(2) The simulation approach is based on the selection of a set of historic "training storms" extending

over the range of storms that actually occurred at the site location. The training events are selected from the

data set of historic storms as well as from storms that could occur, such as a historic event with a slightly

shifted track. The descriptive storm parameters for each of these events (pressure deficit, radius to maximum

wind, maximum wind, forward speed, angle of propagation, and tidal phase) are computed and referred to

as input vectors describing each storm event.

(3) These input vectors define the multi-vector space of input parameters associated with each storm

event. These parameters are input to the PBL storm model to produce a time-varying spatial distribution of

wind and pressure fields subsequently used as input to a numerical hydrodynamic model for computing storm

surge hydrographs over the computational domain. (Examples of numerical modeling of storm surge and

tidal circulation are presented in Part II-5-7.) The maximum storm surge elevation reached at specified gauge

locations is defined as the response vector of the storm at that location.

(4) Other output vectors such as maximum shoreline erosion, maximum dune recession, or maximum

wave height can be described. The output vector(s) represents the environmental response to the storm. This

response is defined at location *X *and is a direct consequence of the storm via the storm parameter values

defined at the point of nearest proximity of the storm eye to point *X*. For the case of stage-frequency

analyses, maximum surge is assumed to occur when the eye of the storm is nearest location *X*.

(5) Input to the EST model is the training set database of storm input and response vector(s) for a

specific location. An example storm surge analysis for the coast of Delaware (Scheffner, Borgman, and Mark

1993) identified 33 tropical events that impacted the study region during the 104-year period of 1886 through

1989. Of these events, 15 were selected to be representative of the full set. All events were extracted from

the NOAA (1981) database of historic storms. These data were used to develop input to the PBL model to

generate input data to a hydrodynamic long-wave model. Storm surge hydrographs were computed for each

of the 15 storm events. This preliminary analysis generated a set of input and response vectors for 15 storm

events for input to the EST.

(6) Each storm surge is computed independent of the local tidal phase. Each event has an equal

probability of occurring at high tide, MSL after flood, low tide, or MSL after ebb. Each of the computed

surge elevations were linearly combined with the four phases of the tide to generate an expanded training set

of 60 events, which included tidal elevation.

(7) A cumulative pdf for the 60 storm events is developed based on the assumption that each of the 60

events has an equal probability of occurrence, and that the number of hurricanes along the coast of Delaware

is defined by the number of historical events, i.e., 33 events in 104 years. Each storm is considered a point

event in time with each storm occurring independently. The empirical simulation technique utilizes a

resampling scheme in which a random number seed from 0.0 to 1.0 is used to select a historic event from the

training set. Each selected event is described according to its respective input vectors and the maximum surge

elevation response vector.

(8) Input vectors corresponding to the randomly selected event are used to define a new set of input

vectors, based on the initial selected storm values but weighted to reflect the value of the "nearest neighbor"

parameter contained in a vector space representing parameters corresponding to all events in the 60-storm

database. In this manner, a new storm is defined according to a new set of parameters that is similar to those

defining the selected event but adjusted to reflect the parameter interrelationships of the full database. Multi-

vector interpolation schemes use the new event input parameters to estimate a new response parameter - the

peak storm surge at point X.

II-5-48

Water Levels and Long Waves

Integrated Publishing, Inc. |