A tsunami warning system is a chain that starts with a seismometer and ends with a person on a beach deciding whether to walk uphill, and the weakest link is almost never the detection end. This piece covers how the warning centers work out that a wave exists and how big it will be, what the four National Weather Service tsunami products actually mean, why a local tsunami and a distant tsunami are different operational problems, the natural warning signs that have to substitute for an official message when the wave arrives in fifteen minutes, when vertical evacuation is the only option left, and why the community education program is the part of the system that does the saving.

Detection: seismic first, water second

Every tsunami warning in the first minutes rests on an earthquake solution, because seismic waves travel through rock at kilometers per second while the tsunami crosses open ocean at the speed of a jetliner. Global and regional seismic networks give the warning centers a location, a depth and a first magnitude estimate within a few minutes, and from those three numbers an analyst can say whether a tsunami is physically plausible, since the destructive ocean-crossing events come overwhelmingly from shallow thrust faulting on subduction zone interfaces rather than from deep or strike-slip ruptures. The hard part is magnitude. The older surface-wave magnitude scales saturate for the largest earthquakes, meaning a magnitude 8.5 and a magnitude 9.2 can look similar in the first data, and the moment magnitude that actually scales with seafloor displacement comes from long-period analysis, including the W-phase methods now in routine use, which takes longer to converge.

That lag has consequences on the record. When the magnitude 9.0 to 9.1 Tohoku earthquake struck on March 11, 2011, the Japan Meteorological Agency issued its first tsunami warning roughly three minutes after the rupture, working from an initial magnitude estimate near 7.9, and the forecast wave heights in that first message were raised substantially in later bulletins as more data arrived. JMA has published its own review of that performance and has since changed how it handles magnitude uncertainty in early warnings. The operational lesson for anyone downstream is that a first bulletin is a first bulletin, that the numbers in it are expected to be revised, and that a revision upward is a normal part of the process rather than evidence that someone got it wrong.

Water confirmation comes second. NOAA’s Deep-ocean Assessment and Reporting of Tsunamis stations, the DART buoys, pair a bottom pressure recorder with a surface buoy so that a passing tsunami wave a few centimeters high in deep water is detected and reported by satellite, and coastal tide gauges add the nearshore picture. NOAA expanded the DART array considerably after 2004, and the current station list and operating status are published through the National Data Buoy Center, which is where to look rather than relying on any count in an article. The centers feed those observations into pre-computed propagation models to tighten the forecast for specific coastlines, which is how a distant-source warning gets refined from a basin-wide possibility into arrival times and expected amplitudes by community. For a near-field event, the water data usually arrives after the wave has already come ashore.

The warning centers and the four messages

Two National Weather Service centers do this work for the United States. The Pacific Tsunami Warning Center at Ewa Beach, Hawaii traces back to the warning system established at the Honolulu Observatory in 1949 after the April 1, 1946 Aleutian earthquake sent a tsunami into Hilo, an event for which about 159 deaths are commonly published for Hawaii, plus the five Coast Guardsmen lost at the Scotch Cap lighthouse on Unimak Island, which is why some totals for the event appear as 165. The center at Palmer, Alaska opened in 1967 after the magnitude 9.2 Alaska earthquake of March 27, 1964, took on West Coast responsibility in the 1980s, and was renamed the National Tsunami Warning Center in 2013. Areas of responsibility between the two have been reorganized more than once and both also serve as international Tsunami Service Providers for other basins, so confirm at tsunami.gov which center issues for your coast instead of working from memory.

The National Weather Service issues four tsunami message types, and the distinction between them is operational rather than cosmetic. A Warning means damaging inundation is expected or occurring and the response is evacuation of the hazard zone. An Advisory means strong currents and dangerous waves are expected in the water and immediately at the shore without widespread land inundation, so the response is clearing beaches, harbors, marinas and the immediate waterfront while leaving inland areas alone. A Watch means a distant event may produce a tsunami and the situation is still being evaluated, which is the cue to staff up and get ready rather than to move people. An Information Statement means an earthquake occurred and no tsunami threat exists, or one exists only far from your coast. Read the current product definitions and the message text conventions on the warning centers’ own site, because the wording and the thresholds have changed over the years.

