Case Study · Tsunami · 2018
September 28, 2018. A M7.5 earthquake struck Sulawesi, Indonesia on a strike-slip fault — the horizontal-motion fault type that scientists did not associate with tsunami generation. Five minutes later, a 10-meter tsunami wave slammed into Palu Bay. The Indonesian warning system issued an alert. Then cancelled it, based on early gauge readings. People who had evacuated began to return. The tsunami arrived as the warning was expiring. 4,340 dead. Not all tsunami-generating earthquakes are the ones warning systems are built to watch for.
Palu Bay, Sulawesi, Indonesia · September 28, 2018
At 6:02 PM on September 28, 2018, a M7.5 earthquake struck the Palu-Koro fault in Central Sulawesi. The fault is a left-lateral strike-slip fault — the type that moves horizontal plates past each other, rather than the subduction zone thrusts (where one plate dives under another, displacing water vertically) that are the standard source of large destructive tsunamis. The EBSCO Research account of the event documents the scientific community's assessment: "For scientists this was unexpected because strike-slip faults are not typically associated with tsunami generation and it usually requires quakes of even larger magnitude to trigger one." Costas Synolakis of the USC Tsunami Research Center said directly: "This type of earthquake is not the standard mechanism that triggers a tsunami."
Indonesia's meteorological and geophysical agency, BMKG, issued a tsunami warning approximately 5 minutes after the earthquake — acting on the seismic data available. But the tsunami was already arriving. The NOAA CTDR account documents that the tidal gauge at Pantoloan, along Palu Bay, recorded the tsunami arriving 6 minutes after the earthquake with a maximum wave height of approximately 2 meters at the gauge location — but 8-meter watermarks were observed near the coast, and run-up heights as high as 10.7 meters were measured in Donggala Regency. The gauge reading, which underestimated the actual wave heights, informed the BMKG's decision to cancel the warning approximately 30 minutes after the earthquake. By then, the most destructive waves had already struck. People who had evacuated from the beach after the initial shaking were, according to multiple accounts, returning to the beach area when the waves arrived.
The Wikipedia account of the 2018 Sulawesi event documents the final toll: 4,340 dead; 10,679 injured; 667 missing; 70,821 evacuated. Total damage approximately $1.71 billion. The event produced significant scientific reappraisal of the mechanisms that can generate local tsunamis — specifically expanding the recognized tsunami-generation hazard beyond the conventional subduction zone model to include strike-slip fault systems with specific geometry, and the landslide-triggered tsunamis that can amplify the earthquake-generated wave.
Sep 28, 2018
Date
M7.5
Strike-Slip Earthquake
5–6 min
Tsunami Arrival Time
10.7 m
Max Wave Height
4,340
Deaths
The scientific investigation of the Palu 2018 tsunami has produced peer-reviewed analyses in multiple journals. The consistent finding is that the observed tsunami heights cannot be explained by the earthquake fault mechanism alone — the waves were too large for a strike-slip fault of this magnitude. The Wiley Earth and Space Science analysis identifies the tsunami sources as a combination: the strike-slip fault included localized normal (dip-slip) faulting components that contributed to vertical displacement; submarine landslides triggered by the earthquake in Palu Bay significantly amplified the wave heights near shore; and Palu Bay's narrow, funnel-shaped geometry focused and amplified the wave energy as it propagated toward the city. The University of Washington event analysis (SEAC) quotes Dr. Vasily Titov of NOAA's Center for Tsunami Research: "About 5-10 minutes later the first of what are believed to have been three tsunami waves crashed ashore." Three waves, in a local tsunami, from a strike-slip fault, amplified by a bay. The warning system that exists to watch for the "standard mechanism" could not fully characterize what was happening in real time.
The Science
Think of the two ways a fault can move. A thrust fault (like the Cascadia Subduction Zone or the Japan Trench) moves one plate vertically over another — like a table tilting and dropping one end. That vertical movement displaces an enormous volume of water, creating the large ocean waves that make up most historically deadly tsunamis. A strike-slip fault moves two plates horizontally past each other — like two ice floes sliding sideways. Horizontal motion doesn't normally displace water vertically. The standard tsunami generation mechanism requires vertical seafloor displacement. Strike-slip faults generate very little of it. This is why, before 2018, earthquake engineers and tsunami scientists generally assigned low tsunami-generation probability to strike-slip earthquakes unless they were very large. A M7.5 strike-slip earthquake was not a recognized high-risk tsunami source. The Palu disaster proved this category exclusion was wrong — under specific conditions.
