Académique Documents
Professionnel Documents
Culture Documents
APPENDIX J
Dynamic Response Analysis
Appendix J
Dynamic Response Analysis
TABLE OF CONTENTS
J1 INTRODUCTION............................................................................................................................. 1
J2 EARTHQUAKE TIME HISTORIES ..................................................................................................... 1
J3 MODEL AND INPUT DATA FOR SITE RESPONSE ANALYSIS ........................................................... 4
J3.1 General ............................................................................................................................. 4
J3.2 Shear Wave Velocity ......................................................................................................... 6
J3.3 Modulus Reduction and Damping .................................................................................... 9
J4 RESULTS OF SITE RESPONSE ANALYSES ...................................................................................... 10
J4.1 Weathered Rock Profile at Samarco Office Site ............................................................. 10
J4.2 Tailings Dam Profiles at Fundo Dam Site ...................................................................... 12
J5 CYCLIC LIQUEFACTION TRIGGERING ........................................................................................... 14
J6 SEISMIC DISPLACEMENTS ........................................................................................................... 16
List of Tables
Table J2-1 Summary of time histories provided for analysis by Atkinson (2016) ............................. 2
Table J2-2 PGA values in bedrock of estimated November 5, 2015 time histories at Samarco ....... 3
Table J4-1 Computed amplification factors from SHAKE analyses .................................................. 12
List of Figures
Figure J2-1 Estimated time histories of November 5, 2015 earthquakes from Atkinson (2016).
(BC1 median-level motions on left; BC2 median-level motions on right) ........................ 3
Figure J2-2 Response spectra of BC1 (mainshock) and BC2 ground motions provided by Atkinson
(2016) ................................................................................................................................ 4
Figure J3-1 Aerial image showing locations of Samarco office and Fundo Dam .............................. 5
Figure J3-2 Shear wave velocity data in weathered phyllite at Samarco office site........................... 6
Figure J3-3 Selected soil columns at Fundo Dam for 1D site response analysis ............................... 7
Figure J3-4 Shear wave velocity data in tailings at Fundo and Germano ......................................... 8
Figure J3-5 Shear wave velocity data in compacted sand .................................................................. 9
Figure J3-6 Three soil columns used in SHAKE2000 analyses ............................................................. 9
Figure J3-7 Modulus reduction and damping curves for tailings sands, slimes and soft rock ......... 10
Figure J4-1 Computed peak accelerations from SHAKE2000 analyses of weathered rock column . 11
Figure J4-2 Comparison of output and input response spectra from SHAKE2000 analyses ............ 11
Figure J4-3 Results of SHAKE analyses for crest and toe soil columns at Fundo Dam.................... 13
Figure J4-4 Comparison of cyclic stress ratios induced by CD1 and CD2 ground motions ............... 14
TABLE OF CONTENTS
(continued)
Figure J5-1 Pore pressure development in laboratory test following lateral extrusion mechanism
and then cyclic loading (test ID TX-31) very loose sample (=+0.05) ......................... 15
Figure J5-2 Laboratory test following lateral extrusion mechanism and then cyclic loading (test
ID TX-31) very loose sample (=+0.05) ....................................................................... 16
Figure J6-1 Calculation of yield acceleration for Newmark-type displacement analysis ................. 17
Figure J6-2 Estimated displacements ................................................................................................ 18
List of Attachments
Attachment J1 Ground Motion Time Histories used in Newmark Displacement Analysis
J1 INTRODUCTION
This appendix presents the results of dynamic response analyses of the November 5, 2015
earthquakes at Samarco, using the one-dimensional ground response analysis software SHAKE2000.
The November 5, 2015 earthquake time histories were estimated by seismologist Dr. Gail Atkinson
(Atkinson 2016) and used as input ground motions in our site response analyses. The analyses were
performed at two locations, namely:
1. at the Samarco office site to evaluate potential amplification of ground motions through the
weathered rock profile; and
2. at the Fundo Dam site to estimate the likely earthquake-induced shear stresses in the tailings
dam profile.
Section J2 summarizes the input earthquake time histories. Section J3 describes the soil models and
input data used in our site response analyses. Section J4 presents the results of the SHAKE2000
analyses. Finally, Sections J5 and J6 present the results of assessments to identify whether pore
pressures or displacements significant to the triggering of liquefaction could result from the
computed ground motions.
