Offshore Wave Statistics

Ideally, we would like to transfer the full offshore six hourly time series to the nearshore location P. With a perfect model, and given all the necessary local information, this would correspond to having had a wave measuring instrument operational at P for the corresponding length of time. For various reasons a simply transformation of the offshore series in this way leads to some difficulties. First, practically all the input data have been provided under the condition that the time series data are not to be transferred to third parties (i.e., the actual time series cannot be included with the Eurowaves package, although some form of scrambling in time, retaining the actual parameter triplets is another alternative being considered which would seem to be acceptable). Secondly, there is the problem of the computer time required for the transfer, at least in the case when the more advanced model is required. Presently, on a fast PC, one run of the SWAN model with about 300 grid points takes about 20 seconds. With six hourly data, about 1500 wave conditions have to be transferred each year, and, as a result, a time series spanning a few years may take several days to complete.

The alternative is to work directly with the offshore wave statistics. Given a suitable discrete distribution of the input parameters significant wave height, H, mean or peak period T, and wave direction, q , we have at the offshore input location a three-dimensional statistics, represented, for example, by a trivariate probability distribution (see Athanassoulis et al., 1998b, for a discussion of this subject). Instead of the single values of the time series, we now only need to transfer the triplets (H, T, q ) of the single cells, reconstructing the corresponding statistics at point P. However, in reality it is not as simple as this. First, each triplet must be expanded into a full two-dimensional spectrum F(f, q ) in order to carry out the transformation to the coast. As we have discussed in the previous section, this already implies some approximation. Next, the transfer of discrete values, rather than the original ones, implies a smoothing of the distribution. Additionally, the discrete distribution needs to be quite coarse. Otherwise, we will end up with a number of cells, which are so large that the computer time would again be very large. Note, however, that in this case many cells would be empty, which would in part reduce the computational time. The report by Athanassoulis and Belibassakis (1998) gives some practical examples. As the "best" solution depends on the specific problem, on the geometry of the area, the local bathymetry, on the wave climate offshore and obviously on the requirements of the user, we are presently considering the possibility to offer to the user several approaches, with guidelines as far as the implications in terms of accuracy and computer time.

Various wave statistics will be made available to the user. We show here some examples of univariate and bivariate wave statistics for the input offshore conditions and the output target location.

National Technical
University of Athens

Oceanographic
Company of Norway

Istituto Studio
Dinamica Grandi Masse

This site is best viewed with Internet Explorer 4+ or Netscape Navigator 4+. Javascript must be enabled for the navigation menu to operate. Please note there may be linked pages also under menu items having sub-menus.

[javamenu.htm]