The peer-reviewed analysis of the Palu 2018 tsunami identifies a convergence of three factors that together produced a catastrophic wave from a "non-standard" earthquake mechanism. First, the Palu-Koro fault included localized dip-slip (vertical) components within its primarily strike-slip motion — the Wiley paper shows "the presence of some normal faulting component" that contributed to vertical seafloor displacement. Second, the earthquake triggered multiple submarine landslides in Palu Bay — landslides on the seafloor generate tsunamis directly, and these are now documented as significant contributors to the wave heights observed near shore. Third, Palu Bay is a narrow, funnel-shaped bay oriented to direct and concentrate incoming waves toward the city. The ResearchGate analysis notes that "the long and narrow shape of the bay played a role in intensifying the wave." Each factor alone might have been manageable. All three together produced waves that killed thousands.
Indonesia's tsunami warning system is designed around the standard tsunami-generation model: large subduction zone earthquakes generating waves detected by offshore buoys and coastal tide gauges. The Palu event was not a standard event. The gauge at Pantoloan recorded waves of approximately 2 meters — significantly lower than the 8–10 meter waves observed in eyewitness accounts and field surveys near the shore. The discrepancy likely reflects the gauge's location within the bay and the complex wave pattern generated by multiple tsunami sources (fault mechanism, landslides). The BMKG, seeing low gauge readings and no significant open-ocean tsunami signal, cancelled the warning approximately 30 minutes after the earthquake. This was a reasonable decision based on the available data from systems designed for conventional tsunamis. It was fatally wrong for the specific, locally-generated, bay-amplified, multi-source tsunami that Palu Bay was experiencing. The lesson for warning system design: local tsunamis generated by non-standard mechanisms can arrive and cause damage before conventional detection systems can fully characterize them.
Timeline
01
September 28, 2018, 6:02 PM: M7.5 earthquake on the Palu-Koro fault. Strike-slip mechanism. Shallow depth (~10-20 km). Surface rupture mapped across 30 km of the fault — runs through the west side of Palu City. Fault ran directly under the center of the city. Extreme ground shaking: MMI X (Extreme) at some sites. Liquefaction destroys three villages entirely. Fault line visible on satellite imagery. A beach festival was underway in Palu Bay at the time of the earthquake.
02
~6:07 PM: BMKG issues tsunami warning — approximately 5 minutes after the earthquake. Some people evacuate to higher ground. But ~6:30–6:32 PM: BMKG cancels the warning based on coastal gauge readings that show waves of approximately 2 meters — within the moderate range. The gauge at Pantoloan underestimates the actual wave heights in the bay. People who had evacuated begin to return. The warning that may have saved some lives is no longer in effect when the most destructive waves are striking.
03
5–10 minutes after earthquake: First of three tsunami waves reaches Palu Bay coastline. Waves up to 10.7 meters in Donggala Regency; 4-7 meters at Palu Bay. 300 meters inland inundation. Multiple sources: fault mechanism + submarine landslides + bay geometry amplification. Beach festival area: videos show wave arriving while people still present. 4,340 confirmed dead; 10,679 injured; 667 missing; 70,821 evacuated. The fault that was "not supposed to generate a tsunami" had generated waves large enough to destroy a city.
04
2018–present: Major scientific revision of tsunami generation hazard assessment. Strike-slip faults near bays and coastal areas now recognized as potential tsunami sources if they include localized dip-slip components or if they can trigger submarine landslides. Indonesia reviews and upgrades warning system to improve detection of locally-generated tsunamis. Research published in multiple peer-reviewed journals: AGU, Wiley, ResearchGate. The Palu event changed how scientists categorize tsunami-generating fault types globally.
Human Decisions
What the warning system got right — and what it missed
BMKG issued a tsunami warning within approximately 5 minutes of the earthquake — faster than many conventional warning timelines. This response was reasonable and likely saved lives among people who evacuated. The failure was not in the warning issuance but in the cancellation. The cancellation was based on gauge readings that significantly underestimated the actual wave heights — not because the gauge malfunctioned, but because the gauge location and the specific multi-source nature of the Palu tsunami produced readings that didn't reflect what was happening at the shoreline. The warning system worked as designed. Its design assumed a different type of tsunami.
The BMKG's decision to cancel the tsunami warning approximately 30 minutes after the earthquake was a decision made on the best available data from systems designed for conventional tsunami detection. But the Palu tsunami was not a conventional tsunami. The cancellation, based on gauge readings that underrepresented the actual wave heights, communicated to people in the evacuation zone that it was safe to return. Some did. In the context of a 5-10 minute tsunami arrival time, the cancellation was issued after the most destructive waves had already struck — but for a second round of waves and for people who were in the process of returning, the cancelled warning may have reduced the urgency of evacuation behavior.
