The literature on forecasting and early warning systems is extensive. This section sets out a few general principles of good practice and discusses some of the most important issues in making warnings effective. The aim of early warning systems (EWS) is to enable individuals and communities threatened by hazards to act effectively and in sufficient time to reduce the likelihood of death, injury and damage to property and the environment. EWS vary greatly in size, structure, management and technological sophistication, according to the extent of their coverage, the nature of the hazard(s) and the human and material resources available. But they have many features and issues in common.
Early warning systems must be understood as working systems with inter-connected components (see Figure 16.1: Flood forecasting/warning system and Figure 16.2: Key components of an early warning system). A weakness or failure in any one component of the system (technical or human/organisational) can potentially undermine the whole.
(Olaf Neussner, GIZ)
|Risk knowledge||Monitoring and warning service|
Systematically collect data and undertake risk assessments.
Develop hazard monitoring and early warning services.
|Dissemination and communication||Response capability|
Communicate risk information and early warnings.
Build national and community response capacities.
UNISDR, Developing Early Warning Systems: A Key Checklist (Geneva: UN Office for Disaster Reduction, 2006), http://www.unisdr.org.
16.5.1 Management and resources
Large-scale early warning systems require considerable resources: people, infrastructure, technology, data and funding. They have to operate continuously. They are complex to manage, needing to integrate multiple actors (scientists, civil authorities, the media and the public) and different levels (international, regional, national, local). They must also be linked to disaster preparedness and DRR programmes. There must be strong links throughout the system and between its stakeholders: warning system failures often occur when these links are weak or break down (see Case Study 16.3: Early warning failure).+C. Garcia and C. Fearnley, ‘Evaluating Critical Links in Early Warning Systems for Natural Hazards’, Environmental Hazards, 11 (2), 2012. Institutional arrangements for coordination and communication have to be worked out carefully and agreed, and responsibilities defined. Setting up a system can take a long time, therefore, according to its scale and degree of complexity. Systems should always be undergoing testing, practice, review and refinement (warning systems for frequent events tend to be more effective than those for infrequent ones because they are used more regularly). Facilities and equipment have to be maintained and where necessary repaired; staffing and volunteer levels also have to be maintained. However, it is certainly not the case that only rich societies can have effective forecasting and warning systems.
16.5.2 The ‘last mile’
Large-scale, centralised systems tend to achieve broad geographical coverage but can fade out as they get closer to vulnerable communities and more marginalised groups. Information can be transmitted accurately and effectively through different levels in the system, but may not reach communities at risk (what is often called the ‘last mile’). This problem has been highlighted on a number of occasions. In most systems, the bulk of effort and expense goes into transmitting detailed, clearly presented information to decision-makers and emergency management services. Less effort and funding go into disseminating this information right down to individual communities or households through accessible messages that will warn them and help them to make decisions about how to respond.
Warning systems need to be end-to-end, therefore, connecting those who need to hear messages to those who prepare and deliver them. Here it is particularly important that they reach the most vulnerable and marginalised (the ‘last mile’ is as much social as spatial), and trigger local evacuation and protection mechanisms. The vulnerabilities, needs, roles and capacities of different groups in society must be taken into account. Messages reach people in different ways, they may interpret them differently and they have different responsibilities in response.
16.5.3 Local and community-based systems
There is an important role for small-scale, local and people-centred early warning systems that rely on the participation of those exposed to hazards. These can utilise local capacities and technologies to a greater extent than larger systems, reducing the need for sophisticated, expensive equipment and external technical experts. They can deal with the local incidence of hazards, which larger systems cannot usually manage, and are better aligned to local needs and capacities. Communities are more involved in running them and more likely to respond to their warnings.
The effectiveness of such systems is particularly evident in community-based monitoring of drought/famine (see Chapter 14) and flood warnings (see Case Study 16.3: A community-managed flood warning system). Local warning systems can sometimes be free-standing, but for comprehensive, integrated outreach it is better if they form ‘sub-systems’ of larger-scale programmes. It can be a challenge to establish and maintain sufficient local capacity (particularly of local government organisations) to connect local EWS to larger-scale, end-to-end systems and the institutions that manage them.
