Drivers of sea level variability using neural networks

Abstract

Understanding the forcing of regional sea level variability is crucial as many people all over the world live along the coasts and are endangered by extreme sea levels and the global sea level rise. The adding of fresh water into the oceans due to melting of the Earth’s land ice together with thermosteric changes has led to a rise of the global mean sea level with an accelerating rate during the twentieth century. However, this change varies spatially and the dynamics behind what forces sea level variability on a regional to local scale are still less known, especially the high-frequency variability causing extreme sea levels. Finding a straightforward approach to understand the dynamics behind local variations is beneficial for decision makers who want to mitigate and adapt with appropriate strategies. Here I present a novel approach using machine learning to identify the dynamics and determine the most prominent drivers forcing coastal high-frequency sea level variability. I use a recurrent neural network called Long Short-Term Memory (LSTM) network, with the ability of learning data in sequences and capable of storing long memory and finding temporal dependencies in the data. As input data in the model I use hourly ERA5 10-m wind, mean sea level pressure, sea surface temperature, evaporation and precipitation data between 2009-2017 in the North Sea region. I use data from the entire North Sea basin, to be able to capture the larger climatic patterns forcing the sea level variability. The target data in the model are hourly in-situ sea level observations from West-Terschelling in the Netherlands. My results show that the dominant pressure pattern over the North Sea, coupled with a zonal wind pattern, is the most important driver of high-frequency sea level variability in my location of interest. The model also found a strong relationship between the sea level variability and another zonal wind pattern that could not be detected by classical correlation analysis, which indicates that the LSTM network has the ability to capture more complex relationships. This approach shows great potential and can easily be applied to any coastal zone and is thus very useful for a broad body of decision makers all over the world. Identifying the cause of local high-frequency sea level variability will also enable the ability of producing better models for future predictions, which is of great importance and interest.