attributed to the differential movement occurring in the embankment. In order to protect against internal erosion of the embankment in the cracks, it was recommended that a sand beach be maintained against the face of the embankment. The purpose of this sand beach was to provide for sand to be carried into any cracks that developed, thereby preventing internal erosion from occurring. The fact that sand was carried into the cracks was confirmed by the observation of sand in cracks on the walls of the breach after the failure.
Just prior to the failure, freeboard on the embankment had been decreased by pond filling operations to the point where tailings solution came into direct contact with the embankment and a sand beach was not maintained. In that configuration, the sand beach, because of its level below the tailings fluid, was ineffective and the cracks within the embankment could become filled with tailings solution. The owner's consultants then proposed a probable failure mechanism consisting of internal erosion of the embankment under the action of the fluid in the cracks.
The embankment was constructed on soil deposits having depths extending up to about 100 feet that included collapsible soils. Some of these soils exhibited collapse in excess of 10% upon wetting.
As the impoundment was filled with water (tailings solution) a seepage front would progress downward through the foundation soils. In addition, capillary pressure in the water in the foundation soils would cause some lateral migration of moisture from the zone of high water content above the wetting front.
The amount of water required to collapse such soils is generally small and collapse could be accomplished by capillary water. As the wetting front moved laterally, the upstream toe of the embankment would settle first, followed by settlement of the downstream toe when the wetting front had reached that point. Furthermore, the lateral migration of the wetting front would increase in depth as time continued.
After the failure, piezometers were installed along the northern portion of the embankment before operations were resumed in the southern pond area. Water levels in these piezometers have indicated that over a monitoring period of approximately six months, ten to fifteen feet of rise in piezometric level could occur. This piezometrlc behavior indicates that a groundwater mound has been established in that region within a short period of time after resumption of operation. Consequently, it is apparent that wetting of the entire depth of foundation soils beneath the impoundment could have occurred in the breach are the beginning of operations until the time at which the breach occurred. Due to lateral migration of the wetting front, the upstream side of the embankment would settle first thereby causing cracking. Longitudinal cracks had been observed in the embankment and had been plugged with bentonite and kerosene slurry on two occassions prior to the failure.
Transverse cracks formed as a result of differential settlement along the axis of the embankment. In addition, the convex alignment in the location of the failed section would have encouraged the development of transverse cracks. Sand was observed in the cracks at various locations on the face of the breach indicating that the tailings sand was carried into the transverse cracks prior to the failure. These cracks would provide a seepage path resulting in a high phreatic surface being developed in the embankment.
Thus, the cracking of the embankment as a result of the large settlement would have a two-fold effect. One would have been to provide a high phreatic surface in the embankment and the other would have been to provide a pathway for seepage allowing for potential piping of the embankment.
After the failure, a number of test pits and test borings were located in the area of the breach. Unconsolidated undrained (UU) and consolidated undrained (CU) triaxial tests with pore pressure measurement were performed on samples taken from the test pits and borings. These data were reported in late 1979 (Ref. 3). Shear strength parameters were determined from these data by the authors of the paper for use in stability analyses. Those shear strength parameters are presented in Table 1.
Stability analyses were conducted in an attempt to evaluate whether slope instability may have been a factor in the dam failure. These stability analyses were conducted for a variety of conditions including the following:
• Analyses that adopted the effective stress shear strength parameters listed in Table 1. As shown in Fig. 1, this analysis assumed essentially complete saturation of the downstream slope from the pond level that existed just prior to failure. This conservatively assumed level of saturation attempted to allow for the effect of extensive cracking. The critical failure circle as shown in Fig. 1 for these conditions reflected a factor of safety less than unity.
• On the basis of field investigation and the test pit logs completed after the breach, the cross-section of the breach and the actual final failure surface was located by the owner's consultants (Ref. 4) as shown in Fig. 2. Also shown in Fig. 2 are several water content readings taken from borings in the vicinity of the breach. The water content values marked by asterisk in Fig. 2 indicate that a zone of unusually high water content extended from the upstream side of the embankment down to the downstream toe. This zone of high water content exists in lose proximity to the proposed actual failure surface. Analysis using the effective stress shear strength parameters listed in Table 1 and the failure surface shown in Fig. 2 indicated a factor of safety of approximately 2 for even a relatively high phreatic surface. It is not expected, therefore, that a general shear failure would have occurred along that entire plane.
• A limited number of unconsolidated, undrained triaxial tests (UU)