A flood and cyclone warning system in the Búzi River basin in Mozambique is a typical example of design for the needs and capacities of local communities. Work on setting up the system began in 2002 with the help of experts from Deutsche Gesellschaft für Internationale Zusammenarbeit (GIZ) and Honduras, and with co-financing from the Munich Re Foundation. Appointed village officials take rain and river gauge readings along the river, passing on messages by radio to a control centre if there is heavy rain or river levels rise. If a crisis threatens, blue, yellow and red flags are raised to indicate different alert levels and volunteers issue warnings using drums, megaphones, SMS and local radio. The volunteers receive training and there are regular system tests and practice drills.
In February 2007 Cyclone Flavio struck the area. Strong winds and heavy rainfall caused considerable damage and river levels rose rapidly. Villages at risk had been warned two days beforehand, and when the order to evacuate came from the district government some 2,300 inhabitants were moved to designated safe areas. The floods caused extensive damage to property and infrastructure, but only four people were killed.
Flood-warning System in Mozambique: Completion of the Búzi Project (Munich: Munich Re Foundation, 2007), http://www.preventionweb.net/english/professional/publications/v.php?id=2919; Mozambique: Disaster Risk Reduction as the Basis for Climate Change Adaptation (Eschborn: GIZ, 2011), http://www.preventionweb.net/files/32970_32970giz20110501enmozambiquedisaste.pdf.
For early warning systems to be sustainable, a wide range of stakeholders need to be involved in their design, set-up and management. This includes producing and verifying information, agreeing operational protocols and selecting appropriate communication strategies.
16.5.4 Official response
When a hazard threat develops, a designated institution or team has to make decisions about when and how to react, taking into account the nature, extent and timing of the threat, the location and vulnerability of people at risk and the local resources and capacities for emergency response. This places considerable responsibility on the decision-makers concerned. Underestimating the danger or reacting too late causes avoidable damage or casualties, but over-reacting can lead to false warnings and undermine the warning system’s credibility.
Official warnings to the public are usually given in stages, using defined warning or alert levels. The alert levels can be increased as the likelihood of disaster becomes more certain or imminent. This ensures that awareness is raised and emergency preparations can be made in good time, although in some cases (e.g. flash floods or landslides) the warning time may be very short, even just a few minutes.
Institutional response to forecasts and warnings of impending disasters is influenced by external factors – political, attitudinal, legal, economic, logistical, ideological and institutional – that are unrelated to the purely scientific data. Where events are seasonal or frequent, such as cyclones or monsoon floods, institutions are familiar with them and it is easier to develop and run effective warning systems. But in the case of infrequent events, officials may not understand the hazard and establishing a warning system is less likely to be a political priority. Volcanic eruptions are a prime example: many potentially dangerous volcanoes have not erupted in living memory, the exact timing of eruptions cannot always be predicted with certainty and volcanoes are complex natural phenomena that are not easily explained to non-scientists. Successful evacuations, such as that of 60,000 people ahead of the eruption of Mount Pinatubo in 1991, owe their success to the effort and ingenuity that went into communicating with non-specialists, including decision-makers, the media and the public.+See Communication during Volcanic Emergencies (London: UCL Hazard Centre, 2003), http://r4d.dfid.gov.uk/Output/5400; R. S. Punongbayan et al., ‘Eruption Hazard Assessments and Warnings’, in C. G. Newhall and R. S. Punongbayan (eds), Fire and Mud: Eruptions and Lahars of Mount Pinatubo, Philippines (Quezon City/Seattle: Philippine Institute of Volcanology and Seismology/University of Washington Press, 1996), http://pubs.usgs.gov/pinatubo. Case Study 16.4 outlines a famous, tragic example of political-institutional weaknesses contributing to a volcanic disaster that could have been avoided.
