How scientists predict tsunamis.
How scientists predict tsunamis.
Answers to your questions about the news.
Sept. 30 2009 4:44 PM

My City Is Under a Tsunami Advisory. Should I Run for My Life?

No. But you might want to avoid the beach.

Damage from the tsunami. Click image to expand.
Damage from the tsunami in Samoa

An earthquake under the Pacific Ocean on Tuesday morning caused a tsunami to rip through the islands of Samoa and American Samoa, killing dozens. Shortly thereafter, the Pacific Tsunami Warning Center issued a warning for the Pacific islands and New Zealand, a watch for Hawaii, and an advisory for the Pacific Coast of the United States. How do scientists decide when to issue an alert?

Ocean-based sensors. In the moments after an earthquake, seismologists consult pressure sensors on the ocean floor to determine whether there has been an increase in the height of the column of water directly above—a change that indicates a tsunami. They then consult topographical maps to predict, based on historical data and mathematical models, how quickly and powerfully the tsunami might impact various parts of the world. Seismologists issue a "warning" if a tsunami has been detected and is likely to head for a given landmass. Evacuation and repositioning of ships are usually ordered. A "watch" means a tsunami has been detected far from the landmass, and forecasters are tracking the wave to see if it has the potential to cause damage. Evacuation is put on hold until more information is gathered. An "advisory" means that damage is likely to be minimal and minor actions like beach closures are advisable.


Up until five years ago, tsunami forecasters relied mainly on data from land-based seismic stations, making predictions based on the magnitude of the quake and its location. This method was highly imprecise. It's tricky to determine the magnitude of a quake in a short period of time—tsunami forecasters have to make a prediction within 10 minutes—because a seismic wave has high and low points, and it's tough to know which is which in the first moments after the shock. And some sea-based quakes—even very large ones—don't cause tsunamis at all.

After the 2004 tsunami took more than 200,000 lives, the federal government invested heavily in detection technology—including the new ocean-based pressure sensors. These sensors, however, are far from perfect. Several mathematically complex calculations make predicting a tsunami's far-flung impacts challenging. Tsunamis tend to spread out and diminish as they move outward from the epicenter, but the waves can converge and multiply in strength at very distant points. (The antipode to the epicenter—located 180 degrees around the globe—is at particular risk of this additive effect, because waves can converge on it from many directions simultaneously.) A 1960 tsunami that originated near Chile killed 122 people in Japan. The 2004 tsunami had a surprisingly strong impact in Somalia. The shape of the shoreline can also complicate the forecast. When small tsunamis strike oddly shaped harbors, the frequency of the waves—usually somewhere between five and 60 minutes—can resonate with the shore, causing unpredictable and significant increases in wave amplitude.

Although relatively inexpensive themselves at about $400,000 each, ocean-based sensors are costly to maintain. The batteries run out every few years, and a ship must be dispatched to haul the sensors up from the ocean floor for replacements. Also, vandals frequently swipe the solar panels off the buoys that receive the sensors' data. But since tsunamis spread out quickly in many directions, we need very few—about 40—all of which are located in the Pacific Ocean.

Got a question about today's news? Ask the Explainer.

Explainer thanks Emile A. Okal of Northwestern University.

AP Video: Tsunami in Samoa

Brian Palmer writes about science, medicine, and the environment for Slate and Earthwire. Email him at Follow him on Twitter.

  Slate Plus
Slate Plus
Oct. 9 2015 3:29 PM Introducing the Best of Slate Podcasts Only have time for one podcast this week? Make it this one.