What the warning centers do not do is order an evacuation. They characterize the hazard and publish the message, and the decision to evacuate a specific zone, open shelters, close a harbor and stop traffic on a coastal highway belongs to local and state officials under state law. That division is the source of a great deal of confusion in tabletop exercises, where participants wait for the center to tell them to move people. Write the trigger into your own plan so that receipt of a Warning for your coastal segment automatically initiates evacuation of the mapped zone without a further deliberation step, and reserve the deliberation for the harder cases, which are Advisory-level events and the awkward interval after a Watch.

Local and distant tsunamis are two different operations

A distant tsunami gives you hours, and in the Pacific those hours are generous. Waves from the magnitude 9.5 Chile earthquake of May 22, 1960 took roughly fifteen hours to reach Hawaii, and modern travel times from Chile to the United States West Coast, from Japan to California, or from the Aleutians to Oregon range from several hours to more than half a day. With that much lead time, the warning chain works the way the textbook describes: a Watch, then confirmation from DART stations and tide gauges along the path, then a refined forecast with arrival times, then an evacuation order issued by local authority with enough notice to move people on foot or in vehicles, clear the harbors, and get a shelter open. The 1960 Hilo event also shows the failure mode of long lead time, because sirens sounded in Hilo hours before the wave and 61 people in Hawaii died anyway, some of them because they did not believe the wave would be large and some because they went to look at the water.

A local tsunami is a different job entirely. When the source is the subduction zone off your own coast, the first wave can arrive in fifteen to thirty minutes, the shaking itself may have damaged roads, bridges, power and radio sites, and no official message will reach the beach before the water does. The 1993 Hokkaido Nansei-oki earthquake put a tsunami onto Okushiri Island in roughly five minutes, about the same time it took JMA to issue its warning, and published tolls for that event run to around 230 dead and missing. For the Cascadia subduction zone, paleoseismic work summarized by the USGS and by state geological surveys puts the average interval between great earthquakes at roughly 500 years with individual intervals ranging from a couple of centuries to about a thousand years, and the most recent full-margin event has been dated to January 26, 1700 using Japanese written records of an orphan tsunami together with tree-ring and tsunami deposit evidence.

The planning implication is that your coastal jurisdiction needs two plans with different logic. The distant-source plan is a notification, decision and transportation operation with a clear chain and time to execute it. The near-source plan cannot depend on notification at all, so its content is pre-event education, mapped and signed routes, vertical refuge where the ground offers no alternative, and a response concept that assumes your own stations, apparatus and dispatch center were shaken first and that outside help will be slow because the same earthquake damaged the roads leading in.

Natural warning signs, and the sources that give none

The natural warning signs are few and worth stating exactly. Strong or prolonged ground shaking at or near the coast is the first. A sudden unusual withdrawal of the sea that exposes reef and seabed is the second, and so is a sudden unusual rise in water, because either drawdown or run-up can come first depending on which part of the wave train arrives and how the seafloor deformed. A loud roar from offshore, often described by survivors as sounding like a train or an aircraft, is the third. The correct response to any of them is immediate movement to high ground, and the guidance published by the National Weather Service and by the Pacific Northwest states generally sets the target at roughly 100 feet of elevation or about a mile inland, with the mapped evacuation zone for your own community taking precedence over any rule of thumb because local bathymetry, river mouths and low spits change the answer.

Some tsunamigenic events produce weak felt shaking and therefore no useful natural warning on land. So-called tsunami earthquakes rupture slowly on shallow parts of the fault and generate far more wave than their felt intensity suggests, and the 1992 Nicaragua and 2010 Mentawai events are the standard examples in the literature. Submarine landslides are worse in this respect, because a slump can be triggered by a moderate earthquake or by nothing obvious at all and can produce extreme local run-up with almost no seismic signature. The 1998 Papua New Guinea tsunami at Sissano Lagoon, attributed in the published research to a submarine slump, killed a number generally reported between about 2,100 and 2,200 people, and the wave arrived within minutes.

