Modelling and mitigation of real-time sea level measurement over the coastal area of Japan

This study investigates the real-time sea-level measurement from six tide gauge stations in the coastal area of Japan. Initially, the six co-located global navigation satellite station sites vertical velocity has been observed followed by estimation of the relative tide gauge velocities. The results showed that the estimated relative sea-level changes abruptly at different sites depending on the location of tide gauge stations. To accurately analyse the regional tide gauge patterns at each site, a continuous wavelet transforms based on Morse wavelet as a default transformation has been applied. Analysis showed that the peaks of tide gauges have not occurred at a static time at each site, rather they were repeating with the variation of time. The width of the spectrum peaks varied with different sites and the overall spectrum frequency was not unique. These characteristics variation with time may be due to the variables that affect the stability of the sea level measurements. Finally, the kernel extreme learning machine (KELM) approach based on variational mode decomposition is used to reconstruct the tide gauge measurements. The correlation coefficients between observed and reconstructed data sets are very high varying from 0.91 to 0.96. The estimated root mean square error between two data sets is varying from 0.04 to 0.06 m at the entire region. This means that the KELM reconstructed tide gauge showed good agreement with the observed tidal waveforms. Thus, it can be concluded that the proposed techniques can be implemented for effective real-time sea-level monitoring changes with the available data.