Note that the nearest Brazilian regional seismographic stations that recorded the November 5, 2015
earthquakes near Fundo Dam were more than 150 km away from the dam, hence ground motion
prediction equations were used by Atkinson (2016) to estimate the likely ground motions at Samarco.
Atkinson (2016) used the above data to develop a time history of motions that represents those that
likely occurred at the Samarco site on November 5, 2015 prior to the dam failure at approximately
15:45 (local time). The time history sequence includes three earthquakes closely spaced in time:
M 2.2 at 14:12:15 (foreshock)
M 2.6 at 14:13:51 (mainshock)
M 1.8 at 14:16:03 (aftershock)
where M is moment magnitude, estimated from local magnitudes reported by RSBR, and all
times noted are local Brazilian time.
As a result of the ground motions analysis, Atkinson (2016) provided time histories on rock that can
be used to evaluate the response of the structures at Samarco due to the November 5, 2015
earthquake. The time histories provided are summarized in Table J2-1.
Table J2-1 Summary of time histories provided for analysis by Atkinson (2016)
The time histories were provided for three directional components, i.e. two horizontal and a vertical,
and for a reference NEHRP B/C site condition (soft rock) with near-surface average shear wave
velocity, vs30, of 760 m/s. Since the vs30 values of the weathered rock at Samarco are lower, at about
340 m/s to 400 m/s based on Multichannel Analysis of Surface Waves (MASW) measurements
(Appendix C), which corresponds to NEHRP site class C/D (stiff soil), Atkinson (2016) proposes an
amplification factor of 1.4 (multiplication factor) to convert the site class B/C ground motions to site
class C/D ground motions.
To account for both aleatory and epistemic uncertainties, Atkinson (2016) used a factor of 3.2 times
the median (50th percentile) ground motions to obtain the 84th percentile confidence-level (mean
plus one standard deviation) ground motions.
Figure J2-1 shows the two sets of median-level time histories provided by Atkinson (2016). The BC1
time histories contained the estimated foreshock, mainshock and aftershock sequence of
November 5, 2015, whereas the alternative BC2 time histories were scaled up from a M3 earthquake
that occurred about 70 km west of Samarco on May 2, 2016, and was recorded on the Nanometrics
local array. Note the very long duration of the BC2 ground motion due to its original recording at
70 km distance from Samarco. As noted by Atkinson (2016), the BC2 time history sequence was
intended to represent a composite of the foreshock-mainshock-aftershock events of November 5.
Even so, the duration of the BC2 record is expected to be longer than the combined duration
expected of the November 5, 2015 earthquake sequence.
Figure J2-1 Estimated time histories of November 5, 2015 earthquakes from Atkinson (2016).
(BC1 median-level motions on left; BC2 median-level motions on right)
Table J2-2 summarizes the peak ground acceleration (PGA) values of the two sets of records, i.e. BC1
and alternative BC2, for both median-level and 84th percentile horizontal ground motions, as well as
the corresponding ground motions for C/D site conditions using the amplification factor of 1.4
proposed by Atkinson (2016). Note: only the horizontal component time histories are used in the
dynamic response analyses.
Table J2-2 PGA values in bedrock of estimated November 5, 2015 time histories at Samarco
Mw used in Equivalent C/D Equivalent C/D
Horizontal B/C PGA (%g) B/C PGA (%g)
Scaling by PGA (%g) PGA (%g)
Components Median 84th %
Atkinson (2016) Median 84th %
BC1-H1 Foreshock 2.2 2.0 2.8 8.4 9.0
BC1-H2 Foreshock 2.2 1.9 2.7 6.1 8.5
BC1-H1 Mainshock 2.5 2.5 3.5 8.0 11.2
BC1-H2 Mainshock 2.5 2.4 3.4 7.7 10.8
BC1-H1 Aftershock 1.8 1.2 1.7 3.8 5.4
BC1-H2 Aftershock 1.8 1.7 2.4 5.4 7.6
BC2-East 2.5 2.4 3.4 7.9 11.1
BC2-North 2.5 2.2 3.1 7.1 9.9
Figure J2-2 shows the response spectra of the BC1 (mainshock) and alternative BC2 median ground
motions, for both horizontal components. As shown, the BC2 records are generally larger in
amplitudes across most of the periods of interest (or frequencies) than the BC1 records.
