Early warning system for non-seismic gravity driven tsunamis

Gravity-driven mass flows represent a significant near-field hazard in volcanic environments and, increasingly, in coastal regions affected by climate-driven slope instability. Despite their destructive potential, quantitative assessments of real-time detection systems operating under natural conditions remain scarce. Here we present the performance and sensitivity of the tsunami early-warning system currently operating at Stromboli Volcano (Italy), based on offshore pressure gauges and a real-time detection algorithm. Sixteen years of sea-level records (2010–2026), including 63 tsunamis events ranging from 0.02 to 2 m height, show that the system issues alert within less than 22 s from the tsunami onset and tens of seconds before its maximum amplitude, with no false alarms. A sensitivity analysis, based on Peak-Over-Threshold statistics and extreme wave modeling, demonstrates that the detection capability remains effective even under extreme sea state conditions also for tsunami with height as low as 4 m. To support regional tsunami warning and reduce the alerts for tsunami with no inundation effects on a distant (> 10 km) coasts, we apply real-time Ensemble Empirical Mode Decomposition (EEMD), which successfully extracts tsunami signals and preserves waveform characteristics. EEMD also isolates longer period sea level oscillations associated with atmospheric pressure waves within the meteotsunami frequency band, demonstrating the applicability of the method to multiple classes of non-seismic tsunamis. This work provides the first long-term quantitative evaluation of an operational near-field early-warning system for volcanogenic tsunamis and proposes a transferable multi-tier framework for landslide-tsunami and meteotsunami hazard mitigation for coastal environments worldwide.