Available via license: CC BY 4.0
Content may be subject to copyright.
Time-dependent Tsunami Source Following the 2018 Anak Krakatau1
Volcano Eruption Inferred from Nearby Tsunami Recordings2
Yifan Zhu, Key Laboratory of Hydrodynamics (Ministry of Education), School of Naval
Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai 200240,
China, zyftop@sjtu.edu.cn
Chao An, Key Laboratory of Hydrodynamics (Ministry of Education), School of Naval
Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai 200240,
China, anchao@sjtu.edu.cn
Teng Wang, School of Earth and Space Sciences, Peking University, Beijing 100871, China,
wang.teng@pku.edu.cn
Hua Liu, Key Laboratory of Hydrodynamics (Ministry of Education), School of Naval
Architecture, Ocean and Civil Engineering, Shanghai Jiao Tong University, Shanghai 200240,
China, hliu@sjtu.edu.cn
3
Preprint submitted to Earth, Planets and Space on May 11, 20204
1
Abstract5
The eruption of the Anak Krakatau volcano, Indonesia, on 22 December 2018 induced a destructive6
tsunami (the Sunda Strait tsunami), which was recorded by four nearby tidal gauges. In this study we7
invert the tsunami records and recover the tsunami generation process. Two tsunami sources are8
obtained, a static one of instant initial water elevation and a time-dependent one accounting for the9
continuous evolution of water height. The time-dependent results are found to reproduce the tsunami10
recordings more satisfactorily. The complete tsunami generation process lasts approximately 9 min and11
features a two-stage evolution with similar intensity. Each stage lasts about 3.5 min and elevates a12
water volume of about 0.15 km3. The time, duration and volume of the volcano eruption in general13
agree with seismic records and geomorphological interpretations. We also test different sizes of the14
potential source region, which lead to different maximum wave height in the source area, but all the15
results of time-dependent tsunami sources show the robust feature of two stages of wave generation. Our16
results imply a time-dependent and complex process of tsunami generation during the volcano eruption.17
Keywords18
2018 Anak Krakatau tsunami, time-dependent tsunami source, tsunami source inversion19
Introduction20
Large-scale volcanic events in the ocean have a great potential to trigger devastating tsunamis that21
could cause severe damage in nearby coastal areas. For instance, the eruption of the Krakatau volcano22
in 1883, which was one of the most explosive volcanic events in history (?), claimed over 36,000 lives23
and most deaths were attributed to the tsunami (?). However, it is generally difficult to determine24
and simulate the tsunami generation process, due to complex phenomena associated with the volcano25
eruption, such as earthquakes, landslides, caldera collapse, etc (?). Previous studies have attempted to26
develop numerical models by categorizing the tsunamis according to the generation mechanism (?????).27
For example, ?investigated three possible tsunami generation mechanisms following the 1883 Krakatau28
eruption – pyroclastic flow, caldera collapse and phreatomagmatic explosion, based on three different29
models – the two-layer shallow water model, the piston-like plunger model and a simple empirical model,30
2
respectively. They found that the tsunami simulations from the pyroclastic flow model matched the31
tsunami data well for this particular event. Generally speaking, the simulation results depend on a32
variety of presumed parameters, such as the flow density, duration of the eruption, and initial conditions33
including the geometry of the source area, the total volume and the time-dependent flux of the pyroclastic34
flow (e.g., ?). In this study, we bypass the complexity of volcano-water interaction during the tsunami35
generation process, and recover the tsunami source purely based on tsunami recordings and well-developed36
tsunami simulation tools.37
Tsunamis are mostly generated by thrust earthquakes in subduction zones, and there have been many38
research works on the inversion of tsunami waves to constrain the earthquake rupture process (???),39
and the associating tsunami warning strategies (????). If the fault geometry is relatively unclear, or40
tsunamis that are generated by other sources instead of earthquakes, it is more applicable to derive the41
initial sea surface profile (???). In the typical inversion method, the possible tsunami source area is42
divided into grids, and the initial water elevation at each grid is obtained by inverting tsunami data.43
Another approach to reconstruct the tsunami source is the time reversal imaging (TRI) technique, which44
basically adopts numerical simulations to let tsunami waves propagate from observation stations to the45
