Principal Component Analysis Of Tsunami Buoy Record: Tide Prediction And Removal

摘要

Principal component or Empirical Orthogonal Function (EOF) analysis is applied to tsunameter records by treating them as two-dimensional signals, where the second dimension is created by breaking a single time series into cycles and treating the cycle number as a second dimension. Under certain conditions, principal components calculated from different records are shown to determine the same functional space. Signal decomposition into pre-calculated principal components is used to predict or extract the tidal component of a record. This work shows that EOF processing allows for short-term tidal predictions at tsunami buoy locations with the precision of more advanced methods and with minimal a priori knowledge about tidal dynamics. Also shown is that filtering in EOF domain is sensitive to the non-tidal component of a record and therefore presents a tool for early tsunami detection and quantification.