Systematic reviews have become an increasingly popular research method in relation to climate change. In such a review a study pool is created based on a well-defined search protocol and a set of inclusion and exclusion criteria (Boland et al., 2013; Ford et al., 2011; Groulx et al., 2017; Lumbroso et al., 2017; Petticrew and Roberts, 2005). The metadata searches for this systematic review were conducted in two large and multidisciplinary academic databases: Scopus and Web of Science. This review emulates the technique used by Groulx et al. (2017), as their analysis of the different learning outcomes of citizen science projects is similar in approach to our aim of investigating communication outcomes of different two-way communication formats. The process followed in the systematic review is outlined in Fig. 1.

Our systematic review was limited to formats that allow two-way communication, where there are two “communicators” who issue and receive information (Bowater and Yeoman, 2012). Two-way communication can take many forms, but for the purpose of this review we accepted anything from face-to-face dialogues, to communication composed through posts and comments in social media. These formats include social media, educational programmes, discussion forums (collaborative encounters where experts work closely with affected communities; Whatmore and Landström, 2011), serious games (those which exceed mere entertainment purposes and also intend to educate; Abt, 2002), and citizen science projects, which can be described as “collaborative research that involves members of the general public (or citizens), and which actively involves them collecting, generating, and analysing data” (Illingworth and Allen, 2016, pp. 5–12). This list of formats was informed by an initial literature review of the field (Amri et al., 2017; Aubert et al., 2015; Fohringer et al., 2015; Gravina et al., 2017; Mani et al., 2016; McCormick, 2012; Pennington et al., 2015; Salvati et al., 2016; Whatmore and Landström, 2011), and in discussion with stakeholders in urban risk management, namely the Association of Greater Manchester Authorities (AGMA). The initial literature review also revealed that the search strategy had to account for terminological synonyms and alternative denominations. For example, “flood” is often used interchangeably with “inundation”. Finally, the literature review elucidated that other terms had to be considered together with “communication”, such as engagement, participation, knowledge, and education. Search keywords were linked using the Boolean operators AND/OR, and search was programmed to retrieve articles containing these search terms in the title, abstract, and keyword sections.

Not all the articles retrieved in these searches were relevant and thus a set of seven inclusion criteria was designed (Fig. 1, box 4):

Criterion 1 (two-way communication) was designed to include only articles addressing two-way communication. For example, articles dealing with flyers and leaflets in mailboxes as educational propaganda were retrieved under the term “educational programmes”, but these formats do not permit interaction between experts and non-experts.

The articles were coded for qualitative data analysis using the NVivo software. All the articles were assigned four independent variables: country, hazard (flooding or air pollution), format of communication (social media, serious games, educational programmes, citizen science, or forums), and academic database (Scopus or Web of Science). The coding exercise was developed in two cycles. Firstly, the data were analysed through a method called “descriptive coding”, which allows the attribution of a label that describes the basic theme of a paragraph. Secondly, the data were later re-coded using a method called “pattern coding” (Saldana, 2009), finding relationships between codes and grouping data into more meaningful units. The coding exercise took place first with the articles retrieved from Scopus, and the articles in Web of Science were used to prove data saturation (Bryman, 2012).

All the articles analysed in the systematic review can be found in Table 1.

Social media can be used for short-term communication in the case of an imminent or ongoing incident, such as flooding (e.g. Bunce et al., 2012) or air pollution crisis (e.g. Zhang et al., 2014). Most examples which involve emergency communication come from flooding and only relatively few from air pollution. Most often, air pollution is an ongoing problem and thus social media is used with a long-term focus (e.g. Fedorenko and Sun, 2016).

