2007 - university of washington
TRANSCRIPT
Parsing the Sandy Onshore Deposits of Modem Tsunamis
Bretwood Higman
A dissertationsubmitted in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy
University of Washington
2007
Program Authorized to Offer Degree:Department of Earth and Space Sciences
University of WashingtonGraduate School
This is to certify that I have examined this copy ofa doctoml dissertation by
Bretwood Higman
and have found that it is complete and satisfactory in all respects,and that any and all revisions required by the final
examining committee have been made.
Chair of the Supervisory Committee:
Reading Committee·~~
IS
Brian Atwater
~fl2Vasily Titov
Dateo f5: :r~Or
,,
In presenting this dissertation in partial fulfillment of the requirements for the doctoraldegree at the University of Washington, I agree that the Library shall make its copiesfreely available for inspection. I further agree that extensive copying of the dissertation isallowable only for scholarly purposes, consistent with "fair use" as prescribed in the U.S.Copyright Law. Requests for copying or reproduction of this dissertation may be referredto ProQuest Information and Learning, 300 North Zeeb Road, Ann Arbor, MI 481061346, 1-800-521-0600, or to the author.
Signature~Date_'b 0",\J\R.. 8-'7fl'1-
, ,
University of Washington
Abstract
Parsing the Sandy Onshore Deposits of Modem Tsunamis
Bretwood Higman
Chair of the Supervisory Committee:Professor Joanne Bourgeois
Department of Earth and Space Sciences
Vertical and lateral changes in grain size of sandy onshore deposits can aid in
reconstructing the shape and sequence of waves in a tsunami wave train. I ascribe
upward fining to decelerating flow, upward coarsening to acceleration, and horizontal
variation in vertical size grading to horizontal gradients in flow. I plot resulting
predictions on diagrams that depict continuity ofdeposition through time for deposits and
for modeled flows. Then I explore three specific applications. First I present evidence
that long·period tsunamis of the largest earthquakes produce proportionally less upward
fining than do shorter·period tsunamis from the smaller earthquakes. This evidence
includes particle·size analysis of 864 slices from deJX>sits of tsunami associated with five
earthquakes (1992 Nicaragua, Mw 7.6-7.7; 1998 Papua New Guinea Mw 7.1; 2001 Peru.
Mw 8.4; 2004 Indian Ocean, Mw 9.15, and 2005 Sumatra, Mw 8.6). Deposits of
tsunamis from the larger among the five earthquakes more often include basal sections of
upward coarsening that imply gradual acceleration, small intervals of upward coarsening
that imply interruptions to gradual deceleration (surges), and sharp coarse-over-fine
contacts best explained by arrival of later waves. Next I deduce the onshore stratigraphic
signatures of a seaward-directed bore that fonns when a steep-fronted tsunami inflow
reflects offan inland slope. I predict an upward·fining, seaward-thinning bed that drapes
local highs near the reflector and begins at a sharp basal contact on coarser, inflow sand
farther seaward. Finally I interpret the 2004 tsWlami deposit on a beach-ridge plain with
• •
a steep slope at its inland limit at Bang Lut, near Khao Lak, Thailand, as a record of a
wave train of at least three waves. Sand from the first wave is preserved only on the
seaward halfof the plain. Next came a larger wave that included two superimposed
surges. The second surge inundated all the plain and formed a reflected bore that draped
the plain with a seaward-thinning, upward-fining bed. The second wave concluded with
still water marked by the only widespread silt layer. Deposits near the beach record a
later wave that included at least two additional surges.
• •
TABLE OF CONTENTS
Page
List of Figures iv
List of Tables vi
Introduction I
Background 3
Tsunanli Sources 3
Tsunami Dynamics 4
Applications ofTsunami Deposit Studies 5
Parsing tsunami deposits 5
Approach 6
Sedimentary Model 6
Caveats 8
Formulation of Derivatives 10
Continuity Diagrams II
Case Studies 13
Sandy Clues to Tsunami Period (Chapter I) 13
Predicted Sedimentary Record of Reflected Bores on Coastal Plains Overrunby Tsunamis (Chapter 2) , 13
Sand-Sheet Stratigraphy Recording the 2004 Indian Ocean Tsunami on aCoastal Plain in Thailand (Chapter 3) 14
Chapter I: Sandy Clues to Tsunami Period 20
Hypothesis 20
Observations 20
Model 21
Caveats 23
Conclusions 24
Chapter 2: Predicted Sedimentary Record of Reflected Bores on Coastal PlainsOvemlll By Tsunamis 31
Introduction 31
Predictions 32
I
••
Simulated Deposits of Reflected Bores 33
Deposits Near Reflector (location C) 34
Deposits at an Intermediate Distance from the Reflector (location B) 34
Deposits far from Reflector (location A) 35
Field ExaDlples 35
Discussion 36
Conclusions 36
Chapter 3: Sand-sheet Stratigraphy Recording the 2004 Indian Ocean Tsunami on aCoastal Plain in Thailand 47
Introduction 47
Methods 48
Eyewitness Accounts 48
Field Surveys 48
Grain Size Analysis 49
2004 Tsunami Deposit in Thailand 49
Propagation and Height 49
Human Records of Multiple Waves 50
Images of Wave Fronts and Surges 51
2004 Deposit at Bang Lut 52
Thickness and Volume 52
Stratigraphy and Architecture 52
Pre·tsunaDli Soil and Beach 52
Unit A 53
Unit B 53
Unit C 54
Sedimentary Structures 54
Inferred Sequence of Flows 55
First Wave S5
Second Wave 56
First Surges 56
Reflected Bore 57
u• •
Still Water 58
Withdnlwal 58
Later Wave or Waves 59
Implications 59
Conclusions 61
References 70
Appendix A: Grading Data 78
Appendix B: Statistical Results 79
Pocket Material:
Stratigraphic Map
Eyewitness Photos and Videos
IU.,
LIST OF FIGURES
Figure Number Page
1. Sedimentary Model 16
2. Explanation ofa Continuity Diagram 17
3. Using a Continuity Diagram to Summarize Model Output 18
4. Comparison Between the Continuity Diagrams 19
5. Example Deposit 25
6. Moment Magnitude vs. Proportion Upward Fining 26
7. Earthquake Rupture Areas 27
8. Inverse Grading That Records a Gradual Acceleration 28
9. Inverse Grading That Records Minor Periods of Acceleration 29
10. Rapid Accelerations Accompanying the Arrival of Later Waves 30
11. Photos of the 2004 Indian Ocean Tsunami 38
12. Idealized Cross-section of a Reflected Bore 39
13. Cartoon of a Depth-integrated Reflected Bore 40
14. Reflected Bore Modeled Using Coulwave 41
IS. Multiple Reflected and Incident Bores From a Wave Train Modeled by Delft3D 42
16. Deposits Fanned by a Cartoon Reflected Bore 43
17. Deposit Fonned by a Reflected Bore as Modeled by Coulwave 44
18. Deposit of a Wave Train Including Reflected Bores 45
19. Profile Inundated by the 1992 Nicaragua Tsunami 46
20. Index Map Showing Relation Between Tsunami Source and Thailand Field Area .. 62
21. Coastline and Bathymetry Near Bang Lut 63
22. Mapped Structure of the Tsunami Deposit at Bang Lut 64
23. Trench 52m From Top of Beach 65
24. Trenches 320m From Top of Beach 66
25. Thickness of Unit B Ploned as a Function of Elevation 66
26. Summary of Observations and Interpretations of the Tsunami Deposit.. 67
.v, ,
27. Sequence of Erosion and Deposition Likely in the First Wave 68
28. Fom18tion ofa Reflected Bore Deposit 69
27. Sequence of Erosion and Deposition Likely in the First Wave 68
28. Fonnation of a Reflected Bore Deposit 69
v, ,
LIST OF TABLES
Table Number Page
I. Transport Properties of Sand in Two Example Tsunami Flows 15
2. Statistical Results 80
VI
••
ACKNOWLEDGEMENTS
Bruce Jaffe provided grain size analyses from the 1998 Papua New Guinea
tsunami and the 2001 Peru tsunami used in Chapter I. Patrick Lynett and Guy
Gelfenbaum provided numerical model output used in Chapter 2. Saidul Alam and
Ulrich Glawe initiated the project described in Chapter 3, and collected the data presented
in Figure 22 during the October 2005 survey, at which I was not present. These people,
and many others, including but not limited to lady Bourgeois, Brian Atwater, Vasily
Titov, Erin McKittrick, Bre MacInnes, Beth Martin, Andrew Mattox, Andrew Moore,
and Jeff Parsons contributed greatly by helping my develop and communicate the ideas I
present here.
vn• •
I
Introduction
On most coasts thai they threaten, tsunamis happen too rarely for their hazards to
be defined by written history alone. Particularly in the aftennath of the 2004 Indian
Ocean tsunami, which lacks known historical precedent, the sedimentary deposits of past
tsunamis are accordingly gaining use as guides to tsunami risk. Such geologic records
can help define tsunami hazards while also clarifying the associated hazards from
earthquakes, landslides, and volcanic eruptions at tsunami sources (Bourgeois, in press).
In addition, as explored in this dissertation, tsunami geology has potential to reveal the
inner workings ofthe tsunami waves themselves.
Tsunamis leave many kinds ofgeologic records onshore. As it moves into
shallow water, a tsunami can generate strong currents that enlrain everything from mud to
boulders (e.g. Eaton et aI., 1961). Satellite images of the effects of the 2004 tsunami
show depressions and channels scoured in coastal plains (Higman et al. 2005). Tsunami
borne mud, sometimes laced with shredded vegetation, can blanket coastal plains (my
own unpublished observations) and settle to the bottoms of coastal lakes (Bondevik et aI.,
1997; Kelsey et aI., 2005). More so than mud or boulders, however, sheets of sand have
provided widespread records of past tsunamis.
Sand is well suited to record a tsunami. The grains are coarse enough to contrast
with the S9ils in which tsunami deposits are commonly preserved, yet they are fine
enough for a sample of a few grams to be statistically representative. Sandy deposits
commonly include grading and layering that probably record temporal variation in the
flow. The settling velocities ofsand grains-in the range 0.3-30 cmls (Ferguson and
Church, 2004}--make sand sensitive to variation in the flow speeds typical of tsunamis
(Table I). In two different but plausible tsunami flows (Table t), very coarse sand can
settle through the height of the water column in less time than the flow takes to move 500
meters, while very fine sand cannot. Also, Rouse numbers for sediment in these two
flows span a range of values where sand is usually suspended, but may include bedload
conditions (Rouse number> 2.5 Middleton and Southard, 1984).
.,
In this dissertation I demonstrate that a simple sedimentary model applying
intuitive ideas about sediment transport can be used to further our understanding ofeven
very complex deposits. In three chapters I explore both general patterns in tsunami
deposition and specific features measured in particular deposits. This work is founded
both in observations of relatively well understood modem tsunamis, and in the output of
hydrodynamic tsunami models.
, ,
2
3
BackgroundTsunami sources
Tsunamis can carry infonnalion about the event that generated them (e.g. Fujii
and Satake, 2007) to coastal depositional settings. Tsunamis relain substantial
information about lheir source as they move from source to shallow water because they
are nearly linear in water depths greater than a few tens ofmeters (Synolakis and
Bernard, 2006). As they move into shallow water, much infonnation should still be
retained despite nonlinear effects (see Tsunami Dynamics) jfthe time required to cross
shallow water is short compared to tsunami period, or if bathymetry and topography are
known and simple enough that their effect on a tsunami is nOI 100 strongly dependent on
small variations in initial wave structure. Tsunamis have their greatest impact. and thus
leave their clearest records, in areas readily accessible by geologists-near shore and
onlandjust above sea level.
If a tsunami carries information from offshore source to onshore sink, the
tsunami's onshore deposits may contain clues to the earthquake, landslide. volcanic
eruption, or bolide impact that generated it. Such tsunami triggers can be difficult to
discern in the geologic record due to inaccessibility or lack of preservation. For example.
the largest earthquakes occur in subduction zones (Kanamori, 1977) on faults that
primarily rupture the earth's surface at ocean depths of several kilometers, whereas the
deposits from their resultant tsunamis are conveniently located onland, in some cases
stratigraphically associated with evidence for tectonic land-level change (review:
Bourgeois, in press). Also, though difficult to measure directly, the rate and sequence of
sliding in submarine landslides have been estimated by comparing hypothetical tsunami
runup to minimum runups from evidence in the fonn of tsunami sediment in lakes
(Bondevik, "2005). Finally, bolide impacts are random on earth and thus lest often occur
on continental crust than over the more common oceanic crust (Hagstrum, 2005) where
their craters are difficult to identify and are usually subducted within a hundred millions
years. The deposits of tsunamis generated by these bolide impacts (e.g. Bourgeois et al.
'.
41988; Dypvik et aI., 2003; Gliksan, 2004) are more likely than craters to be preserved
due to their broad geographic distribution on the edges of longer-lasting continental crust.
Tsunami dynamicsThe abrupt displacement of water that leads to a tsunami can be complex in shape
(see examples ofearthquake sources in Fig. 7 in Chapter I) so the resulting wave train
can be similarly complex (Tilov and Synolakis. 1998; Synolakis and Bernard, 2006)Also,
tsunami waves evolve in shape and interact with one another in shallow water, so the
shape of attacking waves varies (Lynctt et aI., 2002). Wave fronts are sometimes steep
and breaking, other times gradual like a fast-rising tide (see Eyewitness Photos and
Videos). Usually the first positive wave is preceded by a negative wave that can leave
hundreds of meters ofsea floor exposed to the onrushing tsunami (Srinivas and Synolaks,
1996; also Eyewitness Photos and Videos). When they move onland. these long-period
(minutes to tens of minutes) waves (Mirchina and Pelinovski, 1981) produce flooding for
hundreds of meters or even kilometers inland (see for example Eyewitness Photos and
Videos, and Fig. 21 in Chapter 3). Overland water velocilies caused by tsunamis can be
meters per second and probably even tens of meters per second (e.g. Fritz el aI., 2006).
Fault rupture leads to tsunamis by abruptly deforming the seafloor and thus
displacing the overlying water column. The resuhing slope on the surface of the ocean
generates a horizontal pressure gradient that accelerates waler toward areas of lower
pressure. This disturbance propagates as waves that are commonly longer than water
depth even .in the deep ocean; such waves conform to shalJow·water wave theory
(Synolakis and Bernard, 2006). Shallow·water wave theory predicts that a wave moves
with a velocity c (celerity) related to the height of the water column: c::: (gh)1f2 , where
9 is the acceleration due to gravity, and h is the height of the water column.
Wh~n a tsunami is offshore, the height ofthe wave is insignificant compared to
the total height of the water column. Under these conditions independent waves can
cross each other without interacting, so the equations for wave propagation are linear.
The shape of straight-<:rested waves in uniform depth does not change through time
(Titovand Synolakis, 1998).
'.
5As a tsunami approaches a coast, the height of the waves becomes a significant
(0.1) portion of total water depth; the wave motion becomes nonlinear. High waves, by
moving faster, overtake and coalesce with lower waves. Similarly, preferential advance
of the wave crest creates asymmetry that shortens the times of s of rising water while
lengthening the intervals of water·level fall (TilOY and Synolakis, 1998). For example
where wave height is equal to water depth, the wave crest will move at 1.4 times the
speed oflhe leading edge. Moreover, sleeper sections of waves can disperse into
multiple short-period waves or they can break-phenomena not predicted by linear or
nonlinear shallow-water wave theory (Lynen et aI., 2002). The short-period dispersive
waves can increase a tsunami's capacity to move sediment onshore through pulsing of
onland flows and breaking that increases turbulence.
Applications oftsunami deposit studiesTsunami deposits provide evidence ofpast events that could recur. and this
evidence has influenced knowledge of hazards on several coasts (Bourgeois, in press).
For example, in eastern Hokkaido, Japan, tsunami deposits provide evidence of tsunamis
largerthan those that have occurred in this area's two centuries of written history
(Nanayama et aI., 2003; Nanayama et aI., in press). On the west coast of the US and
Canada, where written history is similarly brief, tsunami deposits helped demonstrate the
occurrence ofgreat earthquakes at the Cascadia subduction zone (Atwater. 1987; Witter
et aI., 2003; Kelsey et a1., 2005; Peters et a1., in press). On coasts around the northeastern
Atlantic, ts~i deposits record an landslide-generated tsunami (Bondevik et a1., 1997;
Smith et aI., 2004).
Tsunami deposits can also help to define the boundaries of tectonic plates. In the
southwest comer of the Bering Sea, tsunami deposits have been used to estimate tectonic
convergence rates, and these suggest active convergence in this area where plate
boundaries are uncertain (Bourgeois et aI., 2006). Recent work by Martin and MacInnes
constrains rupture location and geometry using the geographic distribution of tsunami
deposits along the coast of Kamchatka (in progress).
Parsing tsunami deposits
6Details of the shape, number, and height of tsunami waves are important because
these details are a function of the tsunami source geometry and timing. The most readily
apparent details of tsunami deposits are usually grading and layering (e.g. Konno et aI.,
1961; Kelsey et aI., 2005; Higman and Bourgeois, in press). Layers revealed by changing
grain size have been interpreted as records of successive waves, surges, or pulses of
deposition (e.g. Atwater and Hemphill~Haley,1997; Kelsey et aI., 2005; Kilfeather et aI.,
2(07). The layers can even be scrutinized in thin section (Kilfeather et aI., 2007).
