Statistical Characterization of Hazards and Risk in Coastal Areas 26.3 Historical Perspective 575
Part C | 26.3
based on the probabilistic estimation of a design level
associated with a selected value for an acceptable
hazard/risk level. Our review here will show that the
field of engineering has been shifting through the years
from the first approach to the second and that this trend
is continuing.
Examples of the deterministic upper-limit approach
can be found in early concepts used for designs of
harbor protection, primarily based on somewhat subjective extrapolations of visual observations [26.36]. They
have continued to play a dominant role in the development of design specifications throughout much of
the twentieth century, including the estimation of environmental conditions for siting nuclear power plants in
coastal areas, the formulation of design parameters for
large dams, and the design of large coastal structures
for protection from tsunamis. These methods tacitly assumed that there was no epistemic uncertainty in their
estimated maximum values; hence, the estimated deterministic levels could be interpreted as absolute upper
limits which could not be exceeded (i. e., the annual
exceedance probability was zero or, conversely, the average return period was infinite). As will be discussed
subsequently, applications of ultimate limits assuming
no uncertainty can be under-conservative compared to
probabilistic design levels; however, functioning structures based on this earlier approach still exist in many
areas of the world.
Examples of the second approach include design
specifications for most site-specific coastal structures
around the world since the middle of the last century,
when the combination of measurements and predictive
tools attained sufficient skill to permit such estimates.
Initially, these estimates were based predominantly on
historical storms; but with time, there has been a slow
transition from designs based only on events occurring in the historical record at a given site toward
designs that consider all events that could occur at that
site. Early during the development of the probabilistic
approach, the accepted hazard level was taken to be associated with an annual exceedance probability of 0.01
(a mean interval between exceedances of 100 years).
As shown earlier in (26.5) and the discussion following
that equation, such a probability level does not represent
a very low risk for many critical designs or applications,
since there is a 1 in 3 chance that such a value will be
exceeded in about 40 years.
26.3.1 The Development
of Deterministic Methods
Before modern risk concepts, analytical tools, and computer codes existed, it was still necessary to estimate
design parameters for various engineering applications.
For example, in the 1950s, nuclear power began to be
harnessed for commercial applications; and the issue of
plant location quickly became extremely important to
the overall scientific and engineering communities, as
well as the public at large. Initially, a major consideration was given to issues related to national defense; and
although the importance of this issue faded with time,
it influenced the siting of many of the earliest nuclear
power plants. In addition to economic and national defense issues, these plants were expected to be located in
areas where natural hazards would not adversely affect
performance and safe operation over their expected lifetime. The list of potential hazards considered was quite
comprehensive and included the effects of earthquakes
(both direct such as ground motions and indirect such
as the loss of cooling water); atmospheric phenomena
(lightning, tornadoes, tropical cyclones, etc.); flooding
due to coastal storms, tsunamis and precipitation; and
wave action superimposed on top flooding events.
A well-documented example of a deterministic design approach can be found in the methodology for siting of nuclear power plants in US coastal areas affected
by hurricane storm surges. In 1959, the US Army Corps
of Engineers (USACE) contracted the National Weather
Service (NWS) to develop a hypothetical hurricane that
could be used to design hurricane protection projects
along the Gulf and Atlantic coasts of the United States.
At that time the NWS, as part of its National Hurricane
Research Project, set out to define the most severe storm
that is considered reasonably characteristic of a region, defining a storm with such characteristics as the
Standard Project Hurricane (SPH) [26.37]. Since their
storm definition was related to nuclear power plants,
the estimated maximum value possible was consistent
with the concept of a very low allowable risk. Graham
and Nunn’s report [26.37] contains a description of the
derivation of the storm parameters for SPH along US
Gulf and East coasts.
It is clear that all project designs do not require
a level of acceptable risk as low as that used for nuclear
power plants; so in 1979, the NWS Technical Report
23 [26.38] redefined the SPH in a fashion that would
permit it to be more broadly applied. In this report, the
SPH was defined as:
a steady state hurricane having a severe combination of values of meteorological parameters that will
give high sustained wind speeds reasonably characteristic of a given region.