The articles show that when social media is used for short-term communication, it can be done so with three different aims. (1) It can be used to share warnings and information in relation to an incident. For example, during the 2011 Queensland (Australia) floods, the Queensland Police Service used their Facebook page as the main channel for emergency communication, allowing citizens to “access, post and share information about road closures, flood peaks”, etc. (Bunce et al., 2012, p. 37). Similarly, during the flooding in Thailand in 2011, social media was used for sharing information and advice on how to behave and what actions to take either from other more experienced citizens or from official sources: “victims shared the photos of their homes and the flood, and the knowledgeable ones uploaded their advice and analysis about the situation” (Leong et al., 2015). (2) Social media can also be employed by experts (e.g. government officials or response teams) to communicate with affected citizens and neighbours and collect on-the-ground information regarding the situation during and after a flooding incident, which can help provide a more accurate response to the situation (Rizza and Pereira, 2014; Yadav and Rahman, 2016). Social media is also used by affected populations to share emotions, and consequently by government officials to observe citizens' physical and psychological well-being during a crisis. During the 2013 Southeast Asian Haze, for example, social media allowed the acquisition of information from the general population in this regard, which is key in determining how citizens are coping (Zhang et al., 2014). (3) Social media has also proven to be useful for coordinating response to a crisis. For example, it can be used to ask for donations, provide help, or to gather supplies such as temporary accommodation or food (Wan Hussin et al., 2016).

Social media can also be used with a long-term focus, in order to raise awareness about key issues (Roshandel Arbatani et al., 2016). The air pollution movement in China (2011–2012), for instance, used social media for “illustrating the size and impacts of particulate matter in accessible ways” (Fedorenko and Sun, 2016). Social media is also a valuable medium to encourage attitudinal and behavioural change, for example, through sharing adaptation strategies and measures that citizens can adopt to tackle air pollution (Kay et al., 2015). Alternatively, social media can function as an effective platform to encourage debate between experts and community members. In the aftermath of the 2011 Thai flood crisis, social media was used to “share compassionate stories, obtain emotional and physical support from their peers” which helped in the recovery process (Kaewkitipong et al., 2016). Finally, social media can also be employed to campaign in relation to an environmental injustice or problem, fostering collective action, group identity, and a sense of belonging to a community (Xu, 2014). This last usage of social media is especially relevant for countries, such as China, where public gatherings of any political nature are prohibited.

Social media facilitates an important three-way process. In social media, downstream approaches from experts to the public coexist with “horizontal interactions” between citizens (Fedorenko and Sun, 2016), alongside an upstream approach where citizens take the lead: “the power previously contained in the hands of government agencies shifts to the people” (Leong et al., 2015, p. 193). In the face of this democratisation in the communication process, different studies highlight the importance of having a person or a team dedicated to managing social media communications, before, during, and after a crisis. Constant interaction as well as tirelessly responding to questions and comments is key to counteracting misinformation and rumours with real and valid information (Wan Hussin et al., 2016; Xu, 2014). The fact that anyone can post information on social media, and that there is no quality control over what gets posted, can be a limitation inherent to this medium. People receiving incorrect information on how to behave during a flood, or people taking advantage of flood incidents by failing to pass on donations to victims are two examples of such a limitation (Wan Hussin et al., 2016). On a more practical basis, different studies highlight the importance of using hashtags to help channel the discussion and the communication efforts to relevant people or to interested parties (St. Denis et al., 2014). Additionally, social media appears to be a medium for concise and brief information. Twitter in particular has a character limit on tweets and it is commonplace to find messages containing links to external resources and more detailed information (St. Denis et al., 2014).

If the systematic review found that social media is suitable for short- and long-term communication, then educational programmes are used solely with a long-term focus, namely, to raise awareness or promote a certain behaviour. An educational programme implemented in Arizona, for example, was designed to help home-based and hard-to-reach businesses (e.g. a beauty salon or a printers) in becoming green. The aim was to help them transition towards less-toxic substances, as well as to encourage these professionals to reuse and recycle materials (Moreno Ramírez et al., 2015). In the Netherlands, an educational programme had the aim of improving the understanding of hydrological concepts, and enhancing preparedness intentions of 15-year-old students (Bosschaart et al., 2016; Demir, 2014).

