Tsunamis can hit coastal societies very shortly after large earthquakes. To mitigate their impact, tsunami early warning systems can alert the coastal population and civil authorities. This may allow evacuation, and mobilization of emergency services. Immediately after the earthquake notification, the destructive potential of an associated tsunami is highly uncertain. At the start, there is much uncertainty about the exact size, the exact location, and the exact mechanism of the earthquake and generation of the tsunami. This uncertainty means that tsunami early warning systems can occasionally issue false alarms, or may even fail to warn of an actual tsunami.

Read the article published in Nature Communications here.

Current tsunami early warning systems provide single-outcome forecasts that can't account for this uncertainty. For the sake of safety, they will often overestimate the tsunami impact. This can however result in too many false warnings. The opposite could happen with a badly rigged system, with too many missed events.

Managing the risk of false alarms and missed alarms lies in the political sphere. However, decision-makers need to be informed by tsunami warning systems that accurately forecast the uncertainty. The PTF resembles modern numerical weather forecasts work by calculating multiple simulations – an ensemble – each with a slightly different starting point and slightly different model parameters. PTF works by simulating vast numbers of earthquake-tsunami scenarios, covering the full range of possible sources.

The principle is shown below. Using average forecasts, or incorporating a small part of the uncertainty, fewer false alarms are issued. However, the risk of missing events increases. Incorporating a very large uncertainty range, the missed alarms are almost eliminated. However, we thereby risk more false alarms. While the optimal solution may lie between these extremes, this selection process needs to be done through political choices by the responsible stakeholders. The PTF method provides a basis for this optimization, and links the optimal solution to the desired trade-off between false alarms, missed events, and correct warnings.


1) Istituto Nazionale di Geofisica e Vulcanologia, Le Grazie, 2) Italy, Department of Physics “Ettore Pancini”, University of Naples, Naples, Italy, 3) German Research Centre for Geosciences (GFZ), Potsdam, Germany, 4) Norwegian Geotechnical Institute (NGI), Oslo, Norway, 5) Grupo EDANYA, Universidad de Málaga, Málaga, Spain.

Norges Geotekniske Institutt (NGI) er et internasjonalt ledende senter for forskning og rådgivning innen ingeniørrelaterte geofag. Vi utvikler optimale løsninger for samfunnet og tilbyr ekspertise om jord, berg og snø og deres påvirkning på miljøet, konstruksjoner og anlegg. Vi arbeider i følgende markeder: Offshore energi - Bygg, anlegg og samferdsel - Naturfare - Miljøteknologi. NGI er en privat næringsdrivende stiftelse med kontor og laboratorier i Oslo, avdelingskontor i Trondheim og datterselskap i Houston, Texas USA og Perth, Western Australia. NGI ble stiftet i 1953.

Probabilistic tsunami forecasting for early warning