INDIAN TSUNAMI WARNING SYSTEM

Shailesh Nayak and T. Srinivasa Kumar

Indian National Centre for Ocean Information Services (INCOIS), Hyderabad 500055, India –
[email protected]

Commission IV

KEY WORDS: Hazards, Ocean, Database, Modelling, Decision Support

ABSTRACT:

Tsunami is a system of ocean gravity waves formed as a result of large-scale disturbance of the sea floor that occurs in a relatively
short duration of time. The Indian Ocean is likely to be affected by tsunamis generated mainly by earthquakes from the two potential
source regions, the Andaman-Nicobar-Sumatra Island Arc and the Makran Subduction Zone. A state-of-the-art warning centre has
been established at INCOIS with all the necessary computational and communication infrastructure that enables reception of real-
time data from the network of national and international seismic stations, tide gauges and bottom pressure recorders (BPRs).
Earthquake parameters are computed in the less than 15 minutes of occurrence. A database of pre-run scenarios for travel times and
run-up height has been created using Tunami N2 model. At the time of event, the closest scenario is picked from the database for
generating advisories. Water level data enables confirmation or cancellation of a tsunami. Tsunami bulletins are then generated based
on decision support rules and disseminated to the concerned authorities for action, following a standard operating procedure. The
criteria for generation of advisories (warning/alert/watch) are based on the tsunamigenic potential of an earthquake, travel time (i.e.
time taken by the tsunami wave to reach the particular coast) and likely inundation. The performance of the system was tested on
September 12, 2007 earthquake of magnitude 8.4 off Java coast. The system performed as designed. It was possible to generate
advisories in time for the administration and possible evacuation was avoided.

1. INTRODUCTION

“Tsunami” is a system of ocean gravity waves formed as a

result of large scale disturbance of the sea bed in a short
duration of time, mostly due to earth quake (or volcanic
eruption or submarine landslides). As displaced sea water return
by the force of gravity to an equilibrium position, a series of
oscillations both above and below sea level take place, and
waves are generated which propagate outwards from the source
region. Tsunamigenic zones, the Andaman-Nicobar-Sumatra
Island Arc in the Bay of Bengal and the Makran Subduction
Zone in the North Arabian Sea, that threaten the Indian Ocean
have been identified by considering the historical tsunamis,
earthquakes, their magnitudes, location of the area relative to a
fault (ANSS, 2007; HS, 2007; NGDC, 2007, ISG, 2007; USGS,
2007; Rastogi and Jaiswal, 2006), and also by tsunami
modelling (Figure 1).

The Indian coast has recorded few tsunamis in the past.

However, in spite of infrequent occurrence of tsunamis (about 6
events reported in the 20th century) in the Indian Ocean, they
could occur at any time and could be very devastating. The
latest Indian Ocean Tsunami (December 26, 2004) has been the
one of the strongest in the world and the deadliest of all time by Figure 1. Tsunamigenic zones in the Indian Ocean
an order of magnitude.

The east and west coasts of India and the island regions are 2. TSUNAMI WARNING SYSTEM
likely to be affected by tsunamis generated mainly due to
earthquakes in these subduction zones. Hence there was a need The following are major components of the tsunami warning
for developing Tsunami Warning System. system.

In this paper, details about various components of tsunami 2.1 Estimation of Earthquake Parameters
warning centre, development of the application software and its
performance during the Java Earthquake of M 8.4 on September A Network of land-based seismic stations for earthquake
12, 2007 have been discussed. detection and estimation of focal parameters in the two known
tsunamigenic zones is a prime requirement of the warning
1501
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008

centre. INCOIS is receiving real-time seismic data from surge to prepare and update storm surge/tsunami hazard maps.
international seismic networks as well as from India The accuracy of model predictions is directly related to the
Meterological Department (IMD) and has been detecting all quality of the data used to create the bathymetry and topography
earthquake events occurring in the Indian Ocean in the less than of the model area.
15 minutes of occurrence. Necessary software have been
installed for real-time data reception, archiving, processing and Coastal Bathymetry is the prime determinant of the height of the
auto-location of earthquakes as well as for alert generation and tsunami wave or storm surge as it approaches the coast. High
automatic notification. resolution coastal bathymetry is thus the key input for various
tsunami and storm surge prediction models. Preliminary
2.2 Monitoring of Sea Level Surveys have already been conducted to acquire high-resolution
bathymetry for a few vulnerable areas of the coastline. Naval
In order to confirm whether the earthquake has actually Hydrographic Office (NHO) has been providing detailed
triggered a tsunami, it is essential to measure the change in bathymetry data.
water level as near to the fault zone with high accuracy. Bottom
pressure recorders (BPR) are used to detect the propagation of Topography of the entire coastline of the country is required at
tsunami waves in open-ocean and consequent sea level changes. 1:25000-scale with contours at intervals of 0.5 m to 2 m up to
A network of Bottom Pressure Recorders (BPRs) has been 20 m contour interval. Preliminary maps have been prepared of
installed close to the tsunamigenic source regions to detect coastal topography using CARTOSAT-1 stereo data for the
tsunami, by the National Institute of Ocean Technology (NIOT). Indian coast. The National Remote Sensing Agency (NRSA)
These BPRs can detect changes of 1 cm at water depths up to 6 has been mapping the topography of about 15, 000 Sq km area
km. with airborne LIDAR & Digital Camera data in conjunction
with GPS control survey using photogrammetric techniques.
A network of tidal gauges along the coast helps to monitor 3000 sq km area has already been mapped. These products have
progress of tsunami as well as validation of the model scenarios. been used to prepare coastal vulnerability maps.
Near-real time data is being received from national and
international centres has been received. Necessary software for 2.5 Warning Centre
real-time reception, display and archiving of tide gauge data has
been developed. A dedicated 24 x 7 operating Tsunami Warning Centre
comprising necessary computational, communication and
2.3 Tsunami Modelling technical support infrastructure as well as a robust application
software that facilitates data reception, display, analysis,
The TUNAMI N2 model basically takes the seismic modelling, and decision support system for generation of
deformation as input to predict the run-up heights and tsunami advisories following a standard operating procedure has
inundation levels at coastal regions for a given tsunamigenic been established. The warning centre continuously monitors
earthquake (Imamura, 2006). The seismic deformation for an seismic activity in the two tsunamigenic source regions and sea
earthquake has been computed using Smylie and Mansinha, level through the network of national and international seismic
(1971) formulation using the earthquake parameters like stations as well as tide gauges and bottom pressure recorders
location, focal depth, strike, dip and rake angles, length, width (BPR’s). The monitoring of water level enables confirmation or
and slip of the fault plane. At the time of earthquake, only cancellation of a tsunami. A custom-built software application
location, magnitude and focal depth are available immediately. generates alarms/alerts whenever a pre-set threshold is crossed.
For operational quantitative tsunami forecast, there needs to be Tsunami bulletins are then generated based on pre-set decision
a method to quickly estimate the travel times and run up based support rules and disseminated to the concerned authorities for
on the above available parameters. For this purpose, all the action, following a Standard Operating Procedure (SOP).
other input parameters such as length, width and slip are
calculated from the magnitude using global relations
(Papazachos et al, 2004). strike angle, dip angle and slip angle 3. APPLICATION SOFTWARE USING GEOSPATIAL
are considered for worst case scenario (strike angle parallel to TECHNOLOGY
trench or coast, dip angle 45 deg, slip angle 90 deg).
One of the key components of the early warning centre is the
A database of pre-run scenarios has been created for 1000 unit development of application software around GIS technology
sources covering all the tsunamigenic sources in the Indian that performs the following operations:
Ocean region (Kuwayama, 2006). In the current database each
unit source has a length of 100 km and width of 50 km that (i) Acquisition, display and analysis of real time data of
represents a rupture caused by a Mw 7.5 magnitude earthquake seismic sensors, tide gauges and BPRs.
with a slip of 1m. At the time of earthquake occurrence, based
on the location and magnitude of the earthquake, the basic unit (ii) Generation of model scenario database for assumed
source open ocean propagation scenarios in scenario database earthquake parameters as well as Retrieval, Display and
are selected to merge and scaled up/down using scaling Analysis at the time of an event.
relations.
(iii) Generation, Display and Analysis of Bathymetric Data,
Coastal Topographic Data and Vulnerability Maps.
2.4 High Resolution Data Base on Bathymetry and Coastal (iv) Decision support system for generation of tsunami
Topography advisories following a standard operating procedure.

Generating and updating a high resolution database on (v) Data warehousing, Data Mining and Data Dissemination
bathymetry, coastal topography, coastal land use, coastal Details of the individual functions are given below:
vulnerability as well as historic data base on tsunami and storm

