It was a through-the-looking-glass moment for Chris Goldfinger, sitting in a meeting about Sumatran earthquakes on a recent Friday afternoon in Chiba, Japan, on the outskirts of Tokyo. The floor started heaving as if a switch flipped. That terrible shaking turned out to be the magnitude 9.0 Sendai temblor, tsunami-maker and devastator.

"We felt pretty safe," says Goldfinger, director of Oregon State University's Active Tectonics and Seafloor Mapping Lab, "but, oddly, still had time to run outside and ride through four or five minutes of mainshock. That was a very long time for the Earth to feel like the ocean."

At that point the ocean itself was already rearing up from its cracked floor to drown the coast 290 kilometers to the northeast. One month later, Japan is still in crisis.

Even so, one finer point in the wake of that horrible day will be what science can do to improve the odds—to give better, faster tsunami warnings. For Goldfinger—still rattled in Chiba—warning systems are a bill of goods, a chimera. "The earthquake is the warning," he says, describing a "near-field" event like Sendai where the temblor is very close to the coast. "Warning systems have been greatly oversold by those who created them."

One such creator, physicist Jörn Lauterjung, disagrees and draws a different lesson. Japan has decades of planning experience and an established early warning system. On March 11, a tsunami warning went out within three minutes to the three most-affected provinces, providing about 10 minutes to react. Instead of 25,000 dead and missing, with waves humbling nine-meter high seawalls, Lauterjung figures it all could've been much worse, although he and others even see room for new research, for improvement.

A German in Indonesia
Lauterjung's reference point is the 2004 Indian Ocean tsunami which hit Indonesia and southern Asia. Some 250,000 people died that day. There was no early warning.

After that event the German government in Berlin pledged $60 million to build such a system: engineering, data processing and geologic experience were readily available. Lauterjung's team at the Helmholtz Association's Research Center for Geosciences (GFZ) worked with a dozen other German science labs, private tech companies and international research institutions on the new system. It went online in 2008. Researchers have since been tweaking and optimizing it as well as educating system users, operators and the local population. The Indonesians are now poised to take over: Lauterjung was in Jakarta a couple weeks ago handing off the keys to the network.

The volcanic, earthquake-prone Sunda Arc, which forms the Indonesian islands of Sumatra and Java, lies along the boundary of two eastern Eurasian tectonic plates, putting Indonesia at high risk for "near-field" tsunamis like the one at Sendai. In order to issue an early warning, a precipitating earthquake must first be detected; alarm data is then transmitted to the Jakarta main warning center in the Indonesian capital. The resulting wave height and arrival time is determined, evaluated and retransmitted to a wide variety of potentially affected locations along the rough jungle coast of Sumatra. Warnings must be accurate—every false alarm erodes confidence in the system—and very rapid.

The baseline goal is to issue initial warnings and information to the public within five to 10 minutes of detecting risk. Even then, the local population must be notified and know how to respond. It is a tall order.

Hardware and software
The system has a set of main components and one master: time. The GFZ design connects four basic units, including seismic stations; a buoy and pressure-plate system to monitor wave motion; a set of coastal tide gauges; and a real-time GPS lattice network overlaying it all.

The German engineers drew on a 60-year history of tsunami early warning research dating to the National Oceanic and Atmospheric Administration's Pacific Tsunami Warning Center,. The new system incorporates many recognizable elements from the legacy NOAA and Japanese systems, but also features new research and design elements and integrated, newly designed software and hardware.

The Japanese early warning system is considered by researchers to be the most advanced in the world in its scale and distribution. The basic tools are seismometers, tidal gauges, pressure sensors and waterborne buoys—instruments on which the Indonesian system is based. It also links a new earthquake alert circuit that went online in 2007—connecting more than 1,000 land-based and ocean-bottom seismometers under the aegis of the Japan Meteorological Agency—with Pacific-wide tsunami alert coordinating centers in Japan, Alaska and Hawaii. The GFZ-designed system, in comparison, is a bit more compact, more connected to GPS monitoring, and integrated with newly developed software. This SeisComP 3.0 software analyzes seismological and sensor data, and is linked to a local grid of tsunami wave data models. It is all designed to make the system faster and more responsive to the near-field tsunamis that threaten Indonesia.

