Given the massive costs associated with evacuation orders (not only in lost productivity but deaths, injuries and other negative outcomes), government must be the sole authority. Given the certainty of blame if a tsunami does hit, over-use of warnings and evacuation orders is likely. It is important that procedures be established not only to make considered but quick decisions about watch/warning/evacuation messages, but also to counter the bias toward excessive warnings and evacuation orders.

Disaster risk-reduction professionals know that false warnings are an artefact of the inexact art of predicting the onset of hazards: but the general public does not. If they are subject to too many false warnings, they will not respond even to true warnings.

Now that we have gotten over the problem of issuing no warnings, we have to address the problem of false warnings.

But it appears that a crackpot university teacher and a local TV channel have combined to sow panic along the Sri Lanka coast, as documented impeccably by Nalaka Gunawardene:

The latest was on 15 April 2011, when confusion and panic were reported from many coastal areas of Sri Lanka following rumours of an oncoming tsunami. It was attributed to a television channel that had broadcast the views of a Lankan geologist who is speculating on predicting earthquakes with a little help from the heavens. Well, at least certain planets in the Solar System.

Scientific speculation is one thing, but causing public alarm and panic – especially at holiday time – is quite another. The Disaster Management Centre (DMC) was quoted as saying its units in the southern coastal areas had to take special measures to assure the people that there was no threat. The media reported how some people in Matara, Galle, Kalutara, Negombo, Trincomalee and Batticaloa fled their homes fearing another tsunami. Many of these areas were battered by the 2004 Boxing Day tsunami.

The panic prompted the Disaster Management Minister to say that ‘legal action will be taken against astrologers, academics or others who make predictions on natural disasters and thereby cause panic among the people’.

I can recall the suspicions our disaster preparedness work evoked, despite repeated assurances that we would never infringe on the government’s authority to issue warnings. I hope the Minister will initiate legal action.

To understand the following story better, it is useful to compare the budget for dam safety in Sri Lanka  which is LKR 15 million a year or slightly over LKR 1.2 million a month with say the purchase price of a unregistered car (cannot buy one with the budget for dam safety for one month) or the cost of a MIG 27 (USD 2 million plus; or LKR 200 million plus).  It will take 13 years of dam safety budgets to buy a reconditioned MIG 27.

Features | Sundayobserver.lk – Sri Lanka

The Irrigation Department is equipped with the expertise, they have 275 Engineers. When the reports come to Head Office the Dam Safety Unit which is chaired by the Deputy Director and a few Engineers study the repot and if the Director’s intervention is necessary for the matter, it is referred to the Director himself. The annual budget for dam safety is around Rs. 15 million.Unfortunately there is no mechanism to give early warning to people even if a dam breaches, what the Department does is to get the assistance of the Police, Army and with loud speakers mounted on vehicles tell the people.

Kantale Tank Dam episode is a case in point.

This kind of situation is highly unsatisfactory. Therefore in case of dam breach situation the Irrigation Department has no mechanism in place to give prior warning to people.

Powered by ScribeFire.

Lessons from the Sri Lanka tsunami warnings and evacuation of September 12-13, 2007

The tragedy of the 2004 Indian Ocean tsunami was the absence of any official warning. The September 12th Bengkulu earthquake shows that this is unlikely to be the case in the future. We have seen that the new institutions created since the 2004 tsunami have the will and the capacity to act. What we must guard against now is indifference to warning; of populations that refuse to evacuate in the face of real danger.

The prediction of a tsunami is an inexact art that must be practiced in conditions of highly imperfect information and time pressure. In the Pacific Basin, which has had the most experience with tsunamis, 75 per cent of all warnings are false. But this does not cause major harm because the false warnings are contained within the official system and do not get through to the general population for the most part.

Tsunamis are rapid-onset hazards. It took 90 minutes for the 2004 India Ocean tsunami to reach the Arugam Bay-Panama coast in Eastern Sri Lanka. In September 2007, the earthquake occurred at 1110 UTC or 4:40 PM Sri Lanka time. Tsunami Bulletin 001 issued by the Pacific Tsunami Warning Center at 1124 UTC (4: 54 PM Sri Lanka time) projected arrival times of over three hours, with the shortest being to Devundara (3:37 hours after the earthquake, at 8:13 PM Sri Lanka time) and the longest being to Jaffna (5:15 hours). Colombo, which was marginally affected even in 2004, was projected to be made contact with at 8:45 PM Sri Lanka time (4:05 hours after the quake). This was because the origin was well to the south of the northern tip of Sumatra where the 2004 Indian Ocean tsunami originated and because the earthquake was 10 times weaker than that in 2004.