ApproacbSedimentary model
I seek to predict and interpret stratigraphic relationships among sedimentary
structures that result from evolving tsunami flows. To do this, I apply a sedimentary
model developed for interpreting turbidites and ignimbrites, by which a flow's sediment
carrying capacity and competence to carry coarser grain size distributions vary in both
space and time (Kneller, 1995). My model resembles that of Kneller (1995). except that I
applied it to tsunamis and specifically allow for reversal in flow. Also, in chapters 2 and
3, I go beyond this model to consider the time needed for grains to settle or erode. In
both Kneller's model and mine, faster flows carry more and coarser sediment. These
models consider only relative change in grain size and flow speed. This simplification
provides flexibility in applying the models.
My sedimentary model can be applied qualitatively either as a forward model, to
predict tsunami deposits given a particular tsunami flow, or as an inverse model. to infer
flow from deposits. As a forward model it predicts general spatial relationships among
beds, and grading within beds, but it does not specify bed thickness or grain size. As an
inverse model it gives patterns of flow such as the abrupt change in speed associated with
a bore, but not specific flow velocities or time periods.
Other approaches to modeling tsunami sediment transport do manage more
complete forward or inverse modeling, but these approaches rely on simplificatjons that
as yet prevent them from being applied to interpret the internal structure of tsunami .
deposits. Only a handful of sediment-transport models for tsunamis have been published
thus far (Gelfenbaum et aI., 2007; Soulsby et al., 2007; Jaffe and Gelfenbaum, in press).
• •
7Two of these are inverse models that use a deposit's grain size and geometry to infer
properties of the flow (Soulsby et aI., 2007; Jaffe and Gelfenbaum, 2007). None of the
inverse models, however, account for the internal variability I have found in deposits I've
studied (Fig. 6 in Chapter 1. Field Ex.amples in Chapter 2, as well as Chapter 3). The
forward model of Gelfenbaum et aI. (2007) applies sediment transport theory based on
flows that are steady and uniform. However. a real tsunami flow will be neither steady
nor uniform and so may in some cases may not be well modeled using this approach.
Tsunamis include unsteady phenomena such as bores (Yeh, 2006) and generate
nonuniform flows over varied topography (see Eyewitness Photos and Videos). Also, in
order to encompass the large spatial scales of tsunami sources and tsunami propagation in
reasonable computing time, Gelfenbaum et aL (2007) and some other unpublished
modeling efforts are run at resolutions often meters or more, too coarse to resolve
depositional features on the horizontal scale of centimeters. Finally, forward modeling
requires specification of unknown boundary conditions such as the packing condition and
grain-size distribution of the sediment source.
My sedimentation model, though it does not specifically quantify important
factors such as bed shear and turbulent mixing, uses the following concepts in order to
predict and interpret sedimentary structure in tsunami deJX>sits. First. relatively strong
flows (faster flow, higher shear, stronger turbulence) carry relatively coarser sediment
(Komar, 1996). Second, stronger flows also carry more sediment. so accelerating flows
scour the bed, and decelerating flows deJX>sit. Also, over irregular beds, more nearly
steady uniform flows fill depressions with sediment while eroding al local highs. In
chapter 2, where I apply my model quantitatively, I use flow speed to approximate this
general concept of flow strength with flow speed (Fig. I).
My sedimentation model is designed to simulate deposition by flows, like those of
tsunamis, that can quickly change velocity (a vector quantity, having a direction) and
speed (the scalar magnitude of velocity). For example, using an eyewitness video o{the
2004 Indian Ocean tsunami flowing along a city street more than three kilometers inland
in Banda Aceh, Fritz et al. (2006) measured flow velocities that increased by 1.5 mls in
less than 30 seconds. Other flows shown in other eyewitness videos of this tsunami show
, .
8nearly still water accelerated to speeds likely over 10 meters per second injusl
moments, an acceleration more than an order of magnitude higher than that measured by
Fritz et aI. (2006) (see Eyewitness Photos and Videos). As was discussed at the 2007
Tsunami Sedimentology workshop at Friday Harbor, Washington
(http://earthweb.ess.washington.edultsWlami2ldepositsitsunamisedimento]ogy.htm), and
the 2006 Tsunami Fundamentals workshop in Hila, Hawaii
(http://tsunamLorst.edu/workshop/), modeling sedimentation in such flows is a difficult
and open problem. My model seeks to circumvent the difficulties inherent in a complete
description of tsunami sedimentation by only predicting trends in sedimentation rather
than numerical estimates ofgrain size and sediment thickness. The model is likely to fail
where totally accelerating flows (see Formulation of Derivatives) are depositional or
carry progressively finer sediment, or where totally decelerating flows erode the bed or
carry progressively coarser sediment.
Caveats
In this section, I explain how scenarios where totally accelerating flows are
depositional or carry progressively finer sediment, or where totally decelerating flows
erode or carry progressively coarser sediment can arise and how they affect the
conclusions in this dissertation. The scenarios are rare, and their occurrence does not
substantially alter any ofthe results discussed in my dissertation.
In slightly accelerated flows well approximated by a steady and unifonn flow, it is
possible for velocity to increase while shear stress decreases, leading to a flow that
becomes less capable of carrying coarse sediment despite an increase in average speed.
Steady unifonn flows have often been approximated using the Law of the Wall
(Middleton and Southard, 1984, Bradshaw and Huang, 1995):
uz'" ~ 1b In(:JWhere K = 0.41 from experimental data, Uz = velocity at a certain height off the bed (z),
Zo is the roughness height, "t is shear stress, p is fluid density. Depth·avcraged speed can
be calculated by integrating across z through the entire now depth (h):
••
9h
U"(UzdZJ"
Thus, for steady uniform flows u2 is proportional to t; higher velocity means higher shear
stress. However, at face value it appears that u might increase with t held constant if h
were simply increased. But in steady uniform flows driven by a gravitational pressure
gradient under a sloping fluid surface, t is proportional to h:
t = pghS
where 9 is acceleration due 10 gravity and S is surface slope. Thus for increase in both h
and U, t must increase as well unless a decline in S makes hS constant or decreasing.
However, ifhS continues to de<:rease, Uz and thus u will decrease due to decreased t.
Thus there is only a narrow band ofconditions that allow a gradually (nearly steady
uniform) accelerating flow to avoid increasing shear stress on its bed, or conversely allow
a gradually decelerating flow to avoid exerting decreasing shear stress on its bed. Since
any occurrence of these conditions would likely only be brief, since almost any change in
conditions would cause these conditions to cease, I think such occurrences are unlikely to
be a major factor affecting my analysis.
Flows that are strongly accelerated, which are likely poorly predicted by the Law
ofthe Wall, will also rarely enter conditions where decelerating flows become stronger,
or accelerating flows become weaker. Under conditions where the derivative of speed is
continuously positive or continuously negative (flows are neither transitioning from
accelerating to decelerating, from decelerating to accelerating, or reversing direction),
only some narrow band of conditions similar to those described for the gradual
acceleration case can lead to increasing speed with decreasing turbulence.
Flows transitioning from accelerating to decelerating will experience a mismatch
between speed derivatives and change in strength, but this mismatch will not change the
depositional record. When a flow passes from accelerating to decelerating, the
turbulence in the flow will continue to increase until turbulent dissipation is equal to. the
production of turbulence (Middleton and Southard, 1984). This increase in turbulence
will lead to an increase in shear stress, since the turbulence will mix an increasing amount
ofmomentum from high in the flow down to the bed. However, given continued
10deceleration and decreasing production of turbulence, turbulence and shear stress will
soon begin dropping off, so though a deposit should record the change from accelerating
to decelerating at a different time (and stratigraphic position) than it occurred, it should
record it nonetheless. This situation probably occurs under reflected bores (Chapter 2).
Only in cases where the flow briefly transitions to deceleration, but then returns to
acceleration, would a complete record not match the actual variation in speed in the
sequence of changes in flow acceleration and deceleration.
To this point, I've assumed that grain size correlates with flow strength. but it is
possible for a flow to include changes in grain·size that do not record changes in flow. In
particular, if the sediment source changes through time, the sediment transported will
change as welL Such changes are likely as a tsunami erodes (See for example Chapter 3,
Fig. 27). Changes in grain size may also result from winnowing in the sediment source,
as has been inferred for floods on the Colorado River (Rubin et al.. 1998). However, I
assume for tsunamis that these changes are unlikely to explain most panems of
deposition because tsunamis erode over such a broad area that they are Wllikely to be
biased by changing localized sources for sediment eventually deposited at a single
location.
My sedimentation model will not apply to cross-laminated sand ifon-bed sorting
of sediment occurred during deposition. Sediment can sort where sediment flows down
lee slopes of bedforms fonning grain flows (Bagnold, 1956), or where sediment
segregates across low reliefbedfonns that may lead to planar lamination (Komar, 1996).
However, in horizontally laminated sand where lamination is subtle or where within
laminae grain-size variation is averaged by sampling. my model will still provide
accurate results.
Formulation ofderivatives
To quantify change in speed, the model separates the total time derivative of
velocity into a lemporal and spalial tenn:
Total
i Temporal Spatial
du=&I. u Oucit Ol. ox
"
11where u is velocity, t is time, and x is horizontal distance. The partial derivatives in
space and time are defined at or across a single point (Eulerian perspective), while the
total derivative follows an individual water particle (Lagnmgian perspective). Because
shear stress and turbulence do not depend on the direction of flow, I prefer to express my
model in terms of speed <llull), leading to this fonnulation (Fig I):
Ou + U OO foru>O
{
01 51(dllull = Undefined for u = 0
dt-0;: + u~~ foru<O
Although the femporal telill the change in speed at a given point-determines
the change in grain·size distribution present in the flow, deposition or erosion depends on
the temporal and sparial terms together. tfboth are slowing, deposition always occurs,
leaving progressively finer sediment and a nonnally graded bed. If both are speeding up,
erosion occurs. If the temporal and spolial terms act in opposite directions, the results
depend on the relative magnitude of these two terms. If the flow is slowing down
through time, but speeding up through space to such an extent that the water experiences
a total increase in speed, the flow will be erosional, leaving no deposit. Conversely for a
flow that speeds up temporally but loses total speed by spatial slowing, the sedimentation
model predicts deposition of an inversely graded bed.
Continuity diagramsI use a new diagram to portray predicted and observed variability in deposits.
Such variability can include changes in the number, size, and sequence oflayers in a
tsunami deposit as it crosses topographic relief on horizontal scales from centimeters to
tens of meters. A small dip and small rise just decimeters apart can be covered with
visibly different deposits because the flow interacts with topographic variation, leading to
spatial gradients in flow properties. The new diagram, which I call a continuity diagram
(Figs. 2·3), shows variation in grain size and in the extent of deposition, both laterally in
the deposit and vertically through time. Time is plotted vertically from bottom to top, in
stratigraphic order. The horizontal axis can a modeled variable, such as speed, that
causes deposition to vary laterally across the bed. Alternatively, the x·axis can be the
variation in relative lateral extent ofdifferent layers within a deposit. Because continuity
"
12diagrams can be generated from either model output or actual deposit data, they
provide a medium ofcomparison between model results and physical reality.
To explain how continuity diagrams are constructed, I first use a continuity
diagram to represent a simplified tsunami deposit (Fig. 2). Next) similarly use a
continuity diagram to represent model output (Fig. 3). Finally, I compare the two results
(Fig. 4).
Here I show continuity diagrams where variation is defined by grain size. Where
recognizable in the field, structures revealed by changes in grain size can be mapped in
outcrop or trench walls, as illustrated in Chapter 3. In addition, as shown in Chapters 1
and 3,1 commonly quantify vertical variation in grain·size by particle-size analysis.
To summarize this kind of data in a continuity diagram, I need to define
depositional surfaces (time lines), which requires making assumptions about the process
ofdeposition (Fig. 2). First, I assume that laminae and sharp contacts are time lines over
horizontal distances ofa few meters or less. I resist extending this assumption for greater
distances because the arrival time for a change in a flow is likely to vary laterally.
Second, I assume that at a given instant. the same grain size accumulates wherever,
within those few meters, deposition is occurring (see argument in Chapter 3, p. 57). This
assumption allows a time line to be drawn within a graded bed.
With a map of time lines, it is possible to construct a time-stratigraphic diagram
(chronostratigraphic chart or Wheeler Diagram; Wheeler, 1958; Emery and Myers, 1996)
On a time-stratigraphic diagram, absolute or relative time replaces height as the vertical
axis, so where and when the record is present or absent can be clearly shown (Fig. 2c).
When used 10 summarize a deposit, a conlinuity diagram is a generalized version of a
time-stratigraphic diagram. Instead of representing horizontal distance at a specific
location, the horizontal axis represents the deposit's generalized lateral extent in tenns of
depositional environment. Senings where the depositional record is complete are
represented on the left side of the diagram, while senings where the deposit is entirely
missing appear on the right (Fig. 2d).
A similar continuity diagram can be constructed from the output ofa
hydrodynamic model that computes velocity as a function of space and time (Fig. 3). In
, ,
13keeping with asswnptions in the sedimentation model, trends in grain size foHow
trends in speed, erosion results from positive total acceleration, and deposition results
from total deceleration (Fig. 3a). To account for local variation not included in the model
run, I add an additional spatial acceleration tenn (Fig. 3b). This tenn is positive in local
depressions and negative on local rises. Because it influences time series orlotal·
acceleration time-series, it predicts erosion and deposition through time across varied
terrain (Fig. 3c). Finally I populate the depositional domain of the diagram with grain
size that I infer from speed (Fig. 3d).
As for comparing observations with models, Figure 4 illustrates the potential for
using continuity diagrams to compare a complex set of graded beds observed in trench
walls with the complex time histories of flows computed from hydrodynamic models. In
Chapter 2, such comparisons aid in interpreting observed tsunami deposits from Thailand
and Nicaragua.
Case StudiesThrough my model variation in grain size of tsunami deposits can be interpreted
as evidence for temporal and spatiaJ change in speed. In this dissertation I explore three
applications oflhis idea: relating graded bedding to deduced differences in tsunami wave
period (Chapter I); predicting the stratigraphic record of seaward-running bores that fonn
at inland scarps (Chapter 2); and inferring the sequence of waves and surges thaI
produced a deposit of the 2004 Indian Ocean tsunami in Thailand (Chapter 3).
Sandy Clu~ to Tsunami Period (Chapter J)
Tsunami deposits vary in complexity that may, in part. reflect the period of
attacking waves. Using grain·size data from the deposits of five modem tsunamis, I
show that this complexity tends to increase with the size of the earthquake at the tsunami
source.
Predicled Sedimentary Record of Reflected Bores on Coaslal Plains Overrun by
Tsunamis (Chapter 2)
My sedimentation model predicts that distinctive deposits result n seaward
directed bore that results from reflection of an incoming flow by an inland scarp. Input
for the predictions includes output from two hydrodynamic models.
• •
14Sand·sheet Stratigraphy Recording the 2004 Indian Ocean Tsunami on a Coastal
Plain in Thailand (Chapler J)
A tsunami, and the deposit that records it, will vary depending on details afthe
bathymetry and topography it overruns. To glean the maximum amount of information
from any tsunami deposit, the deposit should be examined as an individual record in a
specific place, as I do in Chapter 3. Here I reconstruct the sequence of waves, surges,
and reflected bores in 2004 that draped a coastal plain in Thailand in a complex layer of
sand. At this site the sediment source included a broad range of grain sizes, the tsunami
included multiple waves and surges, and small topographic variations provided diverse
settings for the tsunami deposit.
, ,
rties of sand in two exam Ie tsunami flowsExample I Example 2Flow depth = 5 m Flow depth = 2 mFlow speed "" 7 m1s Flow speed = 2 m1sFroude number;; I Froude number = 0.45
IS
Size range (sand)Senling velocity range (spherical silica sand)Senling distance as a percent ofcolumnheight during 500 m of transport.Rouse number range (grain roughness only)Rouse number range (roughness 10 em)
0.065 mm to 2 mm0.3 to 30 emfs
4.26% to 426%0.03 1020.01 to 1
, ,
0.065 mm to 2 mm0.3 1030 emfs
37.5% to 3750'100.1 to 7.30.037 10 3.67
Oil MODEL ASSUMPTIONS
1. FaSlllf flows deposit coarser sedimenl
o~IIUII
2. Faster nowl carry more sediment.
dIJulI ..O ErosIOn..dlLuU :>0 Depouoon..
