THE Pacific tsunami warnings issued by the Hawaii-based Pacific Tsunami Warning Centre (PTWC) on its network are, in general, based on seismic data and coastal tide observations. Although operational, the system is in a state of evolution and improvement.
A recent newsletter of the Pacific Tsunami Information Centre (PTIC) explains how the system works. Seismic waveform data streams from a network of seismic stations are continually monitored at the PTWC. The `watchstanders' are alerted whenever large and widespread signals are detected from a significant earthquake. The watchstanders quickly locate the earthquake hypocentre and estimate its magnitude. If the earthquake is shallow and is located under or very close to the sea, and if its magnitude is more than a certain threshold, a preliminary warning is issued as the situation has the potential to generate a destructive tsunami.
The only way to determine if a tsunami has been generated is to study the sea level data from nearest gauges. These measurements are combined with historical data and other predictive techniques such as numerical simulation. Based on this evaluation of the level of threat, the warning may be continued, upgraded or cancelled. The procedures are the same for destructive local and distant tsunamis.
According to Charles McCreery of the PTWC, over the past several years, the PTWC's operational capabilities have been considerably enhanced as a result of various improvements in the system. These include vastly improved quality and quantity of seismic data, improved methodologies for rapid earthquake analysis, better spatial coverage of coastal tide gauges and, most important, the locating of newly developed deep ocean tsunami detectors, or tsunameters, at seven strategic points in the Pacific. Operational forecasting now also makes use of pre-computed numerical models into which real-time seismic and sea level data are fed. The utility of this has now been apparently well demonstrated. But it must be pointed out that tsunameter-based forecasting is still in research mode and is yet to be tested in a real tsunami generating seismic event.
Critical to the accurate determination of earthquake parameters is the availability of high quality real-time digital data from the network of seismic stations in the Pacific region and outside. Since 1997, real-time data from the broadband data servers of programmes like the International Deployment of Accelerometers (IDA) are being used towards this end. According to McCreery, the new high-quality seismic data and extensive geographical distribution of data sources, now enable rapid and accurate determination of earthquake parameters. Typically, the final values are based on 30 to 50 independent measurements. As a result of this, the elapsed time between a quake and the issuance of a PTWC warning bulletin has come down from 30 to 90 minutes some years back to 20 to 60 minutes. This improved response time would be critical for areas at risk closer to the source of the event.
As regards sea level data, the system makes use of the network of about 100 coastal gauges and located around the Pacific. The centre receives the coastal sea level data from these via satellite from stations around the Pacific. However, tidal gauge data have a limitation when applied to the problem of tsunami forecasting. As a result, a conservative tsunami warning philosophy has prevailed, leading to an unacceptably high false alarm rate of 75 per cent. Indeed, having experienced false alarms in the past, Thailand, a member of the network, did not respond quickly enough to the warning from PTWC following the Sumatra quake.
The tide gauges are generally fixed to land in the shallow protected waters of harbours and bays to ensure that they last longer and can be easily maintained on a routine basis. But in these environments tsunami waves coming in from the open oceans are highly modified in non-linear ways as they shoal and interact with the coast, thus severely limiting the utility of these signals for prediction. Also the spacing between the gauges is not always such as to sample the wave optimally. The only way to be sure whether a tsunami wave is headed toward a distant coast is to place tsunami detectors in its path and track it across the open ocean, points out Frank Gonzalez, leader of tsunami research programme at the United Sates National Oceanic and Atmospheric Administration (NOAA).
Towards this, the Pacific Marine Environmental Laboratory (PMEL) of NOAA has developed a gauge for measuring tsunamis in deep ocean and sending the data back to the warning centre in real time. These tsunameters, called Deep-ocean Assessment and Reporting of Tsunamis (DART) gauges, can accurately measure the tsunami from the pressure created on the deep ocean floor by the undulating mass of water carried by the tsunami wave as it passes (see diagram). The DART system is said to be sensitive to pressure changes caused by tsunamis with amplitude as small as 1 cm in 6,000 metres of water. This kind of accuracy is essential because tsunamis in mid-ocean have amplitudes of the order of a few centimetres only.
THE DART system consists of an anchored pressure transducer based Bottom Pressure Recorders (BPRs) that communicates with a moored ocean-surface buoy in real time. An acoustic link transmits data, which are 15 seconds averaged measurements of pressure exerted by the overlying water column. From the buoy, the data are sent to the warning centre via a satellite link. At present, seven such systems have been deployed in the Pacific, six of which are operated by the U.S. and one by Chile. Four more are being planned.
Under benign conditions, when the BPR senses regular variations characteristic of normal tides, the buoy transmits 15-minute data every hour. However, when a tsunami passes, the onboard algorithm switches to the tsunami mode and transmits data every few minutes with the wave being sampled at a much higher rate. Since BPRs are sited in deep water, they can record accurately the nature of tsunami waves as they propagate unaltered in the deep ocean. Further, they can be sited strategically so that they are directly between tsunamigenic zones and populated coastlines.
The transmitted data can provide accurate forecasts only if the data can be interpreted using numerical simulations of tsunamis. For this purpose, the NOAA has developed a simulation model system called MOST, which is capable of simulating the three distinct phases of a tsunami event - generation, propagation and run-up over the mainland. But reliable and robust modelling still requires an element of judgment, considerable quality control, iterative and exploratory computations.
Mostly modellers assume that sea-surface displacement is identical to that of the ocean bottom, but direct measurements of sea floor motion have never been made. As a result, points out Gonzalez, even predicting tsunami's initial height requires at least 10 parameters. Seismic data give only the orientation of the fault plane, the quake's location, magnitude and focus.
The rest have to be estimated. According to him, this can result in the underestimation of coastal inundation by a factor of 10. Reliable simulations are achieved only after repetitive runs over months. The first step towards a reliable and robust tsunami forecasting capability is to create a database of pre-computed scenarios that have been carefully analysed and interpreted. From this base, a scenario that corresponds closest to the real-life situation at hand can be picked in quick time.