Shortly after 21.00 on the evening of 13 November 1985 the Nevado del Ruiz volcano in Colombia erupted, throwing out clouds of hot ash that scoured and melted part of the summit’s snow and ice cap, sending torrents of meltwater, slush, ice and volcanic debris down the slopes, where they picked up water, vegetation and other debris to form lahars that raced along the valleys of streams and rivers normally fed by the volcano’s melting snow and ice. Shortly before midnight, the lahars reached the town of Armero: more than 21,000 people were killed.
Yet the eruption was not a surprise. The volcano had been noticeably active for about a year. Early in 1985, government scientists and civil defence authorities were alerted. Civil Defence prepared a disaster plan, but this was done without an up-to-date hazard/risk map. This was the responsibility of the government geology and mines bureau, INGEOMINAS, but it showed little sense of urgency when it came to mapping or monitoring the volcano, and in any case did not have sufficient volcanic expertise. Equipment and experts had to be brought in from other countries to help monitor seismic activity (a key indicator of volcanic activity and likely eruption), but it was not until the end of August that the monitoring system was in place, and even then there were two parallel monitoring sets in operation, one run by INGEOMINAS and the other by an officially sanctioned local Volcanic Risk Committee that had been set up by the local government, universities and businesses to prepare against a possible eruption. Central government officials were offered more expert volcanologists, equipment, training and information by UNESCO, but did not act on the offer for nearly two months.
Nevado del Ruiz increased volcanic activity markedly in September, speeding up preparedness activity. The Volcanic Risk Committee issued a public warning of a serious risk of an eruption and avalanches of rock and ice. A national-level emergency committee was formed, Civil Defence developed its emergency management plan and the Colombian Red Cross assumed responsibility for emergency communication and disaster response. Civil Defence identified populations at most risk along the river systems fed by the volcano, initiated awareness programmes in schools, improved radio communications facilities and provided other emergency equipment and met national and local officials. Provincial emergency committees contacted villages to highlight the need for preparedness and encourage the development of local evacuation plans.
Yet disaster management arrangements remained incomplete. A preliminary hazard map was presented in early October, showing that extensive areas were threatened and some towns would need to evacuate rapidly, but only ten copies were made and distributed. The four provinces likely to be affected were developing separate plans, with little coordination. The seismic monitoring programme was still inadequate and data were not being shared fully. It was felt that the national government was hesitant about action, and some government officials in the capital criticised the hazard map as being too alarming. In an attempt to calm the population, a national newspaper stated that the volcano was not dangerous, as did the Director of the Geophysical Institute of the Andes. The Chamber of Commerce in Manizales, a large town near the volcano, expressed concern that irresponsible reporting would cause economic losses, and an Archbishop criticised the media for spreading ‘volcanic terrorism’. The Mayor of Armero stated that many people there were confused by the information they received: they did not know whether to stay or leave.
Improvements to the scientific monitoring system and public presentation of a revised hazard map were delayed by a national political crisis early in November, when guerrillas took over the Palace of Justice in Bogota and the government sent in troops to recapture it. When Nevado del Ruiz began to erupt in mid-afternoon on 13 November, regional and local emergency structures were alerted but no immediate decision to evacuate was made, although it was known that the lahar flows might be rapid, leaving little time to escape: the people of Armero would have at most two hours’ warning to evacuate to higher ground. In Armero, residents were reassured by a local radio station and the church public address system that there was no immediate danger.
After a new and more serious phase of the eruption began at 21.00, the Governor of Caldas Province called local radio stations to issue red alerts to communities living along the rivers. Officials in the capital of Tolima Province attempted to order the evacuation of Armero from 21.45, but there were power and communication difficulties owing to a torrential rainstorm filled with volcanic ash. Shortly afterwards, the lahar broke through a natural dam created by a landslide 12km upstream. The dam had been holding back 250,000 cubic metres of water, which were now released in a 40-metre-high wave. The Mayor of Armero had stated his concern about the dam on 17 September and government geologists had recommended draining it, but the work had not begun.
Survivors’ accounts suggest that there was no official, systematic order to evacuate, although in some cases representatives of relevant agencies took action as individuals. Many people were reluctant to move having heard the earlier reassurances from the local priest and radio station. Even the Mayor and his family remained. In Armero, most people fled, on foot and in the darkness, only after hearing the first flood waves hit the town.
