Skip to main content

Opti-Alert: Enhancing the efficiency of alerting systems through personalized, culturally sensitive multi-channel communication

Final Report Summary - OPTI-ALERT (Opti-Alert: Enhancing the efficiency of alerting systems through personalized, culturally sensitive multi-channel communication)

Executive Summary:
The goal of the project was to develop concepts and prototypes for an improved alerting of the population in times of disasters. Authorities should be able to communicate with the population more swiftly, consistently, and effectively. This should be enabled by an integration of different message production and message distribution systems, improved personalization and regionalization of alert messages, and tools for the simulation of the impact and effectiveness of alerts. Basis for the proposed improvement was an in-depth, qualitative analysis of the population’s information needs and communication patterns, accompanied by a study on crisis communication via the mass media.
Sociological studies conducted in Opti-Alert indicate that the operators of communication tools should be able to adopt alert content to the recipient group, that alert messages must be embedded in the national trust context (who is a trustworthy source?), and that the alert message should be regionalized to fit into the situational context at the recipient’s area of residence. Aspects to be considered when designing alerts are, among others, frequency and severity of disaster situations in the target area, and thus, experience of the recipients with the respective threat. Personalization is necessary for people with impairments, tourists, and, in many cases, also migrants. Some authorities also wished to be able to address people with children separately.
Communication studies conducted in the project underline the need for simple, consistent and unambiguous messages. In order to align official communication with media reporting on a disaster, authorities should adopt a proactive media strategy, provide additional material on official websites and should participate in the “public” discourse on social media, and with opinion makers from the media. Verified facts should be provided as swiftly as possible to avoid intransparencies and speculation.
Technology development in the project focused on (a) an integrated alerting toolsuite, and (b) the development of simulation tools to analyze the effect of an alert on the population, in general and inside critical infrastructures like airports or stations. Core element of the integrated alerting toolsuite is an alert messages repository that allows alert message producing tools to publish alerts, and different distribution systems to collect relevant alerts and to send those to their recipients. Within the project, two alert message generation tools, and 4 distribution systems have been integrated. These systems cover alerting via pager, SMS, automated telephone calls, E-Mails, a website, and a smartphone app. The alert messages producer allow for a personalization of alert messages based upon a baseline message for everybody, with extensions for different target groups. The actual personalization is then taken over by distribution systems and end-user devices. All attached alerting systems must transmit at least the baseline message – personalization is only provided to the end-user if the distribution system is capable of doing so.
In order to enhance alert planning, a simulator has been developed that shows how different types of people (agents) behave inside critical infrastructures in case of alerting. In addition to general personal characteristics, certain disability types and language skills can also be modelled. The different impact of communication media on the agents can be simulated, as well as how quickly, for example, a critical infrastructure like a station can be evacuated if needed. A second simulator analyses how quickly alert messages are distributed in the general population. This simulator takes into account the availability and type of alerting system (e.g. point alerts or range alerts), situational factors like the time of the day, and regional characteristics like population density and household sizes. Based upon this input both direct communication between authorities and the population, and indirect alerting effects like the dissemination of messages via word-of-mouth, is modelled.

Project Context and Objectives:
1.1.2 Summary description of project context and objectives Project context

In recent years, public authorities and private companies invested heavily in different types of alerting systems. Examples include the alerting system “WIND” for extreme weather conditions financed by the in¬suran¬ce industry or the publicly funded “German-Indonesian Tsunami Early Warning System (GITEWS)”, the “HAZARDOUS” flash flood warning system in Italy, or the more general “Emergency Alert System” (EAS) in the United States. At the start of the project, these alerting systems allowed users to subscribe to different types of alerts and to select relevant alert levels and communication channels. However, alert message composition and design only depended on the user’s functional role (e.g. member of the emergency services, home owner, driver etc.). In other words, the perso¬nal, regional, cultural and social characteristics of the warning recipients were not taken into account during message composition and delivery. This shortcoming was particularly relevant in case of large-scale international disasters (such as tsunamis, storm surges, large-scale nuclear accidents or hurricanes) where an integrated, multi-national warning and alerting strategy for the general public is needed. Due to a lack of cultural and personal sensitivity of existing alerting systems, warning messages were not ideally adapted to the recipients and therefore could not achieve the optimal impact.
Another unsolved yet important issue was the interplay between alerts issued via different communication channels. Little is known, for example, how personalized warning messages issued via e-mail, or SMS interact with more general warnings issued via the mass media, and how such an inter¬ac¬tion may affect the behaviour of the general public in an emergency situation. Simulation tools that allow for a quick assessment of the likely impact of different warning strategies on the general public were virtually non-existing. As a result, decisions on how to alert were usually taken intuitively on an ad-hoc basis.
The initial situation was aggravated by the fact that the “alerting landscape” consists of a plethora of often insulated systems, operated by different countries and companies. Due to both the heterogeneity of the alerting framework and the lack of inter-system interaction, warning strategies frequently lacked a holistic view, and as a result, compliance with recommendations and action requests was often sub-optimal.
Improved regionalisation and personalization of warning messages, as well as a closer coope¬ration and integration of industry-funded alerting systems and state-funded alerting tools, was considered as one way to mitigate the identified problems. In other words, it was time to move away from a “one message fits all” paradigm and to offer more appropriate alerting methods. Project objectives

The objective of the Opti-Alert project was to develop an alerting suite that

(a) allows for a simulation of the impact of different alerting strategies (depending on the selected media-mix and current availability of communication media)
(b) supports the composition of the optimal mix of communication channels (individualized alerting channels and mass media)
(c) improves alert compliance through social and cultural adaptation and personalization of alert messages and communication channels
(d) supports the rapid and automated implementation of a selected alert strategy,
(e) can simultaneously address a large variety of communication channels to facilitate efficient high-throughput alerting and
(f) can be integrated with existing tools and legacy systems via well-defined interfaces.

These objectives were supported by the following key research activities:

(a) in-depth analysis of the impact that social and cultural and regional factors have on risk perception and risk communication
(b) analysis of the influence which the observed socio-cultural differences have on regional alerting strategies
(c) analysis of the impact of individualized alerting (via SMS, E-Mail etc.) and alerting via the mass media
(d) identification of best-practices in alerting via mass media
(e) definition of appropriate algorithms for the simulation of alert propagation within the population (in general, but also inside critical infrastructures such as metro stations), depending on the selected mix of communication channels and communication patterns between humans “in the field”.

One goal of this project was to improve the impact of alerts by developing alert strategies, which take socio-cultural characteristics of the alert recipients and the local alerting infrastructure into account. Based upon the context of an alert situation, the authorities should be enabled to swiftly simulate different alerting strategies (in terms of communication channels and media mix). This should enable authorities to re-assess alert procedures and processes and to improve impact and coverage of alerts.

Another innovative aspect of this project is the adaptation of alert content to the socio-cultural milieu of the message recipients. This refers, e.g. to the wording of the messages, or layout. The idea is to improve the compliance of alert recipients with the proposed protective actions by creating trust and, if necessary, a sense of urgency among those who have been warned.

Project Results:
The discussion of scientific results in this section follows the structure of RTD work packages in the project. For each such work package, key results will be presented, based upon a compilation of most relevant results by the work package leader. Socio-cultural factors in risk and crisis communication (WP2)


Objective of this work package is to address the fundamentals: how are different types of risks (technological, environmental, social etc.) perceived in different socio-cultural and regional contexts, and what is the impact for an efficient risk communication? The perception of risk communication shall be analyzed for different socio-cultural clusters in all participating countries. Additionally, current media repertoires of the different socio-cultural milieus are analyzed, compared, and the impact of the different media repertoires on the risk communication process are reflected. A particular focus within this analysis lies on the perception of individualized alerting (via SMS, E-Mail, personal messaging etc.), “traditional” alerting via mass media like radio and TV, and emerging communication channels on the web (blogs, social networks, etc.).


We have chosen for our study an innovative methodological social scientific approach that combines the information and knowledge gathered by expert interviews and research of official publications, websites, literature, etc. with the information, impressions, knowledge, (information) behaviour and information expectations reported by individuals who were actually personally affected by a disaster or who considered about it within the focus groups. The major advantage of this approach (‘triangulation of methods’) is that we not only (re-)produced an official perspective in regard to the supply of crisis management and crisis communication strategies and methods - of which officials might potentially think and take for granted that the supply will be used by the target groups as intended by the sender - but we also gathered the perspectives of those who did or will use the official offers (users) - or not - and what they actually did and why they did as they did or assumed they will do. The analysis of the explorative interviews with disaster survivors as well as the analysis of the focus group representing “the general public” illuminate strong and weak points of current alerting systems and give several valuable insights for improvements of the current alerting and crisis communication system. The comparison of the two perspectives - expert view/sender of alerting and crisis communication and users of alerting and crisis communication – can contribute to a better understanding of more effective alerting systems.

Description of main S&T results

a.) Summary of findings from the sender perspective (expert interviews):

As this project deals with target-group-specific crisis communication, the question of whether authorities already identify and direct their work to certain socio-economic or cultural groups in a distinctive way has been a major focus during the expert-interviews preceding this report. The basic finding is that, by and large, such strategies exist only to a very limited extent, if at all. While adjusting crisis communication to different stakeholders has become normal by now and a broad range of socio-demographic data like population density is available to the authorities in many countries, the topic of personalized and regionalized alerting still leads a rather shadowy existence in crisis communication.
Still, some modest beginnings in this regard exist. According to the respective answers, five basic dimensions can be identified, where targeting some segments of the audience with special means may become relevant: reachability, perception, comprehension, physical or psychological vulnerability and compliance.
1) Reachability concerns three questions. Firstly, a general reachability: Who are the users of the provided information and which media are appropriate to get messages across to target groups. Different age groups will be reachable via different channels and media: Young people will be accessible by new social media, while elderly people won’t. Secondly, questions regarding spatial reachability: are the available means and channels of communication sufficient to reach individuals even in remote areas like rural regions or in marginalized settlements at the edge of larger cities. One Austrian expert debated this point, coming to the conclusion that in these times of mass media and the Internet, no worries at all exist in this direction. By contrast, an Italian official raised doubts whether this holds true regarding marginalized settlements next to some cities in Italy. However, no special strategies have been developed so far embracing this concern. The third aspect, also related to spatial reachability, is if the infrastructure is adequate to account for perceived higher risk areas, like flood-prone settlements or near certain objects defined as especially hazardous like nuclear power plants. As we have seen, Sweden has de facto defined households near its nuclear power plants as a high risk group, tackling this problem by distributing indoor alarm receivers to each of these for free.
2) The second dimension of special target groups, labelled perception here, leads us to the field of people who receive crisis communication but cannot perceive it due to permanent hearing damage, blindness or other individual characteristics. Particularly people with permanent hearing damage have been referred to by several experts when asked about special groups which are hard to reach. While in some countries, like France, sign-language translation is a standard feature of TV news and alerting messages via television are often conveyed in the form of a news ticker, others point to the prospect of cell broadcast which is deemed an effective way to reach this group as mobile phones include the option of vibration alarm.
3) The next dimension, comprehension, refers to the point that even if an alarm is received and perceived it may still not be understood by the recipient. This matter is brought up in the context of people with a migration background or tourists that do not speak the official language at all or in an insufficient manner. While in almost all countries reviewed here experts mentioned this topic, very little is done so far to address this problem. When this topic was debated in the Netherlands the conclusion was that there is no need for action as tourists are expected to simply ask a Dutchman for translation. At the other end of the spectrum the Hungarian government has decided to publish crisis communication in the language of any ethnic minority in any region where it accounts for 10% or more of the total population. This is for the benefit especially of Romany which is the largest minority in Hungary and who often live spatially and socially isolated. Also Germans, Slovaks and several other regionally concentrated minorities benefit from this measure.
4) The fourth dimension is physical and psychological vulnerability, which concerns groups that receive, perceive and understand crisis communication but are not able to comply with behavioural advice to the usual extent due to their physical or psychological condition, or are very vulnerable to the effects of a disaster for the same reason. Almost all experts referred to one or more groups falling into this category: children, elderly people as well as invalids and disabled persons. Subsequently, some countries like Italy and France include details on institutions hosting these groups, like hospitals, kindergartens and homes for elderly people, in their disaster management plans, defining them as primary targets for additional alerting measures or evacuation. One of the German industrial plants also keeps special lists with the contact data of these institutions in its area in order to give them additional telephone calls in emergencies.
5) The fifth dimension where target-group-specific alerting is potentially relevant is compliance. Even if people have received, perceived and understood crisis communication and are able to follow instructions given by authorities there may be groups that simply refuse to do so for whatever reasons. In general, it can be said that none of the experts mentioned this topic on its own accord, although, for example in Italy, specific groups have been outlined where non-compliance poses a typical problem: farmers, elderly people and owners of houses, as all of them are reluctant to evacuate and prefer to stay in their homes.

b.) Summary of findings from the recipient perspective (biographical interviews and focus groups)

1. Sensible alerting and crisis communication have to take into account special needs of different target groups that require particular attention beyond a ‘one-message-fits-all-message’. We have identified within WP 2 the following target groups, which deserve special attention during the outbreak of a disaster and should clearly be the focus of crisis management: disabled persons, people which are full-time or part-time in social institutions (eg. children in school or day care facilities, retired people living in elderly people’s homes, people committed to hospitals, etc.), people with language deficits (including immigrants and tourists) and families, as we find evidence that particularly families with smaller children have special requirements when it comes to a disaster, since frequently parents and children are not affected in the same location.
2. Multi-channel alerting is a pre-requisite of effective crisis management: there is nothing like a single alerting tool that will be understood and reach the whole population. It is necessary for crisis management to include a variety of alerting instruments, as not all of them may be meaningful in a particular context.
3. Multi-layer alerting should be pursued by authorities: sirens are considered the main alerting tool in all examined countries. But, as sirens give no more additional information, other alerting tools that allow real crisis information and communication need to follow, such as alerting SMS, loudspeaker, megaphone cars, emergency hotlines, etc, and to complement the alerting by sirens.
4. All crisis communication messages released by the authorities should include clear behavioural advice in order to enhance compliance and to avoid public panic.
5. Who the sender of the alerting message should be is strongly dependant on trust in authorities and trust in media. As this varies strongly from country to country, it is advisable to take this into account for enhancing compliance.
6. The formulation of crisis messages should be as concise and simple as possible to be understood by most members of the public. It should include tips for further information, such as links to important websites or the number of the emergency hotline.
7. Public knowledge on alerting and compliance is rather low and often overestimated by the respective authorities. Even if people report being knowledgeable, when looked at more carefully it often emerges to be no more than just purely a feeling of being informed.
8. In the majority of cases, disaster behaviour/coping strategies are not based on knowledge but can be considered as purely intuition.
9. Crisis communication should be, in areas with a lot of people with a migration background or many tourists, performed not only in the national language but also in other relevant languages in order to be understood by these people as well. This request is easy to implement on websites, but could be also done on electronic billboards, and subtitles on TV screens, etc. At the least the language of the respective minority population should be considered – a situation which we found respected in only one of our examined countries: Hungary.
10. The value of personal communication in case of a disaster could not be overestimated for compliance. Authorities should be aware of this and establish structures that allow the inclusion of either local people in the crisis management / crisis communication system or the inclusion of other persons considered trustworthy.
11. Alerting of disasters should not be dependent on prior registration, but should be done as automatically as possible. The Dutch NL Alert system could be considered as best practice here.
12. SMS appear to be the most promising new alerting tool for many people, though not for all.
13. Currently, social media cannot be considered an effective alerting tool, as even the majority of young people would not trust it in case of an emergency. It remains to be seen if this will change in the future.

c.) Concluding general rules for personalized and regionalized alerting

Those who are responsible for the (improvement of) alerting and crisis communication should be aware of the following three general rules that we identified within our empirical research for meaningful alerting:

Rule 1: The choice of the alerting tool should correlate with the age of the recipients.
As a general rule: the younger the recipients of information the more likely they would use new media, like SMS or Internet. Social media, such as Twitter or Facebook, should be currently considered as one additional information channel next to others, but is far from being meaningful for alerting in all examined countries. SMS is considered by the younger and middle-aged generations to be the most popular and powerful alerting tool throughout all countries. The older the recipient of information, the more likely they will use traditional media such as radio or TV. It is important to note that the spread of mobile phones does not automatically say anything about how cell phones are actually used by these people. This is in particular striking for older people: although several of them possess a mobile phone, not all of them are – always or even regularly – accessible by their cell phone or even by SMS. We found evidence that several older people in the countries of our study would only turn their mobile phone on when they want to call someone now – and turn it off afterwards. Elderly people are frequently unaware of the possibility of SMS – and reported that they are either unaware of this function, or that they would not know how to use it and assume they would probably never use this function themselves.

Rule 2: The alert message should correlate with the area of residence.
We found differences in regard to risk perception and risk knowledge between more rural and more urban areas in all countries, on the one hand, and differences between natural and industrial risk areas, on the other hand. In more urban contexts people feel safer than in more rural areas. In addition, people in more urban areas reported a subjective feeling of being well informed more frequently, which not always matched their actual knowledge, but which corresponded to their lower risk perception. Therefore, alert messages should take this fact into account and should be framed accordingly. As a general rule it can be stated that people in recurring natural risk areas (such as frequent flooding, snow falls, avalanches, etc.) are more familiar with and know more about disasters and how to handle the situation, have a better knowledge of where to gather information and where to turn to for help. In general, they frequently have established coping strategies as their risk perception is usually higher than in regions with industrial risks, therefore the alerting message can be much more direct than in all other cases. In industrial risk areas people do not show an increased risk perception for the respective risk source (which could be factory, a nuclear power plant or similar), in fact they regularly show a behavior that could be explained as dissolution of cognitive dissonance. The difference between the risk areas is also an expression of the fact that recurring natural risks appear to be more manageable than spontaneous industrial risk accidents.

Rule 3: The choice of the alert sender should correlate with the respective national “trust” context.
We found huge differences in regard to trust in authorities and trust in media, the two main sources of alerting, in the seven examined countries. It is very important for meaningful alerting to become aware of these relationships and to take these into account when choosing appropriate alerting channels for each country. For example, if trust in authorities is very low in a particular country, it could be advisable to include otherwise trusted people for conveying the alert message (for example experts, well-known actors or respected sports people). Similarly, trust in media: we identified countries where public media are not considered trustworthy by those we have interviewed, they would, for example, prefer information via Internet or via other channels. Frequently, there is also a different perception of public and private media, which needs to be taken into account when releasing crisis communication. Another general rule in regard to the sender: We found evidence in all countries that, in case of a crisis, personal communication should be considered the key for alerting.

Not planned within the scope of the project, but nevertheless an important output of WP2 is the development of a sociological theoretical framework that we called ‘risk culture framework’. It is based on the empirical evidence that we gathered from the seven countries and distinguish between three ideal-typical risk cultures: individual-oriented (the Netherlands, partially France), state-oriented (Sweden, Germany and Austria) and fatalistic (Italy and Hungary). The typology is further elaborated and conclusions for crisis management and communication drawn, which should allow the development of more appropriate communication measures to enhance compliance of citizens in case of disaster. Influences of socio-cultural differences in risk perception on alerting strategies (WP3)

Activities in this work package comprised an analysis of workflows in disaster alerting, an analysis of the potential of a regionalization of alert thresholds based upon socio-economic factors and disaster experience, development of a strategy how to address such differences in an alerting system, and and overview on approaches how to model human behavior with regard to socio-cultural differences. The results of these works are presented in the following.
Workflows in Disaster Alerting
During the analysis of workflows, different alerting systems from four countries and their workflows were analyzed and compared:
• a severe weather warning system in Austria (UBIMET)
• an alerting system for earthquakes in Italy (Regione Sicilia)
• early warning systems for flash floods in Italy and floods in France (THALES)
• an early warning system for major pollution accidents (air, water, radioactive accidents) and major fires disasters in the Netherlands and France (THALES)

Deviation of the the region to be monitored into several subregions
The delineation into alert zones differed and is described as follows:
◦ The countries, which are served by UBIMET's severe weather warning systems, are divided into so called nature zones.
◦ These regions share common physical, climatological and/or geomorphological characteristics. Moreover, they are typically defined by their exposure to a certain meteorological threat as well.
◦ The flood warning system is based on the spatial distribution of the basins of the main river systems in France.
◦ There are 22 different regions which represent the basins of these main rivers. Each one is administrated by a so called SPC (Service de Prévision des Crues), which are part of the nationwide French National Hydro-meteorological and Flood Forecasting Centre (SCHAPI).
• Regione Sicilia
◦ The Italian civil protection department identifies and defines the warning areas for entire Italy.
◦ On the basis of various data like geomorphical data, amount of precipitation or temperature distribution maps homogenous territorial warning areas for sicily have been identified.

Generation of warning messages
◦ The analysis and diagnosis of the current state of the atmosphere and the prediction of future severe weather events is done by specially qualified and/or trained meteorologists in a rotating shift 24 hours a day.
◦ The chief meteorologist assigns forecast and warning tasks to several meteorologists who are subsequently responsible for a certain warning area or warning type.
◦ The warning area may encompass several countries or just parts of one country, depending on the general workload.
◦ After a thorough analysis of all available data and depending on the forecasters assessment of the situation, weather warnings of one or several warning levels may be issued for several communities, nature zones or even large parts of a country.
◦ The early warnings and/or warnings are normally issued with a lead-time of at least 6-12 hours, though in some cases (e.g. Freezing rain, local thunderstorms) lead times of about 20-50 minutes are more common.
◦ The decision to trigger the alert belongs to the Prefects.
◦ To decide, they take in account:
▪ the forecasts elaborated by the SPCs
▪ the thresholds specified in the RIC (Règlement d’Information sur les Crues) or Flood Information Regulations
◦ The forecasts issued by the SPCs are based on:
▪ the expertise of the agents
▪ the use of the forecast models best suited to the type of the considered watercourse
▪ the adaptation of the chosen models to the territory dealt with
• Regione Siciliana
◦ Tasks performed at a national level
▪ Seismic activities are monitored by the Italian Department of Civil Protection
▪ Development of hazard scenarios:
• The affected area is described (anthropogenic aspects, boundaries, geography, seismic monitoring networks)
• possible hazards
• Vulnerability
• preliminary assessment of damages and losses
▪ Definition of event scenarios
• There are four levels of intensity of the event defined
◦ Tasks performed at a regional level
▪ Definition of event scenarios
◦ Tasks performed on a local level
▪ Development of civil protection plans for the specific regions
• characterization of the behavior of earthquakes in the area
• danger analyzes in the area
• acquisition of data about the buildings in the area
• definition of risk scenarios
• development of an intervention plan

Possibilities of individual severe weather alerts depending on geographical location of recipients
An important aspect to be taken into account in crisis communication is the familiarity of the population with certain types of disasters and crises. This significantly affects the perception of the threat posed by a situation and the need for information and advice. While 10 cm of snow may cause chaos in Southern Europe, it would hardly be noticed as exceptional in mountainous Alpine regions or in Northern Scandinavia. As a result, the thresholds used by authorities to define what actually constitutes a crisis or disaster have to be localized. This issue was analyzed exemplary for Austria.

Within the project a telephone survey was conducted in March 2012 by the insurance company UNIQA. An overall number of 501 persons were interviewed. The interviewees were rather evenly distributed over the densely populated regions of Austria. Thus, fewer people were interviewed in rural and in mountainous areas than in urban regions like Vienna, Salzburg, Innsbruck, Linz, or Graz. The interviewees were linked to postal codes in order to associate the telephone survey findings with geographic information about terrain, land cover, and river network. The results of the telephone survey were analyzed in relation to the number of matching answers per zip code, which were imported in the GIS as comma separated files (CSV). In the following, the risk perception of interviewees to each severe weather parameter is described.

Along with floods, hail is the most feared severe weather phenomenon in Austria. Interviewees confirmed several times their fear of hail as a danger to their belongings and even to their lives. Hail can cause heavy damage to field crops especially in agricultural regions. Because of the importance of viniculture in some regions of eastern and southern Austria, and also because of the growing wine tourism, hailstorms can harm the local economies of wine regions. Thus, even non-winegrowers are affected in those regions. Consequently, the results of the telephone survey clearly showed a higher risk perception of hail in wine-growing regions than in other regions.
Snowfall in Austria is common during the winter season. Thus, most Austrians don´t see snow as a threat to their lives or belongings. This was clearly reflected by the telephone survey, which resulted in only three persons mentioning snow as a danger. Nevertheless, there is certainly the need to define different warning thresholds. The snow warning thresholds should be generally lower in urban and flat areas than in alpine regions, where people are used to higher amounts of snow.

Due to an extensive drainage system and the already mentioned high precipitation alongside the Alpine main ridge, Austria shows a higher risk for annual flooding than countries with fewer mountains and a more continental climate like, for instance, Hungary. Due to the high capacity of floods to cause damage, the interviewees of the telephone survey showed more sensitivity to floods than to snow.

Resulting from the telephone survey, the adaption of warning thresholds to geographic regions with different levels of awareness of severe weather events and different potential risks could be demonstrated for different meteorological parameters.
The potential risk of flood events increases with the pure amount of rain, as with the time period in which the rain falls. Different thresholds for heavy rain depend on the time period in which the precipitation occurs. Another criteria for determining different regions for heavy rain alerts could be the neighborhood of bigger rivers and water bodies. Main rivers are monitored by automated water gauges. Potential flooding occurs after a certain time period of precipitation. Municipalities along mountain torrents with smaller drainage basins are more affected by sudden flooding called flash floods. Commonly classified as dangerous mountain torrents are often monitored by visual warning systems. However, advance warnings by weather services are important in order to alert the potentially affected municipalities.
Regarding heavy snowfalls, it is useful to define three thresholds which can be used in different regions that are predefined by GIS-analyses. The lowest threshold is defined for urban regions, which are particularly sensitive to snowfalls. The high volume of traffic in urban regions leads to delays and traffic jams especially when the first snow falls. Particularly in Vienna, private auto drivers are unprepared for snowfall that causes delays in private and
public transport and in the transportation of goods. In mountainous areas, schools sometimes stay closed after heavy snowfalls, for example in January 2014 in Carinthia after nearly two meters of fresh snow. Nevertheless, the ability of people to cope with snowfall differs particularly between mountainous and flat regions of Austria. Although both Vienna and Innsbruck are urban regions, the residents and infrastructure operators of Innsbruck are used to snow while the Viennese infrastructure is not adapted to high amounts of snow. Concluding, three different thresholds could be suggested:
• Threshold 1: more than 5 cm. of fresh snow within 24 hours for urban areas and the surrounding suburban regions below 600 m.
• Threshold 2: more than 15 cm. of fresh snow within 24 hours for regions beyond urban areas below 600 m. and urban and suburban areas above 600 m.
• Threshold 3: more than 25 cm. of fresh snow within 24 hours for all regions above 600 m.
Additionally, the thresholds for warnings can be customized separately for regions with very high annual snowfall, such as the Arlberg region.
Wind speed must reach a threshold between 70 and 80 kilometers per hour to be regarded as the event of a storm. The warning thresholds depend on the geographic location and the ability of the vegetation to cope with high wind speeds. The risk of wind damage for forest stands is a result of tree and stand characteristics as well as of the average and gust wind speed and wind direction in the region. In regions with high average wind speeds, the vegetation can acclimate to it by increasing the allocation of assimilates to the roots and stem taper in order to increase their mechanical resistance. As a consequence, the same wind speed can lead to more severe damages in areas where the forest is not adapted to high wind speeds in comparison with areas where strong winds are common. The inner-alpine regions and the Bohemian Forest (Muehlviertel and Waldviertel), located north of the Danube and in Carinthia, are less adapted to high wind speeds and have a higher risk of wind damage than the southern parts of Styria and the Burgenland. This fact could be accommodated by lower thresholds for storm in the first mentioned regions and higher thresholds in the latter regions or in regions with low forest coverage in general.
Hailstones with diameters that are greater than 0.5 cm are worth warnings in all regions, since these hailstones can cause damages to vehicles, windows, greenhouses, etc. and can hurt people. In agricultural areas hailstones of even smaller diameters can cause heavy damage. Since hail is always related to thunderstorms and predicting the size of hailstones is very difficult, there is no direct possibility of adapting thresholds for hail events. Nevertheless, warnings for thunderstorms differentiate between levels of severity. The arrival of a thunderstorm with the possibility of hail should immediately lead to warning of the highest level in agricultural areas.
Heat warnings are of high interest when defining urban areas by using the freely available CORINE land cover product (CORINE Land Cover 2006). The CORINE land cover product (CLC) contains its own land cover class for densely built-up, continuous, urban areas, called continuous urban fabric. By adding 500 meters of buffer around those urban areas, the regions to be eventually warned first can be isolated. Additionally, CORINE provides a class “water bodies,” which contains the main rivers and lakes. Buffering those areas, it is possible to delineate areas where a cooling factor can be expected. Those GIS-based delineated regions allow the warning personnel to decide to warn earlier or later at the same temperature forecasts coming from the meteorological models. With 4 to 12 km., the spatial resolution of these models is too coarse to address Urban Heat Islands or cooling effects in urban areas.
Nights with temperatures above 20 degrees Celsius are commonly called tropical nights. Tropical nights are considered particularly stressful for elderly or physically infirm persons. Thus, tropical nights should in all cases lead to alerts in Central Europe.
In winter icy temperatures can occur quite often in Austria, particularly in inner-alpine mountainous areas and the Bohemian Forest in the north of Upper and Lower Austria. Since the people living in those regions are more used to cold temperatures than people living in the outer-alpine regions, the thresholds for freeze warnings differ. Resulting from normal diurnal and seasonal temperature variations, the warning thresholds also differ between day and night and between the seasons.
Black ice and freezing ice on streets justify an immediate warning in all regions. A regionalization is not possible because of the high danger, which occurs in such situations for the traffic participants.
GIS-based maps proved not only to be very helpful in analysing risk perception but also to support meteorologists in determining the appropriate warning threshold for different areas. In a Geographic Information System, the areas with different alerting thresholds can be determined and published as vector or raster data. This data can be included as an additional layer in special weather visualization software suites, like Visual Weather, as overlays to support decisions on thresholds for various regions. GIS overlay analyses are especially powerful when different factors play crucial roles in estimating the impact that severe weather can have in specific areas. Taking the example of flooding, it is useful to know where rivers, industrial zones, densely populated areas, or transport infrastructures intersect.
Strategies for the implementation of socio-cultural and geographical differences in the alerting system
The strategy for the implementation of socio-cultural and geographical differences in alerting systems is based on two main principals:
1. adapting warning levels and thresholds as a basis for automation strategies and as support for the alerting personnel,
2. adapting safety advices for the identified socio-cultural and regional alert zones.
Principle 1 was already explained in section 2 (see above). Principle 2 was developed by Partner SIC for the use case of Messina. The warning message should be written in a plan text and it should clearly explain the consequences of the expected hazard indicating the level of risk for each territorial unit involved, the message should be easily understandable from common people as well as from local authorities making possible by this way the effective adoption of those measures provided in the hazard prevention plans.
In Italy the local authority, which is represented by the Mayor, has the task of organizing the municipal structure with the support of all components of the System of Civil Protection in the existing territory. The local civil protection authority has also the obligation and the responsibility to alert people over any existing hazard. Each municipality has to design its own hazard prevention plan, which should include the design of a set of protocols and messages on how to warn citizens quickly and effectively.
The Opti-Alert presentation producer will be integrated in the warning system owned by the municipality of Messina.
Technical approaches on how to simulate human behavior including socio-cultural aspects
In order to analyze technical approaches on how to simulate human behavior including socio-cultural aspects, Thales evaluated and described more than 50 software suites. Evaluations were based on two criteria: the aspect of simulation model and another one for socio-cultural. From those evaluations, we have described different technical approaches used by them. This document has been used to create the architecture of the tools, which were developed within his project.
The algorithms and conceptual design are described in deliverables, which were issued within the project. The main point is: our framework allows to build distributed multi-agent simulation systems dedicated to virtual infrastructure simulation. Its architecture is mainly inspired from the work in “Pedestrians” on simulation of crowd in virtual infrastructure. Mass Media as a Channel for Alerting the Public (WP4)
Media studies within the Opti-Alert project focused on an analysis of editorial processes in times of disasters and crises (in 4 countries), on a content analysis of mass media reporting during an on-going crisis (a swine flu pandemic), on an analysis of media use in crisis situations, and on the elaboration of recommendations for media communication in alert situations.
Strategic alert behaviour through the mass media
Basically, there are great similarities between the media’s function in the four countries surveyed (Germany, France, Italy and Sweden). One difference though is that the authorities in Italy don’t have any agreements with national broadcast media about distributing acute alerting messages to the public in the case of sudden crises.
The authorities in all four countries regard the mass media as the main channels for public information, especially when sudden crises occur. At slowly developing crises some authorities express distrust in media reporting, and wish for channels where they can communicate directly to the public. One such channel, which has gained importance, is the authorities’ websites. Crucial for their effectiveness is that people know of them, and that the authorities succeed in maintaining public confidence.

Previous experiences with mass media in alerting situations
Based on previous experiences the Department of Civil Protection in Sicily identifies mass media as the main channel for information and communication with the population, both in case of preventive activities and during acute crises.
Preventive activities: Customized information addressed to people living in areas at risk serves to implement knowledge about preventive activities and build confidence in authorities. The information should be straightforward, clear and in simple language.
Acute crisis management: Media tend to demand immediate information; even when reliable data are not available. The real effects of an event only appear eventually, so communication must be organised in a progressive way. All information, and communication between authorities and media, should be coordinated trough communication centres in order to minimize the risk of incorrect or misleading information. The information given to the mass media should strive to be uniformed, fixed and accurate. Mass media should be kept informed at regular intervals, also using SMS or e-mail sent by the communication centre.

News reporting when crises or catastrophes occur. An overview of relevant editorial processes in French and Swedish mass media
Both previous research and interviews with journalists and editorial staff from major mass media in Sweden and France confirm that a) news production and news evaluation are similar in most types of media, and in both countries, and b) that when crisis occur, news organisations follow standard procedures, but work quicker and more intensely. The first questions addressed by media when crisis occurs are What has happened, and Who is affected. Their first priority is to gather information, also by using social media as Twitter and Facebook, and to publish as quickly as possible. Authorities are considered as important sources, but usually late in delivering information. Reliable information and confirmed facts are requested to avoid speculations.
Different traditions and legislation influence the relations between the public service media and the political authorities in the two countries. In France newsrooms in public service media are by law (Loi no. 86-1067 du Sep. 1986) imposed to work with the authorities if crises occurs. E.g. authority representatives may control when information is published. Sweden does not have similar regulations but according to a long-time agreement the Swedish authorities have the right to use radio and television for important public announcements. However, they may not interfere with newsrooms’ decisions of publication.

Report on mass media usage patterns
The report is based on focus group discussions and analyses conducted by SINE in Austria, France, Germany, Hungary, Italy and Sweden. The two main conclusions of the report is a) that new media (SMS, Internet) are becoming increasingly important for crisis communication without rendering classic channels redundant, and b) that a multi-channel approach seems best suited to maximize efficiency in all countries.
Specific conclusions on information behaviour:
Mass media such as radio, TV and, with certain limitations regarding age, mobile phones and the Internet are broadly used in all analysed countries for everyday information. In Austria and Germany the TV Teletext system is also quite commonly used, and in France and Hungary local TV channels are regarded as additional important news sources.
When actively seeking information, e.g. in a crisis situation, media habits tend to change and people mainly turn to media that they trust will provide reliable information. In Italy, where thrust in traditional media is poor, people will rather use the Internet for background information. By contrast, in Austria, Germany and Sweden, where trust in media generally is high, the Internet, and especially social networks, are considered to be less reliable. The attitudes towards the Internet are thus ambivalent. On the one hand online newspapers and authorities’ websites are trusted, and false information can easily be identified by cross-checking with other sources. On the other hand, people realise that websites can be manipulated and that the Internet, especially social networks, may be used for disinformation, a view especially common for Austria, Germany and France.
Specific conclusions on information expectations:
For acute alerting the public prefer media that are proactive and accessible, i.e. media that call upon instant attention and that are available to everybody. Classical media as sirens and loudspeakers are rated positively in all countries. Also SMS-alerting is increasingly well-known and supported in most countries with the exception of elderly people. Radio is mentioned in this context throughout all age groups. In Hungary and Sweden social media are considered an efficient way of alerting by several participants. The Swedish example indicates that social networks could become more relevant in the future as an effective multiplier within groups of people.
Specific conclusions on special preferences/needs:
Factors such as age, migration background and size of risk area clearly influence information behaviour and expectations. New media are mainly used by younger people. Migration background may cause language problems. In rural areas personal communication with friends and neighbours is of higher importance than in urban areas.

Pros and cons of mass mediated and personalized communication
Within a workshop, pros and cons of mass mediated and personalised crisis communication were discussed. The main conclusions reached were that:
1. Alerting personalized messages must be considered in relation to mass mediated messages.
2. Personalized alerting messages should/must:
- address people in a personal tone,
- not be contradictory to messages mediated through mass media,
- be sent to key persons who can instruct/guide affected people on what to do,
- be consistent,
- contain operational information, that instructs people on what to do, and inform them on how they are affected.
3. Mass media:
- reach very large and sometimes undefined groups of population,
- are most useful for “one size fits all”-messages,
- must be used according to known media habits of population,
- reach different strata of population at different times of day,
- can be slow, and therefore not always optimal for alerting messages (e.i. print media),
- authorities’ own websites is the fastest way to publish information since they are under full control of the authorities themselves.
4. Authorities should/must:
- establish themselves as opinion leaders when crisis occur,
- establish their websites online, become common known point of information,
- monitor social media (like Twitter and Facebook) to keep updated on public opinion and debate,
- use social media for informing (parts of) the public.

