Tracing tsunami impacts back to their source across ocean basins

Paul Somerville, Chief Geoscientist, Risk Frontiers

Large transoceanic tsunamis have impacts that extend far from the earthquake that cause them, and so it often occurs that tsunami impacts that are recorded in historical documents or in geological sand deposits cannot be easily associated with their source. In the course of preparing this Briefing, I found a possible association of a tsunami deposit in Tanzania and fourteen tsunami deposits in South Asia with the AD 900 tsunami at Nagapattinam, India.

The 1700 “orphan” tsunami in Japan traced back to Cascadia, western North America

On the night of January 27, 1700, a mysterious tsunami flooded fields and washed away houses in Japan. It arrived without the warning that a nearby earthquake usually provides. Samurai, merchants and villagers recorded the event, but nearly three centuries passed before discoveries in North America revealed the tsunami’s source: an earthquake of estimated Mw 9.0 that occurred on the Cascadia subduction zone off the coast of Washington and Oregon. The evidence there consisted of trees that were swamped by the coastal subsidence caused by the earthquake, with tree rings indicating their approximate time of death (Atwater et al., 2016).

Figure 1. The 1700 Cascadia tsunami.  Source: Atwater et al., 2016.

Tracing tsunami deposits in Tanzania back to their likely tsunami source in Sumatra-Andaman

The December 2004 Mw 9.1 Sumatra-Andaman earthquake generated a tsunami whose greatest impact was felt in Indonesia, Sri Lanka, India, and Thailand, where more than 200,000 people lost their lives. Nine hours later, more than 5,000 km from the earthquake’s epicenter, the tsunami reached the coastline of Eastern Africa (Figure 2). Runup heights of almost 10m were measured in Somalia, where 298 fatalities occurred. Reduced impacts were observed farther south, along the African coast of Kenya and Tanzania, probably because it struck on a low tide. There was no tsunami warning system in the Indian Ocean at the time of the 2004 event, but one has now been implemented.

Figure 2. The 2004 Sumatra-Andaman Tsunami. Source: NOAA.

Learning more about tsunami risks has global implications, but the December 2004 tsunami has been predominantly seen as an eastern Indian Ocean event, and as a consequence, much of the work to find ancient tsunami deposits and to understand the recurrence intervals of such catastrophic events has been focused on that region. Tsunami risk has been thought to be low in East African countries, mainly because of the limited damage caused by the 2004 tsunami.

However, in Pangani Bay, Tanzania, Maselli et al. (2020) identified evidence of a deadly tsunami that they conclude occurred about 1,000 years ago, suggesting that the tsunami risk in East Africa could be higher than previously thought. They visited a field site close to Pangani Bay, where they discovered the tsunami deposit. At a depth of about 1.5 meters, they found a sand layer (Figure 3) hosting human remains lacking traditional funerary burial. The bones did not present any evidence of disease or trauma due to battle. Tanzania is subject to cyclones, and as noted by Shanmugam (2012), there can be considerable ambiguity in distinguishing between paleo-tsunami deposits and paleo-cyclone deposits using sedimentological criteria, and considerable uncertainty in the age dating of these deposits.

The mixed fossil assemblage indicative of continental, estuarine, and marine habitats was present within the sand layer. The occurrence of marine shells supported the hypothesis that a tsunami may have impacted the area, although a cyclone could have done the same. Radiocarbon dating indicated that the event that deposited the sand layer in Pangani Bay occurred about 1,000 years ago.

Sedimentary evidence of paleo-tsunami deposits of the same age were reported from Thailand, India, Indonesia, southern Sri Lanka and the Maldives (Figure 4), pointing to an event in about the year 950.

Figure 3. Left: Tsunami deposit in Pangani Bay indicated by the sand layers. Credit: Davide Oppo. Right: Skeletal remains of a victim of the 1,000-year-old Indian Ocean tsunami. Credit: Vittorio Maselli.
Figure 4. Tsunami deposits found in Indian Ocean coastal sites. The horizontal red line highlights the timing of the inferred Tanzania tsunami at 950 AD. Source: Maselli et al., 2020.