Volcanic sources defeat a seismically triggered system in the same way. On December 22, 2018 a flank collapse at Anak Krakatau generated a tsunami that struck the Sunda Strait coasts of Java and Sumatra with no earthquake to detect and no warning issued, and Indonesian authorities reported on the order of 430 deaths. The point for an emergency manager is not to build a program around these rarities, since you cannot detect what the sensors do not see, but to make sure the public education does not accidentally teach a false precondition. A message that says to evacuate after you feel a big earthquake has taught people that water behaving strangely without an earthquake is safe to watch, and it is not.

Vertical evacuation: when there is no uphill

On a low barrier spit, a long sand peninsula or a flat coastal plain, the mapped evacuation route can be longer than the arrival time, and no amount of walking speed fixes it. Vertical evacuation is the response to that geometry, and it means engineered refuge above the modeled inundation depth, reached on foot in the minutes available, in the form of a purpose-built tower, an earthen berm with a wide flat top, or a building with an upper floor or roof platform designed for the loads. FEMA published design guidance for these structures as FEMA P-646, Guidelines for Design of Structures for Vertical Evacuation from Tsunamis, and tsunami loads and effects entered the American Society of Civil Engineers minimum design loads standard with ASCE 7-16 for certain Pacific state applications. Check the current editions of both before anyone designs anything, because the provisions have been revised.

The engineering matters because the failure modes are specific. A refuge has to be high enough to clear the modeled maximum inundation with freeboard, and it has to survive the earthquake that precedes the wave, the hydrodynamic and hydrostatic loading, debris impact from vessels, logs and vehicles, debris damming against columns, scour that can undermine foundations, and in some settings buoyant uplift on enclosed floors. That is why an open-frame lower level and deep foundations show up repeatedly in the guidance. The widespread assumption that any tall concrete building on the beach will serve as a tsunami refuge is a myth, and it is a dangerous one, because an ordinary hotel or condominium designed for wind and seismic loads has not been evaluated for debris impact or scour and may have no exterior stair route to the upper floors that is usable when the elevators are dead.

The United States example most often cited is Ocosta Elementary School in Westport, Washington, rebuilt with a gymnasium roof platform designed as a tsunami refuge and opened in 2016, funded through a local school bond and designed to hold roughly a thousand people above the modeled inundation. Japan has designated tsunami evacuation buildings in coastal municipalities for far longer and expanded the practice after 2011. If your community is considering one, the questions that decide whether it works are unglamorous: who holds the key at two in the morning, whether the access stair is outside the building and unlocked, how the route is signed for someone who has never been there, whether the platform has water, shelter from weather and a way to communicate, and how the several hundred people standing on it are retrieved eight or twelve hours later when the currents finally settle.

The last mile: sirens, WEA, and who pushes the button

The distribution end of the chain is where most local investment goes and where most local failures occur. NOAA Weather Radio All Hazards carries the warning center products directly, the Emergency Alert System relays them through broadcasters and cable systems, and Wireless Emergency Alerts sent through FEMA’s Integrated Public Alert and Warning System reach cell phones inside a targeted area without requiring a subscription. Coastal outdoor siren systems, including the all-hazard alert broadcast sirens along the Washington coast that can transmit voice as well as tone, cover beaches, parks and campgrounds where nobody is watching a television. The United States Coast Guard broadcasts on marine VHF for vessels and harbors. Each of these paths has a known gap, since outdoor sirens are outdoor devices and will not wake a sleeping family indoors, WEA depends on cell sites that a great earthquake may have knocked offline, and broadcast EAS depends on a station that may be off the air.

Alert origination is a local administrative problem that gets discovered at the worst time. Someone in your jurisdiction holds the IPAWS collaborative operating group authority and the credentials to originate a WEA message, and that someone needs to be more than one person, needs current credentials that have not silently expired, and needs pre-scripted tsunami message templates already loaded so that nobody is composing a ninety-character alert under pressure. The scripts should distinguish a Warning from an Advisory in plain words, name the zone in terms residents and visitors can locate, and say what to do rather than what is happening. Test the origination path on a schedule against the IPAWS test environment, and record the date of the last successful test in the plan.

