Synoptic index-based model for reconstructing high-frequency sea level oscillations in the Mediterranean

High-frequency sea level oscillations are gaining prominence in sea level research, as advancements in technology and data collection allowed high-resolution records. Their extreme manifestations, often amplified by interactions with other strong oscillations, can trigger destructive flooding events worldwide, emphasizing the need for in-depth studies of such phenomena and the development of reliable predictive tools. To tackle this, the synoptic index-based model has been designed to reconstruct and predict extreme non-seismic sea level oscillations at tsunami timescales (NSLOTTs). Initially developed for the meteotsunami hotspot Ciutadella, the model was later extended globally, with the strongest synoptic index-NSLOTT correlations observed in the Mediterranean Sea, where NSLOTTs contribute up to 50% of the total sea-level range.

The baseline model, built using ERA5 reanalysis with synoptic variables previously identified as relevant for known NSLOTT hotspots, was subjected to modifications in its configuration in order to evaluate adaptability and robustness in forecasting and detecting extreme NSLOTT events. These modifications included testing alternative reanalysis products, different synoptic variables, and training/testing datasets. Additionally, the impact of changes in NSLOTT series—such as altered temporal resolution, amount of data gaps, and series length—was assessed. Results reveal that stations with higher baseline performance consistently maintain their skill across different model configurations, though their performance variability is greater compared to stations with lower baseline performance. Stations along the eastern Adriatic Sea exhibited the highest performance, highlighting the suitability of the model for this region of Mediterranean. Overall, the model demonstrates higher success in forecasting extreme events than in their detection. These findings offer valuable insights for optimizing model configurations and enhancing predictive capabilities, with the ultimate goal of developing reliable tools for forecasting extreme events, and consequently contributing to coastal hazard and flooding mitigation.