Page 204 - Pipeline Risk Management Manual Ideas, Techniques, and Resources

Segmentation 8/181
not a sample measurement but rather represents an event or tion can also be made regarding the origin and confidence sur-
condition that is tied to a specific point on the pipeline. rounding the collected data. It is entirely appropriate to gather
However, it will he assumed to be representing some distance some data as a simple table-top exercise-for example, field
either side of the location specified. An example is leak or personnel indicating on an alignment sheet their knowledge of
break data. A leak usually affects only a few inches ofpipe, hut ROW condition or depth of cover-with field verification to
depending on the type of leak, it yields evidence about the sus- come later. However, it is useful to distinguish this type of
ceptibility of neighboring sections of pipeline. Therefore, a assumed information from actual measurements taken in
zone of influence, x number of feet either side ofthe leak event, the field. A soil resisitivity measured near the pipeline should
is reasonably assigned around the leak. The whole length of usually have a greater impact on risk perception than an
pipeline in the zone of influence is then conservatively treated assumed regional level of soil corrosivity. Increasing uncer-
as having leaked and containing conditions that might suggest tainty should be shown as increasing risk, for reasons detailed
increased leak susceptibility in the future. Considerations will in earlier chapters.
be necessary for overlapping zones of influence, when the zone One way to account for variations in data quality is to ‘penal-
for one event overlaps the zone for another, leaving the overlap ize’ risk variables that are not derived from direct measurement
region to he doubly influenced. or observation. This not only shows increasing risk with
increasing uncertainty, but also helps to value-show the bene-
Countable events fits of-the direct measurements and justify the costs of such
activities that most agree intuitively are a risk mitigation meas-
Some point events may be treated not as sample measurements ure. Table 8.1 shows an example ofadjustments for data quality.
but rather as countable events. An example is foreign line cross- The adjustment factor can then be used along with an age
ings or one-call reports or ILI anomalies (when an anomaly- (decay) adjustment as follows:
specific evaluation is not warranted). The count or density of
such events might be of interest, rather than a zone of influence. Variable score x (Quality Adjustment Factor) x (Age Adjustmrnt
The number of these events in each section can be converted Factor)
into a density. However, the density calculation derived after a
segmentation process can be misleading because section length to ensure that less certain information leads to higher risk
is highly variable under a dynamic segmentation scheme. A estimates.
density might need to be predetermined and then used as an
event prior to segmentation.
V. Segmentation
Spatial analyses
As detailed in Chapter 2, an underlying risk assessment princi-
The most robust risk assessments will carefully model spill ple of most pipeline risk models is that conditions constantly
footprints and from those estimate hazard areas and receptor change along the length of the pipelines. A mechanism is
vulnerabilities within those areas. These footprints require required to measure these changes and assess their impact on
sophisticated calculation routines to consider even a portion of failure probability and consequence. For practical reasons.
the many factors that impact liquid spill migration or vapor lengths of pipe with similar characteristics are grouped so
cloud dispersion. These factors are discussed in Chapter 7. that each length can be assessed and later compared to other
Establishing a hazard area and then examining that area for lengths.
receptors requires extra data handling steps. The hazard area Two options for grouping lengths of pipe with similar
will be constantly changing with changing conditions along the characteristics are fixed-length segmentation and dynamic
pipeline, so distances from the pipeline to perform house
counts, look for environmental sensitivities, etc. will be con- Table 8.1 Sample adjustments for data quality
stantly changing, complicating the data collection and format-
ting efforts. For instance, a liquid pipeline located on a steep
terrain would prompt an extensive examination of downslope
receptors and perhaps disregard for upslope receptors. Modern
GIS environments greatly facilitate these spatial analyses, but
still require additional data collection, formatting, and model- Measurement 100% Actual measured value or direct
observation
ing efforts. The risk assessor must determine if the increased Estimate 80% Based on knowledge ofthe variable,
risk assessment accuracy warrants the additional efforts. nearby readings. etc. Confident of
this condition, but not confirmed
Data qualityhncertainty by actual measurement; value
proposedwill be correct 99% of
As discussed in Chapters 1 and 2, there is much uncertainty sur- the time
rounding any pipeline risk assessment. A portion of that uncer- Informed guess 60% Based on some knowledge and
tainty comes from the data itself. It might be appropriate to expertjudgment, but less
characterize collected data in terms of its quality and age-both confident: value proposed will be
correct 90% of the time
of which should influence the evaluator perception of risk and Worst case Applied where no reliable info IS
hence, the risk model. A ‘rate of decay’ for information age is default available
discussed in Chapter 2. Adding to the decay aspect, a distinc-