New Algorithm Improves Tsunami Warning Systems

By Lina Sorg

According to the United Nations Office for Disaster Risk Reduction, tsunamis kill approximately 8,000 people each year. While some are relatively minor, others are particularly devastating; at least 227,000 people died from the tsunami triggered by an earthquake in the Indian Ocean on December 26, 2004. Because of the rapidity and volatility with which tsunamis form, advanced warning for coastal towns is of utmost importance.

Current tsunami warning systems rely on localized scenarios that are based on previous patterns and unique to each area. Sensors on the ocean floor detect abnormal tectonic plate movement (as in the case of an earthquake), but cannot accurately predict the size, force, and site of the resulting waves. If a developing tsunami does not follow prior wave patterns, it could catch inhabitants of coastal areas completely by surprise.

Scientists at the Australian National University (ANU) are working to increase the predictability of tsunamis and thus improve warning systems. They created a Time Reverse Imaging (TRI) algorithm and applied it to tsunami waveform data. The algorithm utilizes real-time data from ocean sensors to mathematically reconstruct the tsunami as it appeared upon origination, and uses initial sea surface displacement to predict its trajectory. Identifying its source can then accurately predict the location and force of an impending strike.

The researchers studied the movement of tectonic plates in the Japan Trench, part of the Pacific Ring of Fire and prone to the instability that causes tsunamis. They also looked specifically at the Tohoku-Oki earthquake and tsunami, which hit Japan on March 11, 2011. After dividing the tsunami’s source region into a grid of ‘point’ sources, the TRI algorithm estimated the initial sea surface displacement at each point on the grid. Combing this information with data from ocean sensors allowed ANU seismologist Jan Dettmer to work backward in time mathematically and calculate what the 2011 tsunami would look like once it hit land. Dettmer refined the algorithm by checking his results against the tsunami’s actual trajectory.

Once an earthquake occurs and plates shift drastically in the ocean, there is very little time – sometimes only minutes – before a tsunami may strike, says Dettmer. The TRI method brings researchers closer to accurately predicting a building tsunami’s path and level of destruction without sacrificing speed. Read more of Dettmer’s thoughts here.

Knowledge of initial sea surface displacement is necessary to predict a tsunami’s threat level and effect on the coast. Future tsunami warning systems will likely be based on real-time data, because incorporating such data eliminates the guesswork that accompanies current predictive systems. Dettmer plans to further hone the algorithm by testing his method on other documented earthquakes, and estimates that it could be ready for use in under five years.

Click here to read the corresponding journal paper.

Lina Sorg is the associate editor of SIAM News.