Fast gap filling of the coastal ocean surface current in the seas around Taiwan

A fast, recently developed, Least Square regression method based on Discrete Cosine Transform (DCT-PLS) algorithm has been adapted to the mapping of hourly High Frequency Radar (HFR) data in the seas surrounding Taiwan. HFR is a shore based remote sensing system, and can be subject to unexpected observation failure. This algorithm produces both solution and error estimates of oceanographic data. The method explicitly uses both space and time information to predict missing values. In contrast to previous methods, our approach uses all HFR measurements to provide estimation error statistics while permitting long-range correlations, while allowing arbitrary HFR measurement locations. The approach is demonstrated by reconstructing the Hourly HFR data with a spatial resolution of 9km in the Taiwanese seas. We validated the method during the summer 2015 against typical gap scenarios. A major advantage of the approach is the ability to perform fast and robust computation while requiring a small amount of memory storage, showing the feasibility of a real-time application for filling HFR missing data.