Although Maselli et al. (2020) did not associate the Pangani Bay, Tanzania tsunami deposit with a historical earthquake, the NOAA Global Historical Tsunami Database (NOAA, 2020) contains an event at Nagapattinam, India in 900 AD that is estimated to have killed several hundred people, based on Rastogi and Jaiswal (2006) who note that:

“There is mention of tsunami effect in scriptures at Nagapattinam in 900 AD that destroyed a Buddhist monastery. According to literature available in the library of Thondaiman kingdom in Puduckottai, Tamilnadu, it was during the reign of Raja Raja Chola that waves had washed away the monastery and several temples and killed hundreds of people. There is evidence of this in Kalaki Krishnamurty’s book ‘Ponniyin Selvan-The Pinacle of Sacrifice,’ In the chapter ‘The Sea Rises’, the author explains how the sea had risen very high and the black mountain of water moved forward. The sea inundated warehouses and sheds and began to flow into the streets. Ships and boats seemed suspended in mid-air, precariously poised on the water peaks. The book also describes how an elephant was swallowed by the gushing water.”

It seems likely that this is the event that Maselli et al. (2020) identified in Pangani Bay, Tanzania. Further, considering the tsunamigenic earthquake sources that are present in the Indian Ocean (Jaiswal et al., 2008; Schafer and Friedmann, 2019, Figure 5), it is evident that the Sumatra-Andaman Sea Subduction zone is the only one with the potential to generate large transoceanic tsunamis in the Indian Ocean, so we sought published evidence from other geological  records of large tsunamis in that region to see if the AD 900 event is recorded in them.

Figure 5. Map of maximum magnitude estimates for the world’s major subduction zones. The inset obscures most of the southern Indian Ocean, where there are no subduction zones. Source: Schäfer and Friedemann, 2019.

The AD 900 event does not appear on the lists of events inferred to have occurred at sites in Aceh (Rubin et al., 2017) and the Andaman Islands (Malik et al., 2019), but in both cases it occurred at a time when conditions were such that a record would not have been made.  In the case of Aceh, Rubin et al. (2017) found evidence for at least 11 prehistoric tsunamis that struck the Aceh coast between 7,400 and 2,900 years ago, and state:

“The cave probably contained stratigraphic evidence of recent historic tsunamis from 2,900 years BP to the 2004 Indian Ocean tsunami that have been identified elsewhere in the region, but these were most likely removed by subsequent tsunamis inundating the cave as indicated by the erosional unconformity beneath the 2004 deposit.”

In the case of the Andaman Islands, Malik et al. (2019) describe evidence for 7 events between 1881 and before 5600– 5300 BCE, and state:

“The sequence includes an unexplained hiatus of two or three millennia ending around 1400 CE, which could be attributed to accelerated erosion due to Relative Sea-Level (RSL) fall at ~3500 BP.”

From this we conclude that the absence of observations of the AD 900 event at Aceh and the Andaman Islands does not preclude the likelihood that it occurred somewhere in this region.

How often do tsunamis like the Mw 9.1 2004 Sumatra-Andaman event occur?

The NOAA (2020) historical record of tsunamis in the Indian Ocean is very sparse before about 1681; with only the AD 416 event recorded in Java, the AD 900 event recorded in Nagapattinam, and the 1524 event recorded in Dabhol, India appearing in the catalogue before then. However, since 1681, there have been numerous events: 1 of Mw >9.0; 6 events of Mw 8.5-8.9; 8 events of Mw 8.0-8.4; 25 events of Mw 7.5-7.9, and 18 events of Mw 7.0-7.4, most of which have occurred in Indonesia, with many of those occurring east of Sumatra and not greatly affecting the Indian Ocean. Jaiswal et al. (2008) list 7 events affecting India and the surrounding region before 1668, and 14 events since then.

In Aceh, Rubin et al. (2017) conclude that the average time between tsunamis is about 450 years with intervals ranging from a long, dormant period of over 2,000 years, to multiple tsunamis within the span of a century. In the Andaman Islands, Malik et al. (2019) suggest a recurrence of 420–750 years for mega-earthquakes (magnitude Mw about 9), and a shorter interval of 80–120 years for large magnitude earthquakes (magnitude Mw about 8). Taken together, these studies suggest a recurrence interval of about 500 to 750 years for earthquakes like the Mw 9.1 2004 Sumatra-Andaman earthquake.

