Reassessment of the 1907 Sumatra “Tsunami Earthquake” based on macroseismic, seismological, and tsunami observations, and modeling

Abstract

On 4 January 1907, an earthquake occurred off the west coast of Sumatra, Indonesia, with an instrumental surface-wave magnitude (Ms) in the range of 7.5–8.0 at periods of ~ 40 s. The tsunami it generated was destructive on the islands of Nias and Simeulue, where it killed hundreds and gave rise to the legend of the S’mong. This tsunami was also observed in other parts of the Indian Ocean basin. Relative to its instrumented magnitude, the size of the tsunami was anomalous, qualifying the event as a “tsunami earthquake”. However, unusually for a tsunami earthquake, the shaking on Nias was severe (7 EMS). We revisit the 1907 earthquake with a multidisciplinary approach by extracting evidence describing shaking effects or the tsunami from written documents and by acquiring new seismograms. Combining these, we discriminate two large earthquakes within an hour of each other with clear differences in seismological character. The first we interpret to be a tsunami earthquake with characteristic low levels of shaking, an estimated average seismic moment (Mo) of 2.5 × 1028 dyn cm (Mw ≈ 8.2) in the frequency band 6–8 mHz, and an epicentral location close to the front of the Sunda Megathrust. The seismograms we analyzed also document a regular growth of moment with period, approaching MW ≈ 8.4 at the longest resolvable period (~ 170 s). For the second earthquake that caused damage on Nias, we estimate Ms ≈ 7 based on seismograms and phase data. We also identify two Ms ≈ 6 aftershocks within 24 h of the mainshock. Additionally, we present a dataset of 88 locations within the Indian Ocean basin where the tsunami was observed. Using a subset of these, we forward modeled the tsunami to propose a seismic rupture model extending along the Sunda Megathrust for about 220 km (~ 94.7°E to ~ 97°E) with a maximum modeled slip of ~ 21 m. Our new rupture model provides an acceptable fit to our new dataset of tsunami runup and inundation values from 88 local and far-field locations in the Indian Ocean basin. We also urge caution against an over-reliance on the S’mong legend for tsunami evacuation as its premise, that a tsunami will only follow an earthquake with very severe ground motions, is rendered ineffective for tsunami earthquakes.