source area, and interfere with each other to recover the tsunami source (????). Previous studies of TRI46
on tsunamis assumed a static initial tsunami source, which is shown later in this study to be inferior47
to the time-dependent tsunami source for this event. In this study we will present our results using the48
inversion approach to obtain the water elevation profile.49
On 22 December 2018, the eruption of the Anak Krakatau volcano induced a destructive tsunami in the50
Sunda Strait, Indonesia. According to official statistics on 31 January 2019, the tsunami has caused 43751
death, 14,059 injured and 16 are still missing (?). Satellite radar image acquired ∼8 h after the event reveal52
that the western portion of Anak Krakatau completely disappeared after the eruption, which could be a53
major cause of the tsunami (?). The risk of the southwest flank failure was actually warned by previous54
study (?), yet the model underestimated the wave arrival time to Java and Sumatra coasts. Therefore,55
investigation of the real wave records during this event is important to improve our understanding of the56
mechanism of volcano-induced landslides and the associated tsunami hazards.57
For the Anak Krakatau volcano tsunami, post-event field surveys have been carried out widely in Java and58
Sumatra islands, measuring tsunami inundation and impact (???). Furthermore, the induced tsunami59
waves were recorded by four nearby tidal gauges, allowing for reconstructing the tsunami source process60
3
using well-developed tsunami simulation and tracing tools. ?conducted spectral analysis of the tide61
gauge data to reveal the dominant periods of the tsunami, and thus estimated the source length. ?62
applied a trial and error approach based on numerical simulations to constrain the tsunamis source of63
static initial water elevation. ?inferred the landslide basial geometry using satellite images and aerial64
photography, and assumed granular material and dense viscous fluid rheologies to simulate the tsunami.65
They conclude that a single landslide source can explain the observed tsunami waves recorded at tidal66
gauges. However, seismic and infrasound data show that a separated seismic event occurred 115 s prior to67
the major collapse landslide process (?). Geomophological interpretation based on radar images acquired68
∼8 h after the event shows two discrete failure planes (?). Both studies suggest that the process of flank69
failure and tsunami generation is time-dependent and complex. Here, we utilize the tsunami recordings to70
reconstruct the temporal evolution of the tsunami source. We compare the inverted source process with71
seismic records and geomophological interpretation and suggest a 9-min process divided into two main72
sources. The results could be helpful for further investigation of the complex volcano eruption process73
without knowing the source geometry and mechanism.74
Data and Method75
The four tidal gauges that recorded the tsunami waves in the 2018 Akak Krakatau event are shown in76
Figure ??. We prescribe a possible tsunami source area on the southwest of Anak Krakatau volcano,77
extending from 105.38◦E to 105.42◦E in longitude and from 6.14◦S to 6.09◦S in latitude. The dimension78
of the source area is about 4.3 km ×5.6 km, divided into 6 ×8 grids of size 0.7 km ×0.7 km. We also79
test two larger tsunami source areas, which are less likely to represent the actual tsunami generation80
region since they overlap the surrounding islands. The medium source region covers an area of about81
5.6 km ×6.7 km (105.38E to105.43E , 6.15S to 6.09S), and the large source area is about 13.2 km ×82
11.2 km (105.35E to 105.47E, 6.17S to 6.07S). The three different source areas are shown by rectangles83
of different colors in Figure ??. For each grid, we assign an initial water elevation profile using the84
Gaussian-shaped basis function (?), and then simulate the tsunami waves at the four tidal gauges to85
obtain the Green’s functions. The bathymetry data are extracted from the General Bathymetric Chart86
of the Oceans (GEBCO 2014) bathymetry with spatial resolution of 30 arc sec, which are then refined to87
grid size of 7.5 arc sec by spline interpolation. The linear version of tsunami simulation package Cornell88
Multi-grid Coupled Tsunami Model (COMCOT) (???) is adopted to simulate the tsunami propagation.89
4
The total simulation time is 6,000 s, and the time step is 0.5 s to satisfy the Courant-Friedrichs-Lewy90
(C.F.L.) condition.91
Static Tsunami Source92
We first assume a static profile of water elevation as the tsunami source. It is known that the volcano93
