Natural Hazards and Earth System Sciences (2001) 1: 171–175 c European Geophysical Society 2001 Natural Hazards and Earth System Sciences Detailed analysis of tsunami waveforms generated by the 1946 Aleutian tsunami earthquake Y. Tanioka1 and T. Seno2 1Seismology and Volcanology Research Dept., Meteorological Research Institute, 1-1 Nagamine, Tsukuba 305-0052, Japan 2Earthquake Research Institute, University of Tokyo, Tokyo 113-0032, Japan Received: 17 September 2001 – Accepted: 5 December 2001 Abstract. The 1946 Aleutian earthquake was a typi- tsunami run-up exceeded 30 m in height on Unimak Island. cal tsunami earthquake which generated abnormally larger The tsunami also propagated across the Pacific and struck tsunami than expected from its seismic waves. Previously, the Hawaiian Islands, where the run-up exceeded 16 m. The Johnson and Satake (1997) estimated the fault model of this tsunami was observed at tide gauges around the Pacific as far earthquake using the tsunami waveforms observed at tide as South America and the South Pacific Islands. gauges. However, they did not model the second pulse of Johnson and Satake (1997) used the tsunami waveform the tsunami at Honolulu although that was much larger than data recorded at tide gauges to determine the source param- the first pulse. In this paper, we numerically computed eters of the 1946 Aleutian earthquake. The estimated seis- the tsunami waveforms using the linear Boussinesq equa- mic moment of the earthquake was 23 × 1020 Nm, or Mw = tion to determine the fault model which explains the ob- 8.2. They also compare the tsunami waveforms observed on served tsunami waveforms including the large second pulse a tide gauge at Honolulu for three great earthquakes, the 1946 observed at Honolulu. The estimated fault width is 40–60 km Aleutian, 1957 Aleutian and 1964 Alaska earthquakes. The which is much narrower than the fault widths of the typi- peak-to-peak tsunami amplitude observed at the tide gauge cal great underthrust earthquakes, the 1957 Aleutian and the for the 1946 Aleutian event (Ms = 7.4, Mw = 8.2) was larger 1964 Alasuka earthquakes. A previous study of the 1896 than those for the 1957 Aleutian earthquake (Ms =8.1, Mw Sanriku earthquake, another typical tsunami earthquake, sug- = 8.6) or the 1964 Alaska earthquakes (Ms = 8.4, Mw = 9.2) gested that the additional uplift of the sediments near the (Fig. 2). In detail, the waveform for the 1946 event is dif- Japan Trench had a large effect on the tsunami generation. ferent from those for the 1957 and 1964 events. The initial In this study, we also show that the additional uplift of the pulse in the tsunami waveform for the 1946 event is much sediments near the trench, due to a large coseismic horizon- smaller than the second pulse, although the initial pulses for tal movement of the backstop, had a significant effect on the the 1957 and 1964 events are much larger than later pulses. tsunami generation of the 1946 Aleutian earthquake. The es- This difference can be related to the difference in the source timated seismic moment of the 1946 Aleutian earthquake is process of those earthquakes. Johnson and Satake (1997) did 17–19 × 102020 Nm (Mw 8.1). not use the large second pulse recorded at Honolulu to de- termine the fault model of the 1946 Aleutian earthquake be- cause their tsunami computation scheme was inadequate for the later phase. The large depression wave following the ini- 1 Introduction tial pulse was used in their analysis but it was not explained On 1 April 1946, a very unusual earthquake occurred off by the computed wave (Fig. 9 in Johnson and Satake, 1997). Unimak Island in the eastern Aleutians (Fig. 1). The earth- To understand the mechanism of the unusually large tsunami quake generated one of the largest trans-Pacific tsunamis al- generated by the 1946 Aleutian earthquake, we need to ex- though the surface wave magnitude Ms was only 7.4 (Guten- plain the large depression pulse and the following second berg and Richter, 1954). The tsunami magnitude Mt of the pulse recorded on the tide gauge at Honolulu. earthquake was 9.3 (Abe, 1979). The large discrepancy be- In this paper, we try to model the tsunami waveform at tween Mt and Ms categorized this event as a typical tsunami Honolulu including the large second pulse and also tsunami earthquake (Kanamori, 1972), which generates abnormally waveforms observed on tide gauges along the northwestern larger tsunamis than expected from its seismic waves. The Pacific coast. The result allows us to discuss the source pro- cess of the 1946 Aleutian tsunami