Lucy Pearson looks at early warning systems for disasters, their uses and limits, and what accounts for the gap between warning and action.
Through history disasters have destroyed lives and livelihoods, killing people and damaging homes and businesses. Disasters in the past 35 years have taken an estimated 2.5 million lives and cost more than US$1.5 billion, mainly in developing countries. 
Disasters result from natural and biological hazards (floods or infectious diseases, for example) as well as complex sociopolitical emergencies and industrial hazards (droughts or radioactive leaks).
The extent of the damage caused by a hazard is related not just to its severity, but also to the capacity of people living in disaster-prone areas to prepare for and resist it. Efforts to reduce disaster risk have therefore focused, in part, on developing early warning systems to provide timely and effective information that enables people and communities to respond when a disaster hits.
Early warning systems are combinations of tools and processes embedded within institutional structures, coordinated by international — and sometimes national — agencies. Whether they focus on one particular hazard or many, these systems are composed of four elements: knowledge of the risk, a technical monitoring and warning service, dissemination of meaningful warnings to at-risk people, and public awareness and preparedness to act. Warning services lie at the core of these systems, and how well they operate depends on having a sound scientific basis for predicting and forecasting, and the capability to run reliably 24 hours a day.
Scientific and technological advances (Box 1) have driven marked improvements in the quality, timeliness and lead time of hazard warnings, and in the operation of integrated observation networks. But advances in technology alone are not enough — and in some cases they can even create obstacles to the capacity of vulnerable populations to respond.
Box 1: Technologies for monitoring and warning
Forecasting and modelling technology
Several countries have early warning systems based on seasonal-to-interannual climate forecasts.  These systems are based on using monitoring data, including temperature and rainfall values, and state-of-the art climate models. Climatologists analyse the observations and model-based predictions to predict climate anomalies one or two seasons ahead.
Remote sensing and geographic information systems (GIS) applications
Remote sensing and GIS applications have significantly advanced famine early warning systems. The Regional Centre for Mapping of Resources for Development (RCMRD) has been using remote sensing-based regional early warning systems for food security to supplement national initiatives in eastern African countries. RCMRD predicts harvests half way through the growing season to give advance warning on food security before the end of the season. In addition, flood monitoring is now regularly informed by remote sensing that obtains information on soil types, water resources, settlements, cropped areas and forests.
Satellite communication technology
Improvements in satellite communication have helped decrease the lag time between data collection and warning. For example, the Pacific Tsunami Warning System works by a recorder on the seabed relaying data on anomalies to a buoy on the surface. This data is then transmitted via satellite to ground stations every 15 seconds.
Mobile phone technology
With the global spread of mobile phones and networks, this technology is now increasingly used to communicate warnings and coordinate preparation activities — particularly SMS alerts for disseminating mass messages. For example, upon detection of p-waves that precede earthquake shaking, Japanese agencies send out SMS alerts to all registered mobile phones in the country. However, some obstacles can arise with this technology — phone pylons can be damaged or networks can be overburdened during hazards, for example.
ICTs for crowdsourcing
The use of 'crowdsourced' data is gaining traction with increasing Internet connectivity and use of information and communication technologies (ICTs) such as mobile phones. Crowdsourcing was used extensively in the response to the 2010 Haiti earthquake, allowing local people, mapping experts and other stakeholders to communicate what they saw and heard on the ground, and to produce information that could be used by humanitarian workers. This was particularly useful in locating survivors who needed assistance, but it is increasingly recognised that crowdsourcing could also help with pre-disaster activities, specifically risk identification and early warning.
Through initiatives such as Ushahidi and Google Crisis Response, crisis mapping utilises crowdsourcing as well as satellite imagery, participatory maps and statistical models to power more informed and effective early warning. It can provide real-time information on an upcoming crisis in times of uncertainty and confusion. The vast amount of data that can be produced from such systems can be analysed through networks of stakeholders (such as Crisis Mappers).
Early warning systems: what are they good for?
Early warning systems are increasingly considered to be an integral component of disaster preparedness and involve a broad spectrum of actors. Figure 1 shows just some of the key events in the development of such systems.
But early warning systems do not exist in every part of the world. A quarter of the countries assessed in the 2011 Global Assessment Report for Disaster Risk Reduction reported that communities did not receive any timely warnings for impending hazards. 
And while some early warning systems are better than others, existing ones are still in need of improvement. Discussions on how to improve effectiveness can be informed by critical analyses to determine what early warning can realistically achieve, and what is outside its limitations (Box 2).