As much as possible of the limited time must be allocated to (a) decision making by authorized persons at the national level about the issuance of watch, warning or evacuation messages; and (b) decision making and action at the level of communities, including evacuation if appropriate. This means that the time taken to communicate the watch/warning/evacuation messages must be minimized.

This does not mean, however, that evacuation orders should be given as quickly as possible. Ideally, such strong measures will be targeted, for example to the Matara district which was foreseen to be the first point of contact. Even there, closing roads would be done as close to 8 PM as possible, because people need the roads and public transport to effect an orderly evacuation.

Decision-making at the national level

As stated above, the issuance of tsunami watch/warning/evacuation messages is an inexact art, based on imperfect information. Given the massive costs associated with evacuation orders (not only in terms of lost productivity but also in terms of deaths, injuries and other negative outcomes that can be caused), government must be the sole authority. Given the certainty of blame if a tsunami does hit and an evacuation order had not been issued, there is likely to be a bias toward over-use of warnings and evacuation orders especially in a culture that does not value economic activity highly.

It is important that procedures be established not only to make considered but quick decisions about watch/warning/evacuation messages, but also to counter balance the bias toward excessive use of warnings and evacuation orders. This includes clear decision rules and assignment of responsibilities. A multi-stage and multi-party decision process (for example having one center for hazard detection and monitoring (Sri Lanka Meteorological Department) and another for public warning (Disaster Management Center) may not be the best way to improve decision making and minimize time.

When time is short and ambiguity can be deadly, it is best to stick to prepared texts and formats. Standard message templates in all three official languages preferably generated using the Common Alerting Protocol (CAP) software have been recognized as a high priority by the Ministry of Disaster Management and Human Rights and the Disaster Management Center.

Communicating messages to first responders and media

It appears, from anecdotal evidence, that the tsunami hazard information center (Met Department) received large numbers of phone calls from the media when word got around that a potentially tsunamigenic earthquake had occurred on the Indian Ocean side of Indonesia. In many cases, senior officers who should have been communicating the scientific evidence to key decision makers at the DMC and the Ministry were being called directly.

This practice is problematic for two reasons:

1. It eats up the time that should be used for considered decision making on whether or not to issue national-level watch/warning/evacuation messages. Time spent on the phone is time not spent on analyzing or communicating the evidence to the relevant authorities;
2. The unstructured format of a journalist-initiated phone call can lead to misunderstanding. For example, some journalists may not know the difference between an alert and a warning. This format also does not leave a record in case there is a need to review it at a later time.

Expressing concern about talking to journalists in the aftermath of a tsunamigenic earthquake does not equate to a recommendation that no one should talk to the media. What is required are:

1. A reliable and fast method of communication (e-mail, fax, telex, or even a taped telephone voice message) with journalists in all three official languages. Messages should be sent to designated numbers and e-mail addresses, preferably using automated procedures.
2. Journalists who call the designated number (preferably in a hunting configuration that can handle multiple simultaneous calls) should be able to hear a taped message.
3. If journalists require additional information they should be able to call a designated spokesperson, whose sole responsibility is to talk to the media. All conversations should be taped.

The point is to give more information faster and in better formats. Shifting away from the current unstructured modes to a structured mode will allow this to be done. Involving journalists in the process of restructuring the communication system of the Met Department will help improve it and will also serve to educate journalists about it.

Multiple media such as fixed and mobile phones, SMS, fax and the Internet should be used. Redundancy improves reliability. Using CAP enabled media will help streamline the process and expedite the messages. Where character limitations exist, as in SMS, the short message can be used to direct the recipient to the place where the complete message can be obtained.

The LIRNEasia-Sarvodaya Last Mile HazInfo pilot project showed that the most efficacious method of communicating to first responders would be through addressable satellite radio sets plus GSM/CDMA phones. If the government wishes, these instruments can be temporarily loaned to government for trials as decided at a recent meeting convened by the Ministry and the DMC.

Decision making at the community level

The Sarvodaya-LIRNEasia community centered disaster risk reduction approach places greatest weight on decision marking and prior preparation and training at the community level. In this model, each Sarvodaya village will have an ICT guardian who can communicate bidirectionally with the Hazard Information Hub (HIH) of Sarvodaya. The objective of the next stage of work within Sarvodaya is to equip 1000 advanced villages belonging to Sarvodaya with the knowledge and training to devise and implement emergency response plans at the village level and to have the ability to communicate information to and from the HIH.