Educational programmes appear to follow a downstream approach, meaning that although communication flows from expert to public and vice versa, it is the experts who initiate the communication process, that is, the ones to design, organize, and set up the educational programme. Different studies agree on the benefits of implementing these programmes with the help of opinion leaders, that is, figures who are respected, perceived as being knowledgeable, and who hold higher levels of trust than, for instance, government officials. Opinion leaders are usually close, or have access, to the community or segment of the population at which the programme is aimed. In the programme implemented in Arizona, female Hispanic community health workers (called “promotoras”) were invited to run the programme. They possess leadership skills that allow them to effectively influence their community (Moreno Ramírez et al., 2015). They are trusted because they have been trained in public health issues, and because they are “indigenous to the community” (Moreno Ramírez et al., 2015). Sometimes, the figure of the opinion leader is interchanged with that of a mediator. In cases where the programme is designed to be implemented in schools, for example, school teachers become mediators, whose job it is to guide the learning (e.g. Bosschaart et al., 2016), and the experts behind the programme design (e.g. scientists or local governments) remain as an available resource throughout the programme. It is commonplace that these intermediaries receive specific training on how to run the programme. In the case of the “promotoras”, for instance, they received expert training and necessary materials to provide real-world application for specific business needs that can translate into positive change (Moreno Ramírez et al., 2015).

Educational programmes are characterised as being highly interactive. This interactivity is sometimes achieved by incorporating virtual reality technology, allowing visualisations, simulations, and animations that allow people to observe, say, different flooding scenarios and their consequences in real time (Demir, 2014), which would be impossible in real life but are important for learning (e.g. Demir, 2014). In other examples, educational programmes include inquiry-based education where students are provided with equipment and training for air sampling, followed by modules addressing air pollution and health outcomes (Ward et al., 2016). Interactivity was also at the heart of another educational programme based in a school, in which pupils used old newspapers, toys, bottles, etc. from their homes to create an artwork and work with the concepts of reusing and recycling (Elnokaly et al., 2008). Another key feature of educational programmes is that often they employ real information, such as real-time flood conditions or inundation maps (Demir, 2014). Additionally, the systematic review shows that educational programmes need an element of fun, have to be easy to develop, and should offer the appropriate training when more technical knowledge is involved.

Just as with educational programmes, serious games are reserved for long-term communication, to increase awareness and understanding, especially around flooding. For example, the aim of the game “Stop Disasters” is to work with concepts of resilience and resistance (Felicio et al., 2014). In “FloodSim”, players implement a selection of strategies for addressing the risk of flooding based on a pre-defined budget. The game is designed to encourage players to think about what type of barriers to build, which regions to concentrate on, how much funding to allocate to maintenance, what warning systems to establish, etc. (Breuer et al., 2017).

Although serious games can be a vehicle for learning and communication in themselves (Felicio et al., 2014), they are sometimes accompanied by discussion and debate and played in classroom settings, where the teacher guides debate, answers questions, and explains concepts (Rodriguez Bermúdez et al., 2015). But the usage of serious games goes beyond schools and face-to-face workshops or events, where the debate is guided by experts (e.g. scientists). Sometimes, debate can also take place in a mediated environment, such as a blog or a wiki (Rodriguez Bermúdez et al., 2015). Although serious games allow two-way interactions between experts and non-experts, they tend to follow a top-down approach to communication. That is, experts are behind the game design and decide what content and information is included, which will, in turn, guide the concepts to be discussed with the non-experts. The United Nations, flood risk management professionals, and scientists in different universities are some of the people involved in designing serious games who may or may not be involved later in playing the game (e.g. Felicio et al., 2014; Savic et al., 2016). Some of the studies discuss the need for closer collaborations between game developers and knowledge partners so that games more closely fulfil the objective of raising awareness (Rebolledo-Mendez et al., 2009).

