The three models were run on three different computer
shoaling and dissipation are included in the source
platforms. The WIS Gulf hindcasts were run on
function analysis. Propagation is accomplished using
personal computers; WAM and WW3 were run on two
a first-order finite-difference scheme. WISWAVE can
different high-performance computer (HPC) system
be run with nested boundary conditions. Details on
platforms resident at ERDC. All results were run
the wave theory can be found in Resio, 1981, Resio
using single processors although parallel processing
and Perrie, 1989, and Resio et al, 2001. User manual
options are available for all these models. This paper
information on WISWAVE can be found in Hubertz,
is focused on model results; computational speed can
1992.
be attained using suitable compiler and parallel
processing options for all three models.
4.2 WAM
5. GULF LEVEL 2 COMPARISONS
The WAM model used for this hindcast was WAM
Cycle 4.5 which is an update of the WAM Cycle 4
The Gulf Level 2 hindcast grid covers the entire Gulf
model (Komen et al, 1994, Guenther et al, 1992) using
of Mexico (see Figure 1). NDBC 42001, 42002, and
Fortran 90. WAM originated from the Wave
42003 in the central Gulf of Mexico have
Modeling Group directed by Klaus Hasselmann, who
measurements available for comparison for the entire
developed techniques to calculate the non-linear
year of 1995. Comparison line plots will be shown for
source function (Hasselmann and Hasselmann, 1985;
October, 1995, which includes Hurricane Opal. Wave
Hasselmann et al., 1985). This method is known as
direction information is also available at 42001 for
the discrete interaction approximation (DIA). Wave
October and directional comparisons will be shown.
spectra in WAM are not tied to a specific spectral
Summary tables of statistics for only NDBC 42001
shape. WAM's basic physics have not changed but
will be shown since results are similar at NDBC 42002
the source function integration scheme by Hersbach
and 42003. Statistics used for comparison are bias, the
and Janssen, 1999, has been included in this latest
root mean square error (RMS), scatter index, skill
version. WAM does not have options to use different
score, and correlation. Positive bias values indicate
source functions and uses the same physics in all
that hindcast results are higher than measured. Scatter
applications. WAM includes an option for sea ice and
index values should be low indicating little scatter in
can be run in nested applications. WAM has been
the results. Skill scores are reported in decimal values
used in USACE hindcast applications (Jensen et al,
with 1.00 being a perfect match. Correlations are also
2002) in the Alaska area.
reported in decimals with 1.00 indicating perfect
correlation. Monthly statistics using significant wave
4.3 WAVEWATCH III
height from the WIS, WAM, and WW3 hindcasts in
comparison to the measurements at NDBC 42001 are
The third-generation wave model, WAVEWATCH III
shown in Tables 8-10 in the Appendix. Tables 11-13
(WW3) Version 2.22 includes some of the most recent
in the Appendix show the monthly mean wave period
advances in third-generation modeling. Tolman
statistics for this same location. Tables 14-16 in the
(2002) presents a user manual that describes the
Appendix show monthly peak period statistics for
physics and gives directions on installing and running
42001. Tables 3, 4, and 5 are included within this
the model. This model was developed at the Marine
paper's text to summarize the statistical means for all
Modeling and Analysis Branch (MMAB) of the
the buoys discussed in this paper.
Environmental Modeling Center (EMC) of NCEP and
has its roots in Dr. Tolman's Delft University
5.1 WAVE PARAMETER COMPARISONS
research. WW3 has some similarities to the WAM
model, and the default set-up uses the DIA for the
Table 3 gives a summary of the mean wave height
non-linear source term. WW3 allows the user to select
statistics for all three models for NDBC 42001. Bias
different sets of source functions, and the Tolman and
statistics for significant wave height in Table 3 show
Chalikov (1996) source functions and default set-ups
that WAM slightly under-predicts with a mean of
were used in this application. WW3 also has options
0.08 m for the year. WIS over-predicts with a bias of
to include sea ice and currents and has the capability
0.28 m, and WW3 shows a mean bias of 0.02 m for
of running in nested mode. WW3 is currently being
the year indicating slight over-prediction. RMS
used as the operational model at NOAA/NCEP. See
statistics for all models are slightly over 0.2 m with
the following website for an extensive reference listing
WAM having a slightly lower error than the others.
on WW3: polar.ncep.noaa.gov/waves/reference.html
The WAM wave height statistics show the lowest
mean scatter index and best mean skill score. All
4.4 COMPUTER PLATFORMS
Previous Page