In this model, the community will be alert and ready to respond in the optimal way to any watch, warning or evacuation message issued by the government. The HIH will amplify the government’s message and prior training and preparation will enable to community to respond in the most productive manner to a government message. The objective of those working at the national levels and designing and operating the communication systems should be to give the community unambiguous information and the longest possible time to take appropriate action.

Community preparedness also helps the national-level decision makers. If they know that the community is prepared and capable of responding quickly and in an orderly manner to an evacuation order, they can delay the “strong measures” until absolutely necessary. It is when that confidence is lacking that there is a tendency to issue warnings and evacuation orders too early. It is perhaps because of this that Thailand did not join Sri Lanka and Bangladesh in issuing evacuation orders after the Bengkulu earthquake.

Minimizing the false

In the disaster risk-reduction field, false warnings are not seen as being a result of malice or incompetence. They are simply artefacts of the inexact art of predicting the onset of hazards. Professionals know this, but the general public does not. If they are subject to too many false warnings, the general public will not respond even to true warnings.

Now that we have gotten over the problem of issuing no warnings, we have to address the problem of reducing false warnings. Effective design of protocols at each stage of the warning chain and committed implementation combined with a passion of continuous improvement will help us reduce the false and realize the promise of early warning to save lives.

Rohan Samarajiva, Ph.D.
Project Director, Last Mile HazInfo Project &
Executive Director, LIRNEasia
13 September 2007

Mr. P.D. Amarasinghe, Secretary of the MDMHR, opened the session by discussing the particular disaster role of each of the government agencies asked to attend. He acknowledged the government’s important role in disaster warning. The role of the Disaster Management Center (DMC) will be to disseminate warnings to first responders through various means – RANet, SMS, fax. First responders will be media, police and armed forces, district coordinators of DMC, and other organizations (NGOs, community). Currently, the MDMHR is in the process of developing a proposal for a network of dissemination towers at Hikkaduwa, Kalmunai and Point Pedro. There will be a total of 50 by the end of this year (latest, Jan/Feb 2008). With further funding, the MDMHR will establish 10 EOCs, 4 EMC, in addition to 4 managed by police & armed forces. Thus, he declared that the basic warning system is in place and asked how might the telecom operators contribute to a public warning notification system.

Major General Gamini Hettiarachchi, Director General of the DMC gave a presentation on the Disaster Early Warning and Dissemination Strategy of Sri Lanka. He reviewed the progress of disaster management since the tsunami, through the Disaster Management Act No. 13 of 2005, and the establishment of the National Council on Disaster Management in the same year. He stated that there are currently 3 early warning towers, but by the end of next year there should be a total of 150. Emergency response committees have been formed including SLT, municipalities, 25 district secretaries and 9 province secretaries.

Mr. Anjula Godakumbara from Dialog Telekom made a presentation on Dialog’s involvement in early warning dissemination. Dialog and the Ministry of Disaster Management and Human Rights have signed an MoU. Dialog along with MicroImage and the University of Moratuwa Research Lab looked into using GSM communication in disaster early warning dissemination by setting up the Disaster and Early Warning Network (DEWN). This has SMS and cell broadcast capabilities. The benefits of the DEWN system is that it is low-cost and uses existing communication infrastructure. Dialog reiterated the fact that only the DMC has the authority to send disaster warnings. For more information on this initiative see www.dialog.lk/dewn.

Prof. Rohan Samarajiva, Executive Director – LIRNEasia, gave a presentation on “Effective use of telecom & electronic media in disaster risk reduction”. He asked how will the DMC communicate with the media and the phone companies. The government must have internal protocols which must be double-checked with government officials. There must a technically sound system in place to get to all media and must be capable of showing if people got the message and in the right language. LIRNEasia recommends the Common Alerting Protocol (CAP) system with additional software which is single input, multi-output, multi-language. This can be developed by LIRNEasia within a few months. Essentially this means that once a button is pressed, the software will translate the message accordingly and send the message by SMS, cell broadcast, fax, etc. The government is clearly interested in standard procedures and templates. Therefore, it would be in its best interest to approve a standard template in the language and simply insert the proper word as necessary. Prof. Samarajiva posed yet another question: How might the government ensure that the person issuing the warning is reliable and not someone who plans to sabotage the process? There must be some sort of encryption or verification method so that receivers know that the warning is official.

