SLIDE 1 The 2011 Tohoku Earthquake — from Disaster to Knowledge a Presentation by Professor Jonathan Stewart of UCLA
Chengdu, July 1, 2011
1 Preface
These are the notes I took during Professor Stewart’s presentation. I give no guarantee for accuracy or correctness. Any mistakes and errors in these notes are my own and probably result from my misunderstanding what Professor Stewart said. Please send any corrections to ute.platzer@durham.ac.uk. Ute Platzer, Chengdu, July 2, 2011
2 Introduction
Professor Stewart is a professor of Civil engineering at the University of Cali- fornia in Los Angeles. He is working on geotechnical engineering, especially seismic engineering. He is part of the GEER team, a team of scientists that is based in the United States and goes from there to those parts of the world where disasters happen, to study them and learn from them. The first group from the team went to Japan 2 weeks after the earthquake, and the last group
- nly left a few weeks ago. They always work together with local people and
local experts. They also did that during then Wenchuan earthquake. They never go into the field on their own. 1
SLIDE 2 3 The earthquake
The Tohoku earthquake happened on March 11, 2011. There was foreshock activity: on March 9 there was one foreshock with MW 7.2, and on March 11 there were more, three of them with magnitudes > 6. Not all earthquakes have foreshocks, and you know only after the mainshock happened whether an earthquake is a foreshock or a mainshock. The Pacific plate moves north-east, subducting beneath the North-Ameri- can plate (which actually extends from North America and the Arctic towards Japan, see figure 1). The convergence rate/movement is up to 8 cm year−1, which is very much. Due to the earthquake, there was very high horizontal displacement of up to 4.5 m eastward movement on the east side of Japan, while the west side of Japan moved only about 1 m. Therefore, Japan got wider by 3.5 m during this earthquake. Vertical displacement was only observed in the east of Japan; it moved down by about 1 m, making the effect of the tsunami worse. Along the slip there was horizontal displacement of up to 32 m, which is really huge.
4 Anticipation and Prediction
Could this earthquake and its consequences have been foreseen? Since 1973, the beginning of instrumental measurement, there were 9 earthquakes on the Japan trench with MW > 7. There are also well-docu- mented events in recent geologic history in the Sendai area:
- 1896, a MW 7.6 event created a tsunami with a run-up height of 38 m.
- 1933 a MW 8.6 event resulted in a tsunami with 29 m run-up height.
Both of these events happened north of Sendai, and Sendai itself was pro- tected from the full force of the tsunamis by a small peninsula. Therefore, the expected tsunami height in Sendai was only 4 m. This was obviously a misinterpretation of the available data.
- In the year 869, a MW 8.3 event created a tsunami with similar extents
to the one observed in 2011. See the paper “Entire coastal plane dev- astation in Sendai city” (Okumura 2011). Tsunamis always leave de- posits because they carry a lot of mud and debris. Shortly after the 869 tsunami, there was a volcanic eruption, covering the deposits with ash 2
SLIDE 3
North−American plate Pacific Ocean Japan North America A B Hanging wall Footwall North−American Plate West East Pacific Plate Figure 1: A, location of the North-American and Pacific plates. B, sub- duction zone of the Japan trench. 3
SLIDE 4 inundation zone run−up height inundation depth tsunami wave normal sea level
Figure 2: Run-up height, inundation depth and inundation zone of a tsunami. and thereby protecting and preserving them. That is why we can use them today to determine the extent of that tsunami.
- In 1611, there as another big earthquake which is not well-documented.
It created a 6-8 m tsunami in Sendai and Fukushima. Experts concluded that the worst earthquake to be expected in the region was MW 7.4 – 8.2. Historical data was not taken into account.
5 Tsunamis
5.1 How is a tsunami generated?
It is created when the sea floor (the hanging wall of a fault) moves upward. This movement happens in several small steps, and that is why a tsunami always consists of several waves and not only one single wave. The speed of the waves is comparable to that of a Boeing 747 airplane. Some definitions (see figure 2):
- The run-up height is the elevation at the inundated point which is furthest
away from the shore, minus the sea level elevation.