The natural warning that was available
For local tsunamis (generated by nearby earthquakes) like the Palu event, the available warning time between the earthquake and the first wave is measured in minutes — not hours. The 5-6 minute arrival time at Palu left essentially no time for a warning to be issued, communicated, and acted upon by everyone in the tsunami zone. The Japanese Kamaishi lesson applies directly: when you experience strong shaking in a coastal area, run to high ground immediately without waiting for an official warning. For the Palu community, the M7.5 earthquake — which they felt violently — was the warning. The USGS/NOAA guidance documents this as the first natural warning sign requiring immediate response without waiting for official confirmation.
Multiple accounts report that a beach festival was underway in Palu Bay at the time of the earthquake. Amateur video footage posted after the event shows people at the waterfront when the tsunami arrived. The combination of the non-standard earthquake mechanism (not recognized as a tsunami source), the cancelled warning, and the concentration of festival-goers at the waterfront created conditions where a large population was in the immediate impact zone when the waves arrived. The Palu disaster documents what happens when community tsunami preparedness focuses on the "standard" tsunami scenario and doesn't account for the possibility of a local tsunami from a fault type not previously associated with such events.
The cascade lesson
The 2018 Palu tsunami is the most scientifically consequential tsunami event in recent history in terms of what it changed about how scientists understand tsunami generation. Strike-slip faults near coastal bays are now recognized as potential tsunami sources in specific conditions. Submarine landslides triggered by earthquakes are now understood as significant tsunami amplifiers. Bay geometry is understood to concentrate and amplify local tsunamis well beyond what the earthquake magnitude alone would suggest. And the warning system lesson is that locally-generated tsunamis can arrive before any conventional warning system can fully characterize them — which is why the natural warning of strong coastal shaking must be treated as an immediate evacuation trigger without waiting for official confirmation. The earthquake felt by Palu residents at 6:02 PM was the only warning that could have saved everyone. For local tsunamis from any source, anywhere in the world, that is always the case.
What You Can Do Now
Palu's lesson is the most local and most urgent of all the tsunami case studies: for an earthquake that you feel, near the coast, you may have only minutes before the tsunami arrives. The earthquake is the warning. These five actions apply that principle.
The Palu tsunami arrived 5–6 minutes after the earthquake. No warning system — however well-designed — could have fully characterized a multi-source local tsunami and communicated a reliable all-clear or sustained warning in that window. The felt shaking is the only tsunami warning available for most local tsunami events. NOAA, FEMA, and every major tsunami preparedness program now explicitly state: strong shaking near a coast = immediate evacuation to high ground, without waiting for a siren or official notification.
Local tsunami immediate response guideThe BMKG cancelled the Palu tsunami warning based on gauge readings that underestimated the actual wave heights. Warning systems can fail or underestimate local tsunamis generated by non-standard mechanisms or amplified by bay geometry. A warning cancellation is based on the best available data — and that data may be incomplete for complex local events. For a local tsunami generated by felt shaking, the principle is: stay on high ground until you can clearly see the water has returned to normal and has remained normal for at least an hour. Not until the warning is cancelled. Until you can see.
Tsunami warning interpretation guideMost tsunami preparedness messaging focuses on the large subduction zone earthquake as the tsunami generator — the Cascadia Subduction Zone for the US Pacific Northwest, the Alaska-Aleutian Chain for Hawaii and Alaska, the Japan Trench for Japan. These are real, high-priority risks. But Palu 2018 documents that local tsunamis can also be generated by: strike-slip faults with specific geometry; submarine landslides triggered by earthquakes; and volcanic activity on ocean islands. For any community near water and near any type of active fault, strong local shaking is a local tsunami warning regardless of the fault type involved.
Tsunami source types guidePalu Bay's narrow, funnel shape concentrated wave energy as the tsunami propagated toward the city. The same bay geometry that made Hilo Bay disproportionately dangerous in 1960 created disproportionate wave heights in Palu Bay in 2018. If you live near a narrow bay, inlet, harbor, or river mouth, your local tsunami risk may be higher than the underlying earthquake or open-ocean wave height would suggest. Check with your local emergency management office about whether bay amplification is a documented factor in your community's tsunami hazard assessment.
Bay tsunami amplification guideThe Palu event included a beach festival at the waterfront when the earthquake struck. Concentrations of people at sea level in seismically active coastal regions create compounded risk: high population density + proximity to water + minimal vertical evacuation options. For travel to coastal areas in seismically active regions — particularly in Southeast Asia, the Pacific Ring of Fire, and the Mediterranean-Himalayan belt — knowing the nearest high-ground evacuation routes before attending any waterfront event is standard preparedness practice.
Coastal travel tsunami preparednessTsunami case study series
The 2004 Indian Ocean Tsunami covers the world without a warning system. Hilo 1960 covers people coming back after the first waves. Crescent City 1964 covers the fourth wave after the quiet period. Tōhoku 2011 covers the seawalls that didn't hold.
Full tsunami case study seriesSources