16
b DERIVATIVES ofvelo<:lIy
Tolal
Temporal SJ)alial
du • iSu + uliu... ..velooty varin in!me and space
C DERIVATIVES of speed on Ierms of velooty denvatlVes
dliuli IitI!:!II ~Idl Ox 61
For u:o 0 ~ +u: : t;:Fat u: 0 Un6efwled Undefined Undefined
d IMPliCATIONS
.... ..•
The interJ)lay 01 spabaland temporal c;:tlange inspeed delllmwws slZegr;tlllng in (leposits
i' l""el'Sollly QI"~ cIe9OSI\S
Normally graded del)OSth
Irregularrties in the bed lead tovariation In the spallal denvahve 01speed Oepo!.l\lOl'llS greatest I!IoealJOns whIIfe I\ow$ sloW lhrough$p<ICII suc;h as swales
•
SYMBOLS
u. cleplh avllfaged velooty (on x (tirllCloon)Hull- spee<l (saJlilr magmu.lde. at norm of veloaly)o• Ql"aa'l SlZII,.-x • hOrizontal disQnce
"",.Pre_tsunami
surlac:e
Swa!e
Figure 1 Sedimentary model. a Faster flows carry more and coarser sediment. When water isincreasing in' speed, it will both erode and carry progressively coarser sediment. When it slows, itwill deposit progressively finer sediment. b To use these assumptions to predict or interpretdeposits it is necessary to relate the total change in velocity of water moving through space to thetempor-al and spatial change in the velocity of water crossing a point. c Derivatives in speed aresimply related to velocity derivatives. d The interplay between temporal and spatial change inspeed leads to either inverse or nonnal grading in deposits that val)' in thickness across thelandscape.
, .
17
II DEPOSIT (VertICallyengg8l'1lled)
b DEPOSIT WITHTIME-LINES .r.owongIhe $l.Ilf1Ice of thedepo$Il thrO!.9h lml!
....... n,wslille
,
-Hon:~l dislan<:e-
d CONTINUITY DIAGRAM~ IlOW' oepos.tlOfl andgriJlI'l SIZe vaned tt1rough lme
Figure 2 Explanation ofa continuity diagram that summarizes variation in a hypotheticaldeposit. a Deposit, showing grading and visible evidence oferosion such as truncated laminae orheavy.mineral1ags. b Time-lines mapped onlo a. C Time-stratigraphic section for this deposit.d continuity diagram. The continuity diagram is faded where it is not constrained by thehypothetical deposit. This constraint narrows through time because the deposit flattens the bed,decreasing bed variability that would make it possible for the deposit to record variation indeposition.
"
18
• NUMERICAl MODELOUTPUT with implicall()llSfOf gl'llin-su:e, erosion anddeposition on • I,Iniform pl.in
L~~,- .111,111 (mIs)
Acceleration
_10-1 tlQ..l 10-1
DlIl,lllmls7JIX
,." '00
..."" ...~ridge
r \,I,
"-.....--I • ,
,,.,~."
Bl,lfrow
I,
~i
C cr2 I ,
I i I,
Spabitl ~IderiIIaM &t'
E>cample:
b PERTURBATIONSOF MODEL 0UTPtJTdue 10 IocII <:tIanges.-
C EROSION andDEPOSITION whefeItIe spabal de~1Jv'evanes ;K:fOSS 0f0ersof magnitude
I DepOSlloonI U~lf1,..-
d CONTINUITY DIAGRAMlhowing dom"ns 01IIOSlOn and depos~ion
along with grit'n-su:e -,~,
Figure 3 Using a continuity diagram 10 summarize model output. a OUIPUI from a numericalmodel (Coulwave: lynen et aI., 2002). b Examples of pelturbed 101a1 derivative time-series. cPeriods of deposition from b overlain on a map of the domains of erosion and deposition whenthe spatial derivative of speed is allowed to vary conlinuously. d continuity diagram.
"
19COMPARISON
'"
A
c
8I
J
./
~ • ~o··f."""'~''''1"':""';;;;:'''''
Drape Lens-ContinuJly- .101 1:10·
dliUllmlsl)dt
Figure 4 Comparison between the continuity diagrams from Figs. 3 and 4. Both show twonormally graded beds (A, and B grading up into C) exhibiting similar histories. A is fairlyextensive, and becomes more extensive with time. The base of B is only deposited in local lows,while erosion is happening elsewhere, but as it fines upward it becomes more extensive.Following a period of increased erosion that only leaves deposits in local depressions, C isdeposited in depressions and later onto flat areas as well.
"
20
Cbapter 1: Sandy Clues to Tsunami Period
HypothesisIn the sandy onshore deposits of tsunamis from subduction~zone earthquakes, the
abundance of upward fining (normal grading) varies inversely with earthquake size
through two intermediate quantities: fault·rupture width and tsunami period. The largest
subduction zone earthquakes, which tend to have the greatest rupture width, warp tens of
thousands of square kilometers of seafloor. Seafloor deformation lifts or drops a similar
region of the ocean surface to initiate a tsunami. The periods of waves composing a
tsunami are set by the initial wavelength ofdeformation, and thus by the width of
deformation. so wider deformation produces a longer·period tsunami. Longer-period
waves are more likely to accelerate gradually, carrying first finer sediment and then
coarser sediment, and so are more likely to leave deposits that coarsen (increase in grain
size) upward. Also longer-period tsunamis probably include more waves and more minor
fluctuations in flow, both of which can lead to small intervals of upward coarsening.
ObservationsI measured vertical variation in particle size to explore relationships between
tsunami deposit characteristics and earthquake size (moment magnitude Mw). Some
deposits have relatively simple variation in particle size; size generally decreases from
the bottom to the top of the deposit (upward fining or nonnal grading). Other deposits
include sections where the size increases upward (upward coarsening or inverse grading).
I quantified the degree to which deposits fine upward (Fig. 5) by using particle-size
analysis of 864 sand samples, in 39 sets of vertically contiguous samples from a total of
25 sample sites among five tsunamis (Appendix A): Nicaragua (1992; Mw 7.6·7.7
(Satake, 1994)), Papua New Guinea (1998; Mw 7.1) (Tappin et al., 2001; Gelfenbaum
and Jaffe, 2003), Peru (2001; Mw 8.4 (PerfeUini, 2005)), Indian Ocean (2004; Mw 9.15
(Chlieh el aI., 2007)) and Sumatra (2005; M. 8.6 (Kone" 2007)).
Among the 39 measured sets, I chose 25 sets·- one from each site _. for plotting
in Figure 6. 1eliminated the others to make each of the plotted sets statistically
independent (Appendix B). To chose among multiple sets from a single site, I selected
••
21the sample set from the thickest deposits, on the premise that the thickest deposits offer
the greatest opportunity to deviate from simple normal grading, as inferred below (p. 22).
The results, tabulated in Appendix A and summarized in Fig. 6, show an inverse
relation between degree ofupward fining and earthquake size. Statistical significance
can be shown for a comparison between deposits of tsunamis from merely large
magnitude (Mw 7·8) earthquakes and tsunamis associated with great (Mw 8-9) and giant
(Mw 9+) earthquakes (p-value < 0.05; Appendix B).
ModelTsunami period should be related to source earthquake magnitude via the width of
seafloor deformation. Larger earthquakes result from wider bands offault rupture and
thus deform a larger area of seafloor (Dorwick and Rhodes, 200I) (Fig. 7). This
deformation sets the initial shape of the tsunami (e.g. Synolakis and Bernard, 2006; Titov
and Synolakis. 1998). The width of defonnation corresponds to the width of the initial
wavefonn, and thus detennines the waves initial wavelength and corresponding period,
as is observed for 33 Japanese and Russian tsunamis recorded on tide gauges (Mirchina
and Pelinovsky, 1981). Though wavelength shortens as the wave moves into shallower
water, period is approximately conserved. Thus larger earthquakes on average produce
longer-period tsunamis. This negative correlation is supported by observations of
tsunami period by Abe (2005) showing that the Mw 9.5 (magnitude from Kanamori,
1977) 1960 Chile tsunami had a much longer·period tsunami (47-49 minutes) than three
tsunamis frpm smaller earthquakes (22 minutes for the tsunami associated with the 1968
Tokachi-oki earthquake of Mw 8.2 (Schwanz and Ruff, 1985), and 10 minutes for the
Sanriku tsunamis of 1896 (Mw 8.0; Tanioka and Satake, 1996) and 1933 (Mw 8.4;
Kanamori,1971».
To explain vertical grading as a function of wave shape. I assume that change in
the grain size of sediment being deposited records the changing speed of the tsunami flow
(See Introduction). At a given locality. faster flows carry coarser sediment, while slower
flows carry finer sediment. Deposition occurs if water slows ~ause of weakening of
the flow through time, or because ofexpansion of the flow or frictional damping as it
moves through space (Kneller, 1995). Conversely, when water accelerates it erodes.
-.
22This model can predict change in relative grain-size at a single location, but do not
make predictions about absolute grain-size.
To detect inverse grading in deposits where it occurs, lhicker deposits, which will
generally be more complete records, should be studied. I interpret inverse grading as a
record of flow that is speeding up. and times when this occurs are likely to be the most
sparsely recorded, because in many places the flow would erode (See Introduction).
Thus the signal I describe here is expected to be weaker for thinner deposits (Figs. 8f,
ge). In general, the samples analyzed from this study were from trenches where the
deposit was relatively thick, as these included the most variation. Where muhiple sample
sets were collected at the same location, only the thickest was included in my analysis.
Asswning that faster flows carry more and coarser sediment, several factors can
explain a negative correlation between tsunami period and upward fining. Longer-period
tsunamis can generate more-gradual accelerations when they flood coastal plains, and
gradual acceleration can be recorded as upward coarsening. Also,longer-period tsunamis
will generate flows that slow more gradually, so even minor increases in flow can result
in upward coarsening. Finally, longer period tsunamis retain more energy as they reflect
or refract (e.g. Liu et aI., 2005), so it is more likely that multiple large waves and of
onland wave reflections would lead to more, sharp coarse-over-fine contacts and thus less
upward fining.
In principle, grading varies with tsunami period through wave shape (Fig. 8).
Because each part ofa shallow water wave like a tsunami moves at a speed (c)
proportional to the square root of water depth (c = (gh)ll2), the tsunami's leading edge
slows relative to the wave crest. Consequently, if a tsunami wave has a short period, the
wave crest can overtake the wave front, which will then be nearly vertical. As a short
period tsunami runs onshore, the inundation begins with abrupt acceleration as the steep
front passes. Thereafter the flow gradually decelerates. The accelerating water at the
front of the wave has only seconds to be recorded by a sedimentary deposit, and the.
acceleration is so strong that it will almost always erode and thus not be recorded. The
ensuing deceleration produces upward fining that dominates the deposit ofa short-period
tsunami.
23By the same token., upward fining need not predominate if the tsunami period is
long. The more gradual rise of water at the front of a long-period wave causes gradual
acceleration which can increases the grain size of sand the flow carries. This increase,
where recorded depositionally, produces upward coarsening. Although an accelerating
flow on a flat surface will erode, upward coarsening can be preserved where topography
spatially decelerates the flow, as happens where a flow expands into a swale between
beach ridges (Fig. 1). This process explains continuous sections of inverse grading in
tsunami deposits, such as shown in Fig. 8g.
Thin bands of upward coarsening interrupting upward-fining trends are also more
likely in the deposits oflonger-period tsunamis (Fig. 9). The more gradually slowing
water in longer-period tsunamis is more likely to be briefly accelerated due to minor
surges than the quickly slowing water in shorter period waves. If accelerated, this water
can carry coarser sediment, depositing it in swales and leading to upward coarsening in
the deposit. This process probably explains some ofcoarse-over·fine samples in overall
nonnally graded deposits (Fig. 9).
Coarse-aver-fine transitions are also more likely for longer-period waves because
they will tend to include more episodes of inundation (waves) than shorter period
tsunamis (Fig. 10). Even successive waves that are recorded by nonnally graded beds
will leave a stacked record where coarser sediment at the base ofone bed overlies finer
sediment at the top of a previous bed. And all else being equal, longer wavelength
tsunamis tend to include more waves because longer waves wiJl retain more of their
energy as they refract and reflect during their approach to a given coastline. For
example, the 2004 Indian Ocean tsunami apparently reflected off the Maldives to arrive
as a late large wave along the west coast of Sri Lanka (Liu et al., 2005). As a result,
long-period waves can leave a sedimentary record that includes coarse-over·fine
transitions (Fig. 7f). However, this process differs from the wave·asymmetry and flow
deceleration effects in that not only longer period waves will generate more upward.
fining, but also higher amplitude waves. By including more waves that overtop the
beach. taller tsunamis also produce more inverse grading.
Caveats
24The approach in this chapter does not consider several factors that could be
incorporated in order to more reliably associate grading and source-earthquake size.
Water depth at a tsunami source should also be inversely correlated with wave period,
and so tsunamis generated in deeper water might appear to have been generated by
smaller earthquakes than those generated in shallower water. Also, complex bathymetry
in a tsunami path can lead to more complex waveforms, resulting in deposits that include
more upward coarsening than deposits on simpler coastlines. Ancient deposits may be
mixed by plant roots and burrowing animals during the time after their formation, and
thus subtle intervals of upward coarsening might well vanish. Finally, the effects of wave
period will be less where deposits drape local rises, so in ancient deposits where
geomorphic setting is unknown. a lack of inverse grading may either indicate the tsunami
was short period or that the deposit is only observed where it is locally thin and doesn't
include inverse grading.
ConclusionsLonger-period tsunamis will tend to generate more gradual flow accelerations,
more interruptions to flow deceleration, and more periods of inundation. Each of these
effects lead to deposits with more inverse grading than deposits of shorter period
tsunamis. This relationship may explain some of the variability shown by particle-size
analysis of five modem tsunami deposits.
"
25
a DEPOSIT b GRAIN SIZE C UPW\RO CHANGE
C
0 ,, FIningI,
I",,, '," ,.
, \ d PROPORTION,
10 ,I Upwald-c;oarselWlg
K " ,,,.,I%
~
" 62%,,20 '. Upward-Iintng
", '( I
""""".. (,""'" "30 0'
11lS [4] 1/813] 1/4 [2J 1/2 Il]Mean grain !llZe (mm [~1J
Figure 5 a Example deposit (2004 tsunami deposit at Lhok Kruet, Sumatra) b Measuredgrading where each black rectangle represeniS a single sample that is contiguous with the samplesabove and below. The height of the rectangle covers the sample's depth range, and the width isan estimate of measurement error. C Grading types assuming grading is continuous between themidpoints of individual samples. Each colored rectangle shows the apparent grain-size trendbetween sample midpoints. d By this definition, this deposit is 62% upward fining. Wheregrading is probably too subtle to be reliably defined by my methods, I still use the trend.assuming that any inaccurate categorization is equally likely for either grading trend.
, ,
26
• •• I• I
" ••
• 9
lAw
••I
1998 Papua New Guinea
1992 Niearagua
20"_2005 Sumatra
2OCI4 lndllln
""'"
1.0 •
r~ 0.8 •
~i o.
"
FigUR 6 Moment magnitude vs. proponion upward fining for five recent tsunamis.· The 1998Papua New Guinea tsunami may have been generated by a landslide rather than by the Mw 7.1earthquake (Tappin eta!., 2001).
• •
E•
27• 2004 Indian Ooe.,.Mw9.1-9.3
b 2005 Sumatr.Mw 8.6-8. 1
C 2001 Peru..."~d 1992 NicaragUilMw76-77,e 1998 P,IPUil New GuineaMw 7.0-7.'or 5 /{m wide landslide (rooSIIlall to show)
Figure 7 E:irthquake rupture areas: a Inferred from coseismic deformation and seismic inversion(Cblieh et aI., 2007); b Inferred from coseismic deformation and seismic inversion (Konca clsl..2(07); c Inferred from seismic inversion (Perfettini et aI., 2005); d Inferred from tsunami records(Satake 1994); e Based on oceanographic mapping ofa fault and a landslide in the area where thetsunami likely originated (Tappin el aJ., 200 I).
, ,
28
longerpo""
9 0
20
•
G".-...• •,
", ,•",'
:~"
GradualacceleralioJ'l
", ,
t"\
",""
1132 [51 1/16 [oi) 118 (3) 1/4121Grall'l-IIZ. (1Tm (oJ)
Figure 8 Inverse grading that records a gradual acceleration: a Two simple waves as they mightbe al their source. b Waves from a, after developing asymmetry as they pass into shallow water.c Flow speed is in phase with tsunami height. d Grain-size correlates with speed. e Depositwhere thick. f Deposit where thin. Only where thicker does the deposit record progressivecoarsening through time. g Example grading data showing a promincnI inverse grading trend inthe lower section ofa 2004 tsunami deposit from Simeulue (95.7577 E. 2.8322 N).
'.
29
..-.g1lC_ Vel
fa I,f_
) Minor acceleration
~,I
"a10• ,
'.~
•,- • '. ,.'I
1132 [S) 1/16 [41 1/8 [3] 1/4 [2J
Grain·size (mm [llll
Figure 9 Inverse grading that records minor periods of acceleration that interrupt gradualdeceleration. a-«= The link between changing water height and changing grain·size for waves ofdifferent period (similar to bod in Fig. 4). d and e Thick and thin deposits. fExample data fromKhao Lak. Thailand (98.2676 E. 8.8330 N). Because the standard error in the mean grain·sizemeasurements is not established for settling column measurements like this, it is difficult todistinguish minor upward coarsening from analytical noise. bUI it is nol unlikely thaI in a sectionof such gradual change in grain-size there would be narrow bands of upward coarsening.
"
30
• Abruptacceleration
G,,'.-~
.,
••
,
bed e
L l.", ~~,"l
•
:-••
, .• •• ••..
f 0
•Shofler I .......~
1
!",n,,,......
1/16141 1/8 [3] 1~ 12] 112 [lJGf..n-~e (nm JOD
Figure 10 Rapid accelerations accompanying the arrival of later waves can be recorded as abruptupward coarsening: 8-e are arranged identically to Fig. 5. f An example dataset from north ofKhao Lak, Thailand (98.2667 E, 8.8333 N). Two abrupt upward coarsening points are marked inthe deposit. These corresponded to laterally continuous coarse-over·tine contacts observed alonghundreds of meters oftransect, and probably record successive waves.
"
31
Chapter 2: Predicted Sedimentary Record of Reflected Bores onCoastal Plains Overrun by Tsunamis
IntroductionThe sedimentary records of tsunamis commonly fonn and endure on beach-ridge
plains and coastal marshes. Where such lowlands have been building seaward in the late
Holocene, they commonly terminate inland at former sea cliffs or other steep slopes.
If, during a tsunami, fast-moving water does reach a steep slope at the landward
end ofa plain. it will likely fonn a reflected bore similar to those observed on a beach in
Thailand during the 2004 Indian Ocean tsunami (Fig. 11) and at a macrotidal Alaskan
estuary subject to tidal bores (Fig. 12b-d). In both of these cases, the incoming flow
crossed hundreds of meters of exposed, nearly horizontal seafloor before reflecting.
Similarly at coastal plains, a tsunami can cross exposed, nearly horizontal land before
reflecting.
Nonlinear, numerical tsunami-run-up models predict the formation of reflected
bores for a tsunami that hits a steep slope at the landward limit of a horizontal plain. The
reflected bore causes an abrupt change in velocity-an event that would likely be
recorded in a resulting deposit (Fig. 12a).
In this chapter I use my simple sedimentary model (Fig. I) to predict deposit
grading for a simple reflected-bore scenario (Fig. 13) and for output from numerical
tsunami models (Figs. 14-15), to gain insights into the internal structures of real tsunami
deposits that blanket beach-ridge plains and marshes. The sedimentary structure of
tsunanli deposits vary for a number of reasons. Relatively simple deposits, such as those
from the 1998 Papua New Guinea and 1992 Nicaragua tsunamis (Gelfenbaum and Jaffe,
2003; Higman and Bourgeois, in press) consist ofjust one or two normally graded beds
that vary in thickness, basal grain size, and rate of grading (change in mean grain size per
depth). In more complex deposits such as those from the 1960 Chile, 200 I Peru and.
2004 Indian Ocean tsunamis (Chapter I; Kon'no, 1961; Jaffe et al., 2003; Moore et al..
2006; Jaffe et aI., 2006), multiple layers are common and grading within layers can be
either inverse or normal. The number ofJayers in these deposits varies across hundreds
32ofmeters, where the variation might be explained by differences in the number and
structure of waves affecting widely separated points, and over just a few meters, where
the variation more likely reflects local factors such as small irregularities in pre-existing
topography.
My ultimate objective, beyond the scope of this paper, is to use evidence of
reflected bores in tswuuni deposits to constrain tsunami flow conditions. At a minimum,
evidence ofa reflected bore shows that a paleo-tsunami did actually reach whatever slope
produced the reflection. Furthermore,longer-period tsunamis will produce more
sustained, strong oDland flow, and thus increase the lime and distance over which
reflected bores evolve as they move against the incoming flow. So if the evolution ofa
reflected bore is recorded in its deposit, the distance over which this evolution occurred
would provide information about the period of the anacking tsunami.
PredictionsHere I apply my sedimentary model (Fig. II; based on Kneller, 1995; see
Introduction) to output from two depth-integrated, nonlinear, shallow-water wave
models-Coulwave (Lynett et al., 2002) and Delft3D (Stelling and van Kester, 1994}
which I implement for tsunami inundation. Coulwave uses a Boussinesque
approximation of the Navier Stokes equations to handle dispersion and includes effects of
wave breaking. Delft3D, though it can be implemented for fully 3D flow, was run in this
case with only one horizontal dimension and with depth-integrated velocity. These
hydrnulic models improve on my cartoon model (Fig. 13) at least by quantifying the flow
geometry and by including specific formulation for dissipation and conservation of
kinetic energy and momentum. The model runs I use here incorporate unrealistically
low friction which makes horizontal trends in flow weaker than in real tsunamis crossing
undulating ~egetated plains. However, I am using these trends merely to resolve general
pattems of flow. In each case I run modeled waves across a horizontal plain with a slope
(reflector) at its landward limit. Specific parameters for each run are provided
graphically in Figs. 14 and 15.
I now describe three simulations ofa reflected bore. The first and most simplified
of these qualitatively presents bore reflection where speed varies linearly with time
••
33except when passage of the bore front causes sudden change (Fig. 13). The second
presents numerical results from Coulwave for a single positive wave reflecting off a 2%
slope (Fig. 14). The third presents numerical results from Delft3D for a wave train that
reflects offa vertical wall (Fig. IS).
Simulated deposits ofreflected boresModeled variation in speed can be combined using my sedimentary model (Fig.
I) to estimate the structure ofa tsunami deposit (Figs. 16-18). I represent the deposit's
internal structure in continuity diagrams (Chapter I) that show, for three places between
the beach and a reflector, the depositional sequence of an incoming wave and its reflected
bore. Near the reflector, where the bore starts fast and decelerates abruptly, the abruptly
slowing flow drapes the sediment surface with normally graded sediment (Fig. 16,
location C). At intermediate distances from the reflector. the direction ofthe flow
reverses, but its speed, and thus capacity and competence, do not change (location B).
Also, because seaward flow can be augmented by the landward-moving withdrawal, the
flow following the bore may be speeding up and more erosional. Finally at locations far
from the reflector the deposit records two waning flows, one from the inflood and one
from the reflected bore, each as a normally graded bed.
Some of the predictions made using this model can be modified to account for the
time it takes for sediment to be eroded or deposited, and for situations where sediment is
deposited, and then subsequently eroded. At medium and long distances from the
reflector(Fig. 16, locations A and B) the flow after the reflected bore is erosive (Fig. 16c)
removing previously deposited sediment. Also, when the flow slows suddenly (Fig. 16,
location C) or speeds up suddenly (Fig. 16, location A) the resulting deposition or erosion
takes some time (Fig. 16c).
The .continuity diagrams illustrate how the deposit varies depending on local
topography. In lows, a more negative spatial derivative of speed leads to deposition even
at times when the flow is erosional elsewhere. Overall, the deposit of the incident flow
and reflected bore includes two normally graded beds far from the reflector, one normally
graded bed with a erosional contact in the middle at intermediate distances from the
"
34reflector, and a normally graded bed that drapes the sediment surface near the reflector
(Fig. 16<1).
Similar diagrams for the Coulwave and Delft3D model runs (Figs. 17, 18) are
more realistic than the cartoon in Figure 13 and 16, but are also likely to include
complexities that are a function of the particular realization of fluid dynamics used in
these models. However, these and the cartoon in Figure 13 show the same abrupt
seaward change in velocity as the reflected bore passes-the feature that my
interprelation rests upon.
Deposit near reflector (loco/ion C)To form a reflected bore, a tsunami must be flowing quickly when it reaches the
reflector, and thus be capable ofcarrying relatively coarse sediment. However, any
deposit recording fast inflowing water will have only begun to form when the water is
abruptly Slopped by the passage oflhe reflected bore. Relatively fine sediment rains out
of this stilled water (Figs. 16b, 17b, location C).
The results ofmy simple model can be improved by accounting for time lag in
deposition after the bore passes (Figs. 16c, 17c). Time lags-from settling time,
dissipation of turbulence, and reversal of the boundary layer-spread deposition out over
some time. The result is nonnally graded deposit that drapes topography, unlike deposits
of more nearly steady flows that lap onto, but are absent from, local high spots. In the
absence of larger incoming waves or strong reflector·parallel currents, it is likely that this
drape will survive later reworking because later currents will usually be slow at this
position far from the beach. After the fonnation of this drape, gradually withdrawing
water will leave very little deposit (Figs. 16c, 17c.location C).
Where a wave moves through standing water, such as a lagoon or waters ponded
from a previous wave, the inrushing wave will generally leave little deposit near the
reflector. Unlike water crossing dry ground, a wave crossing water outruns sediment·
laden water as it moves into sediment·free standing water. Though the inrushing wave
may rework the sediment surface near the reflector by briefly shifting the standing water
landward, it is unlikely to suspend enough sediment to yield a substantial deposit.
Deposit at an intermediate distance from the reflector (location B)
35The abrupt decrease in landward velocity that results in a distinct deposit near
the reflector also occurs in more seaward locations. However, in the time it takes for the
reflected bore to move seaward the incoming flow may slow, so the velocity change can
be great enough to reverse the flow (Figs. 15 and 16, location B); the flow's speed (the
scalar rate of motion relative to the substrate) decreases less than the flow velocity (the
degree to which the flow is moving landward). There exists some point where the abrupt
decrease in landward velocity as the bore passes corresponds to no change in flow speed.
At this point I expect no abrupt change in grain size as lhe reflected bore passes.
However, the flow may still experience a change in the total derivative of speed that
changes the spatial pattern ofdeposition on the bed. For example in Figure 8b, the
reflected bore results in a period oferosion that would be recorded as an abrupt fine-over
coarse lransition. Allowing for lime lags in deposition and erosion scarcely alters this
expected stratigraphic sequence (Fig 16c, 17c.location B).
Depositfar from rejiector(/ocation A)Far from the reflector, in places where it arrives long after the passage of the
leading tsunami front, a reflected bore can meet inflow that has already slowed, stopped.
or even reversed. The slowing of the inflow yields a nonnally graded bed by reducing
the grain size left in suspension. The reflected bore yields a second nonnally graded bed
by slowing after its own abrupt arrival (Fig. 16b, 17b. location A).
The sudden increase in speed upon the arrival of the reflected bore should also
erode the upper portion of the first nonnally graded bed before deposition of a second·
nonnally graded bed. Because it takes time to mix sediment up into the flow, some time
will pass where the speed (and shear stress) ofthe flow is decreasing but erosion is still
occurring. This time lag in erosion is not captured by the simplistic sedimentary model,
which I modify accordingly in Figs. 16c and 17c (location A). Most of the sediment thus
eroded comes from the pre-existing tsunami deposit and pre-tsunami soil-provenance
that may provide aid in distinguishing the reflected-bore deposit from thal ofa su~ssive
incoming wave or surge. Otherwise. an overlying nonnally graded bed from a reflected
bore may closely resemble the deposit ofa second surge.
Field Examples
--
36A deposit from the 1992 Nicaragua tsunami (Higman and Bourgeois, in press),
in a lagoon at Playa de Popoyo with a 11 % slope at its landward limit, includes the
features expected for a deposit recording a reflected bore (Fig. 19). The landward edge
of the lagoon there has Iinle deposit, as expected given the standing water in this lagoon
at the time of the tsunami. In the middle of the lagoon, a sharp contact marked by a
dense-mineral lag between finer sediment and overlying coarser sediment may record
erosion from passage of the reflected bore. At the seaward end of the profile a local low
spot includes two stacked normally graded beds-a feature not present in deposits in
nearby profiles that lack a tsunami reflector.
A deposit of the 2004 tsunami in Thailand also shows features consistent with a
reflected bore (Chapter 3). One fine·grained bed, the only one that thins toward the
ocean, drapes a sharp ridge near a steep slope at the limit of inundation. Farther seaward,
as I saw in Nicaragua, this bed unconformably overlies coarser sediment. In Thailand,
there are no stacked normally graded beds near the seaward limit that could correspond to
a reflected bore. This can happen ifa reflected bore is still encountering landward
moving water (as in location B, Figs. 16, 17) even as it moves seaward of the beach.
DiscussionIn trains of large tsunami waves, later waves that encounter standing water from
earlier waves may form landward-moving bores. In addition, reflected bores can form if
these waves reach inland reflectors. This abundance of flows is capable of producing the
stratigraphic complexity shown in Figs. 15 and 18. Further stratigraphic complexity can
be expected of tsunami waves that encounter curved reflectors, approach the shore from
more than one direction, or enter coastal lakes.
CondusionsIn a tsunami deposit that includes a record of a reflected bore, depositional
patterns are likely to vary with location along a profile between the sea and the reflector:
near the reflector, a thick, normally graded drape; at intermediate distances from the
reflector, a normally graded deposit possibly including a sharp basal contact; and far from
the reflector, two normally graded beds. These predicted features should be detectable in
37paleo-tsunami deposits. and similar features have been found in modem deposits in
senings where reflected bores are likely to form.
The methods employed here could be applied 10 many features of tsunamis and
their deposits. as well as to other sandy deposits. By providing a link between tsunami
dynamics and complex deposits, these tools for reading the paleo-tsunami record can also
improve understanding of tsunami hazard and of the events that generate tsunamis.
39
a DEPTH-AVERAGED b MORE REAlISTICREfLECTeo BORE REfLECTED BORE
Water lIUl'face / ..,) ./s - A'---- - - - -
/ r. r/• Depll...I,.'lffiJged ~e1ocit1
i - - :.> ~~A-- ----- -/ /~
~ - --/ --~/
- - -----~ - --" ~--/ /--~
- --- ----Distance (meters to 100, of meters) D,stance (meters to 1005 of meters)
C REFLECTED BORE FORMING shOWl'l in a composite of 5 photos
Tome3 Tme4 Tome 5~ R~~ R~
e REFLECTED BORE "lOvong agaonst tIow
-
1_1 Tme2FIoocIIi og FIoodiI'II
d COMPlEX BREAKING OF REFLECTED BORE
,'- -
Time 6 Reflecting Time 7 Photos by Andrew Mallo~:used wilh permission
Figure 12 a Idealized cross-section of a reflected bore considering only depth-averaged flow. bMore realistic flow based on observations ofactual bores reflecting. c-e Reflection ofa tidalbore al Turnagain Ann, Alaska (e, a similar bore one day later than c*d).
40
-
~=--I -//
C WAVE SHAPE read from bottom
"',..5
·~'-.l.3 /~------'/
Ou,"ow
A B
• •/~-
a PROfiLE withlocations for Figure 8
b WATER PARTICLEPATHS. moving allhedepth-avereged velocity
o"""ftow
=-----'/;...---"----------'.' /-/
Dislanoe along ~ /_---------lprofile (102 _1a" m) Vertically e~eggerated
Figure 13 Cartoon ofa depth-integrated reflected bore. a Simple coastal plain profile with asteep slope at the distal end. b Motion of the water as the tsunami inundates the plain and reflectsoffthe slope; purple shows inflow and red shows outflow. The water reverses abruptly as thereflected bore moves seaward. C Shape of the water surface at a few points in time.
41
C WAVE SHAPE !lad from bottomto top
•,o
Beach 2% sloperellec:tor
Coastal plain
ABC
Dlslance Inland (km)
o~
20
80
60
a PROFILE wilhloca60ns ror Figure 8
b wo.TER PARTIClEPATHS, Il"IOVWlg a11hedeptIl-averaged ~ell::dy
0""" ...
j
SOx vertlCill exaggeration
Figure 14 Reflected bore modeled using Coulwave (lynen et al.. 2002). Color scheme andorganization duplicated from Fig. 13. This tsunami wave, with a period ofabout 15 minutesforms a reflection on a relalively low-angle slope (2%).
• •
42
C WAVE SHAPE reiitdfrom bottom to lop
Beadl Vertic;alrel\e1;lor
Co.stalplain
•
'4L~_2 0 2 •OlStance inland (Iun)
'"
j 60
a PROFILE dOl malksIoation 01 detail in Figure 9
b """TER PARTICLEPATHS.lTlCMl'lg 8t thedepl:1l-.veraged velocilyof the IIow 80
l00x vertICal exagger,l-on
Figure 15 Muhiple reflecled and incident bores from a wave train modeled by Delft3D. Withthree -I S minute period waves moving on the plane at lhe same lime, individual water paniclesfollow complex paths. The first wave barely reaches the reflector, forming no significantreflection, but the second wave overrides this wave and fonns a strong reflection, as does thethird.
43Loc",rIOHAl__G) LOCATION I lOC"TlOtII CI_...al
• TNE-$£RIU
JJ• • • •
! • -0 0 0 '-'-
! 0 0- -
~....--~--•
• ....... tlOt4o· _.""'"
.[o
b CONTINl..UY
I~-<ri\o .... FS-I!•""'- --............-.-REAUSTIC
~_n
l)!4Gf'''''' .. "\I •-~ .~,-... ,
I! 0•
"-
d Il'EStAI-..G00'0511......~.... IO<.--- I
tb_
I:•..;;."--;;;....'"~_,(11"0'.. _
~--Figu~ 16 Deposits formed by a cartoon reflected bore (Fig. 13). a Time series for velocity andlotal speed derivative. b continuity diagrams delineating the pattern of deposition implied bythese lime series, assuming the flow-sorting approximations (see Introduction). On this plol.time (venical) is ploned against spatial acceleration (horizontal), which determines how laterallycontinuous a given ponion of the deposit is. c continuity diagrams that improve on b by allowingfor time lags in erosion and deposition, marked in gray and black. d Deposit expected from cacross a ridge and swale.
'.
• T_-SERES
•
. , .-,
lOCA'1l)IH C......
.l~• ~.
• • • 0 ,-0 •, , • , • , • • , • , •..-
44
b CONTINUITY~.~
0Ilf\' 'hi FS I- ~I!~
°~e=.=:,e.,c.c-"7.~.O~ Co "";00'
Sf'OI"'~.-
• AI,... """
• 8el0'.1>o<,
,-REAliSTICCON' .....TY~ ....--~... I
!'J.;~=..o,~l... 0._l_.~
01 ClOOO.,
•0L-__ •
',,",,'.-c •
"">JO.....
d R(SUlTlNGOC""',......-...................
S,,·... R,Cl9"_IO'lO·m_0"..- ""'.... ,_
Figure 17 Deposit formed by a reflected bore as modeled by Coulwave(Fig. 14); a quantitativeversion of Fig. 16.
• •
45a TlME-5ERIES
60 • Il,,,.• • (rellectedl
'" '"tr tr tr Ilo"l-iO •
:f(_Ic:omlllg)
~0 Front
20 '" Revel'$al
~ 0 8 0 Il"'- .....
b CONTINUITY60OIAGRAM USing E_o
only the F$ •-,I~ • <N-
o" • ,,-E-tr Relall~egrain size
~ F,~
'" '10.-,; eo.,~
·10-' !.10"" I~'ONergenl Convergent
Spatial changein speed
Deposrtedthen eroded
~-;:.>'''.........'''''"••..--. -j?.,.."',....p;~ /1..,,1 ~I
SIZe grading w~h
anootahon
I!;tr...........~
~'""o~=~::-
Lens Drapelaleriill c:.onllnulty
of depoSItSIZe grading .... depost
d RESULTINGDEPOSIToIlI.JWiIted for a•wlle and ndge
cMOREREAliSTICCONTINUITYDIAGRAM talung11'1'0 eon~f'lJ()tl
tme lags
Swale R~
c::-~ ,f'IO.ln1m __figure 18 Deposit ora wave train including reflected bores (Fig. IS) as modeled by Delft3D;another quantified version of Fig. 16.
46
Foeld notes
e)(aggfIflltlon )(10
Wrack~ne
....Rip-ups
'"Foeld sketch •gram-sIZe
Pre-lsunami water level (lagoon)
J[ Tsunamiw,ter.mark
"
2 beds I bed
"" mb oesERVATlONS
[
a PROFilE
~ PACIFICOCEAN
very little depositas expected fOf, w,veCl"OS$lng standing wale'.
Two normJly graded bedsJS predlCled (Figs 1.8 Ioeatl()l'l A)
C CONTINUITYDIAGRAMS
Normally graded bedInterrupted by erOSIOn(fine~HX)afse contact)as preOded (Fl!ils 7 81ocahon eJ
figure 19 a Profile inundated by the 1992 Nicaragua tsunami (Higman and Bourgeois, in press,and Bourgeois, unpublished notes, March 1993). b Observations of the deposit across a natlagoon. Below each sketch is nOle<! the type ofdata documenting these observations. Cconlinuity diagrams based on b. This deposil is consistent with my predictions.
• •
47Chapter 3: Sand-sheet Stratigraphy from the 2004 Indian Ocean
Tsunami on Coastal Plain in Thailand
IntroductionParsing a tsunami deposit to identify the records of individual waves is important
for sedimentologic estimates of flow depth and velocity and for using tsunami deposits to
teach the public-safety lesson that the first wave isn't necessarily the last, or the biggest.
In addition, tsunami modelers need geologic constraints on the number, sequence, and
shapes of waves to help evaluate their numerical simulations of the wave trains of
hypothetical tsunamis. Such modeled tsunamis may be the basis for hazard assessment
like the time histories that accompany tsunami-inundation maps in northwestern North
America (e.g. Walsh et aI., 2000). Ahernately models can produce sets ofdistinct
hypothetical tsunamis from a variety of different possible sources. such as earthquakes of
different size, slip distribution or location, or such as tsunamis generated by different
sources such as earthquakes, landslides, volcanic eruptions, or bolide impacts. The
predicted tsunamis from these sources can then be compared to geologic records of real
tsunamis in order to narrow the range ofpossible tsunami sources.
The 2004 Indian Ocean tsunami provides unprecedented opportunities to
understand the geologic records of individual waves. For the first time, a wealth of video
footage and photography shows a tsunami in action, even to the point of yielding
esti":!ates of its onshore velocity (Fritz et aI., 2006). Also for the first time, geologists_
have surveyed the deposits of a modem tsunami at locations distributed along thousands
of kilometers of coast. Results of these surveys are now available from India (Chadha et
aI., 2005; Nagendra et aI., 2005; Peterson et aI., 2005; Singarasubramanian et aI., 2006;
Murari et aI., 2007); Indonesia (Moore et aI., 2006; Jaffe et aI., 2006); Maldives (Kench
et aI., 2006); Sri Lanka (Goff et aI., 2006; Morton et aI., 2007); and Thailand
(SzcZllcinski et aI., 2005; Fujino et aI., 2006; Szczucinski et aI., 2006; Hon et aI., in
press; Hawkes et aI., in press; Umitsu et aI., in press) with more to come.
In this study, I start by summarizing the 2004 tsunami wave train as recorded in
video and still photographs from Thailand. These images show various combinations of
waves, defined as inflows separated by outflows, and surges, defined as accelerated
--
48inflows superimposed on waves. Ilhen interpret the details of a sand sheet from 2004
as the geologic record of a series ofsuch waves and surges. I show how a stratigraphic
record of the 2004 tsunami deposit in Thailand varies across the landscape, and I trace
marker layers to identify a stratigraphic record of individual waves and surges .
I find enormous variability in the appearance and completeness of this record
across a coastal plain. This variability resulted from deposition and erosion that were
both patchy and intermittent. The tsunami's stratigraphic record is least incomplete
(most continuous) in local swales. The swales show that the tsunami's stratigraphic
record on flat ground includes gaps, where erosion evidently interrupted deposition.
Such erosion has not yet been incorporated in methods for estimating tsunami size from
deposits (Soulsby et aI., 2007; Jaffe and Gelfenbaum, in press). By documenting the
variability in readily observed deposits from 2004, I lay groundwork for more complete
interpretation of the internal stratigraphic records of ancient tsunamis.
MethodsEyewitness accounts
Because the 2004 tsunami reached Thailand in the late morning on a calm, clear
day, at the height of tourist season, plenty of people were on hand to photograph and
videotape it (Fig. 21, also Eyewitness Photos and Videos.). I compiled the resulting
photos and videos from the Internet. Where possible, I contacted the eyewitnesses
directly and asked them for higlHcsolution files and personal recollections. The file of
Eyewitness Photos and Videos associated with this dissertation contains such data with
the owners' pennission, as well as photosets and videos I found to be widely available on
the internet.
Field surveysIn October 2005 a group led by Alam and Glawe mapped and trenched the 2004
tsunami depOsit along seven approximately shore-normal transects at Bang Lut, 25 km
north ofKhao Lak (Fig. 22). An additional survey, including Higman and Maxcia,
added detail along one profile in March 2006. Elevation profiles (Fig. 21, also
Stratigraphic Map) were measured relative to mean sea level approximated by the
average of the high and low tides on 27 October 2005. With respect to this datum, a
Thailand Public Works Department benchmark at N s<'50.025' E 9s<'16.007' is at
49elevation 4.0 m. The vertical closure errors did not exceed IO em.
In the initial survey we studied the 2004 deposit thickness, stratigraphic sequence,
and aerial extent by digging 68 shore·normal trenches, each about 2 m long (Fig. 22). In
addition, I made spot checks of sediment type and thickness between trenches and
between transects. At each trench, I measured total thickness of the deposit and the
thickness ofconspicuous layers defined by changes in particle size (Fig. 22).
During the work in March 2006 I reopened trenches along one of the transects to
explore further the internal structure ofthe sand sheet (Figs. 23-26, also Stratigraphic
Map). At each of these trenches J measured vertical and horizontal position using a line
level in order to sketch layering and sedimentary structures at field scales of 1:4
(simplified examples, Figs. 23 and 24).
Grain size analysisI made particle-size analyses (via settling column) of vertically contiguous
samples collected in 2006. Samples of various thicknesses, commonly 0.5 em or less,
were analyzed in a 190--cm settling column filled with 21°C fresh water. Settling
velocities are reported as hydraulically equivalent silica sphere size calculated by
inverting Fergusun and Church's (2004) fonnulation for settling velocity as a function of
size. Examples of settling-column grain-size analyses are provided in Figs. 23 and the
Stratigraphic Map.
2004 Tsunami deposit in TbailandPropagation and height
The 2004 tsunami crossed the 450-600 Ion wide Andaman Sea before striking
Thailand's coast (Fig. 20). The tsunami's source area, controlled by a curved fault
rupture, is concave to the east and helped focus the waves toward this part of Thailand
(Grilli et aI., in press). The incoming tsunami shoaled across a continental shelf that
slopes on average 0.2%. Although isolated islands more than 50 Ion offshore likely
refracted the tsunami, the shelffor 30 Ion west of the shore lacks reefs or other obstacles
(Fig. 21).
At Bang Lut the tsunami ran across a coastal plain with less than a meter of relief
except for a few local depressions and hummocks left by tin mining, which ended at
Bang Lut years before the tsunami in 2004. The plain is thick with grass and bushes and
50is doned with trees. At the eastern edge of the beach-ridge plain, the tsunami
terminated against a former shoreline scarp 5 m tall, at the foot of a rubber plantation
(Fig. 22, also Stratigraphic Map).
Along 50 km ofThailand coast approximately centered on Bang Lut, post·
tsunami surveys documented evidence of the tsunami up to 21 meters above sea level,
and more typically between 5 and 15 meters (Fig. 2Ia). Inundation distances, though
widely limited by high ground, exceeded a kilometer in many places. At Bang Lut
measurements include a height of7.4 m at the tsunami's landward limit about 650 meters
inland (Urnitsu et ai., in press).
Human records o/multiple WQyes
The wave train at Bang Lut likely comprised multiple waves that each flooded
ORland, the first or second ofwhich was the largest. I am uncertain of the sequence
because I know of no eyewitness account, photo, or video from Bang Lut itself. Instead,
the eyewitness constraints must be imported from areas about 25 Ian to the south (Khao
Lak),3 Ian to the nonh (Ban Nam Khem), and 32 Ian to the north (Hombill Hill).
Eyewitness documentation shows multiple waves also struck other parts ofThailand's
Andaman Sea outer coast.
Evidence from south of Bang Lut includes a radio interview and several videos.
In the radio interview, an eyewitness who was just south of Khao Lak at the time of the
tsunami described a destructive first wave followed by smaller waves (Smothers, 2005).
The first wave he describes is probably similar to the violent first wave captured on video
at southern Khao Lak.
At Ban Nam Khem, an eyewitness interviewed by Hori et al. (2007) described
three waves. In this account the second wave was largest, and all three overtopped the
beach. Similarly, farther north at Hornbill Hill, on the seaward edge ofPhra Thong
Island, Kimina Lyall saw an initial wave that did not overtop the beach, and a large
second wave. Her photos also show a small additional wave after that. The time of
inundation for the largest wave, based on her account and time-stamps on photos, was for
about 25 minutes between 10:40 am and J 1:05 am (see Eyewitness Photos and Videos).
The smaller wave that preceded this wave was probably less than J0 minutes earlier.
51Along the Thailand coast farther from Bang Lut, eyewitness accounts, photos,
and videos also show multiple waves, as many as three of which overtopped the beach.
At Phi Phi Island (location, Fig. 20), Caroline Malatesta photographed two waves, the
first larger than the second, and each preceded and followed by a withdrawal (see
Eyewitness Photos and Videos). Videos and accounts from Patong Beach on Phuket
show that the first wave there was not as large as a later wave. Videos from Patong
Beach on Phuket also show two waves, and there the second was larger.
The nearest tide-gauge record, from an inlet 45 km to the north at Kuraburi
(location, Fig. 21), records an initial withdrawal followed by four positive waves more
than 50 em above ambient tide level (Fig. 21, also Eyewitness Photos and Videos). The
highest of these, cresting less than I m above ambient tide, was the third, and it followed
the inilial withdrawal by three and a half hours. In height, timing, and relative size, the
positive waves in this gauged sequence have little in common with the open-coast wave
trains recounted above.
Images ofwavefronts and surgesEyewitness photographs and videos show three styles of attack for the 2004
tsunami in Thailand: a single, steep breaking front; multiple fronts that break at or near
the shore; and a rapid but non-breaking rise of water level. Mosl commonly. the tsunami
came ashore as a single, steep, foaming wall of water that originated hundreds of meters
to kilometers offshore. At Phi Phi Island, this steep front only barely overtopped the
beach but was followed by further rise that produced flow deplhs of2 m or more on land.
In other cases the wave arrived as a series ofcurling wave faces that broke at or near the
shore. Videos of the second (and largest) wave at Patong Beach show a series ofsuch
breaking fronts, each one raising the water level a step. The lasl two breaks, spaced
about 12 seconds apart, ran across the top of the beach and continued hundreds of meters
into the city. In contrast the first wave at Patong Beach resembled a flood tide on fast
forward. In about 3.5 minutes, the sea advanced at an increasing rate up the beach (S:ee
Eyewitness Photos and Videos). Though non-breaking, this flow was still rapid; in about
30 seconds at the end of one video it rose about 2 meters to overtop the road and pour
into an underground parking garage, pulling cars along with it. Photos by Helmut Isseis
, ,
52(see Eyewitness Photos and Videos) 12 km north ofPatong Beach show a similar first
wave that lacks any breaking front.
2004 Deposit at Bang LUIThickness and volume
The 2004 tsunami deposit in the study arca varies considerably in thickness
across the coastal plain (Figs. 22a,c, 23, 24). On average, the deposit thickens somewhat
inland because the thickest unit thins seaward (unit 84, Figs. 23, 24). Thickness appears
unrelated to elevation (Fig. 22c). The thickness varies more with local topographic relief
over a horizontal scale ofa few meters than it does with elevation or distance inland. For
example, the deposit thickens from millimeters on top ofa low mound to 30 em on the
floor of a tin-mining pit only a few meters away (Fig. 25).
To estimate the volume of sand left onshore by the tsunami, I contoured the
deposit thickness measured at al168 trenches from October 2005. In the contoured area,
which totals about half of a square kilometer and excludes the channel and ponds plotted
in Fig. 22b, the tsunami deposit totals about 44,000 mJ, for an average thickness of 8.3
cm and, along the roughly IOOO·m length of the study area's coast, an average onshore
tsunami sand volume of nearly SO m3 per meter along the coast. The actual onshore
volume is greater if the tsunami left thick deposits in the excluded channel and ponds, as
it did in a small depression I surveyed (Fig. 25).
Stratigraphy and architectureI identified three main stratigraphic units (A, B, and C) in the 2004 sand sheet at
Bang Lut (Figs. 23, 24, 26, also Stratigraphic Map). Unit A, present directly above the
pre-tsunami soil, is thin, discontinuous, and finer than the lowest part ofunir B. Unit B
extends along the entire transect and makes up most of the deposit inland of the first 100
meters. Unit C is distinct in the most seaward trench (Fig. 23) but becomes thinner and
indistinct inl'and. PosHsunami soil processes are incorporating and reworking the upper
portion of the tsunami deposit (usually unit C) except in swales where the deposit is
capped by muddy sand fanning out below rills presumably washed in by posHsunanti
rain (Unit X, Fig. 24).
Pre-tsunami soil and beach
'.
53The pre-tsunami soil is dark, organic sand bound by roots. I found little
evidence that the tsunami eroded any part of it. The soil grades downward into coarse
sand.
The post-tsunami beach deposits, sampled to a depth of 15 em near the high-tide
line in March 2007, consist primarily affine and medium sand, with less than 0.5% silt.
The sample included one layer with granules that were too coarse to measure precisely by
settling column. Overall, the grain-size distributions in the beach sample lack a mode
typical of the nearby 2004 tsunami deposit: a peak near the silt-sand transition (Figs. 23,
also Stratigraphic Map). Paleo-beach sediment exposed by tin-mining and burrowing
animals also appears to lack this fine population, which I ascribe below to a hypothetical
offshore source.
Unit AUnit A rests directly upon, or fills pre-tsunami burrows within, the pre-tsunami
soil. Its boundary with overlying unit B is everywhere a sharp upward increase in grain
size. Unit A is most common near the shore (Fig. 23); 320 m inland it is reduced to
sparse pockets and lenses (Fig. 24). Nowhere did I find it thicker than 3 em. and
typically it is present only as a thin smear of fine grains at the base of the deposit.
Though in the field I recognized no grading ofA, and though in most cases it is 100 thin
for more than a single grain-size sample, one location yielded two superposed samples,
analysis ofwhich shows upward fining (Fig. 23f).
Uni{BUnit 8, the dominant part of the 2004 tsunami deposit at Bang LUI, overlies either
unit A or pre-lSunami soil. Where it does not extend all the way to the top of the tsunami
deposit, 8 is overlain by a sharp contact with coarser sediment. Where nOI eroded or
disturbed, the upper portion of 8 is the only sand· free silt (85) in the tsunami deposit. It
is typically 5- 10 cm thick and systematically thickens in local lows to a maximum
thickness of about 25 centimeters (Fig. 25). I divide unit 8 into 5 partS: BI. B2. B3. B4.
and 85 (Figs. 23, 24).
Units BI, B2, and B3 form lenses low in B. B1 overlies either unit A or pre
lSunami soil. Though I mapped B2 as a distinct unit in the field, grain-size analysis
54suggests it is just the finer normally graded top to Bl. 81 and 83 are very fine to
medium sand, normally graded in their upper portions and, here and there, inversely
graded near their base. They vary substantially in thickness-in some places B3 or BI is
completely absent-but they generally thin inland. An abrupt contact separates the
coarser Bl and 83 from the overlying 84.
Unit 84, composed of coarse silt and very fine sand, is the only unit that thins
toward the sea, instead of toward the land. Though normally graded in places, it shows
complex grading in olhers. Just landward of a trench 464 m inland, 84 drapes a 2-m-tall
ridge as a deposit 15 em thick. Where not eroded on top, 84 fines upward to a silt layer
(85). Because the transition from 84 to 85 is gradational everywhere, 85 does not have a
readily defined thickness, but the sand-free portion is typically less than 5 mm thick.
Though completely absent in a trench SO m inland of the beach, 85 is present at nearby
test pits (Fig. 23, also Stratigraphic Map I).
UnitCInto unit C I lump any tsunami deposit above unit 8. Near the shore I recognized
two upward-fining beds (CI and C2) (Fig. 4). In some places C2 sharply overlies CI, but
in others the fine upper part ofCl grades, through upward coarsening, into the lower part
of C2. Farther inland, tsunami deposits above unit 8 consist ofa single inversely graded
layer that I correlate with CI by stratigraphic position alone (Fig 5).
In one location just seaward ofa local rise I found a lens of sand, CO, that
sepaiates 85 from Cl (Fig. 24). This lens is distinct because it was lighter in color and
fined up to mud before being capped by the more extensive CI (Stratigraphic Map 4c).
S~dimt!ntary structuresThe 2004 tsunami deposit at Bang Lut includes more normal grading than inverse
grading. Among 377 particle·size analyses at 11 localities, finer sediment overlies
coarser sediment in 63% ofadjacent sample pairs. Normal grading is common in the
upper parts of 8J and 83, and is locally dominant in B4. as well as in CI and high in C2
near the beach. I noticed inverse grading strong enough to change mean grain size by a
factor of J.S or more (O.S~) in the lower parts of Bl, B3, and C2 only. Where I could
55trace it in the field, this inverse grading continues only lOs ofem laterally and passes
into sharp coarse-over-fine contacts (e.g. Fig. 23, contact between CI and C2).
I found lamination, exclusively parallel, in dried trench faces dusted lightly with a
paint brush. J noticed no hint of ripple-drift cross·lamination. In a few locations, subtle
coarse laminae in B dip at less than 5 degrees landward (e.g. Stratigraphic Map 2b, 3b,
I.7b).
The clearest indicators of flow direction are consistently inclined leaves and stems
of herbaceous plants. These plant remains. rooted in the pre·tsunami soil and entombed
in the lSWlami deposits, point in the likely direction afOow. For example, 120 m inland,
grass low in unit B lies nearly horizontal and points landward (Stratigraphic Map 2b).
However, 320 m inland a clump of grass shows signs ofhaving flopped landward during
deposition of unit B3 and seaward after deposition of B5 (Stratigraphic Map 4d).
Flow direction may also be indicated by preservation of lower units. Unit A
appears to have escaped erosion more commonly on landward-facing slopes than on
slopes that face the sea (Figs. 23, 24, also Stratigraphic Map). Similarly unit Bl survives
more often on landward-facing slopes (Figs. 23, 24, also Stratigraphic Map).
Inferred sequence of flowsI interpret the 2004 tsunami deposits at Bang Lut as a record of five surges
(increases in landward flow), two or three successive waves (inland flow followed by
seaward flow), and a reflected bore (seaward-moving wave not necessarily associated
with 'seaward flow) (Fig. 26). Unit A represents either a distinct wave that preceded
deposition of unit B (as inferred below) or the leading surge of the same wave that
deposited unit B. Unit B marks the arrival of two surges in quick succession. The first
of these removed much ofunit A, and the second formed a reflected bore that moved
back across the beach-ridge plain, recorded by B4, but probably did not reverse the flow.
Thereafter the water stilled, as shown by the silt that caps unit B, and it withdrew enough
to bend grass and bury B in a lens affine sediment (Fig. 24). An additional wave,
smaller than its predecessor, included two surges and left behind unit C.
First wave
-.
56Unit A must record the first part of the 2004 tsunami to deliver sand to the
coastal plain at Bang LUI, since it fills burrows in the pre·tsunami soil. If some other
flow bearing sand had preceded this flow, then sand it carried would presumably be
caught in these burrows, and the flow that deposited A would have been unable to erode
it. Unit A is finer than overlying unit B, which may mean that it was deposited by a
smaller wave than the one that deposited B.
Alternately, the grain-size contrast may reflect differences in the sediment source
(Fig. 27), if I assume that the nearshore sediment was finer than the beach sediment. As
the water withdrew, it scoured the nearshore and became loaded with sediment. Then, as
this water pushed landward, it encountered hundreds of meters of nearshore seafloor
exposed by the leading withdrawal. Some time after the flow moved onto exposed
seafloor it reached its maximwn capacity to transport sediment. and stopped eroding. By
the time the water moved across the beach it had slowed due to friction and increasing
bed elevation, losing sediment capacity. As this water flooded inland. it deposited this
finer nearshore sediment. The asswnption that the nearshore sediment was finer than the
beach sand is supported the lack of silt seen in the tsunami deposit, in the post-tsunami
beach sand, and in the paleo~beach sand underlying the pre~tsunami soil. Subsequent
waves encountered water left by this first wave, and so they could erode the beach and
leave coarser deposits.
Second waveFirst surges
The second wave arrived as two closely spaced, powerful surges. The first surge
probably eroded much of the deposit of the first wave; unit A is present only in local
pockets or as a millimeter thick smear, but can be over 2 cm thick on the landward side of
mounds or vegetation where it was presumably more protected from erosion (Figs 23, 24,
also Stratigraphic Map). Deposition of medium to coarse sand (BI) then began, but this
deposition or subsequent erosion was patchy, leaving BI in lenses (Figs. 23.24, also
Stratigraphic Map). Periods of erosion are in places recorded as distinct but
discontinuous laminae composed of the coarsest grains within BI, which I interpret as a
lag (Stratigraphic Map 2, 3, 7). Following the first surge the flow slowed, as shown by
••
57the very fine sand of B1. But then another surge delivered coarser sand to the beach
ridge plain (B3).
Because these three subunits are discontinuous. I infer that deposition was
repeatedly interrupted by erosion, and that deposition was patchy on the bed. Alternately
the lateral variability in these basal layers might be evidence of sorting on the bed
coarser grains can preferentially collect where other coarse grains are present on the bed
because they are trapped in the higher roughness there (Komar, 1996). However, if this
were the case I would expect to see transitional zones along lamina between coarse and
fine sediment where coarse and fine sediment interlayered. Instead I found the coarse
sediment as lenses (Figs. 23, 24, also Stratigraphic Map).
The m'o surges marked by units B/ and 83 probably came in quick succession.
the second climbing atop the first, as was seen for the second wave at Hombill Hill and
Patong Beach. In some locations. normal grading from Bl to B2 passes upward smoothly
into inverse grading across the contact with B3 (Stratigraphic Map 3e). Had the first
surge slowed and reversed before the second surge. B3 should instead overlie B2
disconformably.
Rejlected boreA reflected bore may explain the seaward-thinning, relatively fine unit (B-1) that
overlies 81-83 (Fig. 28). umericaJ models show simulated tsunamis reflecting ofT
slopes (Chapter 2). and photos show the 2004 tsunami reflecting orr Phi Phi Beach
(Higman et aI., 2007). At Bang Lut, the fonner shoreline scarp 600 m inland provides a
reflector that the 2004 tsunami. at its highest run-up, slightly overtopped (Fig. 2b). This
highest run-up was probably attained during the surge marked by 83 because unit 8 is the
only part of the tsunami deposit that extends to the scarp and because 83 probably
represents a.surge that piggybacked on and thus achieved greater height than the surge
marked by 81. Because the flow was energetic enough to carry sand all the way to the
scarp, and because it immersed this scarp in over 4 m of flow. it is likely that a reflected
bore formed.
Several architectural features of unit B4 are consistent with a reflected bore
(Chapter 2): First, the bore's onset may be marked by the anomaJous presence of 84 as a
"
58drape on a sharp ridge 470 m inland. This ridge-crest drape contrasts with the tapering
ofall other units, both here and farther seaward, across ridges of any kind (Fig. 24).
Because this typical tapering implies shear and erosion across high points, its absence for
84 at the sharp ridge implies an abrupt halt in the tsunami's landward advance. Second,
lhe seaward thinning of 84 is consistent with a reflected bore that encountered
progressively slower moving, more sediment-poor water. Third, the characteristically
sharp contact between B3 and 84 (or between HI and 84 where 82 and B3 are absent)
implies flow that was erosional or non-depositional during a transition from strong flow
(marked by the relatively coarse sand of Bl and B3) to weaker flow (the finer sand of
84). I infer that the turbulent water at the front of the reflected bore reworked the
sediment surface, then deposited finer sediment (Fig. 28). Such a hiatus would not be
expected from a landward flow that simply wanes; instead, across an infinite plain, the
2004 tsunami deposit would steadily fine upward from B3 to the silt of 85
Still waterfor the silt cap (85) to be deposited, withdrawal of the second wave must have
been preceded by enough slowing for all sand·size sediment to settle out of the flow.
Because the silt cap, though thin. is present almost to a ridge trenched 320 m inland (Fig.
24), it must have settled out of water that covered the entire plain, not jusl isolated pools.
However, very fine sand falls I min 5 minutes, and it is unlikely that the period of the
wave was so long that the time ofhigh water was the 20 to 30 minutes needed for the
highest sand grains to reach the bed. Instead of a time ofcompletely still water, the silt
cap must record a time of slow withdrawal that resulted in thinning the flow. A thinning
flow advects sediment downward along with the water, adding this advective rate to the
settling velocity of sediment.
Withdrawal,Following this slow withdrawal, stronger withdrawal of the second wave flopped
grass to seaward at 320 m inland and deposited lenticular unit CO (Fig. 24). Because I
found CO at just one place, immediately seaward ofa local rise where B5 had been
eroded, I ascribe it to seaward flow that locally scoured and re-deposited the upper part of
59B. Sparse lenses of sand interpreted as recording withdrawal were also described by
Fujino et al. (in press) along the coast near Khao Lak.
Lour Kla~'e or waJ.'esAfter the waves and surges recorded by units A and B, another wave carried very
fine sand from a source seaward of the beach~ridge plain to sites hundreds of meters
inland. as marked by unit Cl. Although a withdrawal preceding this wave can explain
some of the deposit following unit B (unit CO, Fig. 24). 85 shows too little sign of
erosion for reworking alone to explain unit C, which is continuous in my trenches 52 m.
112 m. and 160 m inland, and mostly continuous in my trenches 320 m inland (Figs. 4. 5.
2). Furthermore. on this part of the beach·ridge plain. the preservation afuoit 85 rules
out unit B as a source for C. Even in the trench 52 m inland, where B5 is absent B-1
escaped complete erosion and C consists of sand coarser than any in B-1. Consequently.
I infer that much of the sand in C originated seaward of this trench.
Two closely spaced waves are implied by unit C in the trench 52 m inland. Here.
C contains two subunits, C I and C2. wilh inverse grading rather than a sharp contact
locally present at the contact between them (Fig. 23). This lack of sharp contact implies
two surges rather than waves separated by stilling or withdrawal.
ImplicafionsThe record of the 2004 Indian Ocean tsunami at Bang Lut is no! laterally
continuous. even over just a few meters. so any single section fails to describe the
struc~ of the deposit. All layers vary in thickness and in some cases arc present only
as lenses. This variability in the Isunami's stratigraphic record probably reflects
differences in the way the flow is recorded. rather than diOerences in the sequence of
waves and surges that affected points only a few meters apart. Barring sorting on the bed
(as discussed and ruled OUI for deposition of units BJ and Bl). this horizontal variability
shows that the deposits at din"crent parts of the beach-ridge plain record the flow for
different spans of time.
This variability probably reflects how the sequence of waves and surges respo·nds
to variations in the bed. Where a flow expands. such as this tsunami would have at the
depression 320 m inland of the beach. deposition is enhanced. leading to thicker deposits
60and thus to a more complete record ofa flow (Fig. 24). Conversely where a flow thins
over a local rise it accelerates, which leads to increased shear on the bed and erosion.
This interplay between topography and flow has the greatest effect on the deposit when
the flow is quasi-steady. Ifthe flow rapidly accelerates, it will tend to erode everywhere,
and ifit rapidly decelerates, as presumably happened over the ridge landward of the
trench at 464 m inland, deposition will occur uniformly and drape the whole bed.
Regardless of whether this process controls the variation in the deposit, the type
of variability observed implies that deposition, not just erosion, was patchy and
intermittent. Units A and 85 can be explained by patchy erosion alone; they probably
fonned nearly continuous sheets before erosion by subsequent waves and surges. Both
of these units are overlain by a sharp and irregular contact and by coarser sediment.
evidence that a subsequent flow accelerated enough to erode and to transport coarser
sediment. In contrast for BI and B3, discontinuity is more easily explained by patchy
and intermittent deposition and re·entrainment (brief periods oferosion), because these
two units lack erosional tops and are overlain by finer sediment. Suppose that each unit
accumulated as a continuous sheet that was eroded by a later flow. Though this sequence
of events best explains units A and B5, it fails to explain units that are limited to coarse
lenses, such as BI and B3. Had Bland B3 originated as continuous sheets, their
widespread erosion should have been followed by widespread re-deposition ofcoarse
sand, before finer sand settled out. The absence ofsuch widespread, relatively coarse. .
sand implies that deposition was spatially discontinuous.
Patchy deposition implies that some grains that reached the bed must have been
re-entrained, at least for long enough to move to another location on the bed. Unit B3 is
present over only about halfofthe bed, so during its deposition, at least half the grains
that reached 'the bed were not deposited and must have been transported farther. The
vegetation that covered the beach-ridge plain, including trees, shrubs, and grasses, did not
create enough roughness to trap grains on the bed and prevent re-entrainment.
Re·entrainment presents a potential problem for tsunami sedimentation models
currently under development. Re·entrainment is treated as insignificant in two
quantitative models for tsunami sediment transport. In a model based on studies of
•
61fluvial sediment transport, Jaffe and Gelfenbawn (in press) assume that the sediment in
a tsunami deposit reflects the sediment in the flow at one point in time. However if
deposition (e.g. in a local depression) occurs at the same time as nearby erosion (on a
local rise), then different portions of the deposit record different spans of time and thus
are not equally representative of the flow. Advection-trajectory models (e.g. Soulsby et
aI., 2007) also include the assumption that no re-entrainment occurs. The uncertainties
that re-entrainment adds to these models need to be identified and quantified for these
models to be accurate tools for estimating paleo-tsunami hydraulics.
ConclusionsA deposit of the 2004 Indian Ocean tsunami in Thailand contains horizontal and
vertical variability recording varying inflow. Outflow appears to explain only one
isolated lens. However, part of the deposit may represent a seaward-moving, reflected
bore that probably stopped the flow without reversing it.
The deposit likely records three waves. The first wave deposited fine sand across
much of the beach-ridge plain (unit A). This deposit was then panially eroded by a pair
of surges from the second wave (units BI-B3) that were followed by a reflected bore (B4)
and still water (B5). The water then withdrew at least partially. before the last wave
recorded in the deposit anived as two surges, the first (C/) stronger than the second (Cl).
11le deposit's lateral variability implies interplay between deposition and erosion.
Only unit B4 near the landward end of inundation appears to have been deposited in a
laterally continuous drape. Elsewhere the deposit is characterized by tapering beds and
lenses that suggest a quasi-steady flow that deposited in one place while it overpassed or
eroded the bed only tens of centimeters or meters away. Lateral variability is an
important part of the information recorded in tsunami deposits, which brings into
question models that assume deposition is spatially and temporally continuous.
62
THAILAND
INDONESIA
CAMBODIA
MALAYSIA
'.•
100" E
EutalllnPlate
N
t
lndo-Auatr.llian II...~
60 """"
Okm,
INDIANOCEAN
0"
"" N
2O" N 9O"E
& Eyewtne$$ • 1+ m upliIl :. -; Rupture areaobservillJ::ln • 1. m subsldence
Figure 20 Index map showing relation between tsunami source and my Thailand field area (Fig.21). Rupture, uplift, and subsidence inverted from near-field vertical and horizontal deformationmeasurements and from far-field seismic records by Chlieh et aI., 2007.
, ,
63
•
•
•
•••
•
.....- ......
•
bW,i ow j21lO1.n 01__.-
~-.,.....
•
~? ._~,--•
,, •-, !.,.t7 .1 './oIt-- • •• BIn _ """" •, PiIiit .... _8a'lgW I~
•1000000..,f'll it •
•, ,,
,.~• ", •._~
•,• • • •- •
I : ,• -, ,
:"t
. ...
, .-,! ~-.-..
,~,t
,
. ,,~~_tom_1Ol'
- T,-'!",_,,<l'IQj(f'9". ~.Sli
Figure 21 a Coastline and bathymetry near Bang LUI with cartoon shapes for attacking waves.Wave shapes are based on eyewitness description (Ban Nam Khem: Hori et al.,2oo7; Khao Lak:Smothers 2005; Eyewitness Photos and Videos (Hombill Hill», video and photos (EyewitnessPhotos and Videos (Khao Lak and Hombill Hill», and a tide gauge (KuraburL Thaicharoen etal.,2005). Solid line indicates the portion of the cartoon described or documented by theeyewitnesses, and numbers count the observed wave crests. Light gray along the coast shows theapproximate area of flooding and does not include flooding along mangrove-lined estuaries(Chuwong 2005). b Tsunami high-water marks are from Umitsu et al. (in press), and Choi et al.(2006). c Offshore and onshore profiles. Bathymetry from bathymetric maps checked withhandline and sonar. .
, ,
64
•
'~m
--• •- .., :-..
Cl~UA$'"~~~
EI.OIUIOH
:r2W~-"-
J:I~l '. :oi~,.. All
;. " r~ :.' ~'.
0, ~ J ~E ...r.",, __
.. , .. • .. OS'....... (.. l
LL~m
_:lO ":'I!j • ','- .,
i " . ~mi1O • ':.:.;.
"
tJ
.... . .
•.'•
~ •'.
• •1'.
• ~• 'i' .
--''"
• Ilt:_ I'HOIO'MTH 1Rt;NCl1lOCAIlONS
......
A J1oIIICKNf.$$A$" r...FUNt1lQN Of' !DIS'IAN(% N..AHtI.
- """"" <--
,
•• 5o""'o<r... o.' .........
, ~ """"'0 IlC"<l
O<u><>or~~_ lOCooe> •
"'...,,:1006~_-. •
1I......c.Ill'"""'Q ""lUI''''''' []~..._d ....~..... , 1"
Figure 22 Mapped struclure of the tsunami deposit at Bang Lut.
'.
65
•
•
•
'" ,.,,
,
••
..,
.".
..,
M FI"<*>o •• '/3~!:::I•
C<_ .......... ,.IIS':<IgIW'Io .. ..,.,,-.., ,_ fI;-,l_.... _S............0<1 "" • ....,.. ........
..... 1'>;1,- "'dO"","""'''''''Cloy""" 'flOw> .' 0"'0-"'0'1'........~ In<_ comoo....O.,,_=.-k>~_....._ -..-<1_._ ad' ....,
__••~iOCI'to_ ....... __ ......... w.-............ed ... _m;r""'" ""._......."..",,-_ ....__ Tho U"J'__ .
1.... __, '" _ C*n, ....,'" tI
,""'Ore.. ;or<> _00AIIt 0<>"''1'"''
•
•
..,Q~
(''''''''''OCIII'_ 0011_..._ ,*Figure23 Trench 52 m from top of beach (also Stratigraphic Map la-h) For a complete key tounits see Fig. 26. a and b show traced and simplified field sketches of a 4 meier trench. c and dshow photos taken of this trench. and e and f show the grain-size distribution of units in thedeposit.
•
66-,
t
I, Looo-.-.,clo"
~ 1 CJIIICtOJ.1_01__>
~. '--No_••~_
Figure 24 Trenches 320 m from lOp of beach (also Stratigraphic Map 4a-e and 5a-<l) For acomplete key to units see Fig. 26. Sketches of two trenches only a few meters apart, one on alocal rise and the other in a depression left by tin miners. Green deJX)Sit is inferred to have beendeposited by post-tsunami rain-scouring afthe slope into the depression. The life position Donaxgenus bivalve, commonly found in the intertidal zone, presumably survived after transport anddeposition long enough to reorient itself.
gJO; 2O~~.r''v-../'-''\
=> L_trench<;; (depreuoonj
J10
~ 0 '----~,'••-- ,',ElevatIOn of base of Un.. B (m)
Figure 25 Thickness of unit B plotted as a function ofelevation for the two trenches shown inFig. 24, In the depression, B drapes uniformly over topography, while higher up it laps ontoexisting topography, thinning with elevation.
,.
67
-'. " --.-e-__ - --- - ....... ~t< lS0111».:l5O! />'3501 - ........ _....... lap • ,__-..d ...__ ..-.c._d....., • • ., - ---_.....--...__or..-• ' '......__0110_....., ...-~
. It _Cl~_-. __ol___
• ••• • - -I
, ,__OIl "" ... ___ 0nIt .... , _ .. """' _ .., , •• • Iig/'W n "'*"_ .......-... 0>«1y0'Q CI, _w.......... _af .... _ ..... 1~_..,..
i • • , - ...__dlr__ t ...~ 0"''''' ... dldur..........
__ "'" t:u"""~' C•• .11I ,lloflKl«Il>o<o tro.""" II .... only "'" "'.. ",.'$ """"...., ....
! • , • 7... -. -. -$Koncl _ of "'" _ .............
I• I , ., - --. -.• , • Anl ....,.. of .... __..... TN _ 01 .... ClOjlOOoI"
• - - ---.... _"po_,,~.- """~• , • ••• 1-.lI ....~t_"' ...............•
_ ........._01 ... _• • ., -._-
.--- JI._-
._~• .....-....g Ii._ .._~![ -Co:<" °4----Figure 26 Summary ofobservations and interpretations oflhe tsunami deposit.
, ,
••
Focused .:.c:.:.:~:.:-.::t="=:lerosion
v.Il1en Kstills befol'e retreal,the fQt wave deposits itssediment. The resultingcleaner witer can erodeeithef a, it runs down thebead'! Of as illS ac • !etatedby the next wave
68
d
b_
Fin! wave has litHecapacity 10 erode beach
•First wIve aooelerlltessediment nch waler fromwithdrawal
Withdrawal erodesnearshore sediment
•
•
•
•'. Neal$lIore
.....Bl!!fore tsunami
figure 27 Sequence oferosion and deposition likely in the first wave. Clean water presentbefore the tsunami (a) flows down the sloping near-shore seafloor eroding sediment (b). Thissediment-laden waler is incorporated into the incoming wave (e) which slows as il climbs acrossthe exposed seafloor and is depositional when it arrives at the beach (d). Water can lake hours todays to drain off a beach-ridge plain, so when the next wave arrives it will incorporate watersflowing down across the beach and standing on the beach-ridge plain (e). In this case erosion ofthe beach is more likely because waler near the beach will be sediment poor, and will haveunfilled capacity after it is accelerated.
• •
-
o STU .. GItAl:llWlT-,.,..--...0_01_
C 80IIE loICM:S SEAwo.RD_ ...llI'4._"'II '''''''"'ll-
/1" IlIOFlECfW IIORf fORMS
-.g ......fI_~
~. -.-1
A INCOMING SUIlGE c.o .............. """'II'~cc......... "" .... _
,--",__/ .'
.'••
-
--
·S._ _..Ii' _-..,_
( 851
· ColI,_~..........Gllf>I""",. "' .._
·l~""'of.,. ......",..,._..:'
Fe ..I
· SOl a~,_,""
1.... ,."'4-1:
· l-.., """'1"(1 cor........ '_"'fI"
.................. ~I..'T~.:_••, ..-, II''''''''*'' ._ "l"..
.... ~ 00...11,•
69
__ ,co. ", ..__
figure 28 Formation ofa reflected bore deposit. Time goes from bottom to top. At left is acartoon showing the changing water surface, while at right is a cartoon of the deposit, depictedwith extreme vertical exaggeration.
• •
70
References
Atwater, B.F. 1987, "Evidence for great Holocene earthquakes along the outer coast of
Washington State", Science. vol. 236, no. 4804, pp. 942-944.
Atwater, a.F. & Hemphill.Haley, E. 1997, Recurrence ifllervafsjor great earthquakes of/he past3,500 years 01 northeastern Wiflapa Bay, Washington, U. S. Geological Survey ProfessionalPaper /576 United Stales (USA).
Atwater, B.F., Musumi·Rokkaku, S., Satake, K .• Tsuji, Y., Ueda, K. & Yamaguchi, O.K. 2005,
The orphan tsunami of /700: Japanese clues to a parent earthquake in North America,
United Stales (USA).
Bagnold, R.A. 1956, "The Flow of Cohesionless Grains in Fluids". Philosophical Transactions of
the Royal Society ofLondon Series A-Mathematical and Physical Sciences. vol. 249, no.
964, pp. 235.
Bondevik, 5., Svendsen, J.I. & Mangerud, J. 1997, "Tsunami sedimentary facies deposited by the
Storegga tsunami in shallow marine basins and coastal lakes, western Norway",
Sedimentology, vol, 44, no. 6, pp. 1115-1131.
Bourgeois, J. in press, "The geologic effects and records of tsunamis", The Sea, vol. 15. Harvard
University Press.
Bourgeois,J., Hansen, T.A., Wiberg, P.L. & Kauffman, E.G. 1988, "A tsunami deposit at the
Cretaceous-Tertiary boundary in Texas", Science, vol. 241, no. 4865, pp. 567-570.
Bourgeois, J., Pinegina. T.K., Ponomareva, V. & Zaretskaia, N. 2006, "Holocene tsunamis in the
southwestern Bering Sea, Russian Far East, and their tectonic implications", Geological
Society 0/America Bullelil/, vol. 118, no. 3-4, pp. 449-463.
Carey, S., Morelli, D., Sigurdsson, H. & Bronto, S. 2001, "Tsunami deposits from major
explosive eruptions; an example from the 1883 eruption of Krakatau", Geology (Boulder).
vol. 29, no. 4, pp. 347·350.
Cellino, M. & Graf, W.H. 1999, "Sediment-laden flow in open-channels under noncapacity'and
capacity conditions", Journal o/Hydraulic Engineering-Asce. vol. 125, no. 5, pp. 455-462.
, ,
71Chadha, R.K., Latha, G., Yeh, H,t Peterson, C. & Katada, T. 2005, "The tsunami ofthe great
Sumatra earthquake ofM9.0 on 26 December 2004-lmpact on the east coast of India~.
Current Science, vol. 88, no. 8, pp. 1297·130 I.
Chlieh, M., Avouac, J.P., Hjorleifsdonir, V_, Song, T.R.A., Ji t C., Sieh, K., Sladen, A., Hebert,
H" Prawirodirdjo. L. t Bock, Y. & Galetzka, J. 2007, "Coseismic slip and aftersJip of the
great M-w 9.15 Sumatra-Andaman earthquake of2004", Bulletin a/the Seismological
Society ofAmerica. vol. 97, no. I, pp. $152-5173.
Chai, B. 2006, Analysis and modeling ofthe distribution functions ofrunup heights ofthe
December 26, 2004 earthquake tsunami in the Indian Ocean, Hanrimwon Publishing Co.,
Lid.
Chuwong, M. 2006, 2004 Tsunami Event: Geologic guide bookfrom Ihe Andoman Coast
southern Thailand, Thailand.
Dowrick, D.J. & Rhoades, DA 2004, "Relations between earthquake magnitude and fault
rupture dimensions: How regionally variable are they?", Bulletin o/the Seismological
Society 0/America, vol. 94, no. 3. pp. 77fr788.
Dypvik, H. & Jansa, L.F. 2003, "Sedimentary signatures and processes during marine bolide
impacts: a review", Sedimentary Geology, vol. 161, no. 3-4, pp. 309-337.
Eaton, J.P., Richter, D.H. & Ault, W.U. 1961, "The tsunami of May 23, 1960, on the island of
Hawaii", Seismological Society 0/America Bullerin, vol. 51, pp. 135-157.
Emery, D., Myers, KJ. 1996, Sequence stratigraphy. Blackwell Science, london, United
-Kingdom (GBR), United Kingdom (GBR).
Ferguson, R.I. & Church, M. 2004, "A simple universal equation for grain senling velocity",
Journal o/Sedimentary Research, vol. 74, no. 6, pp. 933-937.
Fritz, H.M., Borrero, J.C., Synolakis, C.E. & Yoo, J. 2006, "2004 Indian Ocean tsunami flow
velocity measurements from survivor videos", Geophysical Research Letlers, vol. 33, no.
24, pp. L24605.
Fujii, Y. & Satake, K. 2007, "Tsunami source of the 2004 Sumatra-Andaman earthquake inferred
rrom tide gauge and satellite data", Bullerin o/the Seismological Society 0/America, vol. 97,
no. I, pp. S 192-S207.
72Fujino, S., Naruse, H., Fujita., K., Suphawajruksakul, A. & Jarupongsakul, T. in press,
'"Tsunami sedimentation: an example from Indian Ocean Tsunami in southwestern Thailand"
in Tsunamiites-their features and implications, eds. T. Shiki, K. Minora, Y. Tsuji & T.
Yamazaki, Elsevier.
Gelfenbaum, G. & Jaffe, B. 2003, "Erosion and sedimentation from the 17 July, 1998 Papua New
Guinea tsunami", Pure and Applied Geophysics. vol. 160, no. 10·)1, pp. 1969-1999.
Gelfenbaum, G., Valani, D., Jaffe, RE. & Dekker, F. 2007, "Tsunami inundation and sediment
transport in the vicinity of coastal mangrove forest", Prot:. Coastal Sediments '07,
Glikson, A.V. 2004, "Early Precambrian asteroid impact-triggered Tsunami: Excavated seabed,
debris flows, exotic boulders., and turbulence features associated with 3.47-2.47 Ga-old
asteroid impact fallout units, Pilbara Craton, Westem Australia", Astrobiology. vol. 4, no. I,
pp. 19-50.
Goff, J.R., Liu, P., Higman, 8., Morton, R.A., Jaffe, B.E., Femando, H., Lynen, P., Fritz, H.M.,
Synolakis, C.E. & Femando, S. 2006, "Sri Lanka field survey after the December 2004Indian Ocean Tsunami", EarthquaJce Spectra. vol. 22, no. Special Issue III, pp. 155-172.
Grilli, S.,loualalen, M., Asavanant, J., Shi, F., Kirby, J. &. Watts, P. in press. "Source constraints
and model simulation of the December 26, 2004 Indian Ocean Tsunami", Watenl'Oys. Port.Ocean and Coastal Engineering. Special issue on Tsunami Engineering.
Hagstrum, J.T. 2005, "Antipodal hotspots and bipolar catastrophes: Were oceanic large-body
impacts the cause?", Earth and Planetary Science Lellers. vol. 236, no. 1-2, pp. 13-27.
Hawkes, A.D., Bird, M., Cowie, S., Grundy-WaIT, c., Horton, 8.P., Hwai, A.T.S., Law, L.,
Macgregor, C., Nett, J., Ong, J.E., Rigg, J., Robinson, R., Tan-Mullins, M., Sa. T.T., Yasin,
Z. & Aik, L.W. in press, "Sediments deposited by the 2004 Indian Ocean tsunami atong the
Malaysia-Thailand peninsula", Marine Geology.
Higman, B. & Bourgeois, J. in press, "Deposits of the 1992 Nicaragua tsunami" in TsunamiiJes-their features and implications, eds. T. Shiki, K. Minora, Y. Tsuji & T. Yamazaki, Elsevier.
Higman, B., Gelfenbaum, G., Lynen, P., Moore, A. & Jaffe. B.E. 2007, "Predicted sedimentary
recoro of reflected bores", Proc. Coastal Sediments '07.
Higman, B., Lynen, P., McAdoo, B., Borrero, J., Ruggiero, P. 2005 "Geomorphic imprint of the
2004 Sumatra tsunami", Abstracts with Programs-Geological Society of America, vol. 37,no. 7, pp. 93.
73Hari, K., Kuzumoto, R., Hirouchi. D., Umitsu, M" Janjirawuttikul, N. & Patanakanog, B.
2007, RHorizontal and vertical variation of2004 Indian tsunami deposits: An example of
two transects along the western coast of ThaiIandR, Marine Geology. vol. 239, no. 3-4, pp.
163-172.
Ishii, M., Shearer, P.M., Houston, H. & Vidale. J.E. 2005, RExteot, duration and speed of the
2004 Sumatra-Andaman earthquake imaged by the Hi-Net array", No/ure, vol. 435, no.
7044, pp. 933-936.
Jaffe, D.E. & Gelfenbaum, G. in press, "A simple model for calculating tsunami flow speed from
tsunami deposits", Sedimenlary Geology.
Jaffe, D.E., Borrero, J.e., Prasetya, G., McAdoo, 8.G., Gelfenbaum, G" Morton, R.A., Ruggiero.
P., Higman, B., Dengler, L., Hidayat. R. t Kingsley, E" Kongko, W., LUkijanto, Moore, AL.,
TiIOV, V. & Yulianto, E. 2006, "Northwest Sumatra and offshore islands field survey after
the December 2004 Indian Ocean Tsunami", Earthquake Spectra, vol. 22, no. Special Issue
III, pp. 105-135.