Figure J2-2 Response spectra of BC1 (mainshock) and BC2 ground motions provided by Atkinson
(2016)
J3.1 General
We used the computer program SHAKE2000 to perform one-dimensional equivalent-linear site
response analyses at the following two sites in Samarco:
1. weathered rock profile at Samarco office site.
2. tailings dam profiles at Fundo Dam site.
Figure J3-1 Aerial image showing locations of Samarco office and Fundo Dam
The key inputs needed for one-dimensional dynamic response analysis of a site profile, or soil
column, in addition to the input earthquake time histories, are:
shear wave velocity profile of the ground; and
shear modulus and damping variations with shear strain.
These input data for SHAKE2000 analyses are described in the following subsections.
20
40
60 Vs-Section 3 - MASW
80 Survey
Depth (m)
Vs-Depth Relationship
100
Vs= 1500*0.17D0.21
120
140
160
180
200
Figure J3-2 Shear wave velocity data in weathered phyllite at Samarco office site
At the Fundo tailings dam, we modeled a typical soil profile at the crest of the dam and one at the
toe. Figure J3-3 illustrates these two soil columns relative to a cross-section of the dam. The crest soil
column is 88 m deep, consisting of a surface layer of compacted sand, overlying uncompacted tailings
sands and slimes. The toe soil column is 17 m deep and comprises compacted sand overlying only
sand tailings. Both soil columns overlie weathered phyllite (C/D soft rock condition) at the original
ground surface.
Figure J3-3 Selected soil columns at Fundo Dam for 1D site response analysis
For characterizing the tailings, we compiled the vs data measured by Fugro in 2015 at Fundo (see
Appendix C), and recent measurements carried out by ConeTec (see Appendix C, Attachment C2) at
the Germano Dam and Germano Pit Dam tailings sites. Figure J3-4 shows the compiled vs data from
various test locations at Fundo and Germano, and the average vs trend line used to characterize the
tailings deposit for our dynamic response analyses. Note the narrow band of data from the various
sets of vs measurements in the tailings at Samarco.
20.00
Fundao Dam - F04 - Sand
Depth (m)
60.00
Figure J3-4 Shear wave velocity data in tailings at Fundo and Germano
An average vs of 265 m/s was estimated for the compacted tailings sand at the crest, based on
measurements in test hole GSCPT16-06 by ConeTec (Appendix C, Attachment C2) at the Germano
Buttress, as shown on Figure J3-5. For the residual soil/soft rock that underlies the tailings deposit, an
average vs of 400 m/s was estimated based on measurements in GSCPT16-03 by ConeTec
(Appendix C, Attachment C2) at the Germano Pit Dam.
10 GSCPT16-06 -
Compacted Sand
Depth (m)
15
20
25
30
Figure J3-6 shows the three soil columns and corresponding input shear wave velocity profiles used in
our dynamic response analyses.
For tailings slimes, with measured plasticity index values between about 7 and 11,
relationships proposed by Vucetic and Dobry (1991).
For weathered rock, relationships proposed by Silva et al. (1997).
A total unit weight of 22 kN/m3 was used for all tailings and soft rock in the dynamic response
analyses.
Figure J3-7 Modulus reduction and damping curves for tailings sands, slimes and soft rock
-25
-45
-65
Depth (m)
-85
-105
-125
-145
-165
-185
BC1-H1-MS-Median BC1-H2-MS-Median
Figure J4-1 Computed peak accelerations from SHAKE2000 analyses of weathered rock column
Figure J4-2 compares the response spectra of the input BC1 ground motions and the output or
computed motions at ground surface (labeled as -surface layer 1 on Figure J4-2) from the site
response analyses, for both median and 84th percentile mainshock events. Note the site period of the
180 m deep soft rock column computed from the SHAKE2000 analyses is approximately 1.14 sec.