eruption time is around 14:00 (UTC time), but the exact initial time of the tsunami source is unclear.94
Therefore, we conduct a search from 13:50 to 14:05 with an interval of 1 min to find the optimum initial95
source time. For each initial time, we adopt the non-negative least squares method in the inversion96
(?), on the basis that volcano-induced landslide is assumed to cause positive sea surface displacement97
due to mass entering the ocean. For the three different source sizes, results consistently show that98
the initial time of 13:54 best matches the tsunami recordings, which is similar to the origin time of a99
low-frequency earthquake on the regional seismic network (Mw 5.1, 13:55 UTC, ?, GEOFON Program,100
https://geofon.gfz-pots-dam.de/eqinfo/event.php?id=gfz2018yzre) (Mw 5.3, 13:55:49 UTC, ?).101
The optimum static tsunami sources of different source sizes are shown in the top panel of Figure ??,102
and the predicted tsunami waves are plotted in the bottom panel. More details about the tsunamis103
sources at different initial times and the corresponding wave fit are provided in Figures S1 to S6 in the104
supplementary materials. From Figure ??, it is found that the initial water elevation is located to the105
west of Mount Anak Krakatau, which is consistent with satellite image interpretations (?). The total106
volume of elevated water of the small, medium and large source sizes is estimated to be 0.03, 0.05 and107
0.17 km3. Nevertheless, it is also found that the results of the optimum static tsunami sources are not108
satisfactory. The large source has the best tsunami predictions, but the initial water elevation extends to109
the surrounding islands, which is unlikely in reality. While the small tsunami source leads to poor tsunami110
waveform fit. Seismic waveforms and infrasound records both indicate that a separated high-frequency111
event ∼115 s prior to the main event, which was interpreted as the seismic precursor or even trigger112
of the main sector collapse by (?). Also, the spectrogram analysis indicates that the collapse process113
includes a 1-2-minute-long low-frequency signal followed by ∼5 minutes of strong emissions (?). Thus,114
it is inferred that the time-dependent evolution of the tsunami source could be non-negligible, which can115
be taken into account to improve the tsunami data fit.116
5
Time-dependent Tsunami Source117
To account for the time-dependent evolution of the tsunami source, the possible duration time of the118
tsunami source between 13:50 to 14:05 is discretized to time segments with an interval of 0.5 min. For119
each time segment, we assume an independent tsunami source. Wave propagation from each tsunami120
source can be linearly added up to obtain the predicted waves at the tidal gauges. The Green’s functions121
for all the time segments are the same except different time delays. Again, we use the non-negative least-122
squares method to constrain all the tsunami sources varying in time. For a given time, the evolution of123
all the tsunami sources before this moment gives the water surface profile at this moment.124
The evolution of the sea surface profile between time 13:50 to 14:05 is plotted in Figure ??. Note that here125
we only show the results of the small source area, which is more likely to represent the actual tsunami126
source. Also note that the time interval is not uniform. A more detailed evolution process is provided in127
Figure S7 with uniform time interval of 0.5 min. Figure ?? shows two clearly separated stages of water128
elevation, which appear approximately at 13:53 and 13:59, respectively. The location and height of the129
water elevation are similar for the two stages. The sea surface to the southwest of Mount Anak Krakatau130
is significantly elevated from about 13:52 to 13:55, with water height of about 3 m. The sea surface131
then rises again approximately from 13:58 to 14:00, with height of about 6 m. It should be noted again132
that the sea surface profile in each plot is the summation of the propagation of all the previous tsunami133
sources, instead of the newly-generated tsunami source at this moment. However, since the two stages134
are clearly separated, it can be inferred that the second stage is not the propagation effect of the first135
stage.136
We also calculate the volume of newly-generated water elevation, shown in Figure ??. From Figure ??, it137
is seen that the volume of the water elevation generated is negligible before 13:52 and after 14:00. Thus,138
the tsunami generation process lasts about 9 min, approximately from 13:52 to 14:00. Additionally,139
it is separated to two stages at about 13:56. The first stage lasts about 4 min from 13:52 to 13:55140
and the second one from 13:58 to 14:00. The volume of water elevation in the two stages is estimated141
to be 0.13 km3. The volume of elevated water in time using medium and large tsunami source areas142
are also provided in Figure ?? for comparison. It shows that, although the spatial distribution of the143
elevated water varies due to different prescribed source areas, the volume of the elevated water and its144
temporal evolution present high consistency. The feature of two-stage water elevation during the tsunami145
generation is observed regardless of the prescribed source area.146
6
In Figure ??, the predicted tsunami waves at the four tidal gauges are compared with the tsunami147
recordings, as well as the predictions from the optimum static tsunami source. At stations panjang and148
kota, the arrival time is better predicted by the time-dependent source. At stations kota, serang and149
ciwandan, the time-dependent source predicts higher first-wave height. Particularly at station serang,150
the time-dependent source also recovers the first trough to some extent; the second crest and some of151
the trailing waves, which are not included in the inversion, are also better predicted. Therefore, the152
time-dependent tsunami source produces improved waveform fitting than the optimum static tsunami153
source.154
Discussion155
Another approach of utilizing tsunami data to reconstruct the tsunami source in absence of source156
mechanism is the time reversal imaging (TRI) technique. For this event, we have also attempted to157
apply the TRI method. The time reversal images from the four stations are first obtained separately158
(Panels (a) to (d) in Figure ??), and then combined linearly to produce the final profile of initial water159
elevation (Panels (e) and (f ) in Figure ??). However, we find that the tsunami waves from the four160
stations do not interfere in the source area, and hence a satisfactory initial profile is not obtained. This161
is mainly because of two reasons. First, the simulation of the tsunami waves is highly affected by the162
local bathymetry. Among all the four stations, only the leading waves from station serang are clear in163
the source area. The waves from the other three stations are contaminated by the coastlines. Second,164
as shown in the previous section, the generation of the tsunami source lasts about 10 min, which is165
comparable to the wave period, so it is not suitable to neglect the time-dependent process of the tsunami166
source and recover a static image of initial water elevation. Thus, the TRI results of the remaining167
stations do not interfere in the source area to construct a focused static tsunami source.168
In this study, we have excluded wave dispersion and solved the linear shallow water wave equations169
in the numerical calculations. Since the size of the tsunami source is relatively small compared to170
tsunamis generated by large subduction earthquakes, it is questionable if the wave dispersion is indeed171
negligible. Here we use the optimum static tsunami source to simulate the tsunami waves, and compare172
the numerical results with and without dispersion. The simulated waves at the four stations are plotted173
in Figure dispersion. The numerical simulation with wave dispersion is carried out using software174
package FUNWAVE (??). Although there are some high-frequency oscillations that only exist in the175
7
non-dispersive simulation, the overall simulated waves are very similar. A possible reason is that the176
water depth in the source area is relatively shallow. The effect of wave dispersion depends on the ratio of177
wavelength and water depth, i.e., kh, where kis wavelength and his water depth. Thus water waves in178
shallow water generally have insignificant dispersion. Another possible reason is that the computational179
domain in this study is small and the four tidal gauges are near the source. The effect wave dispersion180
cumulates with propagation distance, so the cumulated dispersion is negligible in short distance.181
The SAR image acquired ∼8 h after the tsunami reveals two discrete failure planes: one indicates the182
failure of the western flank, and the other indicates a surface break close to the preexisting crater (Failure183
plane A and B, respectively, ?). Seismic and infrasound records show a separated seismic event followed184
by a Mw 5.3 event of longer duration (?). Based on the results of time-dependent tsunami inversion,185
the first seismic event could correspond to the slide on the listric fault plane near the preexisting crater186
(Failure plane B), leading to the first stage of tsunami generation. Although the seismic derived origin187
time (∼13:54 ?) is about 2 min later than the first generation of the tsunami (∼13:52, Figure ??), it stays188