- the inundation depth is the water level at a specific point.
- the inundation zone is the area which is inundated.
In contrast to earthquake damage which is spotty, tsunami damage is com- plete: a tsunami destroys everything in its path. 4
SLIDE 5 5.2 The effects of the tsunami
The tsunami reached a run-up height of 10 m in Sendai, in contrast to the predicted 2–4 m. In the north, the tsunami reached a run-up height between 34–38 m. In the south near Tokio, it was only 2–5 m, and Tokio itself did not experience any tsunami. It was believed that reinforced concrete buildings on piles are tsunami-safe, and it was even suggested that people who cannot reach higher ground seek refuge on top of this kind of building instead. However, in this tsunami, many reinforced concrete buildings were thrown off their foundations and toppled
- ver. They ended up on their side (see figure 3 A). In one case, such a building
was moved 23 m by the wave. Perhaps the existing theory on tsunami effects is wrong because experiments are made with clear water, while the water of the tsunami is obviously dirty and laden with debris.
6 The Fukushima nuclear power plant
The plant was built in 1971. It has 6 reactors. In the US, nuclear power plants usually have only 1 or a maximum of 2 reactors per site. The first reactor, built in 1971, was designed to withstand peak ground acceleration (PGA) of 0.18g, based on an earthquake that happened in California in the 1950s. Reactors 3 and 6 were built to newer standards to withstand 0.45g. This plant was protected by seawalls, because the tsunami risk was known. The design tsunami was 5.7 m, which was more than the worst expected tsunami of 2–4 m. It seemed to be very, very safe.
6.1 Pre-event state of the reactors
Reactor 5+6 were down for maintenance, i.e. not producing electricity but they still required cooling. Unit 4 was de-fuelled, and 1–3 were operating normally.
6.2 Safety systems at the plant
Then the earthquake hit at 14:46 local time, it hit with a PGA between 0.4 and 0.5 (the exact values have not been released). This triggered an automatic shutdown in reactors 1–3. After that, external energy was required for cooling and to run the control electronics (usually the energy for this is generated by the reactor itself). 5
SLIDE 6 Figure 3: Effect of a tsunami on reinforced concrete structures. They topple over. They had a plan B for this, and this is to use the National Power Grid to get electricity from somewhere else. However, the grid was damaged by the earthquake and there was no electricity. Plan C then were a set of Diesel generators to produce the electricity. There were 3 sets of redundant generators in case one set failed. At 15:41 however, the tsunami hit and destroyed all of the Diesel generators. The tsunami reached a height of 12–14 m, more than double the expected 5.7 m. For that case, there was Plan D, the use of backup batteries to provide electricity for 8 hours, but 8 hours were too short to repair the Diesel gen-
- erators. Therefore they called for backup generators to be brought in, which
the got there after 13 hours. But they could not be connected because of the damage to the plumbing etc. After that, nuclear meltdown happened in units 1–3 etc. Question: why did the cooling system work with the backup batteries, but not with the generators? The problems at the plant remain unsolved. The reactors still require con- stant cooling. It’s a difficult situation because the radiation makes it nearly impossible to do any repairs.
7 Risk assessment
7.1 Types of uncertainty
- Aleatory uncertainties: known unknowns. Example: PGA values. We
know there’s some uncertainty when predicting PGA values, but we know how large this uncertainty is.
- Epistemic uncertainty: we just don’t know.