Different studies agree on the importance of having a binomial fun–knowledge combination. That is, the player has to be engaged, but acquiring new knowledge needs to be a requirement for success in the game: “the game can neither be a simple funny game without any learning, nor only involve difficult concepts without any incentive” (Rodriguez Bermúdez et al., 2015, p. 162). Including random features so that the flow of events can not be predicted or offering a diversity of scenarios are some of the strategies employed to keep the player engaged (Rodriguez Bermúdez et al., 2015). In order to make sure that the player is learning and making informed decisions, supporting information can be integrated into the game. For example, in eCity, players can get a short explanation of the advantages and disadvantages of possible moves (Rebolledo-Mendez et al., 2009). It seems as if educational programmes are designed to educate but need to have an element of fun, and serious games are fun but need to educate. Additionally, serious games appear to be an effective tool for improving understanding, as they offer continuous feedback to the player, avoiding the formation of misconceptions (Savic et al., 2016). They also encourage active and experiential learning, allowing players to handle datasets, modify values, and experience simulations which would be impossible or very expensive in real life (Breuer et al., 2017).

Citizen science is also used for long-term communication, but appears to have a broader usage than that of educational programmes and serious games. Citizen science is helpful in raising awareness, but is also employed to facilitate knowledge exchange between experts and the public. “AirProbe” is a project that aims to raise awareness about air pollution by involving citizens in measuring air pollution in their daily life. The AirProbe project shows that involving citizens in taking measurements can be very effective in producing a positive change (Sîrbu et al., 2015). When citizen science projects aim at raising awareness, they tend to follow a downstream approach, initiated by the experts.

Citizen science projects, however, are also employed to facilitate knowledge exchange between experts and the public, or to complement or challenge expert knowledge. For example, they can be useful in gathering knowledge that would be impossible for experts alone to collect in terms of amount and accessibility. In the case of the Deepwater Horizon oil spill, a citizen science project was developed to allow citizens contribute with information of what they saw and smell, such as “smells, smoke, and other potential risk factors”, which can in turn be used in decision-making and in complementing expert science (McCormick, 2012, p. 2). Furthermore, as experts and the public often work together in these projects, exchange knowledge, and discuss scientific data – as opposed to just receiving the outcomes generated by experts alone – these projects have become a means of improving communication and the levels of trust in each other (Ngo et al., 2017; Thiel et al., 2014).

Across studies we find that affected or concerned citizens, who suffer the consequences of flooding or air pollution first hand, are more inclined to participate in these projects. For example, a community that is affected by a flood has more at stake, and therefore, is more inclined to produce citizen scientists and share flood observations (Le Coz et al., 2016). Similarly, residents who live nearer facilities that emanate or contain environmental contaminants are more likely to perceive the benefits of participating in a citizen science project aimed at improving their situation (Jiao et al., 2015). Learning about the environment is also another motivation to join citizen science projects, which broadens the spectrum of participants from affected communities to society at large. Schools are sometimes also involved in these projects, integrating them in their science curriculum (e.g. Sîrbu et al., 2015). All segments of the population, however, are never equally empowered to participate. For example, projects involving technology, such as social media or Dropbox (Le Coz et al., 2016), will inevitably add to the digital divide.

There are two factors that appear to be essential in citizen science projects. First, most of the studies concur that the first stage of any of these projects should involve training, and finding common ground between all the participants. Citizen science projects have a greater chance of fulfilling the established aims, if participants understand the ultimate aim of the project, how they can contribute and gain the most, how to use the necessary tools appropriately (e.g. air pollution measuring devices), and who to address if they encounter setbacks or need clarifications. Citizen science seems to work better when there is a clear and common goal for every participant involved, when beneficial outcomes for all are set (Jiao et al., 2015). Learning about data collection, creating bonds with other community members, or simply having an enjoyable experience are some of the examples that exceed the most obvious strategic goals usually coupled to these projects (e.g. improving flood management). Second, most of the studies highlight the importance of feeding back the results to the group after the completion of the project. Participants must understand the overall impact of the project and how their contributions fit within it, as well as how the project, the data, and the results are going to be employed (Le Coz et al., 2016).

Discussion forums are used for long-term communication. They are the only medium solely used for knowledge exchange practices, and to bring together local and scientific knowledge. Slightly more examples of forums are found in relation to flooding than to air pollution. In Kenya, for example, a forum was organised by climate experts in order to seek traditional knowledge and build flood early warning systems: “local farmers are witnesses of their own environment and the first to notice changes and potential risks” and thus, “open dialog is necessary to build climate products that reflect farmers' needs in terms of warnings” (Senaratna et al., 2013, p. 11). Another air pollution forum, also in Kenya, explored how academics and local residents together could design action points to tackle the air quality issue (Ngo et al., 2017).

Forums can follow a downstream or upstream approach: they are sometimes initiated by experts who believe in the benefits of exploring the ground reality and local knowledge, or by the public who feel threatened and ignored, and consequently take action. Discussion forums are not always representative of the whole population, as participants usually belong to segments of the population that feel more empowered to participate (e.g. people with higher levels of education). Therefore, community members who participate in these forums are then in charge of passing on the information to the rest of the community. For example, if a forum deals with how to tackle flooding at the community level, the group is then in charge of preparing their community to face a possible flooding event, by organising meetings and developing training to raise awareness and prepare the community for evacuation and rescue plans (Senaratna et al., 2013).

Discussion forums are, however, a challenge, and on many occasions they begin with “palpable tension apparent” (Lo et al., 2013, p. 9). This tension can be attributed to a lack of trust in the experts, an incapacity to break down hierarchies of power and to encourage non-experts to contribute, or to the all too often technical language employed by the experts, which distances them from the public. Therefore, different studies deal with how to ease tensions. We have, consequently, identified four strategies for doing so. The first strategy involves experts and non-experts making something together, the benefits of which are two-fold: making something together works as a process, as a facilitatory medium taking the focus away from anything that generates tensions; but it also works as a product in order to engage the wider community when this is presented, for example, in an exhibition at the community centre, or is showcased in the community hall (Leon et al., 2015). The second strategy is found in one of the forums in the UK, which encourages participants to bring objects (e.g. maps, photos, satellite images, and even a piece of mouldy carpet) that show their connection to a flooding event, which works toward highlighting each member's connection to it (Whatmore and Landström, 2011). The third strategy is inspired by citizen science projects, through which the non-experts collect data that later guide the discussion and debate process. Asking participants to take personal exposures to particulate matter (PM), can be a starting point for the latter discussion and can aid the non-expert in building and supporting their arguments (Ngo et al., 2017). The last strategy consists of initiating the forums with an informative session. In one of the forums, for instance, specialists were invited to set the grounds for the debate, sharing information about the scientific, economic, and political implications of climate change and carbon tax. This was followed by an opportunity to respond to specific queries, and lastly participants were able to design a preferred carbon pricing policy (Lo et al., 2013).

A visual comparison of the results for all five communication formats can be found in Fig. 2.

Finally, these five formats are sometimes combined in order to facilitate their implementation and offer a more positive engagement. Sometimes educational programmes and serious games employ social media as a platform to enable discussion (e.g. Rodriguez Bermúdez et al., 2015). Citizen science initiatives also use social media, not only to facilitate discussion during the project, but also to recruit participants, answer questions, and feedback the results to the group (e.g. Le Coz et al., 2016). In some instances, citizen-science-style activities can be developed as part of educational programmes to improve learning (e.g. hands-on activities, place-based learning) (e.g. Ward et al., 2016). Similarly, citizen science can also be employed as a strategy to empower citizens in their discussion and collaboration with the experts (e.g. Ngo et al., 2017). The way these can be combined also points out that social media and citizen science projects are the two most versatile formats. A visual representation of the combination of formats can be found in Fig. 3.