Fundão Tailings Dam Review Panel Report on the Immediate Causes of the Failure of the Fundão Dam

slimes were present to impede downward drainage and that the sands remained unsaturated,
resistance to liquefaction―a well-known vulnerability of upstream construction―could be assured.
By 2005, the existing tailings facilities at Samarco’s Germano operation were nearing capacity, and a
new third pellet plant would increase production of both sand and slimes. The adjacent Fundão
Valley was chosen as a new tailings site. In the layout that emerged, the sands and slimes would
initially be physically separated, with sands deposited behind Dike 1 and slimes behind Dike 2, as
represented on Figure 2-2. Retention of the slimes required that the sands deposited between the
two dikes always remain at a higher elevation throughout the raising process. This was a matter of
reservoir geometry, and the dikes in Figure 2-2 had been strategically positioned for sands and slimes
in 70% and 30% proportion of the total received from all plants.

Figure 2-2 Fundão Dikes 1 and 2

Two alternative methods were considered for raising Dike 1 after filling the space between the two
dikes with sand. One was centerline raising depicted on Figure 2-3 using compacted sand tailings in
the downstream slope. This alternative was not selected, with the drained stack concept shown on
Figure 2-4 adopted instead. The Dike 1 Starter Dam would be a conventional earthfill structure
constructed of compacted saprolite soils to crest El. 830 m, with subsequent upstream raising with
sand tailings to El. 920 m.

August 25, 2016 Page 5

Fundão Tailings Dam Review Panel Report on the Immediate Causes of the Failure of the Fundão Dam

Figure 2-3 Centerline raising of Dike 1 considered but not implemented

Figure 2-4 Upstream raising of Dike 1 by the “drained stack” concept

Thus, the Germano Buttress structure became the prototype for Fundão. Like its predecessor, the
Dike 1 Starter Dam for Fundão would be underlain by a high-capacity base drain of gravel and rock.
This would connect to another drain on the Starter Dam’s upstream face, along with other
complimentary drainage features—all to minimize saturation in the sand deposit behind it.
A remaining design consideration was how to evacuate surface water inflows from ordinary
precipitation, floods, and discharged tailings slurry. This would be accomplished by two concrete
galleries, 2 m diameter decant conduits of reinforced concrete extending beneath the tailings deposit
and Dike 1 itself. The Main Gallery would be beneath the right abutment and the Secondary Gallery
beneath the left as indicated on Figure 2-5.

August 25, 2016 Page 6

Fundão Tailings Dam Review Panel Report on the Immediate Causes of the Failure of the Fundão Dam

Figure 2-5 Main (Principal) and Secondary Galleries

In the Panel’s estimation, this design concept for Fundão offered several advantages. With the dam
located in a narrow valley constriction, the site was efficient, requiring a modest amount of dam fill
for the storage volume achieved. Once above the valley floor, the reservoir expanded to provide large
capacity relative to the area it occupied. But the concept also had certain vulnerabilities. The design
was not adaptable to variation in the proportion of sands and slimes received. And most importantly,
it depended on achieving adequate drainage of the sands.

2.2 The Piping Incident (2009–2010)

Construction of the Dike 1 Starter Dam, with its requisite drains and galleries, was completed in
October, 2008. Shortly after full-scale discharge of sand tailings began on April 13, 2009, large
seepage flows carrying fines appeared on the downstream slope above the main underdrain as
shown on Figure 2-6, conditions symptomatic of the process of piping or internal erosion.

August 25, 2016 Page 7

Fundão Tailings Dam Review Panel Report on the Immediate Causes of the Failure of the Fundão Dam

Figure 2-6 Internal erosion effects on downstream slope of Dike 1

An Emergency Action Plan in place for the dam at that time was immediately implemented. The
reservoir was lowered, a berm was constructed over the affected portion of the dam slope, and
provisions were made for holding the reservoir’s remaining contents in the downstream Santarem
Dam should failure occur. Engineering investigations later revealed serious construction flaws in the
base drain and its filters, including a portion of the drain’s outlet that had never been completed. This
allowed water pressure within it to build until causing the slope to erode and slump.
As these investigations continued, the impending rainy season made it too late to fully restore the
drainage features to their original condition, making it impossible to repair the damage. Instead, all of
the drains were sealed. With this, the most important element of the original design concept became
inoperative.
Additionally, the balance between sands and slimes crucial to the dam raising plan was changed.
Filling of Dike 2 had begun earlier than anticipated, making its slimes level higher, not lower, than the
projected sands in Dike 1. At the same time, reduction in pellet production reduced the amount of
sand available while delivery of slimes continued. This required construction of yet a third dike
between Dikes 1 and 2, designated Dike 1A, to provide additional slimes capacity. It was November
2010 before all of the measures made necessary by the piping incident were finally completed.
It remained to devise a new design concept to replace the old one.

August 25, 2016 Page 8

Fundão Tailings Dam Review Panel Report on the Immediate Causes of the Failure of the Fundão Dam

2.3 The Recovery (2011–2012)

A revised design for raising Dike 1 to El. 920 m was first described in the 2011 Operations Manual,
then updated in the 2012 version when the dam had reached crest El. 845 m. The central feature was
the addition of a blanket drain on the surface of the tailings to replace the inoperative base drain
below them. As shown on Figure 2-7, the new blanket drain was at El. 826 m just below the Starter
Dam crest. Figure 2-8 depicts how the blanket drain would become embedded within the tailings
during raising of the dam, intercepting seepage that could otherwise emerge on the slope and reduce
its stability. In order to augment capacity for discharging the collected seepage flows, the blanket
drain also contained slotted pipes called “Kananets®”.

