You highlight, in a News story, the controversy over the reliance of tsunami early-warning systems on model-based forecasts (Nature 464, 14–15; 2010). Important advances have been made in near-real-time model-based tsunami forecasting since 2004, by scientists such as those at the US Pacific Marine Environmental Laboratory, the German Research Centre for Geosciences and the European Commission's scientific and technical reference arm, the Joint Research Centre (JRC).

These advances make the models important contributors to tsunami-warning systems. The Intergovernmental Oceanographic Commission (IOC) of the United Nations Educational, Social and Cultural Organization can help to guide the optimal use of such forecasts to improve the effectiveness of existing early-warning systems.

The JRC has been developing a global tsunami-modelling system, which is integrated into the joint United Nations/European Commission Global Disaster Alert and Coordination System (GDACS; http://www.gdacs.org). It is based on a global pre-event calculated tsunami scenario grid and near-real-time simulated forecasts. This combined approach gives an almost instantaneous estimate of wave height and arrival times, as well as enabling a warning to be issued within 20 minutes of an earthquake-inducing tsunami.

The system worked effectively after the Chilean earthquake, with alerts transmitted via SMS and e-mail to some 10,000 GDACS users within 30 minutes; the JRC accurately predicted the height of the tsunami near Concepcion. The risk for the Pacific islands was considered low because of the wave's conditions at the source, with a sea-level rise of less than a metre expected in Hawaii.

The Portuguese Institute of Meteorology is already testing the reliability of forecasts with an early-warning system that will provide a framework for the IOC's future northeast Atlantic regional centre.

It is time for forecasting models to be tested systematically in operational conditions by the tsunami-warning centres. These will complement the arrays of tidal and deep-sea gauges by filling information gaps where gauges are scarce, to minimize false alarms. The challenge is to identify how to formalize the integration of model-based forecasts into the process of issuing warnings.