Analysis of mass media content and its impact in the case of the swine flu pandemic in 2009
This study consisted of two parts; I) A comparative quantitative analysis of news reporting on the swine flu in Italian and Swedish news media in 2009, and II) A qualitative analysis of media content on the swine flu in France in 2009.

Results from the comparative quantitative study from Sweden and Italy
Both Italy and Sweden decided on mass vaccination programs, however of different magnitudes. In Italy the vaccination goal was set to 40 per cent of the population, with a main focus on risk groups, while in Sweden was set to 95 per cent. Both countries failed to reach their goals. The vaccination rate in Italy ended at 1,47 per cent, and in Sweden at 60 per cent.
The content analysis revealed substantial differences in how Italian and Swedish news media reported on the swine flu pandemic.
In Italy there are close ties between media and politics, and journalism is by tradition partisan with a strong political bias in the coverage of current affairs. Both the responsible authorities and the media thus treated the swine flu pandemic as a political issue. The main focus of the reporting was on topics regarding Infected and dead people and the Public preparedness, measures taken. The most frequently appearing actors in the news were Government politicians and Healthcare authorities, quite often expressing contrasting views on the risk evaluation and the measures to be taken. The predominant reassuring tone of the reporting was mainly set by politicians, and the overall framing was that the situation was not alarming and under the government’s control.
The general Italian political culture has a parochial orientation, and citizens have little confidence in official authorities, especially on a national level. Also, in this kind of conflict oriented political culture, citizens are sharply divided, often on both the legitimacy of the regime and the best solutions to major problems. These are probably the best explanations of why so few Italians were vaccinated: the information about the situation, as reflected in the news reporting, was inconsistent and characterised by political logic, and it did not offer reliable guidance. Also, the sceptical approach of citizens to decisions and recommendations deriving from official authorities contributed to the general opinion of vaccinations not being necessary.
In contrast to Italy, Swedish journalism is characterised by a high degree of professionalization, including strong public service orientation and lack of party affiliations. It is strongly market oriented, which leads to an emphasis on conflicts, personalisation, sensation and drama, especially in tabloids and on television. The swine flu pandemic was handled by national administrative authorities, which are basically independent of political control and have extensive juridical responsibilities. The decision of mass vaccinations was communicated in unambiguous messages by all authorities involved in the handling of the pandemic, a communication strategy very much consistent with the consensual political culture of Swedish society. This was also reflected in the media reporting. Authorities' representatives and experts were by far the most common actors in the news, and the dominant topic was Vaccine/vaccinations. Hardly any opposing or conflicting opinions were reported. The limited critique voiced came from the journalists themselves and aimed at the delayed delivery of vaccines, the poor capacity of healthcare centres, and the order in which different groups were vaccinated, not the decision od mass vaccinations. The media took sides with the ordinary citizen who wanted to be vaccinated, and it was the journalists that mainly set the predominant alarming tone. However, the overall frame of the reporting was that despite the seriousness of the situation it was still under the authorities' control.
In a consensual political culture, citizens tend to agree on the appropriate means of making political decisions and on the major problems facing society and how to solve them. This is, as in Sweden, accompanied by a generally high confidence in authorities, and trust in the instructions and recommendations given. This is one explanation of why 60 per cent of the Swedes were vaccinated. Another is that the vaccination campaign never became a (controversial) political issue. The polemics raised, were the expressions of a media logic that assigns the control over how news is selected and presented to journalists.
The overall findings of the study are that the following three conditions are critical to the implementation of effective crisis communication:
Crisis management organisation
Authorities handling a crisis have the power to frame it, i.e. to define it as dangerous or not dangerous, as something that concerns everybody or only a few. They also have the power to decide what measures to take and how to inform people about them. Authorities may act in different ways, depending on constitutional, political, social and cultural traditions, degree of responsibility, and on explicit as well as underlying political and administrative interests.
When crisis management is the responsibility of politicians it will normally be embedded in a political context, and treated as a political issue with potential impact on public opinion. In contrast, when crisis management is the responsibility of administrators or experts it will normally be embedded in a factual context, and be treated as an issue of factual and experience-based knowledge.
Political culture
The political culture of a country influences whether the political and administrative systems will strive to find a consensus solution that is best for everybody, even if it involves a certain degree of compromise, or if they will deal with the crisis in a polarised and conflict-orientated way, where different solutions are confronted and only one can be the best.
Crisis management requires consensus on crisis management to produce adequate output. When there is consensus the probability of consistent crisis communication and of appropriate actions being taken increases. In contrast, when there is conflict on how to handle a crisis communication is likely to be inconsistent, and uncertainty about which actions to take will follow in consequence.
Political culture also includes trust in authorities. When trust is high the citizens greet the authorities with confidence and are likely to follow their instructions. When it is low, they greet the authorities with scepticism and are not likely to follow their instructions.
Journalism culture
When journalism is characterised by political logic media content is politicised in the sense that every issue is evaluated by political standards, and the reporting is polarised in terms of support or criticism of opposing political interest. The media are politically biased and the journalists are mediators of political messages that are formulated outside the media. Crises are likely to be described and interpreted in ways that favour or disfavour the responsible political actors. The citizens’ trust in the media is related to their political positioning.
When journalism is characterised by media logic the content is commercialised, in the sense that it strives to affect and attract the audience, and to be profitable. The reporting is based on objectivity and impartiality, and news is selected and presented on the basis of professional criteria. The media are likely to exploit all the alarming and sensational aspects of the crisis, and to accentuate conflicts, in order to attract the public’s attention. The citizen’s trust in the media is related to journalistic quality and the degree of tabloidisation of news.
Results from the qualitative study in France
When the swine flu pandemic threat became a reality the French government reacted rapidly by implementing ad hoc structures such as the Crisis Interministerial Cells (Cellules Interministérielles de Crise) and daily press briefings at the Health Ministry. This structure was implemented during spring and summer 2009 in order to prepare for the pandemic peak in the autumn.
Nevertheless, although the government followed a strategic crisis plan, it failed for several reasons to communicate the need for vaccinations to the public. Firstly, when introducing the vaccine to the physicians and health care workers, the responsible ministry underlined the controversial aspects of the vaccination, rather than its benefits, which forwarded medical professionals to question the suitability of the vaccine. Secondly, the government’s decision not to include the general practitioners in the vaccination campaign was ill received by both the practitioners and the general public. Both these controversies were reported in the news media.
Also, the government failed to make use of Internet as a channel of communication, the second source of health information among the French after health professionals. Messages discussing the relevance and quality of the vaccine, through a medium allowing information to be broadcast quickly, were not countered by targeted, educational, and convincing official information, answering each criticism in detail.
Moreover, this lack of information, and the increasing visibility of the messages that opposed the standard concepts, created a new attitude concerning the medical choice: a personal understanding of the health risk, based on individual evaluations of benefits versus risks of the recommended treatment (vaccination).
In conclusion, it is of paramount importance that in the future the French government’s communication strategy builds on targeted, well-argued, and convincing messages. Predictably, the H1N1 pandemic increased opposition to vaccination. Indeed, the rate of flu vaccination among the population in 2010 fell back to the level of 2006.
Specification document on “best practice” altering procedures
We summarize the results of the deliverables in the Work packages to the following 17 specifications of “best practice” alerting procedures:
1. Install an official administration point of contact for the media where they can as for
2. Provide unambiguous non-partisan (non political oriented) and consistent information to the
media to fulfil their needs
3. Provide accreditations to journalist to have access to specific crisis centres
4. Provide behavioural advice or guidelines
5. Give the adequate priority to the communication management
6. Select credible authority to communicate on crisis
7. Implement adequate official dedicated web sites to transfer information
8. Favour exchange with the web surfers rather than one-way information
9. Use professional communicators as facilitators
10. Provide information to media before they ask for it
11. Do not forget local media which are closer to people and more aware of the specifies of a
12. Use traditional media as they are the most favoured by a majority of the population: radio, TV,
newspapers (and their web sites)
13. Take in account some specific media use, linked to the concerned area (for instance, use of
14. Use new technologies as they provide interesting capabilities (area selection through mobile
phones disaster alert apps, RDS capability for radio listeners, specific crisis channel, social
media but with cautiousness)
15. Broadcast alert messages in several languages adapted to the specificities of the local
population (including tourists)
16. Where the legislation entitles it, use national public service media to broadcast alert messages
17. Send a message making official the end of the crisis Personalized, adaptive multi-channel alerting (WP5)

Within the work package 5 an alerting system (Opti-Alert system) has been designed and developed. In the first phase we designed the system. We defined interfaces and models. In the second phase, using the prepared documentation, we implemented the warning system. This system has some special properties. It can warn via different communication channels (using warning systems pro-vided by the project partners). This alerting system can also personalize the messages. This means, it can send different messages for different recipient groups. In following subchapters the components of this system and the definition of interfaces are presented briefly. The developed warning system has been tested within field tests. In following subchapters also the results of these field tests are presented briefly.
Personalized and regionalized alerting strategies
A catalogue has been prepared with an overview of typical characteristics of different target groups in all countries which have been analyzed (Sweden, the Netherlands, Germany, France, Italy, Austria and Hungary). The catalogue also includes design recommendations for alert messages. The idea of the alert message design catalogue is to allow software developers to convert specificities of and recommendations for different target groups into machine-readable format so that they can be used for automated or semi-automated processing.
We developed additionally a (xml-based) language for definition of personalization rules, that addresses the following two issues:
• Which socio-cultural specific target groups should be warned?
• How should the message for socio-cultural specific target groups look like?
Opti-Alert System
The Opti-Alert system consists of the following components:
• Repository: the repository is a central component of the Opti-Alert system. Each message that should be send via the Opti-Alert system has to be uploaded to therepository. The repository provides access to messages via a specially defined interface.
• Presentation Producer: the presentation producer is an GUI application that gives users the possibiliy to create messages and upload them to the repository. The presentation producer supports two modes. It can generate fully personalized messages (manually) and can generate draft message. Draft messages are messages scheduled for further processing (like automated personalization). Drafts are sent to repository (same as the fully personalized messages) but are marked as not-ready messages. Two presentation producers have been developed or adapted to suit Opti-Alert, one for general alerts from e*Message, and another one speficially for severe weather alerts from UBIMET.
• Personalizer: the personalizer is a GUI application that gives users the possibility to personalize messages (the messages that needs to be personalized are the drafts; they are loaded by the personalizer from the repository). The personalizer creates a proposal of the personalization. The user can accept the proposal and send it to the repository as ready to send, fully personalized message, or can modify the proposition. The propositions are generated by the personalizer on basis of predefined rules. Initially we developed the personalizer as an application without GUI that personalized the draft fully automated. Later we decided that this approach may be to dangerous and we decided to give the users control over the personalization process.
• Requestors: the requestor is a software component that each project partner (that shares a warning system with Opti-Alert project) had to develop. Requestors load messages from the repository (it sends a request to the repository in short time intervals), adapts the message and forwards the message to the corresponding warning system.
• Adapted warning systems: the systems that have been adapted are:
o e*Warn (a warning system provided by e*Message)
o a warning system provided by Fraunhofer Institute
o a warning system provided by Regione Sicilia
o a warning system provided by UBIMET
Thanks to adapted warning systems, Opti-Alert system can warn via following communication channels (easy integration of more channels is possible):
• SMS (e*Message’s, UBIMET’s and Fraunhofer’s systems)
• VoIP (e*Message’s system)
• E-Mail (Fraunhofer’s system)
• E*Warn siren (e*Message’s system)
• Android application (Regione Sicilia’s system)
• Public information website (e*Message’s system, UBIMET’s system)
The following section describes the interface used for accessing the Opti-Alert repository.
Interface Definition
Within the the project work, we also developed a standard interface. We used this interface to integrate the Opti-Alert system. It is based upon the Common Alerting Protocol (CAP). The „Common Alerting Protocol Opti-Alert Profile“ (CAP OAP) extends the specification of CAP but remains compatible with the pure CAP protocol.
The main two problems that this profile solves are:
• messages in our system are catalogue-based.
• messages in our system are personalized (contain a part that is scheduled for all and parts that are scheduled for socio-cultural specific target groups).
The most important parts of CAP OAP (that are not contained in pure CAP protocol) are:
• text-identifiers are transfered within a message. The test-indetifiers corresponds to entries in the catalogue. The catalogue contains texts like categories (of incident), instructions, hazards. The identifiers of these texts are transfered within the message in parameter’s field (with special parameter name). The text remains in the message at proper position (in event / instruction / description field). Thanks to this, the message remains CAP compatible.
• audience must be one of values from a predefined table.
• It is possible to distinguish the base-message (message scheduled for all) from the target-group specific messages. The base-message has a predefined value of audience field. According to the CAP OAP specification the target-group specific messages should be considered as extensions of the base-message. That means they are using all values from the base-message and additionally values that are contained in the target-group specific message (for example additional instruction or hazard). The target-group specific messages can also limit the warning area (if the target-group specific message does not define an area, the warning area from base-message should be used).
This interface was used while building the Opti-Alert system. The repository uses CAP OAP as input (presentation producer, personalizer) as well as output (requestors).
Field Test
A field test of the Opti-Alert system has been executed. The presentation producer (and other components) have been tested by 4 project partners (e*Message, Fraunhofer, UBIMET, Regione Sicilia). Within this test each of participants collected a group of volunteers and sent them some warnings.
Field test participants described then their experiences with the presentation producer. The more complex results of these field tests are contained in the report of work package 8. As part of the work package 5 only the presentation producer (and cooperation work with other components) has been evaluated.
The feedback about the presentation producer and the idea of personalization of messages was positive except some minor points:
• Initially it was not possible to indicate date, time and duration for which the alert was valid. Especially for severe weather warnings, that are issued in advance, this is crucial (we added this feature during the field test).
• The catalogue used in the presentation producer did not cover all requirements of field test participants. For example, the prefined weather messages don‘t cover the whole spectrum of severe weather situations.
The results of this work package can be used in the future to build a marketable product (or can be used as a part of a marktable product). The Opti-Alert system implements some new ideas that area not common in other warning systems (e.g. detailed personalizations of warning messages and support during this process).
As the results of the tests (experience of field test participiants) shows, that the ideas tested within this work package can considerably improve a warning system. The results of the tests can also help to design better warnings systems in the future. Alert Impact Simulators (WP6)
Within the simulation work package, two software tools were developed to simulate the impact of alerts. The first one was used to show how passengers in buildings of critical infrastructure react to alerts issued via different communication media. The second simulator analyses how quickly communication tools and devices alert the general population and how alert messages spread among ordinary people.
Alerting inside critical infrastructures
The simulator is agent based. The agents simulate different types of passengers. As an example, a French train station was modelled. The different types of travellers were modelled as follows:
- French passenger using the underground. They represent around 15% of the global population.
- French passenger using the train. They represent around 80% of the global population.
- Foreign passenger using the underground. They represent around 3% of the global population.
- Foreign passenger using train. They represent around 2% of the global population.
Foreign passengers are assumed not to understand the local language properly.
Agents have the following character traits:
- Stamina: use to manage life points.
- Vision: use to manage blind people.
- Audition: use to manage deaf people.
- Civism: use to manage cheating.
- Courage: use to manage for example: capability to use fire extinguisher in case of fire.
Agents have an inventory:
- Valid tickets: use to pass ticket barrier.
- Money: use to buy tickets or food.
They have motivations:
- Care need: to manage amount of damage to stamina.
- Security need: to start fleeing when level is high.
- Food need: to manage hungry.
- Leave station (exit or subway) or take a train
Only French passengers, in a very few amount, were modelled to have deaf or blind character trait. Communication media were also modelled. In particular, electronic billboards, loudspeakers inside the train station, and, randomly, passenger receiveing text messages (e.g. alerts).The following figure shows different types of agents in front of an electronic billboard:

Figure 1: Different types of agents in front of an electronic billboard

In the figure, deaf agents are indicated with a blue sign with an ear, foreign tourists with colourful flags, and blind agents with the blue sign and the white, crossed eye.
In an alert simulation, all these agents show different behavioural patterns (e.g. they only evacuate when helped, or they take more time to understand). The simulator was successfully tested on a model of the Gare du Nord Station in Paris, and can handle simulations with several thousands of actors.
Simulation of alert propagation in the general population
To figure out how many people can be reached in a warning area, a simulation algorithm was developed which is adaptable for different scenarios. The simulation algorithm is designed in a general way. It can be used for every type of communication channel, however, it requires and depends on statistical and dynamic data which has to be available or estimable. The warning channels are not predefined; they have to be entered for each region. Each channel has a different alerting efficiency. An interesting question is therefore how efficient the different systems can be and on what factors they depend. The general efficiency for each channel is based on two steps. First, how many people can be reached by official warning, which means through a siren, personal message or other notification from an official organization. Second, how many people are reached by unofficial warning through multipliers that distribute the message. Thereby each step is influenced by individual factors like technical properties or situational circumstances that are presented in the following. As last part the distribution over the time is analyzed.
Official warning channels
Official warning channels are all communication ways that are triggered by an official authority and try to warn the public or registered users. In the following this group of people is also called “directly informed”, because these are the persons warned first and directly from the officials. The number of people that could be alarmed depends on the maximum number of people a channel can reach and the number of people listening depending on situational factors like the time of day and their location.
The maximum number alarmed depends on the device used because every one has a different reach. Three categories of devices can be discerned:
- Devices issuing point alerts, that is, alerts for one specific place, like a building. One example for this type of alert channels is the personal announcement system in a school, which allows the headmaster to talk to all persons who are currently in the school.
- Devices that alert people within a certain range, such as sirens. Sirens can alert almost everybody in the area around (except people with a hearing disability). Depending on the siren model used, a typical alerting range is between 600 m and 1000 m.
- The last type of alerting devices are personal devices that are not fixed to a certain location, such as mobile phones or smartphones which receive alerts via SMS or apps. Such alerting devices typically require prior registration of the recipients (due to data protection law issues).
The theoretical maximum number of people who are reachable (the “reach”) depends on the availability and the technical characteristics of a device, but not everyone who could be reached really receives and listens to the alarm. Therefore, additional factors are taken into account – in particular, the wake-up effect of the channel, the amount of users typically registered to the channel (if registration is required), and the attention towards the channel. These factors are based on situational characteristics, such as the time of the day.
Unofficial dissemination of warnings
Initial feedback obtained from an online survey on personalized alerting indicates that a majority of respondents would inform their family and friends about an alert, thus creating a substantial multiplication effect. These multipliers are unofficial warning channels that distribute the alert message to others. The number of newly informed persons depends on the number of people who have received the official information and would distribute it but also the amount of still uninformed people.
Whether people would distribute the warning they have received depends on their willingness to do so and their location. The location is important because there are different ways of multiplication. There is a difference if the people are at home or outside on the way. So from the daytime it is possible to determine where the people are and whom they can additionally warn. The number of possibly warned people depends on the average number of people around the multipliers who may be in the warning area. It can be assumed that multipliers would first warn the people in their immediate environment. For example household members are very likely to be also affected by the potential disaster for which an alert has been issued. In addition the warning messages are also distributed to people in the neighborhood and to friends and relatives who may also be affected by the disaster. To assess the number of people contacted, regional statistics are used, such as population numbers and household sizes. The number of neighbors to be informed is estimated based upon settlement patterns (rural settlement, major city) and population density. Based upon the directly informed people and the just mentioned parameters it is possible to estimate the number of people informed by multiplication effects.
Share of population noticing the alert over time
For modeling purposes, it is assumed that the multiplication patterns are similar for all channels once the recipient has noticed the alert. There will not be multiple notifications and each channel is considered separately. It is assumed that after informing a first group of people, the message spreads so that saturation approaches 100% over time. Previous research indicates that alert message diffusion follows an S-shaped pattern, due to the fact that only few people notice the alert initially, but soon after dissemination gains speed due to multiplication effects before approaching alert saturation. This exponential growth pattern is assumed as the basis for the simulation. The dissemination function is calibrated based on intitial effects of official and unofficial warnings.
Technical implementation of prototype
A prototype that uses the presented algorithm was developed and implemented for the cities of Hamburg and Messina. The simulator can flexibly react to different situations and can process dynamic data. A client-server-architecture is used to implement the proto-type, where the location of the warning area can be defined over a web interface and then sent to the server. On the server, there is a database with all the statistical data about the area, population density and whereabouts of persons, but also facts about the connected warning channels. The modular architecture applied makes it easy to change and add warning channels or get more detailed information about the location. It’s also possible to have different client systems using the simulator at the same time. For the communication a standardized CAP object is used and the calculated results are sent back within an SVG file so that they can be displayed to the operator. In Figure 1 the output of the simulator is shown for a flooding alert in the German city of Hamburg. The graph on the left shows the distribution of the warning message over 60 minutes. In the chosen area three channels were available and for each one the ratio of successfully alerted people is shown. The curve for “public sirens” already starts approaching saturation after one hour, showing the S-shaped dissemination pattern. On the right, input data is shown like the size of the warning area, the number of inhabitants or the time of the day.

