Environmental Engineering Reference
In-Depth Information
Ben-Haim (2006) have been applied in the context
of flood risk management by Hall and Har-
vey (2009). Sniedovich (2007), is critical of info-
gap approaches suggesting the approach is based
upon analysis in the neighborhood on a point es-
timate of the system state (the uncertain phenom-
ena) and the output of the analysis is sensitive to
this decision. The method makes the assumption
that the future system states become increasingly
unlikely as they diverge from the point estimate
(Hall (2009)). The method assumes that the most
likely future system state is known a priori. Given
that the system state is subject to severe uncer-
tainty, an approach that relies on this assumption
as its basis appears paradoxical and this is strongly
questioned by Sniedovich (2007)'.
Amore traditional method that involves Bayes-
ian type probabilistic weighting for future scenar-
ios and incorporating these into analysis of options
has been explored through application to the River
Thames, including:
.
Intervention scenarios/decision pipelines
(Fig. 15.11) - This includes analysis of a limited
range of expert-derived decision pipelines that de-
scribe a logical progression ofmanagement choices
that are constrained by the preceding choices. Each
decision point is constrained by previous actions
and as such ismore or less suited to different future
states. Such analysis provides a pragmaticmeans of
developing and exploring future assetmanagement
options (McGahey and Sayers 2008).
.
Formal optimization of the asset intervention
investment strategy - More automated methods
to optimize an asset management strategy have
recently started to appear in the context of flood
risk management (Philips. 2006; Woodward
et al. 2010). These methods draw on various fields
within civil engineering (including bridge mainte-
nance, trussdesignandpipenetworkdesign)where
optimization methods are more widely used. The
most promising methods are based around genetic
algorithms (GAs), reflecting their ability to opti-
mize performance across the many criteria of in-
terest associated with flood risk management
decisions.
Genetic algorithms work by seeking to com-
bine the desirable qualities from solutions already
To use these criteria in assessing the likely
impact of the investment, theymust be considered
not only in broad terms reflecting socioeconomics
and climate change, but also in terms of budgetary
and legislative constraints and environmental im-
pacts and opportunities. Understanding the ro-
bustness, flexibility and adaptability of an asset
management strategy in quantifiable terms, how-
ever, remains the more elusive aim of sustainabil-
ity in this context.
Supporting tools and techniques in aiding
robust option choices
In selecting the best investment strategy the de-
cision-maker is faced with choosing between
many possible options of physical intervention,
further data collection and analysis. Underlying
this choice is a desire to maintain the flood risk
system's ability to perform reasonably well in the
context of all plausible futures that may be en-
countered throughout the appraisal period (i.e.
funding changes and future affordability, climatic
conditions, changes in anticipated performance).
In this context performance is typically mea-
sured in terms of efficiency (e.g. risk reduction,
opportunity benefit) and effectiveness (e.g. benefit
to cost ratio). Determining the preference order-
ing, assuming prefect information, would be a
straightforward ranking process. But a multiple
of both aleatory and epistemic uncertainties com-
bine to complicate this process.
Classical decision theory (e.g. French 1988) dis-
cusses two widely considered approaches to deal
with such uncertainty. One, based upon Laplace's
Principle of Indifference or Insufficient Reason,
involves assigning an equal probability to uncer-
tain quantities; and is therefore fundamentally
probabilistic. The other, Wald's Maximin model,
makes the assumption that the worst case of the
uncertain quantity will always arise, and seeks to
choose the option that maximizes the reward
given this assumption - the approach does not
therefore involve assigning any likelihood to un-
certain quantities.
'More recently Info-Gap approaches, that pur-
port to be non probabilistic in nature, developed by
