Abstract Coastal sea level reflects not only global trends but also complex regional and local processes acting across multiple temporal and spatial scales, which are often missed by large‐scale climate models. To address this, we use a data‐driven approach to examine potential drivers of interannual to decadal sea level variability in northern Europe. We train neural network and linear regression models to simulate monthly mean sea level from 45 tide gauges using 11 potential drivers and quantify their contributions using permutation feature importance. To include possible lag or memory in the system, models include forcing history. Models explain approximately 70% of observed variability, except in the Danish Straits, where lower skill (25%–50%) suggests missing drivers. In the Baltic neural networks perform best, revealing highly nonlinear relationships between sea level and its drivers, while in other areas linear regression works better, indicating predominantly linear driving mechanisms. Most locations reveal a memory of at least one previous month, often longer. Primary sea level drivers are local wind and atmospheric pressure, followed by the North Atlantic Oscillation and sea surface temperature (both local and globally averaged), with minor influence from precipitation. Greenland and Antarctic mass loss, as well as regional runoff and evaporation do not affect sea level variability on these time scales. Regional analysis reveals clear spatial patterns, with different driving mechanisms in the North, Baltic, and Norwegian Seas, and the Danish Straits. These findings enhance our understanding of regional sea level variability and offer additional tools for improving coastal flood risk assessments.
Poropat et al. (Mon,) studied this question.