Figure 2: Output of the simulator (left) the graph and (right) additional information, both for an alert in Hamburg Data Protection and Privacy (WP7)
System development activities also included the development of privacy concepts in order to make sure that the system is compliant with legal requirements.
Two different system architectures following the privacy-by-design principle were considered, to identify a solution which has on the one hand a low privacy impact but on the other hand guarentees effective and robust communication in crisis situations. The two approaches discussed were (i) a server-centric system architecture and (ii) a client-centric system design, which are influenced by data storage location, degree of centralization (e.g. of message creation and adaptation components), system scalability, network communication and data security. These will be described in more detail below.
Server-Centric Design
Although a more client-centric solution has benefits regarding data handling, it is evident that a server-centric system is an effective solution for targeted alerting. Assumed that the user has been subscribed within the distribution system and personal data required for sociocultural message adaptation has been received, the system can, in case of an emergency or hazard scenario, instantaneously prepare and dispatch personalized messages to the user. The communication channel used for message transmission, which can be Short-Message-Service (SMS), email or a push notification service, depends on the individual user preferences. Two cases have to be differentiated here:
1. If the user has selected a communication medium that supports the maximum number of characters required for a detailed, personalized alert message, such as email, then the full message content will be sent directly and no further client-server interactions are necessary.
2. If the user has selected a communication type that is restricted to a specific number of characters (e.g. SMS or push notification), supporting no or only restricted personalization, then a brief notification will be sent to the user with the instruction to open the alert application for detailed information. In this case another request will be sent from the client to the distribution system to obtain the full message content.
It has to be noted that the distribution system has a requestor component which polls for new CAP alert messages from the message repository in short intervals. If a new message is received it will be mirrored on the distribution system for processing purposes. In the next step, performed by the personalization component, the alert location geometry will be matched against the location of the subscribed users (e.g. the postcode area or the geographic coordinates relating to the subscriber’s current location) and analyzed for intersections. If an intersection is detected, an alert message will be prepared, consisting of a general base message for the location of the user and a personalized message extension with more details for the respective audience group, to which the user belongs. The different audience specific message extensions are represented individually by message content elements (i.e. “info” element) of the CAP-message. Finally the dispatcher component determines the preferred communication channel and sends the alert messages to users accordingly.
The personal data to be stored and processed encompasses information about the person’s age group, nationality, language, home country, potential disabilities, phone number, email-address and postal code.

Figure 3: Server-centric alert dissemination

Client-Centric System Design
In the client-centric approach, except communication information (device-id, phone number and email-address) no personal data will be stored centrally. Instead, the user profile will be established on the client device as part of the alert application. The process is structured in a way similar to the previous solution with the main difference that the client decides about the required personalization parameters. Upon retrieval of the alert information from the repository, the distribution system informs the client about the new alert. The user subsequently has to activate the alert application and a request, including individual profile parameters, for a personalized alert message will be submitted to the distribution system. In a last step the message will be presented to the user in text format, and in cases of visual disability via voice output. The communication between client and distribution system is realized as a trusted channel to ensure confidentiality and protection of transmitted data.
In a variant of this approach, the message components may be replaced by text component codes, which are defined in a catalogue stored within a relational database. This would not decrease the number of request-response cycles between client device and distribution system, however it would reduce the data to be transmitted to a set of message codes and some base information, such as time and location of the event. Advantages would be the reduction of the network traffic and a reduced duration of the client-server communication. In crisis situations where the capacities of mobile networks can be exhausted quickly, this could be of decisive importance. In addition the scalability of the distribution system would profit, as the system could manage more connections due to the reduced communication times.

Figure 4: Client-centric alert dissemination

Implementation and Privacy Impact

For the prototypical implementation, which was also applied as part of the field test, the server-centric architecture was chosen, as a client-centric approach would have implied high development efforts for a client application, which was not in scope of the original project planning. As mentioned the server-focused architecture requires centralized storage of personal information, which was listed in the respective section. Hereby data sparsity was applied to data processing and storage for privacy impact reduction. A privacy impact assessment revealed that some of the abovementioned information were assigned a high identification potential value. These include email address and mobile phone number, as this information must be regarded as unique data. However for sending alert message via SMS or email this information is imperative. Some information was further reduced such as the a person’s age to a age group, reducing the privacy impact to some degree. Information with a medium identification potential include city and postcode, information with a low potential the person’s language, disability group and parental status. Although all personal information is stored as transformed encrypted information accessible only after passing authentication and authorisation steps, there still is some overall privacy impact of the system. Therefore, informed consent of all alert recipients has to be obtained before they are included. Demonstration (WP8)
This work package joins results of many work packages (WP3, WP5, WP6). The result of this work package is a system that not only can send personalized warnings via multiple channels but also supports the users in the composition of the warning (/ alert) in many ways.
Another important result of this work package are the results of field tests. The tests were done in 4 different areas using different communication channels.
In following subchapters we will briefly present the result of field tests and the demonstrator. The demonstrator is a system that integrates all software components developed within different other work packages. The components that were integrated are:
• Presentation producer
• Simulator (indoor & area)
• Geographical Information System (prepared by UBIMET)

The demonstrator is an extension of the Opti-Alert system that has been developed within the work package 5. The demonstrator is additionally integrated with simulator and (regionally) with the geographical information system. The simulator and geographical information system are implemented only as a proof of concept. They contain only small amount of data (proper simulations for areas within predefined area for that data is present are possible and geographical information system contains only data for Austria about area height and rivers). In order to be useful, these systems will have to be completed by adding more data. The simulator and geographical information system have been integrated as part of the presentation producer. The presentation producer makes it possible to define areas. Now while the user is defining a new warning area, he has a possibility to display the additional information provided by the geographical information system on the map. While composing the warning, the user has a possibility to generate a simulation of the distribution of this warning.
Field Test
4 project partners took part in the field tests. Each of them organized and executed his own field test. The project partners that were involved in tests are: EMESS, FHSS, SIC, UBIMET. Figure 5 Tests organization summarizes all tests that have been executed.

