Item

Abstract

Sporadic misses in the tide gauge records occur frequently for a number of reasons, which hinders understanding of ocean physics in coastal regions. In other words, general-purpose time-series analysis and prediction methods require continuity of the data. To fill the sea-level data gaps, scientists have been working hard, but the existing reconstruction techniques have a disadvantage when missed sea-level records are longer than the timescales of coastal processes. To solve this problem, artificial neural network model is also being used to fill the gaps, but this method has a chronic problem that the shape of the target is fixed. To overcome this obstacle, we designed a model, named one-step prediction operator, which predicts the sea level after a unit of time. A data assimilation technique is additionally applied to merge seamlessly the model-predicted sea level with observed one. The recursivity of this model makes it possible to reconstruct missing data even longer than 72 hours successfully. The reconstructability of sea level records at 16 tide gauge stations around the Korean peninsula confirms that it can successfully reconstruct missing values with root-mean-squared errors of 0.5–1.3 cm on average.