1502
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008

3.1 Data Acquisition and Display (ii) generate automatic alarms based on preset decision rules
for one or many of the input parameters and
The warning centre currently receives seismic data, tidal data
and BPR data through VSAT links, Internet, and INSAT from (iii) carry out criteria-based analysis for one or many of the
multiple sources. All such data are centrally collected by high above mentioned input parameters to generate online
end servers where, it is processed by ETL jobs for loading into advisories (Figure 2).
staging database followed by archiving staging database into the
central database periodically. The criteria for generation of different types of advisories
(warning/alert/watch) for a particular region of the coast are to
The requirement is to display observational data (seismic, sea be based on the available warning time (i.e. time taken by the
level) from selected platforms and modelled data (travel time, tsunami wave to reach the particular coast). The warning criteria
run up heights, inundation areas, DTMs, etc.). Other need is to are based on the premise that coastal areas falling within 60
plot real-time sea-level data overlaid on predicted tidal values at minutes travel time from a tsunamigenic earthquake source need
each specified location. The application software is capable of to be warned based solely on earthquake information, since
performing all the above functions. enough time will not be available for confirmation of water
levels from BPRs and tide gauges. Those coastal areas falling
3.2 Model Scenario Database outside the 60 minutes travel time from a tsunamigenic
earthquake source could be put under a watch status and
Tunami N2 model has been used to estimate travel time and upgraded to a warning only upon confirmation of water-level
run-up height for a particular earthquake. Since the model data.
cannot be run at the time of an event, due to large computing
time as well as due to non-availability of required fault The application software is capable of doing all the desired
parameters in real-time, a database of pre-run scenarios is functions in addition to generating alerts (as and when each
essential. At the time of event, the closest scenario is picked individual parameter exceeds a critical value) in the form of an
from the database for generating the warning. The output from alarm at the warning centre as well as SMS alerts to specified
the modelling exercise is a huge database (approximate size is 6 persons on a mobile phone. Provision in the software is made to
Tera bytes) consisting of spatial maps depicting the water level modify threshold value, as and when need arises.
in the Indian Ocean region at each time-step for about 5000
simulations. The application software has an interface to store, 3.5 Data Warehousing, Data Mining and Dissemination
retrieve, analyze and display the spatial maps from the database.
The spatial layers currently being handled by this application This module is capable of organizing the data and storing it in a
include fault lines, fault segments for different earthquake centralized storage server with appropriate load balancing. The
magnitudes, travel time maps, directivity maps, simulation entire data is organized using state-of-the-art user friendly
results for about 1800 coastal forecast points, graphs of model database for storing, analyzing and quick retrieval at the time of
and observed tsunami wave profiles at each coastal forecast the event. This database is linked to a dedicated tsunami website
point, etc. Application Software has a user friendly GUI/control through which data/information/advisories are made available to
panel depicted on a spatial canvas of the Indian Ocean Region the users. The website support information in the form of text,
through which user can perform GIS operations like navigating maps and multi-media.
to a desired location, zoom, pan, query, analysis, etc.
The application software enables dissemination of bulletins to
3.3 Vulnerability Maps designated contact points using SMS, e-mails, fax, telephone,
VPNDMS, etc. The entire warning centre infrastructure is
Tsunami run-up causes flooding of seawater into the land up to hosted on highly reliable hardware designed for mission critical
few km resulting in loss of human life and damage to property. applications with necessary redundancies.
To minimise such losses, it is imperative to prepare coastal
vulnerability maps indicating the areas likely to be affected due
to flooding and rending damage. Tunami N2 Model has been 4 PERFORMANCE OF THE SYSTEM
customised for the Indian Ocean Region and has been validated
using the December 2004 Tsunami observations. This model The system was fully tested on September 12, 2007, when
has been run for five historical earthquakes and the predicted earthquake of magnitude 8.4 struck near Java.
inundation areas are being overlaid on cadastral level maps of
1:5000 scale. The maps are to be used by the central and state The performance of the end-to-end system is assessed based on
administration responsible for disaster management. These the
community-level inundation maps are extremely useful for
assessing the population and infrastructure at risk. These maps (i) individual performance of the observation sub-systems
will be provided using the web-GIS interface of the Application (seismic, sea-level) and the
Software.
(ii) performance of the application software sub-systems viz.
display, analysis, modelling, DSS, SOP and dissemination.

3.4 Decision Support System and Standard Operating

Procedure

The decision support software is intended to

(i) monitor the online input data from individual sensors,

1503
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008

Figure 2. Standard Operating procedure for generating advisories

1504
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008

The ultimate measure of the performance of the system is the The earthquake parameters for the September 12 event were
accuracy and timeliness of the advisories generated. This can be computed within 13 minutes having initial magnitude 7.9
assessed using performance indicators such as accuracy of (USGS initial estimate being 8.0 in 12 minutes). Based on this
magnitude a scenario was picked up from the database and
(i) earthquake determination; travel time and run up heights at various coastal forecast points
(ii) model scenario results viz. travel time and run-up estimates were estimated. The modelled and observed values are given in
and Table 1.
(iii) real-time sea-level observations.

