Table 3. Sediment transport estimate errors based on FRF 8-m array measurements
Sediment Transport Rate
Bias (m3/sec)
RMS Error (m3/sec)
Year
Method
1997
Parametric
47.6
478
Spectral
25.4
437
1998
Parametric
56.5
1728
Spectral
55.7
1350
5.
DISCUSSION
Use of the WIS hindcast wave spectra to drive the nearshore wave transformation model STWAVE provides
improved results compared to using bulk parameters of wave height, peak period, and mean direction. For 1997, the
bias in transport rate was cut in half using the spectral method. Although the wave height errors are similar for both
the spectral and parametric methods, the errors in peak period and, more importantly, mean direction are reduced by
approximately one-third. Wave direction is a key element in the calculation of longshore energy flux.
The main difficulty in application of the parametric method is that it over simplifies the energy distribution with
direction. An extreme example of this is shown in Figures 8 and 9 for comparisons of the frequency spectra and
dimensionless directional distributions, respectively, for the 8-m measurement and the two simulation results. In
this case, two spectral peaks were present at 25-m depth (a high-frequency peak that was oblique to shore and a low-
frequency peak that was more normal to the shore). In the truncation of the input spectra and the transformation to
shore, the higher frequency peak is nearly eradicated and the lower frequency peak dominates. This can generate
large errors in total energy, peak period, mean direction, and longshore transport volume (see Figure 6). The key
error in the parametric approach is the specification of the mean wave direction. As noted previously, errors in wave
height are similar to the spectral approach, but occasional large errors occur in mean direction (20-60 deg). The
parametric approach used here could be improved by truncating the WIS spectra to a half plane prior to calculating
the wave parameters or by identifying individual wave trains and transforming them independently. These
independent wave trains need to be recombined prior to calculating wave breaking.
Parametric approaches to nearshore wave transformation have been used to reduce computational time. Wave
parameters are used to generate a wave climate. Then, representative wave parameters are chosen for
transformation instead of simulating a full year or multiple years. The nearshore time series of waves/sediment
transport are then reconstructed by matching offshore time histories to nearshore model output. This approach
requires approximately two orders of magnitude less computation effort than running a full-year time history at 3-hr
intervals. Parallel processing now allows us to run simulations with quick turnaround. A one-year simulation for
the 16 by 26 km grid used for the FRF took 3.2 hours on an Origin 3000 using 16 processors (or approximately 6
days on a single processor, 400-MHz Alpha workstation). Computational time could be significantly decreased by
applying a sediment transport threshold (e.g., Gravens et al. 1991) and eliminating wave events that produce
insignificant transport.
Errors in the nearshore wave results come from two sources: input to the model and the model itself. Input error
may include errors in the WIS hindcast (e.g., overestimate of wave height for Hurricane Bonnie), local winds, and
bathymetry. The comprehensive verification effort for the updated hindcast will help identify WIS errors. Errors
within STWAVE may include the truncation to a half plane, linear transformation, and local generation. These
potential errors are being further investigated.
Previous Page