Pseudo Acceleration Response Spectra - Weathered Bedrock -BC1- H1-Mainshock Pseudo Acceleration Response Spectra - Weathered Bedrock -BC1- H2-Mainshock
0.30 0.30
0.25 0.25
H1-MS-Median H2-MS-Median
0.20 0.20
SA(g)
0.05 0.05
0.00 0.00
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Period (s) Period (s)
Figure J4-2 Comparison of output and input response spectra from SHAKE2000 analyses
Table J4-1 summarizes the range in amplification factors, defined as the ratio of the output motion
(i.e. computed surface response spectrum) to input motion (i.e. BC1 response spectrum), across all
periods for the two horizontal components (H1 and H2) of the mainshock. As shown in Table J4-1, the
SHAKE2000 results compare well with Atkinsons proposed amplification factor of 1.4 to convert site
class B/C ground motions to site class C/D ground motions.
Table J4-1 Computed amplification factors from SHAKE analyses
Ratio of Output to Input Motion Response Spectra
Minimum Median Maximum
BC1-H1 Median 0.83 1.25 1.59
BC1-H2 Median 0.90 1.25 1.67
th
BC1-H1 84 Percentile 0.97 1.34 1.64
th
BC1-H2 84 Percentile 0.96 1.28 1.50
Sensitivity analyses were performed with different modulus reduction and damping curves, and the
results were very similar to the above. The B/C ground motions amplified by a factor of 1.4 were
carried forward into the dynamic response analyses of the Fundo Dam. The amplified versions of the
BC1 and BC2 time history sequences are termed CD1 and CD2, respectively.
Depth (m)
Depth (m)
Depth (m)
Depth (m)
-50 -50 -50 -50 -50 Range CSR 0.00139 - 0.0014 CD1- Median
Range CSR 0.0041 - 0.0042 CD1-84th
-60 -60 -60 -60 -60
Depth (m)
Depth (m)
Depth (m)
Depth (m)
-8 -8 -8 -8 -8
-10 Sand -10 -10 -10 -10
-12 -12 -12 -12 -12
-14 -14 -14 -14 -14
-16 -16 -16 -16 -16
-18 -18 -18 -18 -18
-20 -20 -20 -20 -20
Water Table CD1-H1-MS-Median CD1-H1-MS-Median CD1-H1-MS-Median
Sand CD1-H2-MS-Median CD1-H2-MS-Median CD1-H2-MS-Median
Compacted Sand CD1-H1- MS 84th Percentile CD1-H1-MS - 84th Percentile CD1-H1-MS - 84th Percentile
Original ground CD1-H2-MS 84th Percentile CD1-H2-MS 84th Percentile CD1-H2-MS 84th Percentile
Figure J4-3 Results of SHAKE analyses for crest and toe soil columns at Fundo Dam
For the crest and toe soil columns at Fundo Dam, we also ran SHAKE2000 analyses using the
alternative CD2 input ground motions. The cyclic stress ratios induced by the CD2 time histories are
compared to those from the CD1 time histories on Figure J4-4, for both median and 84th percentile
horizontal ground motions. In general, the higher-amplitude alternative CD2 ground motions
generated CSRs about 40% higher than those of the CD1 ground motions. Also, the earthquake-
induced cyclic stresses at the toe soil column are higher than at the crest soil column, due to
amplification of ground motions.
