most common signs of internal erosion (which have led to an incident with sinkhole development) are: i) increase in leakage; and ii) muddy leakage. These signs were also identified by Foster et al (2000) as being the most common.

However, pore-pressure variations in the dam may also indicate internal erosion (Figure 4). Other noted signs are concentrated leaks at the toe of the dam and near-hydrostatic pressures in the core. There are also dams for which there has been no apparent warning sign prior to the forming of sinkholes.

The signs in Figure 4, if observed at an early stage, may indicate the initiation of internal erosion in a dam, and furthermore, if it re-occurs or progresses, they may also indicate the potential continuation of internal erosion. The time-span from the first indication (e.g. leakage-increase, clear seepage turning muddy, or notable pore-pressure variations) to the incident (formation of sinkhole) may provide a possible ‘warning time’. Such a warning is important from a dam safety perspective, especially in terms of monitoring and surveillance periodicity. Based on the 30 dams reviewed in this paper, there is generally a matter of months or longer between the first possible sign of internal erosion to the incident (Figure 5).

If the possible ‘early-warning’ time is subdivided into dam zoning types (rockfill and earthfill dams, and homogenous sections), rockfill dams account for the majority of cases where there is a short period of time between the first sign and the incident of internal erosion (Figure 5). Many of the rockfill dams included in this study are composed of only a single filter zone between the core and the shell, which may partly explain the short warning time to the incident. However, there are still a number of rockfill dams for which there are a number of months between the first sign of internal erosion and the incident.

In terms of the earthfill dams reviewed in this paper there is generally a period of months, in some cases years, between the first sign of internal erosion and the incident. This suggests a longer progression-time from the initiation of internal erosion to the continuation phase (at which sinkholes form).

Location of incident

Foster et al (2000) concluded that conduits through the embankment have the most important influence on the initiation of piping (especially for dams with limited zoning); with other locations including contacts with concrete structures, over foundation irregularities, and steep abutments, but these are less frequent.

For the majority of the cases reviewed in this paper, the location at which internal erosion surfaced is unknown or randomly located over the soil or rock foundation. However, in close agreement with Foster et al (2000), locations with higher incidence of surfacing internal erosion are: i) where embankments abut to concrete structures or sheet pile walls; and ii) where the core is placed on steeply inclined foundation. From experience, these are locations difficult to compact, and are generally more susceptible to hydraulic fracturing and differential settlements (Foster et al, 2000).

Long-term behaviour

In this paper, 30 existing dams comprising broadly graded soils (of glacial origin, i.e. moraines and tills) are reviewed. These are all dams that, to differing extents, have experienced incidents of internal erosion. Based on the reviewed dams, the table opposite summarises the long-term behaviour of internal erosion afflicted dams.

Incidents are most likely to occur early-on, since almost half of the dams reviewed experienced incidents during first filling or within the first year of service. However, without any previous signs, there are dams that experience first signs of internal erosion decades into operation. When not randomly located over soil and rock foundation, the surfacing of the internal erosion incident has been found in this review to occur where dams abut to concrete structures or sheet pile walls, and over foundation irregularities, as discussed by Foster et al, 2000.

The most common signs that precede an internal erosion related incident are: increase in leakage; muddy leakage; and pore-pressure variations in the dam body (although less frequently observed). Such signs may be early-warnings of initiated and progressive internal erosion. Note that there are also dams to which there has been no apparent warning-sign prior to the formation of sinkholes. Generally there is a warning time of a number of months, or longer, between the first sign of internal erosion and the incident. However, rockfill dams account for the majority of cases where there has been a relatively short period of time between the first sign and the incident of internal erosion. Monitoring and surveillance are key features in ensuring reservoir safety. Based on the review of existing dams, this paper shows that there may be a relatively short period of time between the first signs of internal erosion and a potential incident. In order to detect and understand the signs of internal erosion, monitoring needs frequent follow-ups to detect changes in trends and patterns. Furthermore, observations during visual inspections need continuous cross-referencing with the monitored data.

Hans F. Rönnqvist, Post-graduate Student, the Royal Institute of Technology, Stockholm, Dep. of Land and Water Resources Engineering, Hydraulic Engineering.

The research presented in this paper has received financial support from the ”Swedish Hydropower Centre – SVC”. Further information is available on

Full details of this study will be published in Volume XX Issue 2 of Dam Engineering. For further information on this journal contact Tracey Honney via email: