Submitted to Guyana Geology and Mines Commission
by Dam Review Team

November 16, 1995

2.1 Internal dam configuration
A simplified explanation of the internal features of the dam is necessary for a proper understanding of the discussions that follow. The attached Figure 1 illustrates the internal zones of fill material as indicated on design drawings. These include:

Uncompacted mine waste is contiguous with the compacted rockfill zone on its downstream side. This zone is of secondary structural importance to the dam itself, serving primarily as a convenient waste dump area for overburden materials excavated from the pit.

The dam was raised in stages, a common tailings dam construction practice, with fill placement beginning in early 1992 and proceeding intermittently thereafter. No core fill was placed during the last 3 months prior to the failure. The initial Stage 1A, or starter dyke, was constructed of clayey saprolite soils, with subsequent stages also incorporating rockfill, and filter sand zones as shown on Figure 1.

Prior to construction of the starter dyke it was necessary to confine the flow of Captain Mann Creek through the foundation area within a corrugated steel diversion conduit to allow the initial fill to be placed in dry conditions. The conduit was incorporatd into the starter dyke fill, and after serving its temporary purpose the portion indicated on Figure 1 was filled with a cement grout "plug". Except for this grouted section, the conduit was left open over its remaining length.

2.2 Impoundment operation
Tailings are the barren particles of finely ground ore that remain after extraction of the valuable minerals in the mill process. During operation, tailings slurry was discharged from a pipeline at the rear of the impoundment, with the settled tailings forming a delta sloping gently toward the dam and with the overlying (supernatant) water accumulating directly against it. Within much of the impoundment basin, the finest fraction of the tailings, or slimes, settled from suspension to form a near-horizontal submerged surface about 15m below the water level at the time of failure. Some of this water was pumped to the mill for recycle by a floating-barge decant system located adjacent to the dam at the southwest corner of the impoundment.

As shown on Figure 1, the dam crest was at el. 534m and the water level at el. 529.6m at the time of failure. The current water level is about el. 517m, with the slimes surface about 1- 2m below the water level at approximately el. 515m. Thus, virtually all of the impounded water was lost in the failure. In November, 1994 the slimes surface was at el. 511m, and this level would have increased during tailings deposition over the ensuing 9 months. This indicates that a comparatively small volume of tailings solids was released.

2.3 Failure events
Failure of a dam is usually understood to mean breach of the dam embankment, but this did not happen at Omai. Rather, failure in this case resulted from complete loss of dam core integrity, where here we define "failure" as uncontrolled release of the reservoir.

The crest of the dam was observed on the afternoon of Saturday, August 19, and at that time no distress was apparent. Routine water quality monitoring of the Omai River on August 15 had detected no elevated cyanide levels.

At about midnight on August 19 substantial flows across a mine haul road were observed near the south end of the main dam. About one hour later the dam crest was inspected by flashlight, and longitudinal cracking was observed with a few centimetres of horizontal width and vertical offset. At the same time large flows issuing from the north end of the dam were also noted. The elevations of these discharges were later determined to be 512m at the south end and 509.5 at the north, suggesting that discharge may have initiated at the north end of the dam.

The combined outflows were substantial, with peak discharge reported to have been on the order of 30 m3/s. By noon of August 22, the impoundment water level had dropped 12m to el. 518, resulting in an average drawdown of 15cm/hr over an 84-hr period and with reported maximum drawdown at about twice this rate.

2.4 Emergency response
By August 20, OGML had evacuated personnel and equipment from the Fennell Pit and completed a diversion ditch to intercept flow from the south end of the dam and direct it into the pit for containment. At this time discharge from the north end of the dam into the Omai River remained uncontrolled. By August 24 a "cofferdam" structure to contain this discharge was completed near the north end of the dam by dumping rockfill and saprolite materials across the discharge channel. Water rose behind this structure to el. 514m, impounding water internally within the rockfill zone of the main dam, where it was discharged at the south end of the dam into the diversion ditch, and from there to the Fennell pit.

These conditions remain in effect as of the date of this report. Water is currently impounded behind the cofferdam to about el. 512m, which implies that the rockfill zone within the main dam is saturated to this elevation, a level about 5m lower than the current water surface within the tailings impoundment itself.

3.1 Failure mode screening
We have adopted a systematic approach to evaluation of failure causation hypotheses in an attempt to avoid overlooking any plausible mechanisms. This has consisted of first compiling potentially applicable failure modes from known causes of other tailings dam failures. These candidate failure modes have then been screened to eliminate those that are either unsupported or precluded by available evidence and/or analyses. Those causes considered as possible initiators of the failure and currently ruled out on this basis include:

These failure modes will be revisited if further information indicates the need to do so.

3.2 Proximate cause of failure
It is our current judgement that failure of the dam was caused by massive loss of core integrity resulting from internal erosion of the dam fill, a process also known as piping. Internal erosion in this context means simply that finer particles from one soil moved freely under the influence of seepage forces into and through the interstitial voids of adjacent coarser soil due to excessive disparity between particle sizes of the two soils.

We believe this process began at the interface shown on Figure 1 between the filter sand and the compacted rockfill. Loss of filter sand into the rockfill left the overlying saprolite core material unsupported and subject to the development of cavities, softened zones, and cracks as its particles too moved into the rockfill. Cavity development in the core fill is likely to have propagated undetected for some period of time until reaching the reservoir at and above the slimes level. The final breakthrough of these cavities formed "pipes," or tunnels, in the core fill at multiple locations that allowed uncontrolled flow of water into and then longitudinally through the rockfill zone of the dam. These features are now manifested by sinkholes on the upstream slope of the dam, as core fill and riprap have subsided into the open voids.

Longitudinal core cracking now evident on the crest of the dam is interpreted as an effect, not a cause, of this process. Core tension cracks observed on the crest during the first hours of the failure appear to have been produced by loss of filter sand related to elevated internal water levels within the rockfill zone, and resulting core deformations. Additional widening and vertical downdrop of up to 1.5m across these cracks at some locations developed during the days and weeks subsequent to the failure in a manner consistent with continuing subsidence of highly disturbed and softened areas or cavities within the fill.

There are believed to be two primary physical defects in the dam that allowed this process to occur, one related to filter incompatibility between the sand and rockfill zones, and another involving the diversion conduit. Both were produced by known or suspected deficiencies in design, construction, or construction inspection, either singularly or in combination. Moreover, defects related to filter incompatibility and the diversion conduit may not be mutually exclusive, and may have interacted in complementary ways not yet fully understood.

3.3 Filter incompatibility
Internal erosion between zones of adjacent soils is prevented by controlling their particle size distributions (or gradations) according to filter design criteria developed over 50 years ago and little changed since then. These filter criteria were applied in the dam design, which limited the gradation of the rockfill directly adjacent to the filter sand. No dimensions or nomenclature for this finer rockfill zone were provided, and it is designated here as the "rockfill transition zone".

Figure 2 shows the specified particle size distributions (or gradation) for the sand and rockfill, along with the coarse limit restriction for the transition rockfill. The transition rockfill gradation is indicated to have been determined from filter criteria using a media d85 particle size for the sand of 0.85mm. However, the specified gradation envelope for the sand on Figure 2 allows for d85 as low as 0.5mm, material finer than median value, and filter criteria would not be satisfied for sand at the fine limit of its specification.

Current design practice also recognizes that coarse material is prone to particle-size segregation during construction that has often allowed internal erosion to occur even when other filter design criteria have been satisfied. According to supplemental criteria that address this problem, the transition zone rockfill should not have exceeded a maximum specified particle size of about 25-50mm, whereas Figure 2 allows for fragments as large as 600mm. Therefore, the transition zone rockfill as specified would have been highly susceptible to particle size segregation and consequent filter incompatibility.

These design deficiencies notwithstanding, it is apparent that the transition zone rockfill was never included in the dam during construction in any complete or systematic way. Even if properly designed, meticulous adherence to transition rockfill gradation specifications at each and every location within the dam would have been mandatory to ensure its safety against internal erosion. By contrast, construction documentation and existing conditions on the dam crest indicate that pit-run rockfill of essentially unrestricted gradation was placed directly against the filter sand, without adequate construction control of this critical feature.

Rockfill placement was supervised during construction of the initial stages of the dam and is believed to have been inspected or observed by several geotechnical engineers on various occasions. Such gross disparity of particle sizes between the filter sand and adjacent rockfill as can be currently seen on the dam crest should have been visually evident to any experienced geotechnical engineer, along with equally clear implications for filter incompatibility between the two materials. However, we have been provided with no information to indicate that any such supervision, inspection, or observation sufficiently recognized the severity of this condition, adequately warned of its potential consequences, or undertook measures necessary to correct it.

In basic terms then, the rockfill adjacent to the filter sand was simply too coarse to prevent the sand from washing into and through it, and both potential and actual problems this produced appear to have gone unrecognized or uncorrected throughout the sequence of design and construction until the failure occurred.

3.4 Diversion conduit
The pattern, nature, and distribution of surficial damage provides circumstantial evidence to suggest that the corrugated steel diversion conduit was associated with internal erosion processes. Furthermore, problems related to conduits in general have been responsible for a significant proportion of earth dam and tailings dam failures, and in particular the use of unencased corrugated metal culverts through dam cores is considered bad practice.

