On the Application of Machine Learning Techniques to Regression Problems in Sea Level Studies

• Long sea level records with high temporal resolution are of paramount importance for future coastal protection and adaptation plans. Here we discuss the application of machine learning techniques to some regression problems commonly encountered when analyzing such time series. The performance of artificial neural networks is compared with that of multiple linear regression models on sea level data from the Swedish coast. The neural networks are found to be superior when local sea level forcing is used together with remote sea level forcing and meteorological forcing, whereas the linear models and the neural networks show similar performance when local sea level forcing is excluded. The overall performance of the machine learning algorithms is good, often surpassing that of the much more computationally costly numerical ocean models used at our institute.

Ämnesord

NATURVETENSKAP  -- Geovetenskap och miljövetenskap -- Meteorologi och atmosfärforskning (hsv//swe)

NATURAL SCIENCES  -- Earth and Related Environmental Sciences -- Meteorology and Atmospheric Sciences (hsv//eng)

Nyckelord

Europe

Neural networks

Time series

Ocean models

Regional models

empirical orthogonal functions

neural-networks

extreme sea

north-sea

model

variability

rise

tide

Engineering

Meteorology & Atmospheric Sciences

Publikations- och innehållstyp

ref (ämneskategori)

art (ämneskategori)