Risk Assessment Oil and Gas

Download 361.05 Kb.

View original pdf

Page	30/66
Date	24.03.2021
Size	361.05 Kb.
	#56167

1 ... 26 27 28 29 30 31 32 33 ... 66

OILGAS
ADNOC Toolbox Talk Awareness Material 2020, ADNOC Toolbox Talk Awareness Material 2020, TRA-Installation of Field Instruments, Road Maintenance Plan & Status-Map Format

3.5.1. What Is “Risk”?
Formally, a “risk” is a product of the probability of an undesired outcome multiplied by the severity of that outcome, summed over the range of possible undesired outcomes. This is the standard definition of risk in decision theory. See the entry on “Decision Theory” in the
Encyclopedia of Statistical Sciences (Kotz, Johnson and Read, 1982). This definition of risk applies as well in insurance calculations, economic analysis, health risk assessment, or ecological risk assessment.
Under this definition, the “probability of outcomes” is a probability distribution. The spread of the distribution represents the uncertainty as to which outcome actually will occur.
Since the probabilities from this distribution are multiplied by measures of severity of outcome,
and then summed to calculate the total risk, the uncertainty is part of the risk. If the outcomes were not uncertain, we would not call it risk, we would just call the analysis a prediction.
The total uncertainty reflected in the spread of the probability distribution arises from two sources: inherent randomness in future processes and imprecision in our knowledge of how to model those processes. Both sources of uncertainty affect the risk in exactly the same way, so there is no point to separating the two when calculating the risk, and there is nothing to be gained from attempting to remove them from the risk characterization.
This perspective is not new. It was recommended 7 years ago in application to environmental risk assessment (Finkel, 1990). In the intervening years, one component of this approach—simulation of probability distributions for uncertain parameter values in order to propagate the effects of the uncertainty through the risk calculation—has become fairly common in risk assessment, and is generically called “Monte Carlo” (but that term has much broader meaning outside the risk assessment community). A formal approach to quantifying the uncertainty in parameter values is not as common. In recent years a substantial literature has appeared that could contribute to facilitating practical implementation of a formal and rigorous approach. Two important developments are better understanding of empirical and hierarchical
Bayes methods to circumvent the problems associated with subjective theories of probability, and new computational techniques that make the probability calculations much easier. Both are reviewed in Carlin and Louis (1996). Goodman (1997) has shown how these techniques can

28
apply to calculation of risk of population extinction in an ecological risk assessment.
The alternative of treating the uncertainty analysis and risk characterization separately is quite unsatisfactory. If risk as we define it is probability of adverse outcome, weighted by some measure of how adverse that outcome would be, and then summed over the spectrum of outcomes, “uncertainty” must already be factored into the probability. If the uncertainty has not already been taken into account, then the probability is wrong and the risk assessment is wrong.
If the uncertainty has already been taken into account in the probability, separating the consideration of uncertainty creates opportunities for mistakenly using the uncertainty twice.

Download 361.05 Kb.

Share with your friends:

1 ... 26 27 28 29 30 31 32 33 ... 66