In discussing the use of ICTs in public warning (an area not covered by the Last-Mile HazInfo Project), Samarajiva suggested that two of the most difficult cases be addressed: a passenger in a moving train and a tourist at the Yala National Park. If public warning can reach these two individuals, all the other less difficult problems can be solved. In both cases, he suggested that cell broadcasting would be the most important technology, though there may be areas within Yala that would be out of signal coverage, in which case no public warning was possible.

The discussion that followed the presentations. Mr. Hettiarachchi called for a link with service providers to develop automated procedures. Secretary Amarasinghe suggested that there be a technical committee set up for hearing recommendations on early warning from telecom providers and broadcasters. It was decided that there would be two separate technical committees – one for telecom providers and the other for broadcasters. Dates and agendas for the first technical committee meetings of telecom providers and broadcasters were not discussed.

Mr. U.W.L. Chandradasa, Director – DMC, wrapped up the session by reiterating that the DMC is responsible for disseminating warning messages. Regarding first responders, he agreed that other agencies could be involved. He acknowledged that government alone cannot take charge of “Last Mile” warning without the contributions of telecom providers, broadcasters and the private sector.

The presentation is at ministry-of-dmhr-7sep07final.ppt.

What I learned during my visits to the Civil Defense Center and the Tsunami Museum in Hilo and the Pacific Tsunami Warning Center in Ewa Beach in Hawai’i last January greatly contributed to the disaster communication research program undertaken by LIRNEasia in the aftermath of the 2004 Indian Ocean tsunami. Therefore, I welcomed the opportunity to step back and reflect on the research program a year later, also in Hawai’i.

The occasion was a workshop funded by the National Science Foundation of the US. It was organized by Louise Comfort, Daniel Mosse and Taieb Znati, all at the U of Pittsburgh. Louise is from Public Policy and has been working on disasters for a long time. Daniel and Taieb are in computer science and new to the field. I really liked Daniel’s fixation on time lines. That is critical to the whole enterprise of warning.

I had some reservations when I first received the invitation, but was persuaded to attend. All I knew about sensor-based networks was what I learned from Tilak Illangasekera and Anura Jayasumana in the course of our work on early warning systems for dam-based hazards (http://www.lirneasia.net/2005/05/dam-safety-expert-consultation-may-20/). I had been thinking about using advanced sensors and telemetry for hazard detection for a long time. The disasters and ICTs workshop I organized while at the Arthur C. Clarke Centre back in 1986 had this subject on the agenda; and Tilak and Anura responded to a specific request to talk about “electronic dust” in relation to dam safety. But still, that did not qualify me as an expert on the subject.

The agenda explained the rationale: the organizers began the workshop with presentations on the current state of tsunami warning in Indonesia, Thailand, Sri Lanka and India followed by presentations on the design of new systems, if any, in each of the four countries. They had a deputy minister from Indonesia and the head of the Meteorology Department in Thailand. From India and Sri Lanka, they had invited researchers. The idea was to get a good sense of the “demand” side and then develop creative ideas to “supply” the solutions.

In order to do justice to the first presentation I analyzed the Sri Lanka Disaster Management Act (No. 13 of 2005) and the Disaster Management Roadmap published in December 2005 by the Disaster Management Center, created under that Act. Avoiding slipping into the conventional government mode of talking about the future, I analyzed the actual experiences of with tsunami warning in Sri Lanka (December 26^th, 2004; the fiasco of the non-functioning e-mail addresses and fax numbers of February 27^th, 2005; and the Nias Great Earthquake and the false warning of March 28^th, 2005).

The LK presentation was different from the others, because it said pretty much nothing about hazard detection and monitoring, other than to identify it as a necessary condition that LK hoped would be satisfied by international and regional action. This was not simply because of the focus on public warning in our own research; it was because LK is the only country among the four to not have an active program in developing a tsunami hazard detection capability. Shifting the emphasis from tsunami hazard detection to public warning was part of the thrust of LIRNEasia’s and Vanguard Foundation’s work in the aftermath of the tsunami (e.g., http://www.lirneasia.net/2005/02/learning-from-the-tsunami/).

