Page 37 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

III 6 Risk: Theory and Application
facilities pose more risk than others. The former is a frequency- The terms quantitative and qualitative are often used to dis-
based measure that estimates the probability of a specific tinguish the amount of historical failure-related data analyzed
type of failure consequence. The latter is a comparative meas- in the model and the amount of mathematical calculations
ure of current risks, in terms of both failure likelihood and employed in arriving at a risk answer. A model that exclusively
consequence. uses historical frequency data is sometimes referred to as quan-
A criticism of the relative scale is its inability to compare titative whereas a model employing relative scales, even if later
risks from dissimilar systems-pipelines versus highway assigned numbers, is referred to as qualitative or semi-quantita-
transportation, for example-and its inability to directly provide tive. The danger in such labeling is that they imply a level of
failure predictions. However, the absolute scale often fails in accuracy that may not exist. In reality, the labels often tell more
relying heavily on historical point estimates, particularly for about the level of modeling effort, cost, and data sources than
rare events that are extremely difficult to quantify, and in the the accuracy of the results.
unwieldy numbers that often generate a negative reaction from
the public. The absolute scale also often implies a precision that Subjectivity vs. objectivity
is simply not available to any risk assessment method. So, the
“absolute scale” offers the benefit of comparability with other In theory, a purely objective model will strictly adhere to scien-
types of risks, while the “relative scale” offers the advantage tific practice and will have no opinion data. A purely subjective
of ease-of-use and customizability to the specific risk being model implies complete reliance on expert opinion. In practice,
studied. no pipeline risk model fully adheres to either. Objectivity can-
In practical applications and for purposes of communica- not be purely maintained while dealing with the real-world situ-
tions, this is not really an important issue. The two scales are not ation of missing data and variables that are highly confounded.
mutually exclusive. Either scale can be readily converted to the On the other hand, subjective models certainly use objective
other scale if circumstances so warrant. A relative risk scale is data to form or support judgments.
converted to an absolute scale by correlating relative risk scores
with appropriate historical failure rates or other risk estimates Use of unquantifiable evidence
expressed in absolute terms. In other words, the relative scale is
calibrated with some absolute numbers. The absolute scale In any of the many difficult-to-quantify aspects of risk, some
is converted to more manageable and understandable (nontech- would argue that nonstatistical analyses are potentially damag-
nical) relative scales by simple mathematical relationships. ing. Although this danger of misunderstanding the role of a fac-
A possible misunderstanding underlying this issue is the tor always exists, there is similarly the more immediate danger
common misconception that a precise-looking number, of an incomplete analysis by omission of a factor. For example,
expressed in scientific notation, is more accurate than a simple public education is seen by most pipeline professionals to be
number. In reality, either method should use the same available a very important aspect of reducing the number of third-
data pool and be forced to make the same number of assump- party damages and improving leak reporting and emergency
tions when data are not available. The use of subjective judg- response. However, quantifying this level of importance and
ment is necessary in any risk assessment, regardless of how correlating it with the many varied approaches to public educa-
results are presented. tion is quite difficult. A concerted effort to study this data is
Any good risk evaluation will require the generation of sce- needed to determine how they affect risk. In the absence of such
narios to represent all possible event sequences that lead to pos- a study, most would agree that a company that has a strong pub-
sible damage states (consequences). Each event in each lic education program will achieve some level ofrisk reduction
sequence is assigned a probability. The assigned probabilities over a company that does not. A risk model should reflect this
are assigned either in absolute terms or, in the case of a relative belief, even if it cannot be precisely quantified. Otherwise,
risk application, relative to other probabilities. In either case, the benefits of efforts such as public education would not be
the probability assigned should be based on all available infor- supported by risk assessment results.
mation. For a relative model, these event trees are examined, In summary, all methodologies have access to the same data-
and critical variables with their relative weightings (based on bases (at least when publicly available) and all must address
probabilities) are extracted. In a risk assessment expressing what to do when data are insufficient to generate meaningful
results in absolute numbers, the probabilities must be preserved statistical input for a model. Data are not available for most of
in order to produce the absolute terms. the relevant risk variables of pipelines. Including risk variables
Combining the advantages of relative and absolute approaches that have insufficient data requires an element of “qualitative”
is discussed in Chapter 14. evaluation. The only alternative is to ignore the variable, result-
ing in a model that does not consider variables that intuitively
Quantitative vs. qualitative models seem important to the risk picture. Therefore, all models
that attempt to represent all risk aspects must incorporate
It is sometimes difficult to make distinctions between qualita- qualitative evaluations.
tive and quantitative analyses. Most techniques use numbers,
which would imply a quantitative analysis, but sometimes
the numbers are only representations of qualitative beliefs. VIII. Choosing a risk assessment
For example, a qualitative analysis might use scores of 1,2, and 3 technique
to replace the labels of “low,” “medium,” and “high.” To some,
these are insufficient grounds to now call the analysis quan- Several questions to the pipeline operator may direct the choice
titative. of risk assessment technique: