Buoy Data Assimilation in Nearshore Wave Modelling (Assimilatie van boeigegevens in de modellering van golven nabij de kust)
Buoy Data Assimilation in Nearshore Wave Modelling
Portilla, Jesus; M0122308
The study of wind waves deserves major attention at present time as they play a very important role in many human activities. Thanks to numerical modelling, at present wave estimates are available at global scale and with an outstan ding degree of accuracy. Previously, measurements have served mainly to verify and calibrate models, but more recently the abundance of in-situ and remote sensed data have motivated the incorporation of measurements into the models in a process known as data assimilation. The focus of this study is the assimilation of offshore buoy data in the nearshore wave modelling scenario, looking for improvements of wave hindcasts at coasta l locations. The model used is WAM, a state-of-the-art third generation wa ve model. The assimilation approach is the so called Optimal Interpolation (OI). This method has some advantages over other methods as it is rather straightforward to implement and it is computationally very efficient. H owever, two major limitations are inherent in this method. The first is the a-pr iori representation of the errors characteristics of the system, and the seco nd the retrieval of analyzed magnitudes into the wave spectrum. The confrontati on of these shortcomings has largely drawn the experimental line followed in t his study. For the former, it is recognized that the successful implementation of a ny assimilation system requires a proper knowledge of the spatial structure of the syste ms errors (model and observations). However, that knowledge is in general v ery poor and that is the case for the present study area. Therefore, differe nt parameterizations of the gain matrix were tested. The OI scheme was impl emented initially following a conventional approach where the errors of the syst em are considered homogeneous and isotropic. However, these assumptions were fo und to be restrictive in the nearshore. Therefore, experiments with other approximations introducing a number of anisotropic functions were carrie d out. Namely, a parametric 2D Gaussian function, a function based on long-term model o utput, and a function based on short-term model output. The advantage of the la st two is that they are more transferable to other locations. Nevertheless, the results obtained using the three approximations are similar. For the latter, from early data assimilation studies the distinction of different wave systems was believed to increase the robustness of the retrieval algorithm and p alliate this deficiency of the OI approach. In this study, some steps forward ha ve been done in this regard with the development of more robust partitioning and identification algorithms. It is pointed out that their practical implementation in the assimilation system is not trivial. The developmen t of a robust algorithm for cross-assigning model and buoy partitions needs fur ther investigation. From the results of the numerical experiments, the benefit of the assimi lation is evidenced by the capability of the system to correct some of the model deficiencie s. Especially, excessive energy dissipation could be compensated resulting not only in better estimates of mean wave parameters, but also in better spe ctral representations. The most evident enhancements were obtained in conditio ns where the model performance was low, namely, in conditions of moderated wind with the presence of swell. A remarkable aspect is the long duration effect of the assimilation. Cor rections prevail in the system for a time lag of the order of days while the travelling time of waves in the domain is of the order of hours. This effect is attributed to the spectral shape corrections, which act as improved initial conditions for the proc eeding wind events. Low frequency bands (swell) are better represented in the assimilation run than in the normal run. Through the quadruplet non-line ar interaction it is then easier to transfer energy input from wind to lowe r frequencies The simultaneous assimilation of wave height and period did not show an advantage over the assimilation of wave height only. One of the reasons is that by the assimilation of wave height only, the spectrum is subject to variation a nd wave period changes accordingly. However, other considerations should be take n into account for further research as the physical assumptions used for updati ng the spectrum and the simplifications imposed in the applied algorithm. The assimilation of wave height during wind-sea dominated periods did no t show a large impact. On the one hand, this is because in purely wind-sea conditions the model be tter reproduces waves. On the other hand, the performance of the assimilation scheme in those conditions is affected by the fact that a typical model underes timation of wave energy is lower at the coastal locations than at offshore locati ons. This produces conflicting assimilation requirements and lowers the impact of assimilation in wind-sea conditions.