12.2 Dam Breach Consequences
181
A paper entitled “Predicting Tailings Dams Breach Release Volumes for Flood
Hazard Delineation” (O’Brien et al. 2016) attempts to give solutions to the same
problem).
It uses again a set of past failures (with all the problems we have discussed earlier)
and derives three relationships (Eqs. 12.7–12.9) based on the breached dam height
as follows:
V upper bound = 20, 419 ∗ H
1.2821
(12.7)
V mean = 3604 ∗ H
1.2821
(12.8)
V lower bound = 1052 ∗ H
1.2821
(12.9)
As most of the data used by O’Brien et al. were the same than those used by
Azam and Li (2010) and most of the past cases were low-volume, low-height dams,
we would advise to use, for dams above 45–50 m and out of prudence, the upper
bound solution.
Lately a research paper entitled “Tailings Dams Failures: Updated Statistical
Model for Discharge Volume and Runout” by Concha Larrauri and Lall (2018) has
lead these authors to more detailed discussions and to publish a free-access app
1
which has the capability of computing the probability of exceeding a runout volume
specified by the user based on a set of historic failures relationships. In this manner,
the uncertainty regarding the volume can be considered when estimating the runout
distance. The app also provides various percentiles of the results.
Concha Larrauri and Lall used 29 failures instead of the original 22 used by Rico
et al. At the end of the day they also show that the total runout volume is appx. 1/3
of the impoundment volume. They state that this type of analysis should always be
considered probabilistically due to the uncertainties. We believe that introducing in
the sample population failures like Samarco, where tailings were carried hundreds of
kilometres away by the Rio Doce (river transport mechanism), constitutes a serious
flaw intrinsic to the dataset. In fact, although tailings travelled more than 600 km
(due to the river), 90% of the released volume remained within 120 km from the
source (which is already quite significant as well).
For Samarco the median prediction of runout volume by Concha Larrauri and
Lall was approx. 50% of the actual release whereas for Mount Polley the median
prediction of runout volume was approx. 85% of the actual release. Additionally, for
smaller TSF the margin of error can become very large.
Rourke and Luppnow (2015) have recently published “The Risks of Excess Water
on Tailings Facilities and Its Application to Dam-break Studies”, which correlated
recent catastrophic failures which reached the bottom of the TSF to the percentage
of the TSF surface inundated. The resulting correlation is staggering, as shown in
Fig. 12.4.
1 https://columbiawater.shinyapps.io/ShinyappRicoRedo/.
181
A paper entitled “Predicting Tailings Dams Breach Release Volumes for Flood
Hazard Delineation” (O’Brien et al. 2016) attempts to give solutions to the same
problem).
It uses again a set of past failures (with all the problems we have discussed earlier)
and derives three relationships (Eqs. 12.7–12.9) based on the breached dam height
as follows:
V upper bound = 20, 419 ∗ H
1.2821
(12.7)
V mean = 3604 ∗ H
1.2821
(12.8)
V lower bound = 1052 ∗ H
1.2821
(12.9)
As most of the data used by O’Brien et al. were the same than those used by
Azam and Li (2010) and most of the past cases were low-volume, low-height dams,
we would advise to use, for dams above 45–50 m and out of prudence, the upper
bound solution.
Lately a research paper entitled “Tailings Dams Failures: Updated Statistical
Model for Discharge Volume and Runout” by Concha Larrauri and Lall (2018) has
lead these authors to more detailed discussions and to publish a free-access app
1
which has the capability of computing the probability of exceeding a runout volume
specified by the user based on a set of historic failures relationships. In this manner,
the uncertainty regarding the volume can be considered when estimating the runout
distance. The app also provides various percentiles of the results.
Concha Larrauri and Lall used 29 failures instead of the original 22 used by Rico
et al. At the end of the day they also show that the total runout volume is appx. 1/3
of the impoundment volume. They state that this type of analysis should always be
considered probabilistically due to the uncertainties. We believe that introducing in
the sample population failures like Samarco, where tailings were carried hundreds of
kilometres away by the Rio Doce (river transport mechanism), constitutes a serious
flaw intrinsic to the dataset. In fact, although tailings travelled more than 600 km
(due to the river), 90% of the released volume remained within 120 km from the
source (which is already quite significant as well).
For Samarco the median prediction of runout volume by Concha Larrauri and
Lall was approx. 50% of the actual release whereas for Mount Polley the median
prediction of runout volume was approx. 85% of the actual release. Additionally, for
smaller TSF the margin of error can become very large.
Rourke and Luppnow (2015) have recently published “The Risks of Excess Water
on Tailings Facilities and Its Application to Dam-break Studies”, which correlated
recent catastrophic failures which reached the bottom of the TSF to the percentage
of the TSF surface inundated. The resulting correlation is staggering, as shown in
Fig. 12.4.
1 https://columbiawater.shinyapps.io/ShinyappRicoRedo/.