In a near-field event the most crucial element is the seismic network, Lauterjung says. Each node is equipped with a seismometer recording ground motion over a wide range of frequencies and along three axes. The stations are linked to SeisComP 3.0, which uses the inbound data to calculate earthquake location and magnitude. Measurements from the whole system are matched with a database holding thousands of modeled tsunami simulations calculated across the Sunda Arc: some 3,500 now, with another 7,000 in the works.

Not every earthquake triggers a tsunami—it takes a vertical shock to transfer energy to a water column. The GFZ design accounts for this distinction by including GPS units in the seismic stations. These concrete-mounted units transmit data continuously and align with geodetic stationary, or base, points, allowing observers to calculate nearby deformation in Earth's crust.

Offshore, 4.5-meter steel buoys are attached to pressure plates on the ocean floor set with instruments: an acoustic modem for transmission, another seismometer for the seafloor, and gauges which detect sudden changes in water pressure above. The pressure plates can measure changes in sea level within an accuracy of several millimeters. Also, a GPS-referenced data stream from the buoy, overlaid with an algorithm and mathematical filtering, calculates sea-surface heights by measuring the buoy's movement compared with a set of land-based reference stations.

Last but not least are the tidal gauges measuring sea-level changes off the beach. The stations look like orange boxes sitting on dock pilings, and comprise two component sets: One is a set of three different kind of tide gauge sensors, including a radar hanging like a lure at the end of a rod. The other contains data processors and—again—an integrated GPS antenna, which monitors the stability and displacement of the station during earthquakes. Onboard algorithms for the sensors go into motion if there is a sudden tidal outrush and indicate both the timing of a potential tsunami arrival and its relative strength.

The satellite-linked network now stretches like a veil over the coastal face of Sumatra and its smaller islands: 10 GFZ tidal stations dot the coastline, along with 21 seismometers, 10 buoy systems and 20 GPS stations. The Indonesians themselves ordered another 105 seismic stations in addition to the 15 donated by Japan, 10 from China and several from France. The upshot, Lauterjung says, is that Indonesia now has one of the densest seismic networks in the world.

The quake is the warning
And still, time punishes. There wasn't enough of it when a deadly wave careened in October 2010 toward the Indonesian island of Mentawai, a surfer's paradise off the coast of Padang. Alarm data streamed to the Jakarta main warning center. In seconds officials evaluated the results as they filtered through the system algorithms and models, and sent warnings to the Mentawai islands, police, local responders, and TV and radio broadcasters.

But the tsunami was on Mentawai in minutes and 400 people died. Local officials bitterly complained that the warning systems must have been vandalized or switched off—that people should have gotten a warning and did not. "That's wrong," Lauterjung says. Data shows a warning was sent out from the system within four minutes of the quake and broadcast on local radio and TV stations after seven minutes.

Sadly, that's just when the wave hit. As Goldfinger says, sometimes the quake is the warning.

Lauterjung says Mentawai presents the limit of early warning in terms of time. "And there is a limit, for the techniques these days," he says. Lauterjung himself and other researchers point out that science and technology aren't enough—much depends on behavior and luck; on not going back to the house to fetch family photographs; on taking bicycles rather than cars to higher ground to avoid traffic jams; on better education. Even little things like paying attention to animal behavior can help. (Even though they're not part of a warning system, some animals are sensitive to the low-frequency waves issued by earthquake and tsunami.)

He and other researchers also think there's still more to be done in making warning systems better. Georgia Institute of Technology geophysicist Andrew Newman, who is testing new early warning data-processing algorithms, expects improvements down the road in rapidly assessing earthquake source parameters, faster distribution of information, more precision in pinning down earthquake depth, rupture area, and the extent of slippage to determine seafloor displacement. GFZ researchers are also pursuing more experimental avenues for improvement, such as using satellite reflectometry and radar to monitor oceanic trenches for near-field tsunamis. Humboldt State University tsunami and earthquake expert Lori Dengler says it's a cost-benefit problem: What does shaving a minute or two off the warning time cost, and how much benefit does it lend?

Less esoteric (or more whimsical) calculations are in the mix as well: Lauterjung says he's seen families store balloon-rigged escape pods in their backyards, like those used to escape oil-rig emergencies. In case of flood or tsunami, a family could get in the pod and bob along the surface of the water, like James Bond at the end of The Spy Who Loved Me.

Everything has limits, Lauterjung says. "We cannot avoid every victim. The aim is to reduce them," he says. "There is no measure of how many you save."