The Mw 9.1 2004 Sumatra-Andaman earthquake had little impact on Australia, with most of the impact occurring along the northwest coast of Western Australia. Localised inundation and tidal surges lasted for several hours along the Western Australian coast, resulting in boats losing their moorings and being damaged in marinas, and thirty swimmers needing to be rescued. The Mw 7.7 Pangandaran, west Java earthquake of 17 July 2006 generated a tsunami that affected several areas from north of Geraldton to Port Hedland, and inundated and destroyed a camp in the Steep Point area, where one family was fortunately able to move quickly from their camp to safe ground and another held onto their vehicle as it was moved 10 metres by the tsunami. Fortunately, the main sources of tsunami-generating earthquakes are sufficiently distant from Australia that the Australian Tsunami Warning System that is now in place can provide at least 1.5 hours of warning before the arrival of the tsunami onshore, providing adequate time for evacuation in most situations.

Given the locations of tsunami sources in the Indian Ocean (Figure 5) and the estimated recurrence interval of very large earthquakes on these sources, we expect the tsunami hazard in Australia to be concentrated along the northwest coast. This expectation is borne out in the probabilistic tsunami hazard map for Australia for an AEP of 1:500 (Figure 6, Davies and Griffin, 2018). The catastrophic Indonesian volcanic eruptions that occurred in Tambora in 1815 and Krakatoa in 1833 indicate the presence of a potentially more dangerous tsunami source if a larger eruption were to occur. The Krakatoa eruption produced a tsunami whose runup on the coast of Western Australia was in the range of 0.5-2 m (Allport & Blong, 1995).

Figure 6. Probabilistic near shore tsunami wave height in Australia for a water depth of 100m. Source: Davies and Griffin, 2018.


Tsunami impacts that are recorded in historical documents or in geological sand deposits often cannot be easily associated with their source. However, notwithstanding the ambiguity in distinguishing between tsunami and cyclone sources (e.g. Shanmugam, 2012), it seems likely that a 1,000 year old inferred deposit recently discovered in Tanzania can be associated with the AD 900 tsunami at Nagapattinam, India and 14 tsunami deposits in South Asia. This discovery enhances our understanding of what appears to be a major historical transoceanic tsunami in the Indian Ocean. The source of this tsunami is likely to have been the Sumatra-Andaman subduction zone, which is the main source of the moderate level of tsunami hazard in northwestern Western Australia.


Allport, J.K. and Blong, R., 1995. The Australian Tsunami Database.

Atwater, Brian, Musumi-Rokkaku Satoko. Satake Kenji, Tsuji Yoshinobu, Ueda Kazue, and David K. Yamaguchi (2016). The Orphan Tsunami of 1700: Japanese Clues to a Parent Earthquake in North America. University of Washington Press.

Davies, G., Griffin, J. 2018. The 2018 Australian Probabilistic Tsunami Hazard Assessment. Record 2018/41. Geoscience Australia, Canberra.

Jaiswal, R.K., B.K. Rastogi and T.S. Murty (2008). Tsunamigenic sources in the Indian Ocean. Science of Tsunami Hazards, Vol. 27, No. 2, page 47 (2008)

Malik, J.N., Johnson, F.C., Khan, A. et al. Tsunami records of the last 8000 years in the Andaman Island, India, from mega and large earthquakes: Insights on recurrence interval. Sci Rep 9, 18463 (2019).

Maselli, Vittorio, David Oppo, Andrew. Moore, Aditya Riadi Gusman, Cassy Mtelela, David Iacopini, Marco Taviani, Elinaza Mjema, Ernest Mulaya, Melody Che, Ai Lena Tomioka, Elisante Mshiu and Joseph D. Ortiz (2020). A 1000-yr-old tsunami in the Indian Ocean points to greater risk for East Africa. Geology, v. 48, p. 808–813.

NOAA (2020). NGDC/WDS Global Historical Tsunami Database, 2100 BC to Present.

Rastogi, B.K. and R.K. Jaiswal (2006). A catalog of tsunamis in the Indian Ocean. Science of Tsunami Hazards, Volume 25, Number 3, p. 128-143.

Rubin, C. M. et al. (2017). Highly variable recurrence of tsunamis in the 7,400 years before the 2004 Indian Ocean tsunami. Nat. Commun. 8, 16019,

Schäfer, Andreas and Wenzel Friedemann (2019). Global Megathrust Earthquake Hazard—Maximum Magnitude Assessment Using Multi-Variate Machine Learning. Frontiers in Earth Science 7, 136,

Shanmugam, G. (2012). Process-sedimentological challenges in distinguishing paleo-tsunami deposits. Nat Hazards 635–30.