Report on the Immediate Causes of the Failure of the Fundão Dam

The Fundão Tailings Dam failed on November 5, 2015 in a liquefaction flowslide that initiated at the dam’s left abutment. This Investigation was performed to determine its cause.

In structuring its investigation process, the Panel systematically identified and evaluated multiple causation hypotheses. It further imposed hypothesis testing by means of the following three questions that the candidate failure mechanism should be able to explain:

1. Why did a flowslide occur?
2. Why did the flowslide occur where it did?
3. Why did the flowslide occur when it did?

Forensic methods adopted by the Panel integrated multiple lines of evidence: observations from eyewitness accounts; data and imagery in geographic information system (GIS) format; field evidence from subsurface exploration by the Panel and others; advanced laboratory testing; and sophisticated computer modeling. Responding to the above three questions for hypothesis testing demanded a high level of quantification and exhaustive detail in each of these aspects of the Investigation’s evidence-based approach.

To understand the failure first requires understanding the materials the dam contained and their properties. There were two types of tailings, both produced in slurry form and delivered in separate pipelines to the Fundão impoundment. Sand tailings, or simply sands, are a mixture of sand-sized and finer silt particles. The sands are relatively free-draining, but when loose and saturated are susceptible to liquefaction, a process whereby the material loses nearly all of its strength and flows as a fluid. The slimes, on the other hand, are much finer and clay-like in nature—soft and compressible with low permeability. How these two materials interacted is key to understanding the failure.

Another central aspect is how their deposition was influenced by a series of unplanned occurrences during the dam’s construction and operation. Together, these incidents established the conditions that allowed the failure to take place. These included: (1) damage to the original Starter Dam that resulted in increased saturation; (2) deposition of slimes in areas where this was not intended; and (3) structural problems with a concrete conduit that caused the dam to be raised over the slimes.

It was originally planned to deposit sands behind a compacted earthfill Starter Dam, then raise it by the upstream method to increase progressively its capacity. These sands, in turn, would retain slimes deposited behind them such that the two materials would not intermingle. To preserve the free- draining characteristics of the sands, a 200 m beach width was required to prevent water-borne slimes from being deposited near the dam crest where they would impede drainage. A high-capacity drainage system at the base of the Starter Dam would allow water to drain from the sands, reducing saturation.

The first incident occurred in 2009 shortly after the Starter Dam was completed. Due to construction defects in the base drain, the dam was so badly damaged that the original concept could no longer be implemented. Instead, a revised design substituted a new drainage blanket at a higher elevation. Together with the revised design there was a fundamental change in the design concept whereby more widespread saturation was allowed and accepted. This increase in the extent of saturation introduced the potential for sand liquefaction.

The second incident associated with slimes and water management occurred over an extended period of time in 2011 and 2012 while the new design was being constructed. During operation, the 200 m beach width criterion was often not met, with water encroaching to as little as 60 m from the crest. This allowed slimes to settle out in areas where they were not intended to exist.

Another incident occurred in late 2012 when a large concrete conduit beneath the dam’s left abutment, the Secondary Gallery, was found to be structurally deficient and unable to support further loading. This meant that the dam could not be raised over it until it had been abandoned and filled with concrete. In order to maintain operations in the interim, the alignment of the dam at the left abutment was set back from its former position. This placed the embankment directly over the previously-deposited slimes. With this, all of the necessary conditions for liquefaction triggering were in place.

As dam raising continued, surface seepage began to appear on the left abutment setback at various elevations and times during 2013. The saturated mass of tailings sands was growing, and by August, 2014 the replacement blanket drain intended to control this saturation reached its maximum capacity. Meanwhile, the slimes beneath the embankment were responding to the increasing load being placed on them by the rising embankment. The manner in which they did so, and the consequent effect on the sands, is what ultimately caused the sands to liquefy.

As the softer slimes were loaded, they compressed. At the same time, they also deformed laterally, squeezing out like toothpaste from a tube in a process known as lateral extrusion. The sands immediately above, forced to conform to this movement, experienced a reduction in the horizontal stress that confined them. This allowed the sands to, in effect, be pulled apart and in the process become looser.

To replicate this process in the laboratory, the Panel applied these stress changes to the Fundão sand. The saturated specimen completely and abruptly collapsed, losing nearly all its strength—a laboratory demonstration of liquefaction. The Panel then undertook a program of numerical modeling to determine whether stress changes similar to those imposed in the laboratory would have also occurred in the field. Using computer simulation of how the slimes deformed during embankment construction, and tracking the corresponding response of the sands, comparable stress conditions that caused the sands to liquefy in the laboratory were reproduced computationally. Simply put, what is known to have occurred during the failure was replicated in the laboratory, and what occurred in the laboratory is shown to have occurred at the left abutment of the dam.

A related aspect of the failure was the series of three small seismic shocks that occurred about 90 minutes earlier. By then the left abutment of the dam had reached a precarious state of stability. Computer modeling showed that the earthquake forces produced an additional increment of horizontal movement in the slimes that correspondingly affected the overlying sands. Although the movements are quite small and the associated uncertainties large, this additional movement is likely to have accelerated the failure process that was already well advanced.

Hence the failure of the Fundão Tailings Dam by liquefaction flowsliding was the consequence of a chain of events and conditions. A change in design brought about an increase in saturation which introduced the potential for liquefaction. As a result of various developments, soft slimes encroached into unintended areas on the left abutment of the dam and the embankment alignment was set back from its originally-planned location. As a result of this setback, slimes existed beneath the embankment and were subjected to the loading its raising imposed. This initiated a mechanism of extrusion of the slimes and pulling apart of the sands as the embankment height increased. With only a small additional increment of loading produced by the earthquakes, the triggering of liquefaction was accelerated and the flowslide initiated.

Immediately following this Executive Summary is an inventory of structures and their locations to help the reader become oriented to the various features associated with the site.