earthquake. Recently, Tan- Correspondence to: Y. Tanioka ([email protected]) ioka and Seno (2001) indicated that an additional uplift of 172 Y. Tanioka and T. Seno: The 1946 Aleutian tsunami earthquake ft 1946 Aleutian 1957 Aleutian 1964 Alaska 3 60˚ Sitka 1946 Aleutian earthquake 2 Crescent City 40˚ San Francisco San Diego 1 Ms = 7.4 Ms = 8.1 Ms = 8.4 20˚ Mw = 8.3 Mw = 8.6 Mw = 9.2 Honolulu 0 180˚ 200˚ 220˚ 240˚ 16 17 18 18 19 20 8 9 10 hrs, 4/1 hrs, 3/9 hrs, 3/28 Fig. 1. Location of the 1946 Aleutian tsunami earthquake. Trian- Fig. 2. Observed tsunami waveforms for the 1946 Aleutian, 1957 Figure 1 gles show the location of tide gauges where the tsunami waveforms Aleutian, and 1964 Alaska earthquakesFigure 2 as recorded on a tide gauge used in this paper were observed. Tsunami numerical computation at Honolulu after Johnson and Satake (1997). was made in this region. " 2 2 !# 1 ∂ 2 1 ∂ Qϕ ∂ + d + Qθ sin θ (2) sediments near the Japan Trench, due to a large coseismic 3R sin θ ∂ϕ R sin θ ∂t∂ϕ ∂t∂θ horizontal movement of the backstop, had a large effect on ∂Qθ gd ∂h the tsunami generation of the 1896 Sanriku tsunami earth- = − + f Qϕ ∂t R ∂θ quake. In this paper, we also estimate the effect of the addi- " 2 2 !# 1 ∂ 2 1 ∂ Qϕ ∂ tional uplift near the Aleutian trench on the tsunami genera- + d + Qθ sin θ (3) tion of the 1946 Aleutian tsunami earthquake. 3R ∂θ R sin θ ∂t∂ϕ ∂t∂θ where R is the radius of the Earth, g is the acceleration of gravity, d is the water depth, f is the Coriolis parameter, h is 2 Data and method the height of the water displaced from the equilibrium posi- tion, Qθ (= uθ d) and Qθ (= uθ d) are the flow rates in the ϕ- Observed tsunami waveforms on 5 tide gauge stations, at and θ-directions, and uϕ and uθ are the average velocities in Sitka in Alaska, at Crescent City, San Francisco and San the ϕ- and θ-directions. The Alternating Direction Implicit Diego on the west coast of the United States, and at Hon- finite Difference Scheme with a double sweep algorithm (e.g. olulu in Hawaii, are used in this study (Fig. 1). We digitize Kabiling and Sato, 1993) is used to solve the above equa- the records and remove the tidal component. tions (see Tanioka, 2000). The computational area is shown Johnson and Stake (1997) numerically computed the in Fig. 1. The open boundary condition is used at the edge of tsunami propagation by solving the linear long wave equa- the computational area. Total reflection boundary is used at tion. The grid spacing was 5 min of arc except in regions the shoreline. The grid spacing is 1 min of arc except the re- near tide gauges where the grid spacing was 1 min of arc. gion near Honolulu where that it is 20 s of arc (about 600 m). Their computation scheme was inadequate to compute the We use a much finer grid system than that used by Johnson large later pulse of the observed tsunami at Honolulu. In or- and Satake (1997) to compute the later phase of tsunami ac- der to compute the trans-Pacific tsunami more accurately, we curately. solve the linear Boussinesq equation in the spherical coordi- In order to compute tsunami propagation, it is necessary nate system (longitude ϕ and colatitude θ). The governing to estimate water surface initial deformation. In general, equations are the water surface initial deformation is assumed to be the same as the ocean bottom deformation due to faulting of a ∂h 1 ∂Qϕ ∂ large earthquake, because the wavelength of the ocean bot- = − + Qθ sin θ (1) ∂t R sin θ ∂ϕ ∂θ tom deformation for a large earthquake is much larger than ∂Qθ gd ∂h the ocean depth. This assumption may not be made if the = − − f Qθ ∂t R sin θ ∂θ slip is concentrated on a small fault area. We use the equa- Y. Tanioka and T. Seno: The 1946 Aleutian tsunami earthquake 173 observed Unimak Island W=20 km W=40 km 80km W=60 km 1000 m 60km W=80 km L=160 km 2000 m 40km 3000 m 2.4 3.4 6 7 20km 4000 m 5000 m 0 Fig. 4. Comparison of the observedFigure 4 (thick curve) and four computed 6000 m (thin curves) waveforms from the four different fault models that show in Fig. 3. 5000 m 6000 m fault prameters accretionary complex backstop strike=250° dip=6° decollement 10 km rake=90° Figure 5 Fig. 5. A schematic cross-section of the lower trench slope and Fig. 3. Locations of four fault models that have different widths trench wedge seen in the Aleutian Trench. (20 km, 40 km, 60 km andFigure 80 km). 3 wide faults explain the observed waveforms well, includ- tion of Kajiura (1963) to compute the ocean surface initial ing the large second pulse which Johnson and Satake (1997) deformation from the ocean bottom deformation (see Tan- could not model.