Figure 5 Tests organization

The following 4 tests have been executed:
• Field test organized in a more rural German area: this field test has been executed in Lippe (Lippe county is a region in the heart of Germany at the North-Eastern end of the state of Northrhine-Westphalia). The communications channels used were SMS (220 registered participants), Phone-Call (VoIP) (54 registered participants), E-mail (117 registered participants). 93 of participants selected multiple channels (In total 248 persons registered in this test). 3 warnings have been issued: one warning at 11:30 AM, one warning at 6:30 PM, one warning at 2:00 AM. Feedback was collected via SMS / E-Mail (each participant had to send an SMS / E-Mail as soon as he noticed the warning). The effectiveness of each communication channel has been measured.
• Field test organized in a major German city: this field test was executed in Berlin (capital city of Germany). The communications channels used were SMS, Phone-Call (VoIP), e*Warn siren (a provided by e*Message tool for population that uses the e*Message network). 3 warnings have been issued: one warning at 11:30 AM, one warning at 6:30 PM, one warning at 2:00 AM. 51 volunteers took part in the test. Feedback was collected via a dedicated web site. The web site contained few basic questions about the received warnings (like when did they notice the message, what was the first device that they noticed, how many persons could they inform about the warning in their neighborhood). The effectiveness of each communication channel and the multiplication effect have been measured.
• Field test organized in an Alpine region in Austria: this field test was executed in Eastern Tyrol (Austria), in and around the town of Lienz. The communications channel used was SMS. 10 warnings have been issued. 7 within first phase of tests and 3 during the second phase of tests. The second phase has been executed in order to make the UBIMETs tests comparable with tests results of other project partners. Following 3 warnings have been issued within the second tests phase: one warning at 11:00 AM, one warning at 6:30 PM, one warning at 2:00 AM. The feedback from 11 test participants in first field test and 9 test participants in the second field test has been collected. Participants were asked by email to fill out a questionnaire table. Because of the relatively small amount of test participants a more detailed feedback could be collected from each participant: the multiplication effect has been measured, the usefulness of instructions contained in warnings has been evaluated with regard to the gender / age / profession of test participants.
• Field test organized on a Mediterranean island: the field test was executed in Itala, Scaletta Zanclea, Giampilieri and Catania (Sicily). 44 participants took part in the test. The warning messages have been issued similar as in other tests around: 6:30 PM, 11:30 AM and 2:00 AM. The focus of the test was flash floods. Test participants were asked to fill a questionnaire. The effectiveness of each communication channel has been measured. SIC also evaluated the usefulness of the instructions contained in warnings.
From the results of the field tests we could derive many interesting information about the usefulness of our system and about the different communication channels that we used.
In following list we present the field tests results for each warning channel:
• SMS is a powerful warning channel but the disadvantage of it is that is has a very limited wake-up effect (what means that SMS is not a good warning channel for warning in night). The big advantage of this channel is that everyone has his mobile-phone almost everywhere (and can read SMS’, or have a reading tool read the message aloud).
• Phone call has a better wake-up effect then SMS but is also not a perfect tool for warning at night (because it can be muted). The advantage of the phone call is that the response is in most cases almost immediate (immediately when the phone is calling people notice the warning).
• e*Warn siren has (what was proved in field test) the best wake-up effect (what means that it is the best warning channel for alarms in night). The disadvantage of the e*Warn siren is that it is in home (people would not take it with them). Therefore it is not the best communication channel for warnings during work-time.
• E-Mail has the worst wake-up effect and it takes relatively long time until all people warned by e-mail will notice the warning, especially out of office hours. The biggest advantage of E-Mail as warning channel is that E-Mail can contain a lot of information (additional links, more complex description of hazards / instructions).
The field tests results provide us with a lot of information about the warning channels. These data can be used to develop a better warning system that would better satisfy the requirements.
The demonstrator (that is an extension of the Opti-Alert system presented in WP5) could also be used in the future as a marketable product (as a basis for marketable product or as part of a product).

Potential Impact:
1.1.4 Socio-economic impact and societal implications

Impact on employment

In total, 34 people worked on the project during its duration. Of these 34 people, 6 worked on positions that had been explicitly created for the project. The work amount was close to 30.4 person years (calculated on Full Time Employment). This figure does not include additional effects created by subcontracts and / or additional sales generated by project partners due to the project work.

Potential commercial exploitation

During the project, a number of software tools was either newly developed or significantly enhanced. These tools wil be commercially marketable in the short and medium term. These tools include:

- Simulator for behaviour of human beings during alerts in critical infrastructures (this product will enter the market shortly, there is already substantial customer interest)
- Improved alerting device: the e*Warn siren alerting device was developed further during the project and will also be offered on the market. During field tests, the device showed a very high wake-up effect, this now proven effect will be an additional argument for marketing.
- Improved front-end for alerting: the personalization feature of the front end could add a competitive advantage to this component, which increases the market potential of this tool.
- The repository for the integration of different alert system components could also be marketed, but economic impact is expected to be limited for political reasons (in most countries, alerting is organized locally, with limited budget and limited interest in integrated solutions).
- The knowledge gained on crisis communication and alerting infrastructures will be used by several partners (Fraunhofer and SINE in particular) to offer consulting services, workshops and presentations in this field.
- Existing alerting apps could be enhanced with visualizations of expected disaster impacts, so that alert recipients gain a better understanding of the problems they are going to face.
- Improved websites and apps that visualize the expected effects of upcoming disasters (requirement from end-user studies)

Scientific impact of the project

Although it is somewhat early to assess the scientific impact of the project, several results were well-received in the scientific community. Among those are:

- Results from sociological studies on information needs, media use and trust during alerting situations (with data from 7 countries)
- Results from field tests on the dissemination speed of alert messages. The benefit here was that the data is based on realistic alerting exercises.
- Results from media science studies on media behaviour with regard to different types of alerts, and on how media act in different crisis situations. Here again, the strong plus of the studies is that they are based on a substantial amount of real-world data.
- The approach to simulate the dissemination of alerts in the population extends previous approaches to new communication media, and is very flexibly configurable.

The agent-based simulator for behaviour in buildings also follows a very solid and interesting approach (due to confidentiality reasons, however, the dissemination of the exact algorithms etc. to the scientific community will not be possible).

Interaction with end-users

Project results created substantial interest among end users. As part of the dissemination activities (see below), results of WP 2 were widely distributed to end-users and stakeholders at the national level (notably the Nationaler Paging Kongress 2014 with almost 200 attendees from civil protection and disaster management throughout Germany) as well as international level (notably the TIEMS Conference 2013 and the International Disaster and Risk Conference 2012, but also the US Department for Homeland Security 2014). Subsequent to comprehensive newspaper reportage in May 2013 in the Süddeutsche Zeitung (the largest nationwide newspaper in Germany) SINE received a request from the Federal Ministry of Interior in Germany, Department of Crisis Management and Civil Protection, who showed great interest in the Opti-Alert findings on socio-cultural factors of risk and crisis communication. After receiving consent from the Commission Services, SINE has forwarded material explanatory material and deliverables to the Ministry.

Additionally, Kerstin Dressel from SINE was also asked by the Director of the State Office for Occupational Safety, Health Protection and Technical Safety Berlin to give an oral presentation on Opti-Alert WP2 findings at the 15th Civil Protection Day in Berlin in September 2014. Other requests for study output included the Federal Institute of Risk Assessment in Berlin, Department of Risk Communication.

The Federal German Institute for Risk Assessment also approached partner Fraunhofer showing interest in the technical concepts for the integrated alerting infrastructure that had been developed by technical partners in the project. The institute was handed over a scientific paper on this matter.

Lippe county and its disaster management agency did not only participate in the field tests, but used results from the tests in their effort to convince other counties in using state-of-the art multi-channel alerting systems.

The University of Gothenburg disseminated results from the project to end-users in Sweden, namely to MSB, the Swedish Civil Contingencies agency. Regione Sicilia communicated results to the national civil protection agency in Italy.

Our Dutch partner NIFV organized a roundtable with end-users from the Netherlands (mainly firefighters).

The high interest from end-users, stakeholders and policy-makers on results from the study on socio-cultural factors of risk and crisis communication clearly indicates that there is a huge need to know more about the recipients of alerting and crisis communication: how do they perceive the alerting, what does they know about the civil protection structures in their countries and how do people (likely) behave and how they can be reached via traditional and / or new media channels. In all oral presentations and workshops, where we have presented results of the study, it became obvious that there is not much known from the sender side about “the public” and that there is in fact nothing like “the public”, but instead the need to distinguish between different target groups, hence to perform a more personalized and regionalized alerting and crisis communication.

We are therefore planning to work upon the findings acquired in Opti-Alert and to develop these further. In order to do so, we are currently working on a proposal for submission to the Horizon 2020 call within security research: ‘Better understanding the links between culture, risk perception and disaster management’ (coordinated by the University of Gothenburg, with participation of the University of Perugia, Fraunhofer Fokus and the sine-Institute plus several end-users). The whole project idea is based on the findings produced in WP2 and, particularly, the ‘risk culture framework’.

Interaction with the general public

Interest from media has been quite high, in particular in Germany. Highlights of dissemination activities towards the general public include:

- one article and one inverview in one of Germany’s most respected newspapers (Süddeutsche Zeitung, which sells approx. 420,000 copies and reaches roughly 1.4 million people)
- one article in Hungarian weekly magazine “HVG”, which is also well respected and sells about 73,000 copies
- two contributions from the project to professional magazines
- a radio interview with the German science radio (DRadio Wissen)

Several project partners jointly wrote a book on “Current Issues in Crisis Communication and Alerting”, which has been published by Fraunhofer Verlag in May 2014. The book was written in a way so that it also addresses practitioners and not only the scientific community.

The Opti-Alert website attracted substantial traffic. The set target of 5000 visitors was already exceeded by far in the first reporting period.

Knowledge gained in the project was also transferred into higher education with 2 seminars (one held at the FOM University of Applied Sciences in Berlin, and one at the University of Gothenburg), 3 theses coached by project members, and several lectures (again, at the University of Gothenburg).

Interaction with the scientific community

Another highlight was a conference jointly organized with TIEMS, the International Emergency Management Society. This conference attracted visitors from 5 continents (in total, more than 70 participants) and facilitated a lively exchange on alerting and the exchange via social media in crisis situations. Furthermore, the conference served as a platform for exchange with a large number of multinational research projects (mainly from Europe and the US). At the conference, the Opti-Alert project contributed with a system demonstration, 2 keynotes, one presentation and a poster. Furthermore, a project representative participated in a panel discussion on standardization in alerting.

The project also produced several reviewed contributions, thereof one journal paper (plus several more accepted and currently in the publication process), and 4 reviewed conference papers. Several additional non-reviewed presentations and posters complete the picture. Due to the fact that about half of the material is still in the publication process, we will update the publication list regularly on the Opti-Alert website.

Overall, 56 dissemination items were conducted during the project. Key activities include the organization of a scientific conference, publications in journals, conference papers and professional magazines, round table discussions, the preparation of seminars and educational material, and the publication of a book on key project results.