Residue-Correction-Based Data Assimilation In Coastal Hydrodynamics (With An Application To Singapore Regional Model)

Abstract

Singapore Regional Model was developed to predict the water motion in Singapore Straits. However, its accuracy is limited by various factors such as the simplifying assumptions, complex ocean geometry and so on. Residual correction, as one strategy of Data assimilation, can extract the information from the observation to correct the numerical model output directly. At measured stations, the influence of a prior estimate was examined through the method of time lagged recurrent network (TLRN). Besides, a modified local model (MLM) was developed based on chaos theory to yield more stable results over the long horizons. These predicted residues were then distributed spatially to non-measured stations. An Approximated Ordinary Kriging (AOK) which is particularly suited to scenarios with only sparse sample data was resorted to. Then both the space and time lags were taken into consideration in its implementation (ASTOK). In addition to Kriging, Two-sample Kalman filter and Unscented Kalman filter were also applied for comparison. The combined use of MLM and ASTOK was found to be effective in improving the prediction efficacy of SRM, with high computational efficiency.