Day 5

Detailed paper information

Back to list

Paper title 24 Years Sea Level Trend of Baltic Sea Using Resampled Machine Learning Satellite Altimetry Data
Authors
  1. Majid Mostafavi Tallinn University of Technology Speaker
  2. Nicole Delpeche Ellmann Tallinn University of Technology
  3. Artu Ellmann Tallinn University of Technology
Form of presentation Poster
Topics
  • C1. AI and Data Analytics
    • C1.06 Data assimilation and machine learning for the Earth system
Abstract text Satellite altimetry (SA) provides spatial and temporal sea level measurements with respect to an earth-fixed geocentric reference frame. There are however some challenges using the data to retrieve spatio-temporal distribution of sea levels. One of the major constraints is the lack of measured data due to the satellite revisiting time (cycles). This could be overcome by extending SA data temporarily (resampling) to find a continuous trend of sea level. To do this, multiple time series analysis models have been used where the periodic terms and linear trends of sea level variations are fitted and the data gap of SA resampled. The sea level raise trend is significant at the location of studied TGs along the coast in Baltic Sea.

In this study the 24 years SA data are resampled to restive more precise sea level trend in the Baltic Sea. The SA data are compiled by eight SA missions including ERS-2, Envisat, SARAL, Jason-1, Jason-2, Jason-3, Sentinel-3A and Sentinel-3B at a particular radius near each tide gauge. The SA data are resampled by machine learning methods including Autoregression (AR), Moving Average (MA), Autoregressive Moving Average (ARMA), Autoregressive Integrated Moving Average (ARIMA), Seasonal Autoregressive Integrated Moving-Average (SARIMA), GARCH) Generalized Autoregressive Conditional Heteroscedastic (GARCH), etc., to fill the gaps and validated by comparison with tide gauges (TG) observations. The sea level trend at each tide gauge computed from satellite altimetry and tide gauges during 1995 to 2019. This methodology also requires the utilization of high-resolution geoid models to maximize the opportunities in deriving realistic sea level using SA data. Also, careful spatial data selection and outliers’ removal using data screening is prerequisite.