For a near-source event, assume the local alerting infrastructure took the earthquake first. Sirens on poles that survived the shaking still need commercial power or a charged battery and a control path from a dispatch center that may be evacuating itself, and the notification you most want to send is the one your own damaged systems are least able to send. That argues for redundancy of a low-technology kind, including fire apparatus with public address systems on pre-assigned routes, law enforcement and lifeguard sweeps of beach access points, hotel and campground staff trained to move guests without waiting for instruction, and signage that tells a person standing on the sand which way to walk. Put the route assignments in the plan by station and by beat so that they do not have to be invented while the ground is still moving.

What the water actually does, and the myths worth naming

A tsunami is a series of waves with very long periods, arriving over a window that can last many hours, and the first wave is frequently not the largest. Treating the first arrival as the event is the single most common fatal error, and it is why re-entry has to be controlled by the warning center’s cancellation and the local damage picture rather than by the appearance of calm water. Nearshore, the wave behaves less like a breaking surf wave and more like a fast rising flood that keeps coming, carrying vessels, vehicles, timber and structural debris, and then reverses and drains with comparable violence. In harbors and inlets, resonance and channel geometry can produce damaging currents long after the open coast has quieted, which is precisely what the Advisory product exists to communicate.

Several persistent myths deserve to be named as myths in the same sentence as the story. It is a myth that the sea always withdraws before a tsunami arrives, since the leading edge can be a rise. It is a myth that a tsunami is a single towering breaking wave of the kind that appears in illustrations, and that image causes people to dismiss a five-foot surge that is in fact a wall of moving water with enormous momentum behind it. The term tidal wave is a misnomer, because tides have nothing to do with it. It is a myth that you can reliably drive out of a near-field tsunami zone, because a post-earthquake road with a failed bridge approach, downed poles and a traffic jam is a trap, and the guidance from the Pacific Northwest states is to go on foot unless you have a specific reason to believe the route is clear.

The historical record also sets the scale of what these events do, and reputable figures disagree in ways worth acknowledging. For the Indian Ocean tsunami of December 26, 2004, the death toll most often published from NOAA’s historical tsunami database is about 227,900, while other compilations put the total above 230,000, and the number will never be settled because of unrecovered and unidentified victims across a dozen countries. For Tohoku in 2011, Japan’s National Police Agency has published roughly 15,900 confirmed dead with about 2,500 people still listed as missing in its periodic reports. For the 1964 Alaska earthquake, published totals run around 128 to 131 deaths depending on the source, the great majority of them from the tsunami rather than from shaking, including four children at Beverly Beach State Park in Oregon and eleven or twelve people at Crescent City, California, where accounts differ on the count.

Community education is the system

For a near-source tsunami, the detection network, the warning centers and the alerting infrastructure all sit behind the arrival of the wave, which leaves pre-event education carrying the load. The National Tsunami Hazard Mitigation Program, a partnership of NOAA, FEMA, the USGS and the coastal states and territories dating from the mid-1990s, funds much of that work, including the inundation modeling that produces evacuation maps. Those maps come from state agencies, with the Oregon Department of Geology and Mineral Industries, the Washington Geological Survey within the Department of Natural Resources, and the California Geological Survey among the publishers for the West Coast, and your plan should reference the current published version by name and date rather than a laminated copy of uncertain age. The National Weather Service also runs the TsunamiReady recognition program, with requirements covering hazard mapping, signage, redundant alert reception, public education and planning, and the current criteria should be obtained from your NWS Warning Coordination Meteorologist because they have been revised.

