The application of Data Assimilation (DA) techniques in ocean wave modeling using the WAVEWATCH III model and the observation timeseries data of significant wave height at fixed buoy stations was evaluated around Korea coasts. Numerical experiments examined the impact of DA based on the location of the buoy station, whether nearshore or offshore, and compared the model performance with Optimal Interpolation (OI) and Ensemble Optimal Interpolation (EnOI) methods. Additionally, variations based on the influence radius were also investigated. Analysis showed that the OI method notably enhanced performance, especially nearshore, where DA's impact increased with the influence radius but diminished offshore
In ocean modeling, DA is recognized as a critical factor for enhancing model performance. This enhancement is achieved by improving the initial field, which leads to increased accuracy in both forecast predictions and reanalysis outcomes for hindcasting purposes (Bannister, 2017; Kwon et al; 2018; Moore et al., 2019). The effectiveness of DA in wave models has been a subject of study since the works of Thomas (1988) and Esteva (1988). However, the practical operational application of DA in wave models remains limited. For example, the first operational implementation of an ocean wave model at the Korea Meteorological Administration (KMA) was in 1992, with significant advancements occurring in the 1990s following the introduction of supercomputers. (Park et al, 2009). However, by the 2010s, the research began on DA with SWH in operational wave models. And according to Saulter et al(2020), only 4 of the 14 systems that provide for an international wave forecasts from the Word Meteorological Organization (WMO) by the European Centre for Medium-range Weather Forecasts (ECMWF) currently implement DA. This indicates a gap between the theoretical understanding and research on DA and its implementation in operational wave modeling system.
The DA techniques include 3DVAR, 4DVAR, Optimal Interpolation (OI), Ensemble Kalman Filter (EnKF), and Ensemble Optimal Interpolation (EnOI). And for regional wave model, EnKF (Siddons et al., 2009; Almeida et al., 2016;Fujiwara et al., 2019), EnOI and 3DVAR (Siddons et al., 2009), and 4DVAR (Orzech et al., 2013; Song and Mayerle, 2017) have been applied, mainly to the Simulating Waves Nearshore model (SWAN; Booij et al., 1999). In this study, an attempt was made to evaluate DA techniques through numerical experiments applying OI and EnOI methods using the Regional Wave model system (RWW3). This system is operated by the KMA and based on the WAVEWATCH III (Wavewatch III Development Group, 2019), wave model. In contrast to the OI method, where the Background Error Covariance (BEC) is static and uniformly distributed across space, the EnOI method utilizes a BEC that is dynamically calculated based on historical model data (Evensen, 2003). And the EnOI method allows for the incorporation of various types of observations in a single step (Oke et al., 2010). This study focuses on analyzing the performance of the wave model by examining the differences between these two DA methods. Additionally, instead of the typical approach of conducting spatially distributed satellite SWH for DA, this study assimilated timeseries buoy data from fixed points observed by the KMA around Korea coasts. Subsequently, model validation and model performance analysis were carried out based on the SWH in-situ data observed by other organizations, such as the Korea Hydrographic and Oceanographic Agency (KHOA) and the Ministry of Oceans and Fisheries (MOF)