Box 2: What can we expect of early warning systems?
Early warning can save lives
Several countries have significantly reduced deaths by developing effective early warning systems. Cuba's Tropical Cyclone Early Warning System is credited with reducing deaths dramatically for weather related hazards such as tropical cyclones, storm surges and related flooding: five successive flooding events left only seven dead.  Another example is Bangladesh, which now has a 48-hour early warning system in place that allows people to evacuate to safe shelters hours before cyclones make landfall, reducing deaths. In 1970, 300,000 died as a result of Cyclone Bhola, compared to 3,000 in 2007 during Cyclone Sidr, which authorities were able to track as it grew in strength.
...but cannot prevent all damage
While a certain amount can be done at the local level to protect lives and livelihoods once a warning has been received, there is little that can be done to protect infrastructure in a sudden disaster —financial losses from destruction of buildings and interruption of services still occur. However, in slower onset disasters that can be pre-empted days or months in advance, early warning systems can provide enough time for risk reduction measures to be put in place, such as retrofitting buildings and constructing barriers.
Early warning can help in many types of hazard
Warning systems are in place and have proved beneficial for a variety of hazards. In the case of tsunamis, the benefit of an internationally coordinated system was shown in the 2011 earthquake and tsunami in Tohoku, Japan, which threatened many Pacific islands: warnings were more coordinated than in the devastating Indian Ocean Tsunami in 2004, providing time for many people to evacuate to high ground.
Having an impact is more difficult for systems set up to warn of hazards that have complex causes, such as drought. However, some countries have developed systems capable of integrating information from various sources and providing warnings of the imminent onset of drought. And early warning systems for food security have developed significantly over the past few years. The UN Food and Agriculture Organization's Global Information and Early Warning System on Food and Agriculture (GIEWS) is the most globally complete food security monitoring system.
...but is limited when it comes to geological hazards
The signs of an impending volcanic eruption or a landslide can sometimes be detected at an early stage and used for warnings. Regional monitoring systems have been installed in most earthquake-prone regions, and multinational initiatives exist (the GEOFON network of the research institute GeoForschungsZentrum Potsdam, for example). However, picking up earthquake precursors is difficult, and routine predictions remain elusive: the location, magnitude and time of occurrence of earthquakes cannot be forecasted.
Yet even a lead time of a few seconds can make a difference, and some countries are working with the limited information available. In Mexico City, for example, technical systems can identify the first seismic wave following the start of an earthquake that may have happened more than 100 kilometres away — allowing authorities to use this information to shut down critical systems such as gas supply lines.
The gap between warning and heeding
However, improving the effectiveness of early warning systems does not, in itself, lead to reduced risk for disaster-prone communities — early warning does little good unless it is followed by (early) action.
Warnings are still not effectively communicated, and not sufficiently acted upon, even as agencies in developed and developing countries are now more aware of the nature, frequency, locations and intensity of various hazard types, and have advanced technical capabilities for monitoring such as climate models and remote sensing. [3, 4]
A good example is one of the most devastating disasters in history, the 2004 Indian Ocean Tsunami. The Pacific Tsunami Warning Centre in Hawaii picked up the earthquake. But despite the phone calls made by the centre to government agencies in countries such as Indonesia and Thailand, the emergency infrastructure was missing and so the warning was not disseminated to local communities. 
So what accounts for the gap between early warning and response? Identifying the factors that contribute can help countries and the international community to find ways to address them.
The uncertainty inherent in scientific information is one of the reasons for failing to act on disaster warnings. Information from forecasts is often in a language and format that is not easily understood by humanitarian workers or the local communities that need it. Scientific jargon relating to uncertainty regularly causes users not to act.
Statements such as 'there is a 20 per cent chance that rainfall will be above the interannual mean' present information in an unfamiliar language.
However, uncertainty does not have to be a reason for inaction. Two-way exchanges of information can mitigate misunderstanding and help scientists and users of scientific information to appreciate each other's 'language', their respective objectives, and how they might best work together to prepare for a disaster (Box 3). 
Box 3: The need to understand uncertainty
In 2011 the Humanitarian Futures Programme conducted research on the use of climate science in informing livelihood decision making in the context of seasonal flood and drought conditions in Kenya.  It found that although the Kenyan Meteorological Department had been generating useful and relevant information for crop and livestock producers, it was not in a form that they could understand. Questionnaires also indicated that the community had a high level of mistrust towards the agency, largely because it had previously produced predictions that did not materialise. A lack of understanding of the uncertainty of estimations led people to interpret the predictions as wrong, and to believe that estimations could no longer be trusted.