Figure 2-7 Blanket drain (plan view) on tailings surface at El. 826 m

Figure 2-8 El. 826 m blanket drain (section) showing extent behind Dike 1

August 25, 2016 Page 9

Fundão Tailings Dam Review Panel Report on the Immediate Causes of the Failure of the Fundão Dam

Comparing Figure 2-8 to Figure 2-1, it can be seen that the new blanket drain represented an attempt
to replicate the drained-stack concept by providing drainage for the overlying tailings. But the sands
below this drain would remain saturated, as would much of the tailings behind it. Once the base drain
became inoperative, the control of saturation embodied in the original design concept could not be
restored.
A requirement common to both the original and revised designs was that the sands be free-draining.
To ensure that low-permeability slimes would not be deposited where they could impede this
drainage, water containing the slimes had to be restricted from the area of sand deposition. To do so,
a 200 m minimum beach width had been specified in the original 2007 Operations Manual, a
provision retained in the 2011 and 2012 versions.
But as operation proceeded, this beach-width criterion was not consistently achieved. As explained in
greater detail in Section 5.1.3, a new Overflow Channel was conveying water and slimes from Dike 2
to the rear of the Dike 1 reservoir, making beach management more difficult. No longer were the
sands and slimes physically separated; the interface between them could only be controlled by
adjusting the amount of sand spigotted from the dam crest in relation to the amount of slimes-laden
water being introduced. As plotted on Figure 2-9 and documented in Appendix B, during much of
2011 and 2012, beach widths violated the 200 m minimum more often than not, at times encroaching
to as little as 60 m from the crest.
500
Tailings Beach Width
450 200 m Beach Width

400

350
Beach Width (m)

300

250

200

150

100

50

Figure 2-9 Monthly beach width measurements by Samarco, 2011-2012

2.4 The Setback (2012–2014)

Even as recovery from the 2009 Starter Dam piping incident remained underway, new conditions
were developing that would directly affect the left abutment. The galleries shown on Figure 2-5 that
evacuated water from the Fundão impoundment were found to be structurally deficient. This first

August 25, 2016 Page 10

Fundão Tailings Dam Review Panel Report on the Immediate Causes of the Failure of the Fundão Dam

became evident for the Main Gallery at the right abutment when in July, 2010 a vortex appeared in
the reservoir above it, showing that tailings and water were entering. Inspections revealed cracking
and structural damage from foundation settlement and construction defects. Were either of the
galleries to collapse, uncontrolled release of tailings from the reservoir or failure of the dam would be
possible. So in January, 2011 a program of jet grouting was initiated to repair the Main Gallery and
return it to service.
Similar conditions were discovered for the Secondary Gallery, and jet grouting was undertaken there
as well. But by July, 2012, it was apparent that jet grouting had not cured these problems. After a
sinkhole appeared in the tailings overlying the Secondary Gallery in November, 2012, repair efforts
were abandoned. Instead, plans were made to plug both galleries by filling them with concrete from
their outlets to a point beneath the projected crest of the 920 raise in order to prevent their collapse.
Moreover, it was discovered from structural analyses that the Secondary Gallery could not support
tailings higher than El. 845 m, some 10 m lower than the tailings already were at that time.
Because the height of tailings at the left abutment already exceeded the load capacity of the
Secondary Gallery, the dam could not be raised any further over this area until the plugging operation
was completed. As a temporary solution, it was decided to realign the dam at the left abutment by
moving it back behind the portion of the gallery to be filled with concrete so that embankment raising
could continue. This realignment shown on Figure 2-10 became the “setback”.
The setback would move the crest closer to the reservoir water and the slimes it contained, but it was
anticipated that the dam would be quickly returned to its original alignment as soon as the plugging
operations were done. At the same time, as will be explained more fully in Section 5, moving the
crest back from its original alignment would also place it closer to, if not over, areas where beach
encroachment and slimes deposition had already occurred.
Filling of the Secondary Gallery was completed on August 22, 2013. Meanwhile, dam raising had
continued, with seeps that began to appear at the left abutment as early as June 26, 2012, at
El. 845 m. In February, 2013, three-dimensional seepage modeling of the 920 raise showed that
additional drains would be needed at the abutments if seepage breakout were to be prevented. This
analysis was borne out when seepage, saturation, and cracking began appearing at several locations
at the left abutment during 2013. The first such incident occurred in March at El. 855 m, followed by
another seep in June at El. 855 m. Both were treated by constructing a drain. A third seep on
November 15 appeared at El. 860 m and was accompanied by slumping of the slope shown on
Figure 2-11. Another drain was provided to address this condition. On December 26, seepage
occurred at El. 860 m and there was cracking on the left abutment crest at El. 875 m.

August 25, 2016 Page 11

Fundão Tailings Dam Review Panel Report on the Immediate Causes of the Failure of the Fundão Dam

Figure 2-10 Left abutment setback proposed in June, 2012

Figure 2-11 November, 2013 seepage, cracking, and slumping at left abutment El. 860 m

August 25, 2016 Page 12