The strongest evidence that drills change outcomes comes from Japan. In Kamaishi, Iwate Prefecture, where schools had run repeated tsunami evacuation drills under a program associated with the researcher Toshitaka Katada of Gunma University, nearly all of the roughly three thousand schoolchildren evacuated successfully on March 11, 2011, and accounts of the event also record the deaths of a small number of children who were not at school that day. Japanese practice includes the principle described as tsunami tendenko, which holds that each person evacuates immediately on their own rather than waiting to gather family members, on the reasoning that waiting kills more people than it saves. That idea is culturally difficult to teach anywhere, and it is the single hardest piece of tsunami education to get a community to accept.

The practical version of community education is repetitive and local. Signed routes that a stranger can follow, assembly areas at mapped high ground, a published annual drill in which people actually walk the route and time it, school evacuation plans exercised in the fall and the spring, briefing packets for hotel and campground operators to hand to guests, and a standing item on a chamber of commerce or tourism board agenda so that the businesses on the waterfront know what they are expected to do. The timing data from a real walk is the most useful product of the whole exercise, because it tells you which neighborhoods cannot make high ground inside the modeled arrival time and therefore where vertical evacuation or route improvement has to be considered.

What to do at your agency

• Have your emergency manager download the current tsunami inundation and evacuation maps published by your state geological survey this month, note the publication date, and confirm that the maps referenced in your emergency operations plan are the same version rather than an older printing.
• Ask the Warning Coordination Meteorologist at your local National Weather Service forecast office in writing for the current tsunami product definitions, your community’s TsunamiReady status and the outstanding requirements, and put the reply in the plan’s reference file.
• Identify by name every person in your jurisdiction who can originate a Wireless Emergency Alert, verify that their IPAWS credentials are current, confirm that pre-scripted tsunami Warning and Advisory templates are loaded, and run one test in the IPAWS test environment with the date recorded.
• Put the difference between a tsunami Warning and a tsunami Advisory on the agenda of the next shift commanders’ meeting that already happens, and issue a one-page standing order stating what engine companies, patrol units and harbor staff do on receipt of each.
• Walk the mapped evacuation route on foot with a stopwatch from the lowest-elevation occupied building in your jurisdiction and from every school inside the mapped zone, record the elapsed times to the assembly area, and compare them against the modeled first-wave arrival time for a near-source event.
• Ask public works or whoever owns the outdoor sirens for the current inventory, the backup power arrangement at each site, the date of the last full activation test, and a written statement of what happens to the control path if the dispatch center loses commercial power.
• Ask the harbor master or marina operator for their written tsunami procedure covering vessel movement, gangway and dock personnel, and the decision trigger, and if there is no written procedure, get one paragraph added to the plan naming who clears the docks and when.

Takeaways

• The first tsunami warning always rests on an earthquake solution rather than on observed water, so early magnitude estimates and forecast heights are expected to be revised upward or downward as long-period analysis and DART observations arrive.
• The National Weather Service issues Warning, Advisory, Watch and Information Statement products, the Advisory addresses dangerous currents and waves at the shore rather than land inundation, and the warning centers characterize the hazard while local and state officials order evacuations.
• A distant tsunami gives hours and works through the full notification and decision chain, while a near-source tsunami can arrive in fifteen to thirty minutes with no official message reaching the beach ahead of it.
• Strong or prolonged shaking at the coast is itself the warning to move to high ground immediately, and some tsunamigenic sources including slow tsunami earthquakes, submarine landslides and volcanic flank collapses produce little or no felt shaking at all.
• Vertical evacuation is an engineering solution for places where high ground cannot be reached in time, it requires design against inundation depth, debris impact, scour and uplift under current FEMA and ASCE guidance, and an ordinary tall building on the beach has not been evaluated for any of that.
• The first wave is frequently not the largest, the wave train can last hours, harbor currents persist after the open coast quiets, and re-entry should follow the warning center cancellation and a damage assessment rather than the look of the water.
• Published death tolls for major tsunamis genuinely disagree, with the 2004 Indian Ocean total given as about 227,900 in NOAA’s historical database and above 230,000 in other compilations, and the 1964 Alaska total given as roughly 128 to 131 depending on the source.
• Signed routes, mapped assembly areas, an annual walked and timed drill, and briefings for hotels and campgrounds are the parts of the system that operate when the wave arrives before the message does.