B. Voight, ‘The 1985 Nevado del Ruiz Volcano Catastrophe: Anatomy and Retrospection’, Journal of Volcanology and Geothermal Research, 44, 1990.
16.5.5 Community response
As previous chapters in this book have shown, people at risk make rational choices about protecting themselves from disaster. Within communities, there are many different perspectives of risk: these vary according to socio-economic differences in wealth, social standing, education, age, religion, ethnicity and gender. Personal and collective experience plays a significant part. Risk perceptions are likely to vary considerably between different communities, and even within the same community. This diversity presents a challenge to those who have to transmit early warning messages over wide areas.
One of the principal socio-economic factors affecting response to disaster warnings in many low-income countries is the vital need to protect assets and maintain livelihoods. The poorer and more marginalised a household is, the more important it becomes to hold on to its assets, property and income. A household may perceive the risk of evacuation, in terms of losing control of assets and resources, as more devastating than the risk of the hazard, especially where warnings are frequent but do not necessarily lead to disaster. There are many indications that poor people delay evacuation because of this.
Warning specialists often fail to understand how communities perceive and react to hazards and risks. There are several reasons for this. The first is that specialists and communities look at a potential disaster from different starting points. Early-warning systems tend to start centrally, at international and national levels, and then move outwards and downwards towards districts, sub-districts and villages or neighbourhoods. In this perspective, individual villages or neighbourhoods are on the periphery, but for the individual at risk, their home and immediate locality are at the centre of the picture. This means that factors that are of primary importance to the villager or householder at risk are likely to be invisible to the system managers, who work on a much larger scale. Conversely, the manager’s national or regional perspective appears irrelevant to the individual at risk.
The second reason is that the two groups measure and describe risk in different ways. Technical specialists draw upon scientific and engineering methods of analysis to quantify risk, principally in mathematical terms of probability. This specialist technical knowledge is not understood outside the scientific community. It may not even be understood by officials and NGO staff responsible for disaster preparedness and response. It is not easy to translate such mathematical calculations into everyday language (such as ‘high’, ‘medium’ or ‘low’ risk) for operational use; indeed, this may only add to the confusion. Disaster victims and potential victims measure and describe risk in more varied, qualitative terms. Technical words and phrases which seem ordinary to scientists may not be understood by the public. For instance, official warnings of the arrival of Typhoon Haiyan in December 2013 referred to a ‘storm surge’ (the technical term for an abnormal rise in water levels caused by low-pressure weather systems). Yet many residents of the coastal areas affected, as well as local decision-makers, admitted that they did not understand the term and as a result failed to evacuate.+O. Neussner, Assessment of Early Warning Efforts in Leyte for Typhoon Haiyan/Yolanda (Tacloban: GIZ, 2014), Environment and Rural Development Programme, http://www.preventionweb.net/files/36860_36860gizassessmentofearlywarningyol.pdf. Seemingly simple alert level systems, using number or colour codes, are usually designed to be clear to affected communities, but even these can sometimes be misunderstood.
A third reason is the assumption among some disaster professionals that they alone understand and assess risk objectively (i.e. scientifically), whereas the disaster victims’ understanding and assessment is merely subjective, even irrational, perception. There are a number of problems with this attitude. One is methodological: it is not possible to maintain a clear distinction between ‘objective statistical’ and ‘subjective perceived’ risk because ‘objective’ risk estimation itself involves value judgements, such as the definition of what constitutes a ‘hazard event’. This attitude also undervalues the knowledge of those who actually experience hazards on the ground, and overlooks the social and economic forces that make some people more vulnerable than others. A better understanding of such matters requires different approaches to communicating with communities at risk, based on dialogue and community participation (see Chapters 6 and 10).
