- LATEST BLOG POST
- echo $post_date ?>
- Johari window application in risk management is an exploration of possible corporate application of a therapeutic tool developed back in…
- Read More
Hundred years of lessons learned in tailings dams failures is based on the dBase published by CSP.
The infographic below shows the full database of 289 tailings accidents, of any volume, any type of dam, failure mode, release volume, etc. Results are grouped by decade (horizontal axis) and by percentage in each geographic area (in the columns, vertical axis on the left side). The numbers correspond to accident occurrences and on top of each column the sum for that decade is visible. The width of the columns graphically corresponds to the number of occurrences in each decade (the wider the column, the higher number of accidents occurred in that decade).
The figure below uses the same data set to display the number of accidents (the sum of each pie is always equal to 289) by:
First of all remember that causes and incident types may be incorrect as they were established by different people, using different forensic and at different times.
Furthermore the “not available” and “other” categories add to the uncertainties.
That being said and considering the data “as good” (we could hardly do anything differently at this point in time) the following can be inferred:
Thanks to Luna it is possible to go way deeper in interpreting the available data, although we will never stress enough the uncertainties that remain at the database level and around the population of dams in the world.
Back in 2013 we studied the rate of failure per decade using an estimate of 3500 active tailings dams in the world, an assumption that is reasonable based on the fact that active dams represent five times more failures than inactive ones. We derived a rate of failure for major accidents that seems to still be in excellent agreement with what is occurring around the world.
In 2015 we showed that consequences are not necessarily correlated, in one way or another, with dam height or pond volume. Indeed, for example: i) a small (300,000m3) release at Stava killed hundred of people (because the valley downstream funneled the release toward a village a few kilometers away becoming on of the most mortiferous tailings accidents in history), whereas ii) the more recent Samarco huge release ran hundred of kilometres, killing dozens of people and iii) Mount Polley mostly ended in a lake nearby, with no victims.
As in many industries the “scary stuff” is not necessarily the riskier one.
Our practice and research have shown that the probability of failure is, or will be, often way higher in smaller structures than in major ones, simply because more care is taken for larger structures than for “insignificant ones” (or deemed so). Examples like Stava or Bafokeng are there to show that “extreme” consequences can actually occur and that risk has to be evaluated, not be left to “intuition”.
We also demonstrated that the rate of fatalities in the tailings “industry” lies way above the generally accepted “safe” thresholds for hazardous industries. The number of existing, operational, and also closed tailings storage facilities around the world makes it necessary to prioritize the mitigation tasks, if we want to achieve a higher quality, be it at corporate or at national levels.
Finally in 2016 we showed that common misconceptions like considering “insufficient FoS” as a hazard (or a risk) instead than a deliberate choice based on excessive audacity, errors and omissions, insufficient efforts chain-up to yield the current level of probabilty of failure of dams.
We took a rather extreme, but logical, line of thinking, assuming that dams failures find, in the vast majority, their root-causes in human choices and not in natural events and generated a model that actually delivers probabilities of failure in excellent agreement with the performance of the world portfolio.
The fascinating point the above leads-to is the following: as we briefly discussed in our 2016 paper, there are “congenital sins” in dams that remain silent for decades, until, one morning, failure occurs. Of course no failure has one cause only, and it takes a number of factors during the life of the dam and the congenital sins to coalesce into a failure.
No matter what we do today, tailings dams will continue failing, as the congenital sins were made at the moment they were designed unless very costly reinforcements are planned. We know of a few case where these are being built or considered, but in many instances there not even enough physical place to consider them. Only serious risk prioritization cold guide this type of decisions at regional, corporate or higher portfolio levels.
Market conditions (price of metals) may have an influence of present-day conditions (less money to monitor? More distracted operators? Less competent operators? A mix of all of the prior?), but congenital sins were made decades ago.
The tendency to use highly mathematical models and data mining without understanding the processes behind the real long life of a dam, from inception on, can lead to marvelous, but absurd results. Furthermore the database only contains 289 accidents and, happily so, only a few belong to the higher consequences categories.
Using correlations on such a small sample can lead to fascinating, but absurd results.
So, what should we do?
Contact us to know more on how we can help!