In recent years, application and evaluation of the efficiency of different data assimilation methods has been a subject of interest in both wave hindcasting and forecasting systems. The main goal of the current study is to assess the efficiency of an ensemble Kalman filter (EnKF) data assimilation scheme in improving the wave simulation results in Persian Gulf. The so called region plays an important role in the oil and gas industry due to its Geographical and Morphological location and housing a large number of offshore platforms.
A third generation wave model, SWAN, was employed in order to simulate the wave fields in the region. The three hours updated ECMWF wind data were used as the main driving force. The OpenDA toolbox, especially developed for efficient data assimilation purposes, was employed to smooth the chaotic nature of the non-linear wave simulation scheme. The OpenDA utilizes a number of methods that are based on Kalman filter algorithm but do not require the amount of computation efforts that are incurred by the classical filter algorithm. The EnKF is a variant of Kalman filter, where probability density function of a model state is represented by an ensemble of the model state.
Two sets of records for significant wave heights and peak wave periods were used in the analysis process with EnKF to estimate the error covariance matrix. At analysis time, the forecast error covariance was computed by using the model forecasts ensembles. In overall and for the wave climate modeling, the initial conditions of the numerical model were updated using the improved system state, up to the current computing time level. This is achieved by incorporating the previous measurements into the Kalman filter algorithm. The model was then run into the future, driven by the new improved state conditions.
The statistical results and diagrams showed that applying EnKF scheme leads to a noticeable improvement in significant wave heights. However, the accuracy of this technique was subjected to the location and number of observation stations and also ensemble size. With larger ensembles, results of error covariance estimation are more accurate but there is a limitation due to execution time of process and efficiency of the computations.