between of the two water elevation stages, and should be within the uncertainties of both inversions. The189
focal mechanism of the second event demonstrates that it occurs on a south-west dipping fault plane with190
opening mechanism, which probably represents the loss of the western flank (Failure plane A), leading191
to the second stage of tsunami generation. The origin time of the second event obtained by ?(around192
13:56) is consistent with the start of the second tsunami generation in Figure ??. Seismic records suggest193
that the duration of the second seismic event is longer than the first event, and it also radiates more194
low-frequency seismic waves (?), while our results indicate similar duration and intensity of two stages.195
Thus, it is still not fully understood how the first seismic event generates similar tsunami waves as the196
second event. Nevertheless, our tsunami wave inversion together with seismic and geomophology analysis197
is in favor of a time-dependent and complex tsunami source rather than a single puls-like source (e.g., ?).198
Conclusion199
In this study we use the tsunami recordings at four tidal gauges to reconstruct the tsunami source200
following the 2018 Anak Krakatau eruption. If assuming a static initial profile of water elevation, the201
optimum tsunami source occurs at ∼13:54 (UTC time), and the volume of water elevation is about202
0.03 km3. However, we find that the static tsunami source does not satisfactorily predict the tsunami203
waves. By accounting for the time-dependent evolution of the tsunami source, results show that the204
8
tsunami generation process lasts about 8 min, from 13:52 to 14:00. It is divided to two stages, the first205
lasting approximately from 13:52 to 13:55 and the second from 13:58 to 14:00. The volume of water206
elevation in the two stages is estimated to be 0.13 km3for each stage. We have also tested different207
potential source areas in the inversion, and the results of the water volume and its temporal evolution208
are similar. We note that these findings are obtained solely based on the tsunami recordings. The two-209
stage source process in general agrees with seismic and geomophological analysis. Our results can be also210
integrated with more geological, seismic and geomophologcal evidences if one is to extend the results to211
interpret the 2018 Anak Krakatau eruption process.212
9
Declarations213
Availability of data and materials214
The tsunami data are acquired from the Badan Informasi Geospasial (BIG), Indonesia. They are available215
from the corresponding author on reasonable request.216
Competing interests217
The authors declare that they have no competing interests.218
Funding219
This work is supported by the National Nature Science Foundation of China grant No. U1901602 (Y. Zhu,220
C. An), No. 11632012 (C. An, H. Liu) and No. 41974017 (T. Wang).221
Authors’ contributions222
CA designed the concept. YZ conducted the tsunami inversion. TW interpreted the inversion results223
compared to seismic recordings and SAR images. All the authors discussed the results and drafted the224
manuscript.225
Acknowledgments226
This work made use of the GMT software. The authors thank Han Yue and Zhiyuan Ren for useful227
discussions.228
10
panjang
kota
serang
ciwandan
104˚
104˚
105˚
105˚
106˚
106˚
−7˚ −7˚
−6˚ −6˚
−5˚ −5˚
−2.5 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 2.0
Bathymetry
km
0 km 50 km
N
Pringsewu Bandor Lampung
Kalianda
Serang
Pandeglang
105.3˚ 105.4˚ 105.5˚
−6.2˚
−6.1˚
−6.0˚
105.3˚ 105.4˚ 105.5˚
−6.2˚
−6.1˚
−6.0˚
105.3˚ 105.4˚ 105.5˚
−6.2˚
−6.1˚
−6.0˚
Figure 1. The bathymetry near the Anak Krakatau volcano and the location of the four tidal gauges.
Black triangles denote the tidal gauges, and red circles mark nearby cities. The prescribed source area
in the inversion is indicated by the black box, shown in the small panel.
11
105.3 105.4 105.5
-6.2
-6.1
small source area
meter
0
1
2
3
4
5
105.3 105.4 105.5
medium source area
105.3 105.4 105.5
large source area
14:20 14:30 14:40 14:50 15:00 15:10 15:20
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
panjang
observation
static source of small size
static source of medium size
static source of large size
14:20 14:30 14:40 14:50 15:00 15:10 15:20
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
kota
14:20 14:30 14:40 14:50 15:00 15:10 15:20
-1.5
-1
-0.5
0
0.5
1
1.5
serang
14:20 14:30 14:40 14:50 15:00 15:10 15:20
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
ciwandan
Wave Height (m)
Time (UTC)
Figure 2. The optimum static tsunami sources of different sizes (top panel) and the corresponding
tsunami predictions at the four tidal gauges (bottom panel).
