methods cannot be effectively applied in stage-frequency analysis. Thus, numeri-
cal models are often invoked for simulating a larger population of storm-surge
events. Traditionally, modeled tropical storms have been synthesized via a joint
probability method (JPM) to describe storm attributes, such as maximum wind
speeds and pressure deficits. A set of hypothetical storms is built from a
combination of parameter values obtained by statistical analysis of historical
storms.
The JPM requires that all parameters are statistically independent. However,
storm parameters are not statistically independent and the assumption that they
are dependent leads to errors when the JPM approach is taken. Because storm
parameters are related, random grouping of parameters can cause simulation of
storms that may not occur in nature. For example, one parameter may be
assigned a value typical of a weak storm, whereas a second parameter may be
assigned a value representative of an intense storm. Thus, a level of uncertainty
is introduced into the stage-frequency computations. For this study and several
other recent CHL studies, an alternative approach, the EST, has been taken. The
EST preserves the interdependence of typhoon parameters, which is an advan-
tage over the JPM. Details of the EST are given in Borgman et al. (1992);
Scheffner and Borgman (1993); and Scheffner et al. (1999).
Description of technique
EST is a statistical resampling technique that uses historical data to develop
joint probability relationships among the various measured storm parameters. In
contrast to the JPM previously discussed, there are no simplifying assumptions
concerning development of the probability density functions describing histori-
cal events. Thus, the interdependence of parameters is maintained. In this man-
ner, parameter probabilities are site-specific, do not depend on fixed parametric
relationships, and do not assume parameter independence. Thus, the EST is
distribution-free and nonparametric.
For this study, the EST was developed to generate numerous multiyear
intervals of possible future typhoon events for the study site. The ensemble of
modeled or simulated events is consistent with statistics and correlations of past
storm activity at the site. Furthermore, the EST permits random deviations in
storm behavior (when compared to historic events) that could occur in the future.
For example, simulated typhoons are permitted to make landfall at locations
other than those made by historical storms. These random deviations can also
result in more intense storms than the historical events themselves, allowing for
the possibility of a future typhoon being the storm of record.
The simulation approach requires specifying a set of parameters that
describes the dynamics of some physical system, such as typhoons. These
parameters, which must be descriptive of both the physical process being
modeled and the effects of that process, are defined as an N-dimensional vector
space. The parameters that describe the physical attributes of the process are
referred to as input vectors. For example,
28
Chapter 3
Modeling Approach
Previous Page