One of the most critical aspects of tsunami warning system is to Location Arrival Time (h) Water Level (cm)
estimate earthquake parameters with reasonable accuracy in Estimat Observe Estimat Observe
shortest possible time. The elapsed time is generally less than ed d ed d
15 minutes in most cases. The earthquake location and its Padang 1751 1754 80 60
magnitude are two critical parameters to be estimated so that Coco’s Island 1748 1748 40 30
right scenario can be chosen. While it is important to be able to Sabang 1903 1903 20 30
estimate earthquake parameters within acceptable errors (+/- 0.2 TB3 1903 1913 02 01
in magnitude and 30 km in location during initial estimates is TB10A 1931 1941 01 02
considered good for tsunami warning applications), for an TB10 1930 1945 02 02
operational system, it is also extremely important to be able to Port Blair 2010 2013 10 08
do this consistently. To quantify the performance of our system,
Chennai 2105 2110 20 18
a comparison has been made with the earthquake parameters
estimates by the United States Geological Survey (USGS) for
Table 1: Performance of Tsunami Model
abour 20 consecutive earthquakes. Results show that the desired
performance has been achieved (Figure 3 and 4).
It can be seen from the table that system has performed well
within the limits of acceptable error.

The end-to-end system performed extremely well enabling

reception, display and analysis of the real-time and model data
sets as well as generation and dissemination of timely and
accurate advisories following the standard operating procedure.
This information was used to provide necessary advisories to
the concerned authorities, thus avoiding unnecessary public
evacuation for this event.

5 CONCLUDING REMARKS

Geospatial technology has immensely helped in the design of

early warning system for tsunami. The system is capable of
providing tsunami advisories (earthquake information,
estimated travel times, run up heights, threat zones, etc) for the
entire the Indian Ocean.

Figure 3. Comparison of estimation of location errors of REFERENCES

earthquake
ANSS, 2007. Global earthquake composite catalog search (1898
to present), Indian Ocean, http://www.ncedc.org/ anss/catalog-
search.html (Accessed 26th June 2007).

HS, 2007. Harvard Seismology, Global Centroid Moment

Tensor Catalog (1976 to present), Indian Ocean,
http://www.globalcmt.org/CMTsearch.html, (Accessed 26th
June 2007).

Imamura, F. 2006, TUNAMI-N2 (Tohoku University’s

Numerical Analysis Model for Investigation of Near Field
Tsunamis version 2) Manual draft,………………………..
http://www.tsunami.civil.tohoku.ac.jp/hokusai3/J/projects/manu
al-ver-3.1.pdf, (Accessed 16th Dec 2006).

ISC, 2007. International Seismological Centre, Online catalog

rectangular area search, Indian Ocean,
Figure 4. Comparison of estimation of earthquake magnitude http://www.isc.ac.uk/search/bulletin/rectang.html, (Accessed
greater than 6. 26th June 2007).

1505
The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B4. Beijing 2008

Kuwayama, T. 2006, “Quantitative Tsunami Forecast System”, USGS 2007. Earthquake Hazards Program, U.S. Geological
JICA Tsunami Seminar – Unpublished report Survey, Global Earthquake search, Significant Worldwide
Earthquakes (2150 B.C. - 1994 A.D.) & India (1063 – 1984),
Mansinha, L. and Smylie, D. E. 1971. Displacement fields of Catalog search http://neic.usgs.gov/neis/epic/ epic_global.html,
inclined faults. Bull. Seismological Soc. Am. 61, 1433-1440. (Accessed 26th June 2007).

NGDC, 2007. Natioanl Geophysical Data Centre (NGDC), the USGS 2007. Earthquake Hazards Program, U.S. Geological
Significant Earthquake Database (-2150 to 2007) rectangular Survey, Rectangular area search, Indian Ocean,
area search, Indian Ocean, http://neic.usgs.gov/neis/epic/epic_rect.html, (Accessed 26th
http://www.ngdc.noaa.gov/nndc/struts/form?t=101650&s=1&d June 2007).
=1, (Accessed 26th June 2007).

Papazachos, B.C., Scordilis, E. M. Panagiotopoulos, D. G., ACKNOWLEDGEMENTS

Papazachos C. B., and G.F. Karakaisis, G. F. 2004, Global
relations between seismic fault parameters and moment The authors are grateful to Dr. P. S. Goel, Secretary, Ministry
magnitude of earthquakes, 10 th Congr. Hellenic Geol. Society, of Earth Sciences for his encouragement and valuable guidance
14–17 April 2004, Thessaloniki, pp. 539–540. during the course of development of this system. The authors
express their sincere thanks to colleagues of MoES for their
Rastogi, B. K. and. Jaiswal, R. K. 2006, ‘A Catalogue of support in developing this system.
Historical tsunamis in Indian Ocaen ’- Science of Tsunami
Hazards, Vol. 25, No. 3, page 128-143.

1506