This removed the idea that the SPH was related to the
most severe storm expected in a particular area. Since
many coastal projects are examined on an economic
basis over a fixed interval of time, this re-definition of
the SPH offered more general guidance on the relation
Part C | 26.3
based on the probabilistic estimation of a design level
associated with a selected value for an acceptable
hazard/risk level. Our review here will show that the
field of engineering has been shifting through the years
from the first approach to the second and that this trend
is continuing.
Examples of the deterministic upper-limit approach
can be found in early concepts used for designs of
harbor protection, primarily based on somewhat subjective extrapolations of visual observations [26.36]. They
have continued to play a dominant role in the development of design specifications throughout much of
the twentieth century, including the estimation of environmental conditions for siting nuclear power plants in
coastal areas, the formulation of design parameters for
large dams, and the design of large coastal structures
for protection from tsunamis. These methods tacitly assumed that there was no epistemic uncertainty in their
estimated maximum values; hence, the estimated deterministic levels could be interpreted as absolute upper
limits which could not be exceeded (i. e., the annual
exceedance probability was zero or, conversely, the average return period was infinite). As will be discussed
subsequently, applications of ultimate limits assuming
no uncertainty can be under-conservative compared to
probabilistic design levels; however, functioning structures based on this earlier approach still exist in many
areas of the world.
Examples of the second approach include design
specifications for most site-specific coastal structures
around the world since the middle of the last century,
when the combination of measurements and predictive
tools attained sufficient skill to permit such estimates.
Initially, these estimates were based predominantly on
historical storms; but with time, there has been a slow
transition from designs based only on events occurring in the historical record at a given site toward
designs that consider all events that could occur at that
site. Early during the development of the probabilistic
approach, the accepted hazard level was taken to be associated with an annual exceedance probability of 0.01
(a mean interval between exceedances of 100 years).
As shown earlier in (26.5) and the discussion following
that equation, such a probability level does not represent
a very low risk for many critical designs or applications,
since there is a 1 in 3 chance that such a value will be
exceeded in about 40 years.
26.3.1 The Development
of Deterministic Methods
Before modern risk concepts, analytical tools, and computer codes existed, it was still necessary to estimate
design parameters for various engineering applications.
For example, in the 1950s, nuclear power began to be
harnessed for commercial applications; and the issue of
plant location quickly became extremely important to
the overall scientific and engineering communities, as
well as the public at large. Initially, a major consideration was given to issues related to national defense; and
although the importance of this issue faded with time,
it influenced the siting of many of the earliest nuclear
power plants. In addition to economic and national defense issues, these plants were expected to be located in
areas where natural hazards would not adversely affect
performance and safe operation over their expected lifetime. The list of potential hazards considered was quite
comprehensive and included the effects of earthquakes
(both direct such as ground motions and indirect such
as the loss of cooling water); atmospheric phenomena
(lightning, tornadoes, tropical cyclones, etc.); flooding
due to coastal storms, tsunamis and precipitation; and
wave action superimposed on top flooding events.
A well-documented example of a deterministic design approach can be found in the methodology for siting of nuclear power plants in US coastal areas affected
by hurricane storm surges. In 1959, the US Army Corps
of Engineers (USACE) contracted the National Weather
Service (NWS) to develop a hypothetical hurricane that
could be used to design hurricane protection projects
along the Gulf and Atlantic coasts of the United States.
At that time the NWS, as part of its National Hurricane
Research Project, set out to define the most severe storm
that is considered reasonably characteristic of a region, defining a storm with such characteristics as the
Standard Project Hurricane (SPH) [26.37]. Since their
storm definition was related to nuclear power plants,
the estimated maximum value possible was consistent
with the concept of a very low allowable risk. Graham
and Nunn’s report [26.37] contains a description of the
derivation of the storm parameters for SPH along US
Gulf and East coasts.
It is clear that all project designs do not require
a level of acceptable risk as low as that used for nuclear
power plants; so in 1979, the NWS Technical Report
23 [26.38] redefined the SPH in a fashion that would
permit it to be more broadly applied. In this report, the
SPH was defined as:
a steady state hurricane having a severe combination of values of meteorological parameters that will
give high sustained wind speeds reasonably characteristic of a given region.
This removed the idea that the SPH was related to the
most severe storm expected in a particular area. Since
many coastal projects are examined on an economic
basis over a fixed interval of time, this re-definition of
the SPH offered more general guidance on the relation