Jaffe, B.E., Gelfenbaum, G., Pelers, R., Anima, R., Swensson, M., Olcese, D., Beroales, L..
Gomez, J. & Riega, P. 2003, "Tsunami deposits: Identification and interpretation of tsunami
deposits from the June 23, 2001 Peru tsunami", Proc. CoaJtal Sediments '03.
Kanamori, H. 1977, "The energy release in great earthquakes", Joumal o/Geophysical Research,
vol. 82, no. 20, pp. 2981-2987.
Kanamori, H. 1971, "Seismological evidence for a lithospheric noonal faulting; the Sanriku
earthquake of 1933", Physics o/the Earth and Planetary Interiors. vol. 4. no. 4, pp. 289. 300.
Kelsey, H.M., Nelson, A.R., Hemphill-Haley, E. & Witter, R.C. 2005, "Tsunami hislory ofan
Oregon coastal lake reveals a 4600 yr record of greal eanhquakes on the Cascadia
subduction zone", Geological Society 0/America Bulletin. vol. 117, no. 7-8, pp. 1009-1032.
Kench, P.5.; McLean, R.F., Brander, R. W., Nichol, S.L.,. Smithers, S.G., Ford, M.R., Parnell,
K.E. & Aslam, M. 2006, "Geological effects of tsunami on mid-ocean atoll islands; the
Maldives before and after the Sumalran tsunami", Geology (Boulder), vol. 34, no. 3, pp.177-180.
Ki1feather, A.A., Blackford, J.J. & van der Meer, Jaap J. M. 2007, "Micromorphological analysis
of coastal sediments from Willipa Bay, Washington, USA: a technique for analyzing
inferred tsunami deposits", Pure and Appl. Geophys, vol. 164, pp. 509-525 .
• •
74Kneller, 8. 1995, "Beyond the turbidite paradigm; physical models for deposition of turbidites
and their implications for reservoir prediction; Characterization of deep marine clastic
systems", Geological Society Special Publications, vol. 94, pp. 31-49.
Komar, P. 1996, "Entrainment of sediments from deposits of mixed grain sizes and densities" in
Advances in Fluvial Dynamics and Stratigraphy, eds. P. Carling & M. Dawson, pp. 127·181.
Konca, 0., Hjorleifsdottir, V., Song, T.A., Avouac, J., HeImberger, D.V., Ji, C., Sieh, K., Briggs,
R. & Meltzner, A. 2007, "Rupture kinematics of the 2005 Mw 8.6 Nias-Simeulue
Earthquake from the joint inversion of seismic and geodetic data", Bulletin ofthe
Seismological Society ofAmerica, vol. 97, no. lA, pp. 8307-S322.
Konno, E., Iwai, J., Kitamura, N., Kotaka, T., Mii, H., Nakagawa, H., Onuki, Y., Shibata, T. &Takayanagi, Y. 1961, "Geological observations of the Sanriku coastal region damaged by
the tsunami due to the Chile earthquake in 1960", Contributions from the Institute of
Geology and Paleontology, Tohoku University, vol. 52, pp. 40.
Liu, P.L.F., Lynett, P., Fernando, H., Jaffe, B.E., Fritz, H., Higman, B., Morton, R., Goff, J. &
Synolakis, C. 2005, "Observations by the International Tsunami Survey Team in Sri Lanka",
&ience, vol. 308, no. 5728, pp. 1595-1595.
Lynett, Pol., Wu, T.R. & Liu, P.L.F. 2002, "Modeling wave runup with depth-integrated
equations", Coastal Engineering, vol. 46, no. 2, pp. 89-107.
Middleton, G.V. & Southard, J.B. 1984, "Mechanics of sediment movement", SEPM Short
Course, vol. 3, pp. 400.
Mirchina, N.R. & Pelinovsky, E.N. 1981, "The Dependence of Tsunami Wave Period on the.
Source Dimensions", Marine Geodesy. vol. 5, no. 3, pp. 201-208.
Moore, A., Nishimura, Y., Gelfenbaum, G., Kamataki, T. & Triyono, R. 2006, "Sedimentary
deposits of the 26 December 2004 tsunami on the northwest coast of Aceh, Indonesia",
Earth. Planets and Space, vol. 58, no. 2, pp. 253-258.
Murari, M.K., Achyuthan, H. & Singhvi, A.K. 2007, "Luminescence studies on the sediments laid
down by the December 2004 tsunami event: Prospects for the dating of palaeo tsunamis and
for the estimation of sediment fluxes", Current science. vol. 92, no. 3, pp. 367-371.
Nagendra, R., Kannan, B.V.K., Sajith, C., Sen, G., Reddy, A.N. & Srinivasalu, S. 2005, "A
record offoraminiferal assemblage in tsunami sediments along Nagappattinam coast, Tamil
Nadu", Current science, vol. 89, no. ll, pp. 1947-1952.
75Nanayama, F., Furukawa, R., Shigeno, K., Makino, A, Soeda, Y. & Igarashi, Y. in press,
"Nine unusually large tsunami deposits from the past 4000 years at Kiritappu marsh along
the southern KurU Trench", Sedimentary Geology.
Nanayama, F., Salake. K., Furukawa, R., Shimokawa, K., Atwater, B.F., Shigeno, K. & Yamaki,
S. 2003, "Unusually large earthquakes inferred from tsunami deposits along the Kuril
trench", Nature, vol. 424, no. 6949, pp. 660-663.
Nelson, A.R., Atwater, B.F., Bobrowsky, P.T., Bradley, L., Clague, U., Carver, G.A., Darienzo,
M.E., Grant, w.e., Krueger, H.W., Sparks, RJ., Stafford, T.W. & Stuiver, M. 1995,
"Radiocarbon evidence for extensive plate-boundary rupture about 300 years ago at the
Cascadia subduction zone", Nature (London). vol. 378, no. 6555, pp. 371-374.
Perfettini, H., Avouac, J.P. & Ruegg, J.e. 2005, "Geodetic displacements and aftershocks
following the 2001 Mw "" 8.4 Peru earthquake; implications for the mechanics of theearthquake cycle along subduction zones", Journal o/Geophysical Research. vol. 110, no.
B9. pp. 19.
Peters, R., Jaffe, B.E. & Gelfenbaum, G. in press, "Distribution and sedimentary characteristics oftsunami deposits along the Cascadia margin of westem North America", Sedimentary
Geology.
Peterson, C.D., Yeh, H., Chadha, R.K., Latha, G. & Katada, T. 2005, "Flood elevation,inundation distance and flow competence of the 2004 Sumatra-Andaman tsunami, asrecorded by tsunami deposits in thirteen shore-nonnal profiles from the Tamil Nadu
coastline, India", ISET Journal ofEarthquake Technology. vol. 42, pp. 95-95-110.
RubIn, D.M., Nelson, J.M. & Topping, OJ. 1998, "Relation of inversely graded deposits tosuspended-sediment grain-size evolution during the 1996 flood experiment in GrandCanyon", Geology, vol. 26, no. 2, pp. 99-102.
Satake, K. 1994, "Mechanism of the 1992 Nicaragua Tsunami Earthquake", Geophysical
Research Letters, vol. 21, no. 23, pp. 2519-2522.
Satake, K. & Tanioka, Y. 1999, "Sources of tsunami and tsunamigenic earthquakes in subductionzones", Pure and Applied Geophysics. vol. 154, no. 3-4, pp. 467.483.
Schwartz, S.Y. & Ruff, L.J. 1985, "The 1968 Tokachi·Qki and the 1969 Kurile IslandsEarthquakes - Variability in the Rupture Process",Journal o/Geophysical Research-Solid
Earth and Planets. vol. 90, no. NB I0, pp. 8613-8626.
76Singarasubramanian, S.R., Mukesh, M.V., Manoharan. K., Murugan, S., Bakkairaj, D. & Peter,
AJ. 2006, "Sediment characteristics of the M 9 tsunami event between Rameswaram and
Thoothukudi, GulfofMannar, southeast coast of India", Science a/Tsunami Hazards, vol.
25, no. J, pp. 160-172.
Smith, D.E., Shi, S., Cullingford, R.A., Dawson, A.G.. Dawson, S., Firth, C.R., Foster, to.L.,Fretwell, P.T., Haggart, B.A., Holloway, L.K. & Long, D. 2004, '"The holocene storegga
slide Tsunami in the United Kingdom", Quaternary Science Reviews. vol. 23, no. 23·24, pp.
2291-2321.
Smothers, M. 2005, Interview with John M. Thompson on News Talk 1350 KSRO, Santa Rosa,
Califomia.
Soulsby, R.L., Smith, D.E. & Ruffman, A. 2007, "Reconstruction tsunami run-up from
sedimentary characteristics - A simple mathematical model", Proc. Coastal Sedimenls '07.
Stelling, 0.5. & Vankester, JA.T.M. 1994, "On the Approximation of Horizontal Gradients inSigma Coordinates for Bathymetry with Steep Bonom Slopes", IlIternational Journalfor
Numerical Methods in Fluids. vol. 18, no. 10, pp. 915-935.
Synolakis, C.E. & Bernard, E.N. 2006, "Tsunami science before and beyond Boxing Day 2004",Phil050phical Transactions ofthe Royal Society A-Mathematical Physical and Engineering
Sciences. vol. 364, no. 1845, pp. 2231-2265.
Szczucinski, W., Chaimanee, N., Niedzielski, P., Rachlewicz. G.• Saisunichai, D., Tepsuwan. T.•Lorenc, S. & Siepak, J. 2006, "Environmental and geological impacts of the 26 December2004 tsunami in coastal zone ofThailand - Overview ofshor1 and long-term effects", Polish
.Journal ofEnvironmenlol Studies, vol. IS, no. 5, pp. 793-810.
SzcZtJcinski, W., Niedzielski, P., Rachlewicz, G., Sobczynski, T., ZioJa, A., Kowalski, A.,Lorenc, S. & Siepak, J. 2005, "Contamination of tsunami sediments in a coastal zoneinundated by the 26 December 2004 tsunami in Thailand", Envirol/memal Geology. vol. 49,pp.321-331.
Tadepalli, S. & Synolakis, C.E. 1996, "Model for the leading waves of tsunamis", Physical
Review Leiters. vol. 77, no. 10, pp. 2141-2144.
Tanioka, Y. & Satake, K. 1996, "Fault parameters of the 1896 Sanriku tsunami earthquakeestimated from tsunami numerical modeling", Geophysical Research Letters. vol. 23, no. 13,pp. 1549-1552.
• •
77Tappin, D.R., Watts, P., McMurtry, a.M., Lafoy, Y. & Matsumoto, T. 2001, ftThe Sissano,
Papua New Guinea tsunami of July 1998 - offshore evidence on the source mechanism",
Marine Geology, vol. 175, no. 1-4, pp. \-23.
Thaicharoen, C., Weesakul, S. & Das Gupta,. A. 2005. "Tsunami propagation 10 Thailand-A
case study: Phi Phi Island", MTERM International Conference, Bangkok, Thailand, pp. 30]·
312.
TilOv. V" Synolakis.. C. 1998, "Numerical modeling oftKlal wave runup", Journal of Waterway, Pon,
Coastal, and Ocean Engineering, pp. 157·171.
Umitsu, M., Tanavud, C. & Palanakanog, B. in press, "Effects of landfonns on tsunami flow in
the plains of Banda Aceh, Indonesia, and Nam Khem, Thailand", Marine Geology.
Walsh, TJ., Caruthers, e.G., Heinitz, A.C., Myers, E.P. III, Baptista, A.M., Erdakos, G.B. &Kemphaus, R.A. 2000, Tsunami hazard map ofthe southern Washington coast: Modeled
tsunami inundation from a Cascadia subduction zone earthquake, Washington Division of
Geology and Earth Resources.
Wheeler, H.E. 1958, "Time-slraligraphy", Bulletin ofthe American Association ofPetroleumGeologists, vol. 42, no. 5, pp. 1047-1063.
Winer, R.C., Kelsey, H.M. & Hemphill-Haley, E. 2003, ~Greal Cascadia earthquakes and
tsunamis of the past 6700 years, Coquille River estuary, southern coastal Oregon",
Geological Society ofAmerica Bulletin, vol. lIS, no. 10, pp. 1289-1306.
Yeh, H. 2006, "Maximum fluid forces in (he tsunami runup zone", Journal a/Waterway PorI.Coastal and Ocean Engineering-Asce. vol. 132, no. 6, pp. 496-500.
• •
78Appendix A: Grading Data
Tsunami deposit grading dataEach proportion upward fining measurement in Fig. 6 integrates an entire sample
set (Fig. 5). This appendix presents examples of the data included here. I included all
analyzed sample sets for Papua New Guinea, Nicaragua, Peru, and the 2005 Sumatra
tSWlami, along with a subset ofsample sets from the 26 December 2004 Indian Ocean
tsunami. Other sample sets not included are similarly sampled at an interval ofless than
I em and include 10-100 samples each. As discussed in Chapter I, only the thickest
sample set from each locality is included in my analysis and Fig. 5.
.....') _"''0"
.~ :. UPward,.' """II
) Lh••.. K" r-"1:'--
\.>
0'
"'"-
-~ -~ -~_.~~ -
~'".. f ~ .• , ~ , • ,050\_ .: 0155,- II-- <.
•" .... ......,.....\ o.~jO~· "i" ...(l7~~
... , g '-; lJ"*o.~ I
~, ~--" laguftIln
~,
0.J.. (~0" '-'
". 0.f7056 ...., 'r l_
- ~'Y )8fJ~~
."·0.58:.:. eU"on;
,~~" eay
~Ml
\........"'"',-1),73 I (ditWI
0.7'
0,74
,--_Guo>h 1992~_
"""MC IIFFMC\'AI<Ip__Ion0.• ·• _)
'. ~ \'~O.eo :; ..... ~ ...s-. .'.
~\ onlOiIto!l I
0.89 \::::
• •
79Appendix 8: Statistical Results
Statistical comparison of grading in the deposits of longer and sborter periodtsunamis:
I show that the measured proportion normal grading in deposilS of great
subduction zone tsunamis is smaller than measurements from three other more local
tsunamis with more than 95% confidence. Given that grading is a function ofcomplex
geographic and analytic variables, it is not possible to mathematically define a probability
density function (PDF) for proportion nonnal grading. However, it is possible to
compare the relative nonnal grading of pairs of sample sets. Because there are five
sample sets corresponding to local tsunamis and twenty corresponding to the two longer·
period tsunamis, I can generate at most five independent pairs of sample sets. I produced
107 random sets of five independent pairs, and counted in each case how many of the
pairs had more normal grading for a longer-period tsunami than for a local tsunami
(Table 2.) I refer to this situalion, where the random pair has more normal grading in the
longer·period tsunami deposit, as a "reversal."
To determine my 95% confidence, I demonstrate the less than 5% probabilily of
the following null hypothesis:
• The probability p /ha/ a grea/ subdue/ion zone ISImami deposi/
has more normal grading /han local/sunami deposits is a/
least 0.5.
If pairs like this were produced from PDFs where long period tsunami deposit
grading tended to include as much or more normal grading lhan local tsunami deposits,
then the probability of a given number of reversals occurring has a binomial distribution.
The likelihood of a specific number of reversals k given n pairs and a probability of
reversal pis:
The likelihood B that k reversals would occur given p ofat least 0.5 is:
• •
80
Depending on the particular random realization aftive pairs ofsample sets, the number
of reversals k will vary. I distribute B across the proportional occurrence A of varied k to
calculate the total probability that my observed data would occur ifp was at least 0.5:
~Ad31.
Table 2: Statistical Results
"=13 511".( 1,1')" I ~lpe, ''> A.eNumber 01 Occurrences ProportIOnal
f"("I')"'dpRe~ersals occurrence(k) 'A' "0 7,579.802 76% , 56" 1 18%, 2.341,63<1 23% 1094% '56", 78.564 '" " 38% 027%3- 0 0 5313% 0
Total 10,000,000 100% 100% 4.02%
This analysis can be repeated comparing the smallest three earthquakes (Mw 7.0.
7.6, and 8.4) to the largest two (Mw 8.6 and 9.1) and the results are still significant
(3.62%). Increasing the number of analyses on smaller tsunamis would improve these
results, as would developing a more standardized method for sampling.
This analysis depends on the sample sets being representative of the five
tsunamis, and on them representing the five tsunamis in the same way. I believe that my
sample sets are sufficiently representative, but there is significant uncertainty as a result
of this concern. The hypothesis presented here post-dales my post-tsunami surveys, and
there is no established way of selecting sampling locations to be representative.
However, sampling following the 2004 Indian Ocean tsunamis was conducted across a
broad range of settings, including sites on the west coast of Aceh Province in Sumatra;
Simeulue, Nias, and smaller islands off northern Sumatra; Sri Lanka, and Phuket and
Khao Lak in Thailand, and the variation still includes no samples thai have more normal
grading than in the deposits of the 1992 and 1998 tsunamis. There does seem to be some
••
relationship between deposit thickness and normal grading. As discussed in the text
and shown in figure 1, this is consistent with my hypothesis.
.,
81