CSR CSR
CREST
0 0.05 0.1 0.15 0 0.05 0.1 0.15
0 0 0
Compacted Sand
-10 -10 -10
-20 -20 -20
-30 -30 -30
Sand
-40 -40 -40
Depth (m)
Depth (m)
-50 -50 -50
-60 -60 -60
-70 -70 -70
Slimes
-80 -80 -80
-90 -90 -90
Soft Rock Vs=400m/s
-100 -100 -100
CD1-H1-MS-Median CD1-H1-MS 84th Percentile
CD1-H2-MS-Median CD1-H2-MS-84thPercentile
CD2-East-Median CD2-East-84th Percentile
CD2-North-Median CD2-North-84th Percentile
CSR CSR
TOE
CREST 0 0.05 0.1 0.15 0 0.05 0.1 0.15
0 0 0
Compacted Sand
-2 -2 -2
-4 -4 -4
-6 -6 -6
Depth (m)
Depth (m)
-8 Sand -8 -8
-10 -10 -10
-12 -12 -12
-14 -14 -14
-16 -16 -16
-18 -18 -18
Soft Rock Vs=400 m/s
-20 -20 -20
CD1-H1-MS-Median CD1-H1-MS - 84th Percentile
CD1-H2-MS-Median CD1-H2-MS 84th Percentile
CD2- East - Median CD2-East-84th Percentile
CD2-North-Median CD2-North-84th Percentile
Figure J4-4 Comparison of cyclic stress ratios induced by CD1 and CD2 ground motions
calculated at the sand/slimes interface beneath the crest, since that is the region of the dam cross
section where the lateral extrusion mechanism would initiate static liquefaction. The cyclic loading
calculated for the sand slimes interface beneath the crest is a CSR of 0.001 to 0.004. It was not
practical to apply such a low load in the laboratory testing; therefore, a CSR of 0.01 was applied. The
sample did not fail under this load after applying 525 cycles. The load was increased to a CSR of 0.02
and cycled for a further 521 cycles. The sample still did not fail, so the CSR was increased to 0.03 and
the sample failed after a further 209 cycles. Very little pore pressure was developed during the cyclic
loading. This shows that the loading from the earthquake would be insufficient to induce liquefaction
in even a very fragile sample. The results from this test are shown on Figure J5-1 and Figure J5-2.
Refer to the cyclic direct simple shear tests shown in Appendix D for further examples of the
insignificant effect that this level of shaking would have on pore pressure development in other
samples of sand tested along an alternate stress path.
The higher cyclic loads calculated close to the surface occur (Figure J4-3) in compacted material that
would not be susceptible to liquefaction.
CSR = 0.03
CSR = 0.02
CSR = 0.01
Figure J5-1 Pore pressure development in laboratory test following lateral extrusion mechanism
and then cyclic loading (test ID TX-31) very loose sample (=+0.05)
Figure J5-2 Laboratory test following lateral extrusion mechanism and then cyclic loading (test
ID TX-31) very loose sample (=+0.05)
J6 SEISMIC DISPLACEMENTS
Having established that the seismic loading on November 5, 2015 would be insufficient to trigger
liquefaction through development of cyclic pore pressure, an analysis was completed to identify
whether the seismic loading could have contributed to the lateral extrusion triggering mechanism by
generating lateral displacements. We assessed this by completing Newmark-type displacement
calculations using acceleration time histories extracted from the sand/slimes interface in the
SHAKE2000 models. The displacement calculations were made using the software SLAMMER.
The displacement calculations involve the identification of a seismic yield acceleration (ay) from limit
equilibrium analyses. The calculation within SLAMMER then identifies portions of the acceleration
time histories that exceed the yield acceleration. The displacements are then calculated by double
integration of the accelerations > ay, and then summation of the displacements resulting from the
integration.
The yield acceleration was calculated in this analysis for the cross section used in the deformation
and stability analyses (Section 01 - see Appendices H and I) assuming that the dam was on the verge
of collapse due to lateral extrusion. Consistent with this assumption, an su/'v strength ratio of 0.14
was used in the calculations because this is the mobilized strength necessary to initiate liquefaction
due to lateral extrusion (see Appendix I). The ay value calculated in this analysis was 0.01 g.
Strengths: Slimes-Rich Layers = su/'v (0.13); Sand = ' = 33; Compacted Sand = ' = 35 & c' = 5 kPa
Horizontal seismic coefficient = 0.01 g
The displacement analyses were run using the 84th percentile time histories in order to understand
the upper-bound of potential displacements. Analyses were run for both the crest and toe columns,
and using both the CD1 and CD2 time history sequences. For the CD1 time history sequence, the
displacements were calculated as the sum of those from the foreshock, mainshock and aftershock.
The results shown on Figure J6-2 indicate small displacements, ranging from 2 mm to 8 mm, with an
average of 5 mm.
Time histories extracted from the SHAKE2000 models, used in this analysis, are shown in
Attachment J1.
Calculated Displacements
9
8
Ay = 0.01
7
Sliding Displacement (mm)
0
CD1-H1-ALL-84th CD1-H2-ALL-84th CD1-H1-ALL-84th CD1-H2-ALL-84th CD2-H1-84th CD2-H2-84th
Percentile - Crest Percentile - Crest Percentile - Toe Percentile - Toe Percentile - Crest Percentile - Crest
Column Column Column Column Column Column