Listening to the grandiose plans for placing large numbers of very expensive deep-sea sensor devices in Indonesia and the unrealistic deadlines of the peculiar India-only tsunami detection system, I was happy that my government had got it right at least once. In none of the other presentations did I hear what I wanted to hear: systematic analyses of how the systems worked (especially when specific actions were taken) in the “natural experiments” of December 26^th, 2004 and March 28^th, 2005.

I also presented the elements of the HazInfo Project we’re doing with Sarvodaya.

These sessions were followed by some excellent presentations on the state of tsunami detection: Walter Mooney of the US Geological Survey and Rhett Butler from Global Seismic Network talked about the worldwide network of seismometers, using pictures of the Pallekale site in their presentation. The importance of the Pallekale equipment for the Indian Ocean detection system was clearly evident. The Indian bureaucracy paranoidly continues to resist sharing seismic information, making the entire region a black hole, if not for PALK, as the Pallekale unit is fondly known. The Chinese are only slightly less paranoid, imposing a 100-minute delay on their data (thereby ruling that area out as a serious source of seismic information). Isn’t it ironic that the two most populous nations in the world take positions inimical to the preservation of human life in the name of national security?

Two thirds of the earth’s surface is covered by water, but except for one cable-based ocean-floor system near Hawai’i, all the other seismic sensors are land-based. Professor Kanamori from CalTech made a fascinating presentation on the potential of an alternative method of identifying tsunamis using GPS data to do a quick and dirty calculation of lateral displacements of land masses caused by really big earthquakes. Chip McCreery, the Director of the PTWC, talked about how they were planning to use a new tool based on the sophisticated model that gave us that powerful picture of the 2004 tsunami a few days later. Beautiful results from complicated models weeks later are nice, but do not save lives. What Chip described was a simple interface where the PTWC scientists would make some simple entries and the model would spit out information that was practically usable in identifying tsunami risk.

So it was clear that there is plenty of work to be done in improving tsunami detection; in reducing the false warning rate which stands at 75% in the Pacific Basin. Advanced sensor networks, including the motes that are “sown,” can play a very useful role. The NSF was not throwing its money down a drain. We, the people dealing with the gritty reality of public warning and evacuation, could easily tell them how important it was to get the false warning rate down. The hard scientists could get to work on improving the predictability of tsunamis, and of speeding up the communication of the hazard information to the national warning centers.

The tsunami is a peculiar beast. A cyclone/typhoon/hurricane can be predicted days in advance; at the other extreme, there will be a few minutes of unreliable warning on an earthquake, at most. Tsunamis sit in the middle of this continuum. Advance warning of around one hour is feasible for a teletsunami of the type that hit Sri Lanka in 2004. For Kenya and Tanzania, the period can be as high as 8-10 hours. But the same tsunami hit Banda Aceh in twenty minutes, making advance warning difficult to impossible. What was a teletsunami for us was a local tsunami for Banda Aceh.

Sometime early in the discussion this distinction between short-fuse and long-fuse disasters came up. Unfortunately, it did not get picked up as the central organizing theme for the work of the group, despite Daniel’s fixation on time.

In the case of long-fuse disasters, the hazard detection and monitoring system is (and should be) separable from the warning system. They have different qualities. In the case of short-fuse disasters, the two systems have to be unified. This was the distinction we made in the case of NEWS:SL (teletsunami; long-fuse) and the dam hazards project (short-fuse).

In the case of local tsunamis, the prevailing wisdom is “your feet are your signal; run for high ground.” Is it possible for advanced sensor networks to reduce the uncertainty in this situation? I think yes. This is a good case for long-term fundamental research funded by the NSF.

On the public warning side, there is no need for fundamental research at present. Just hard-nosed implementation will do. As we get this piece done, it’s possible that opportunities for fundamental research will appear. But they were not manifest at the meeting in Maui. All we had was a restatement of what everyone in disaster management had known for years. The best practices section of our NEWS:SL report sets out the principles.

But this is not to say that the public warning side will not benefit from fundamental research on the hazard detection and monitoring side. Improvements there will reduce the false warning rate and will also give the public-warning people more time to act.

But, in my view, there is no gain from simply mashing together the hazard detection and public warning systems for teletsunamis. These systems have different qualities and we will save more lives (this is the end objective, we must keep in mind) by improving their performance as separate systems. Unfortunately, the design of the second day’s deliberations did not allow this kind of discussion. The mashing together was a fait accompli.

So, not optimal in terms of intellectual output, but very good in terms of networking and finding out the state of the art on the hazard detection and monitoring side. I hope something good will come of all this.

The conference website is here