A new hybrid model for filling gaps and forecast in sea level: application to the eastern English Channel and the North Atlantic Sea (western France)

This research is carried out in the framework of the program Surface Water and Ocean Topography (SWOT) which is a partnership between NASA and CNES. Here, a new hybrid model is implemented for filling gaps and forecasting the hourly sea level variability by combining classical harmonic analyses to high statistical methods to reproduce the deterministic and stochastic processes, respectively. After simulating the mean trend sea level and astronomical tides, the nontidal residual surges are investigated using an autoregressive moving average (ARMA) methods by two ways: (1) applying a purely statistical approach and (2) introducing the SLP in ARMA as a main physical process driving the residual sea level. The new hybrid model is applied to the western Atlantic sea and the eastern English Channel. Using ARMA model and considering the SLP, results show that the hourly sea level observations of gauges with are well reproduced with a root mean square error (RMSE) ranging between 4.5 and 7 cm for 1 to 30 days of gaps and an explained variance more than 80 %. For larger gaps of months, the RMSE reaches 9 cm. The negative and the positive extreme values of sea levels are also well reproduced with a mean explained variance between 70 and 85 %. The statistical behavior of 1-year modeled residual components shows good agreements with observations. The frequency analysis using the discrete wavelet transform illustrate strong correlations between observed and modeled energy spectrum and the bands of variability. Accordingly, the proposed model presents a coherent, simple, and easy tool to estimate the total sea level at timescales from days to months. The ARMA model seems to be more promising for filling gaps and estimating the sea level at larger scales of years by introducing more physical processes driving its stochastic variability.