Such as forecasting BIG earthquakes: we have not had enough big earthquakes yet to know enough about them to make predictions. We cannot know how big the errors are in predicting them. 6
SLIDE 7 7.2 Lessons for risk assessment
Keep risk assessments and facilities up-to-date with current knowledge. What appears to be safe in 1970 may no longer be safe when you look at it again in 2011 with the new knowledge collected in between. Make sure you don’t have linked vulnerabilities: multiple backup systems that are vulnerable to the same event. Ground motion data of the earthquake is freely available. there were 693 K-Net instruments, 496 Kik-Net instruments and 50 instruments on buildings that measured PGA during the earthquake. On 16 of these instruments, PGA values of more than 1.0g were recorded, and the maximum seen was 2.99g. The astonishing thing about this is that these values were observed more than 50km away from the fault! Normally you expect these values only very close to the fault. But they did not have any instruments closer than 50km to the fault. There were 3 individual rupture events during the earthquake. But each single seismic station picked up only 1 or 2 of these, because of the location
- f the stations – it’s very unlikely that a station will have been in a position to
register all three events. The 2010 Maule, Chile earthquake and the 2011 Tohoku earthquake offer unprecedented and interesting ground motion data on earthquakes with mag- nitudes larger than MW 8.0 Existing theories claimed that ground motion/PGA does not increase a lot for magnitudes > 8.0, but these earthquakes showed that it does indeed increase a lot! Differences in ground motion between soil and rock. There was at least
- n pair of stations within 2km of each other, one on solid rock and one on soil
(10 m deep or so). The PGA on soil was > 1.0g, while that on rock was only 0.4g, so these data offer interesting insights into the differences between rock and soil and the site effects. There is also new data on liquefaction available now. This is only the 3rd dataset available for ground motion at a liquefaction site.
8 Foundation failure through liquefaction
Wide region settlement. Liquefaction took place on a wide area, and individual buildings settled and tilted at different angles because of soil inhomogeneities. Especially heavy buildings with 3 or more stories sink into the ground, for about 20–50cm (see figure 4 B). Regional settlement after liquefaction. Buildings on piles do not settle or tilt, but the soil surrounding them settles, making the building “stick out” (see 7
SLIDE 8
B A
Figure 4: Effect of liquefaction on buildings. A, building on piles. The ground settles around the building, while the building itself re- mains where it is. The ground can settle by 20-50cm. B, buildings without piles sink in, because they move in the shak- ing and the earth softens/liquefies beneath them because of changes in shear stress and pore pressure. Buildings typically settle by about 50cm, and mostly, buildings of 3 stories or more are affected. figure 4 A). All (most) buildings in Japan have a thick foundation of about 1 m concrete. This worked well; the buildings tilted and moved but their structure was unaf- fected, they did not collapse. This is a very important construction principle in earthquake-prone and liquefaction-prone regions. Flotation of buried structures such as sewer pipes, storage tanks... when the soil liquefies, these structures tend to float to the surface. Relationship between PGA and soil density: there’s a curve describing when liquefaction takes place (see figure 5). It is based on very little data for large earthquakes, but this event added some more data to it, which is very good for the prediction of design standards for these big earthquakes.
9 Dam failure
The main shock did not create a surface rupture, but on 11.4.2011, a MW 6.6 aftershock created a surface rupture with 2 m normal fault displacement. There were not many landslides, but 1 dam failure. This is a good result when considering that several 100s of dams were shaken. It is not clear why this dam failed, it was built of clay so liquefaction cannot have been a factor. The most probable explanation today is that repeated cycles of shaking in- duced a landslide on the dam, so part of the dam slid downhill and then the remainder was just too weak to withstand the water pressure. The dam was 8
SLIDE 9 Soil density Ground shaking PGA before Tohoku: not much data available for large earthquakes liquefaction no liquefaction
Figure 5: Relationship between PGA and soil density in the occurrence
totally scoured and the lake behind it emptied, creating a flood wave that killed 9 people. It is believed there are eye witnesses but for cultural reasons they have not been interviewed yet. But it is hoped that more can be learnt on the failure mode of this dam. It did not fail immediately during the earthquake be- cause someone was apparently driving right over it during the quake. There were also subsequent drawdown failures observed at that site. Most levees performed well, mostly because the soils had been dry and there was not much liquefaction. Especially structures reinforced after the 2003 Miyagi earthquake performed well, while there was some disintegration going on in older levees. 9