Many communities draw on their own indicators of impending hazard when deciding how to respond to warnings. These are often based on observation of weather patterns, the action of the sea and rivers and animal behaviour.+See for example P. Howell, Indigenous Early Warning Indicators of Cyclones: Potential Application in Coastal Bangladesh (London: UCL Hazard Centre, 2003), http://www.ucl.ac.uk/hazardcentre/resources/working-papers2; G. Cerdena, ‘Indigenous Know-How on Mayon Volcano’s Lava-Spittle Mysticism’, in R. Shaw et al., Indigenous Knowledge for Disaster Risk Reduction: Good Practices and Lessons Learned from Experiences in the Asia-Pacific Region, European Union/UNISDR Regional Office for Asia and Pacific, 2008, http://www.unisdr.org/files/3646_IndigenousKnowledgeDRR.pdf. Such indigenous knowledge has been shown to be quite reliable on occasion, particularly for frequently occurring events, but there has been little attempt at rigorous scientific validation. More work could be done to correct potentially dangerous errors in understanding and to enable warning systems to incorporate reliable indigenous indicators; famine early warning systems certainly benefit from community participation, as local people are sensitive to socio-economic as well as agricultural indicators of food insecurity.
16.5.6 Controlling information
Emergency planning manuals highlight the importance of officially validated forecasting and warning information issued from a central point. Whilst the use of multiple communication channels is necessary to ensure maximum outreach as well as reinforcing the warning message, disaster managers are often concerned about unofficial sources of information, especially radio, satellite and cable television stations, the internet and social media. Information from multiple unofficial sources, with varying degrees of reliability, is generally reckoned to be dangerous, leading to unjustified alarm and incorrect responses.
There is some justification for this view, but in the modern age command and control of information is unrealistic. The public are increasingly consumers of information from different sources, choosing what information to use and where to obtain it (see Chapter 10). In practice, community members often draw on a variety of information sources, types and messages to decide when to take action and what form of action to take in response to warnings. Social capital, social networks and inter-personal communications are key factors in sharing hazard warning information and motivating people to take responsive action, whatever formal channels or technologies are used for communicating warnings.
It is difficult to strike a balance between the need for authoritative warning information and people’s desire to make their own choices. Disaster managers will have to acquire extensive skills in media management, but the central issue is probably one of public trust in the competence and integrity of disaster professionals. Trust is intrinsic to the success of warning communications. Where the authorities, officials or scientific experts who direct forecasting and warning systems are not trusted, people are more likely to disbelieve forecasts and warnings, or seek out other, informal sources of information.
16.5.7 Science and technology
Recent decades have seen rapid advances in scientific understanding of natural hazards and ways to monitor them. This has greatly enhanced scientists’ ability to forecast the location, timing and severity of events. All forecasting and warning systems rely on scientific knowledge of one kind or another, but scientists’ capacity to predict varies with the hazard studied. For example, in the case of geological hazards (earthquakes, volcanic eruptions, landslides, tsunamis), it is possible to identify where events may take place, but it can be very difficult to indicate when. Short-term predictions or forecasts (over days or hours) are generally much more successful in the case of landslides, volcanoes and tsunamis than they are for earthquakes. Meteorologists have become very skilled at making short-term forecasts of tropical cyclones, predicting their timing and movement, and their seasonal forecasting is also becoming more reliable.
The scientific-technical resource base is the result of many years of investment globally. Knowledge is widely shared among scientific communities. Data from technical devices such as remote-sensing satellites and buoys monitoring sea-surface temperatures are routinely transmitted to forecasters and disaster planners through established global networks. The World Meteorological Organization, for example, has played a significant role in coordinating monitoring and forecasting of hydro-meteorological hazards.
Technological sophistication is not necessarily a barrier to small-scale warning systems or community involvement. A wide range of technologies may be appropriate for particular hazards, localities and needs. Those appropriate at regional or national levels include satellite imagery, GIS maps, computerised networks for receiving and transmitting data, automated gauges and other monitoring devices and radio and television broadcasts. At community level, one might find participatory mapping of hazards and vulnerable households, manual river level or rainfall gauges, signs marking evacuation routes and the use of megaphones, bells and drums to issue warnings. In many cases, vulnerable communities will monitor impending events themselves; for example, communities living close to flood-prone rivers often have people watching water levels at times of severe or prolonged rainfall.