12
105.2 105.3 105.4 105.5 105.6
-6.3
-6.2
-6.1
-6.0
serang
ciwandan
the optimum static tsunami source at 13:54
meter
0
2
4
105.2 105.4 105.6
the time-dependent tsunami source evolution at 13:50
meter
-2
0
2
4
6
105.2 105.4 105.6
13:52
105.2 105.4 105.6
-6.3
-6.2
-6.1
-6.0
13:53
105.2 105.4 105.6
13:54
105.2 105.4 105.6
13:56
105.2 105.4 105.6
-6.3
-6.2
-6.1
-6.0
13:59
105.2 105.4 105.6
14:00
105.2 105.4 105.6
14:05
Figure 3. The optimum static tsunami source (first panel), and the time-dependent tsunami source
(other eight panels) following the 2018 Anak Krakatau eruption. Gray color indicates land areas.
13
105.3 105.4 105.5
-6.2
-6.1
-6.0
105.3 105.4 105.5
-6.2
-6.1
-6.0
105.3 105.4 105.5
-6.2
-6.1
-6.0
13:50 13:52 13:54 13:56 13:58 14:00 14:02 14:04
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
Volume of water elevation (km3)
13:50 13:52 13:54 13:56 13:58 14:00 14:02 14:04
Time (UTC)
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
13:50 13:52 13:54 13:56 13:58 14:00 14:02 14:04
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
Figure 4. The volume of newly-generated water elevation from 13:50 to 14:05 with time interval of 0.5
min. The upper panels show the three different tsunami source areas used in the inversion. Gray color
indicates the land areas. The corresponding results of tsunami source volume are given in the lower
panels.
14
13:50 14:00 14:10 14:20 14:30 14:40 14:50 15:00 15:10 15:20
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
Wave height (m)
panjang
observation
time-dependent source
optimum static source
13:50 14:00 14:10 14:20 14:30 14:40 14:50 15:00 15:10 15:20
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
kota
13:50 14:00 14:10 14:20 14:30 14:40 14:50 15:00 15:10 15:20
Time (UTC)
-1.5
-1
-0.5
0
0.5
1
1.5
Wave height (m)
serang
13:50 14:00 14:10 14:20 14:30 14:40 14:50 15:00 15:10 15:20
Time (UTC)
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
ciwandan
Figure 5. Comparison of the recorded and predicted tsunami waves at four tidal gauges. The black lines
indicate the waveforms predicted by the optimum static tsunami source, and the red lines show the
waveforms predicted by the time-dependent tsunami source.
15
104.5 105.0 105.5 106.0 106.5
-7.0
-6.5
-6.0
-5.5 panjang
meter
(a)
-0.1
0
0.1
kota
meter
(b)
-0.01
0
0.01
104.5 105.0 105.5 106.0 106.5
-7.0
-6.5
-6.0
-5.5
serang
meter
(c)
-0.05
0
0.05 ciwandan
meter
(d)
-0.01
0
0.01
104.5 105.0 105.5 106.0 106.5
-7.0
-6.5
-6.0
-5.5 panjang
kota
serang
ciwandan
meter
(e)
-0.2
0
0.2
meter
(f)
-0.1
0
0.1
Figure 6. Time reversal imaging of the 2018 Anak Krakatau tsunami source. Panels (a) to (d): the
time reversal image from station panjang, kota, serang and ciwandan, respectively. (e): the linear
superposition of the four time reversal images. (f): the same as (e) but zoomed in near the tsunami
source region.
16
13:50 14:00 14:10 14:20 14:30 14:40 14:50 15:00 15:10 15:20
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
wave height (m)
panjang
Observation
Optimum Static Source, Non-dispersice
Optimum Static Source, Dispersive
13:50 14:00 14:10 14:20 14:30 14:40 14:50 15:00 15:10 15:20
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
kota
13:50 14:00 14:10 14:20 14:30 14:40 14:50 15:00 15:10 15:20
time
-1.5
-1
-0.5
0
0.5
1
1.5
wave height (m)
serang
13:50 14:00 14:10 14:20 14:30 14:40 14:50 15:00 15:10 15:20
time
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
ciwandan
Figure 7. Comparison of non-dispersive and dispersive simulations from the optimum static tsunami
source. Blue: data; black: non-dispersive; red: dispersive.
17