Another reason for inaction is that the warnings tend not to reflect an understanding of the decisions people then need to make in response. In developing countries, this means getting a handle on the well-established link between disasters and poverty. For example, a farmer may stay looking after their cattle rather than evacuate because they judge the risk of flood to be lower than the risk of losing their livelihood.
Communicators of early warnings can work more effectively by taking into account how people behave in that crucial period after they receive a warning — particularly how they prioritise different risks. Assessing behaviour after disasters can help to clarify who does and does not heed warnings, and why.
Reducing false alarms
As early warning systems grow in geographical coverage and sophistication, false alarms are rising too. While some believe that they provide invaluable practice, high false alarm rates can undermine public confidence, breed mistrust, dilute the impact of alerts and reduce the credibility of future warnings.
In 2007, a local tsunami alarm was raised mistakenly in Aceh, Indonesia, causing mass panic and injury as residents fled. Anger led residents to later disable the tsunami warning system, causing unnecessary vulnerabilities and long-term risk. And this year, an earthquake measuring 8.7 on the Richter Scale, which hit off the coast of Indonesia, led to the activation of the Pacific Tsunami Early Warning System; but there were no significant tsunamis, and the likelihood of a tsunami was judged to be low based on the characteristics of the earthquake.
One approach to reducing false alarms is to use reliable local hazard indicators, such as animal behaviour or vegetation changes, to verify scientific indicators of upcoming hazards. Another approach is to work with the media to avoid inaccurate, exaggerated or misleading information about potential events.
Monitoring communication tools
Innovative ICTs are being developed and rolled out, playing an important role in disseminating information to organisations in charge of responding to warnings and to the public during a disaster. But their capacity to make an impact is limited by the lack of systematic and consistent monitoring.
Web services, SMS and email, as well as more established technologies such as radio and television, have all been used to communicate warnings. But these tools are created and deployed in various locations and under different circumstances, with insufficient follow up on what does and does not work.
For example, television is not always effective in the most at-risk communities due to mistrust. If follow-up does take place, it often fails to monitor effectiveness over both the short and long term, or may raise questions over reliability if undertaken by the organisation that has implemented it.
Finally, insufficient coordination and collaboration between organisations can hold back efforts to encourage early action because the organisations that produce warnings are not those that disseminate them. For example, in the case of hurricanes, the World Meteorological Organization collects atmospheric data which are then transmitted to the US National Hurricane Centre, which generates forecasts and hurricane advice.
This advice is then transmitted via the Global Telecommunication System, fax and the Internet to national meteorological and hydrological services in countries at risk, where national forecasters use them to produce specific hurricane warnings. These are then dispatched tolocal newspapers, radio and television stations, emergency services and other users.
But communication mechanisms between organisations as well as agencies within countries are limited. And there are institutions with overlapping mandates; for example, both the local agricultural agency and the climate change department may view it as their responsibility to communicate a flood warning to communities, and separate warnings can cause confusion.
Hazards do not abide by the territorial boundaries of countries or districts. And as hazard exposure areas expand due to climate change, the sharing of information is set to become more important. Better communication channels and linked policies that create one authoritative voice can help to address this.
A changing climate means changing needs for developing countries and their capacity to respond to disasters. Shifting rainfall patterns and hurricane paths, and more days of extreme temperature, will bring new hazards to areas that previously may not have experienced them.  In addition, settlements and services are expanding into at-risk locations as urbanisation intensifies along the coasts, increasing exposure to hazards (Figure 2).
Early warning systems — and the technologies and tools that support them — will work best if they are embedded in, understandable by and relevant to the communities they serve.  This will have particular value where communities cannot rely on the government to respond effectively.
And there is a need for local knowledge and practices to be integrated with those of the science community, to improve forecasts and increase acceptance, ownership and sustainability of early warning systems. The UNISDR's Hyogo Framework for Action emphasises the importance of encouraging the use of traditional knowledge.
The idea is that local practice and scientific practice can complement — not replace — each other, because each has its own advantages and restrictions. In the Solomon Islands, for example, integration has occurred with the communication of early warnings on Tikopia Island, where only a few residents received a Radio Australia transmission warning (scientific method) of the coming cyclone in December 2002. The local communication system (indigenous method) then took over with local runners taking the message out to other community members in the local language. [11, 12]
But there is no simple way to improve early warning systems. Their impact will be maximised only when all necessary steps are taken to enhance the effectiveness of technological tools and scientific forecasts that governments and communities rely on, providing more time for appropriate action.