Since the remaining slimes behind the dam and saturation levels within it are likely to preclude any excavation or direct inspection of the conduit, actual damage it may have experienced or produced may never be completely known. Additionally, design details for the conduit were not only ambiguous from the start but also underwent continuing change from the feasibility design continuing on throughout construction, making it difficult for us to fully assess the intent of the design or the actual as- built conditions. One largely unexplained aspect is that the evolution of conduit design and construction appears to have progressed in several important respects from more conservative to less conservative over time, resulting in a number of irregularities. It is known that the corrugated steel culvert was crushed by heavy equipment and repaired at two separate locations and occasions during construction, suggesting the possibility that other latent damage might have gone unrecognized.

The nature of corrugated metal culverts is such that they must deform (from circular to slightly oval shape) in order to develop load-carrying capacity. This raises the possibility that deformation incompatibility between the rigid grouted section and the deformable open section (see Figure 1) may have caused structural failure, or that the combined fill, slimes, and water loads may simply have exceeded the structural capacity of the culvert in the critical region beneath the Stage 1A starter dyke. Any such structural failure would have produced a void or allowed soil to enter the conduit, providing a direct path for concentrated seepage and cavity formation within the fill.

Even so, structural failure of the conduit would not necessarily have been required for concentrated seepage and internal erosion to initiate and propagate along the outer surface of the conduit, a common occurrence without adequate safeguards. The details of conduit backfilling and as-built construction are important in assessing this mechanism, and our ongoing investigation is continuing to evaluate it.

Finally, there is evidence to indicate that sand was used for culvert backfill beneath some portion of the Stage 1A starter dyke. Concentrated seepage within this sand may have produced internal erosion at its downstream terminus with the rockfill as a result of filter incompatibility issues previously discussed.

3.5 Continuing DRT investigations
Compilation and synthesis of information obtained from field forensic studies is necessary to amplify and refine the causation theories advanced at this time. However, initial evaluation of this information suggests nothing to date that would be likely to fundamentally alter the preliminary views expressed herein. Continuing investigations will be required to finalize our conclusions and prepare our final report to GGMC on the failure, including but not limited to evaluation of seepage and deformation patterns within the dam, and analysis of structural capacity and deformation of the diversion conduit.

We view it as important that the mechanisms and processes that caused the tailings dam to fail not be allowed to recur in other structures at OGML. The failure causation theories advanced here have two significant implications in this respect.

First, the "cofferdam" that now prevents contaminated flow from reaching the Omai River was intended to be a temporary emergency structure, not a permanent one, and hence contains no proper filter. Although we have no indication of adverse performance at this time, the seepage integrity and stability of both this structure and the mine waste dump that confines water within the rockfill zone of the tailings dam need to be determined. Additionally, we are aware of no flood routing studies performed to evaluate the security of the cofferdam during extreme precipitation events or seasons. We would urge that a decommissioning plan for the tailings impoundment be prepared and implemented without delay, and that this plan take all necessary measures to ensure the security of the cofferdam and related features.

Second, OGML cannot physically resume operation until a new tailings impoundment is constructed, and lessons learned from the failure need to be addressed in its design. In this regard, our investigation to date provides no reason to believe that the failure was related to any concealed geologic conditions or features, or to any anomalous behaviour or engineering properties of the dam, foundation or fill soils, that would pertain to other structures at the minesite. The failure was caused not by some "hidden flaw" but by inadequate application and execution of sound practices for design, construction, supervision, and inspection that are well understood in current embankment dam and tailings dam technology. Inasmuch as the failure is attributable to defects related to filter incompatibility and/or the diversion conduit, it should be well within the ability of conservative dam engineering and construction practice to eliminate these failure modes altogether for any new tailings impoundment constructed at the mine. An important factor in doing so will be to ensure that construction is controlled by qualified engineers with clearly- defined responsibility, direct lines of reporting, and full authority to stop the work if and when necessary to correct it. This will also require documented plans for quality assurance and quality control to promptly detect and correct any construction deficiencies.

OGML has put forward a design concept for dams confining the proposed Tailings Pond No. 2. These dams contain neither diversion conduits nor rockfill within structural zones that would require filters. Hence, it is not possible for the failure mechanisms described herein to be operative for the new structures.

This design concept is proven in practice, having been successfully applied for many dams constructed of and upon tropical residual soils in Brazil for over 30 years. In addition, the design advanced by OGML contains multiple redundancies for safe control of internal seepage that enhance reliability. The necessary dam construction techniques are not complex or new to OGML, requiring only equipment and procedures routinely used in its mining operations.

This design concept is still at an early stage, and many details remain to be developed. However, we currently see no technical reason why approval of this concept in principle could not be granted that would allow the design process to continue during the initial stages of construction.

Respectfully submitted,
Dam Review Team, November 16, 1995

S. Vick, Chairman

R. Squires, Administrative Coordinator

R. Dundee

R. Pedroni

N. Watson

Glossary of Technical Terms
Figure 1 - Internal Features of Dam
Figure 2 - Sand and Rockfill Gradation Specifications