Visual portrayals of fun in the sun in European news outlets misrepresent heatwave risks

The ways in which news media communicate about heatwaves can influence how society conceptualises and addresses heatwave risks. We examined visual news coverage of the 2019 heatwaves in France, Germany, the Netherlands and the UK, using content and visual critical discourse analyses. Many visuals were positively valenced (in contrast to article texts), framing heatwaves as ‘fun in the sun’. The most prevalent type of images in all countries were photographs of people having fun in or by water. When images did depict the danger of heat extremes, people were largely absent. We conclude that this visual framing of heatwaves is problematic: first, by displacing concerns of vulnerability, it marginalises the experiences of those vulnerable to heatwaves; and second, it excludes opportunities for imagining a more resilient future. We conclude with suggestions to diversify the visual discourse on climate change and heatwaves

management, such as in tornado warnings (Ash et al., 2014). But images can also lead to inaction: the dominance of the politicised and contested visual framing in climate news can leave people feeling powerless to act . Similarly, photographs of climate impacts commonly depict climate change in spatially and conceptually distant places (O'Neill, 2013), decreasing a sense of self-efficacy . Thus, images have considerable power in helping us think, sense and imagine climate change (Schneider & Nocke, 2014;Yusoff & Gabrys, 2011). Images can open up-or foreclose-possibilities, and become blueprints for future realities (Schneider & Nocke, 2014). What reality, which future, then, is encountered in the visual reporting of news about heatwaves and climate change?

| METHOD
Four countries (France, Germany, the Netherlands and the UK) were chosen as they were most impacted by the 2019 European heatwaves . Data were collected from online websites in each country, as they represent the second-most consulted source of information about climate change, after TV (Amdi, 2020). For each country, we analysed the coverage of five of the most visited online news sites (i.e., wide reach), for diversity of type ('up market', tabloid and broadcast) and for political orientation. Each country sample included at least one left-leaning and one right-leaning title, and at least one broadcast company (see Appendix S1, Painter et al., 2021). The sampling period ran from 1 June to 30 September 2019 and captured two periods of extreme hot weather in all countries: heatwaves in France, from 24 June to 7 July and from 21 to 27 July; in Germany, a heatwave from 4 to 26 July and the warmest June since records began; in the UK, heatwaves from 21 to 28 July and from 23 to 29 August; and in the Netherlands, heatwaves from 22 to 27 July and from 23 to 28 August.
Articles were selected for inclusion in the dataset by searching for the term 'heatwave' in UK media and its equivalents in German ('Hitzwelle'), French ('canicule', 'vague de chaleur') and Dutch ('hittegolf') media and collecting all news articles which were substantively about the 2019 heatwaves. Only articles where a link was made to climate change (even if briefly) were included for analysis. Narrowing the sample to these articles reflects the overall justification for this study: responsible visual discourse around extreme heat risks is highly relevant to the increasing impacts of climate change on heatwaves around the world. Articles were excluded where links were broken (as of October 2021). News articles where there was either no image or a video which played automatically (i.e., no static image on opening the webpage) were deleted. The first image on the page (also known as lead or hero image) was collected. This resulted in a dataset of n = 245 images and their associated articles.

| Thematic and valence coding
A reflexive and mixed-methods approach was followed to explore the visual framing of heatwaves in the media. Entman (1993, p. 52) defines framing as the process of 'selecting some aspects of a perceived reality and making them more salient in a communicating text, in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation'. As with critical discourse analysis, framing is concerned with power and the construction of meaning in texts (Van Dijk, 1993), but framing also retains a focus and close reading on particularly salient examples at the analysis stage.
A visual content analysis was designed to capture the image corpus at a broad level, examining the images for their denotative content. This provides an overview of a domain of visual representation. Then, an understanding of the connotative content of images (how the objects portrayed relate to culture, seeking to understand structures of meaning) as well as their ideological content (how the image is read, how intrinsic meanings come to exist) helped to place the images in their socio-political context (Dyer, 1982). Close reading of representative individual images or image groups elucidates these processes.
A codebook was drawn up by the first two authors qualitatively and inductively, with reference to a range of literature on visual (climate) communication (including Hayes & O'Neill, 2021;O'Neill, 2013;O'Neill et al., 2015;Rebich-Hespanha & Rice, 2016;Rose, 2016;Schneider & Nocke, 2017). Valence coding drew on Lang et al. (1990) and Lehman et al. (2019). Valence was used to describe the emotional quality of an image or text-whether it appeared negative (worrying, risky, dangerous, inconvenient) or positive (fun, enjoyable, holiday-like, relaxing) in the context of a news media story about a heatwave. Image valence coding also accounted for images with mixed or neither valence (both positive and negative elements; or no obvious valence signalling, respectively). The resulting themes and valence coding instructions were then refined with reference to an initial sample of images. A spreadsheet was created to record a unique article identifier, the media organisation, article URL, and coding of themes and sub-themes, and image and text valence. The codebook was shared with the wider team, and further iterative refinements were made to clarify meanings and delineate differences between the thematic codes, image valence and text valence; and to ensure the codebook worked across the socio-political contexts of each of the four countries.
Images were multiple-coded, that is, there was no hierarchy of image themes or sub-themes-visual cues which were indicative of a particular sub-code were coded as they arose across all images. Multiple-coding was a way of recognising that images can hold different (sometimes contrasting, sometimes reinforcing) meanings. Valence coding also aided in capturing contrasting meanings ('mixed' valence).
Once agreement was reached on the codebook, four coders (one per country) independently coded a random sample of 20 images and their associated article text as a check for inter-coder reliability. This was calculated as an assurance that visual and textual data collected across the four countries was coded consistently, so national-level results could be cross-compared in analysis (O'Connor & Joffe, 2020). We recognise the arguments around the epistemologically problematic nature of an inter-coder reliability for qualitative work; here, we use the measure of 'reliability' as proxy for 'trustworthiness' (Braun & Clarke, 2013). It is not that this research sought a positivist objective 'truthful' reading of images (recognising that images can have multiple readings; although noting that there does appear to be a dominant reading of climate images in western European nations; Metag et al., 2016, but that, as a team, there was confidence we were analysing the data using a common conceptual framework with reference to both the literature on climate and risk communication, as well as our own experiences and understandings (O'Connor & Joffe, 2020).
Using a distance function calculation that is appropriate for multi-label data, Krippendorff's alpha for thematic coding of the images was 0.74. Krippendorff's alpha for image valence was 0.78. Krippendorff's alpha for text valence was 1.00. These results were regarded as reaching a satisfactory level of inter-coder reliability. However, to further ensure consistency, the team met to work through any disagreements in the pilot image themes or valence, making final refinements to the coding scheme. For the full codebook, containing thematic and valencing coding descriptions, and example images and texts, see Appendix S2. Table 1 describes the overarching codes and sub-codes. Note that this coding schema is limited to the western European context examined here. The four country coders used the codebook to code their country sample. The lead author checked for consistency across each country dataset, with any queries discussed. This resulted in a small number of changes to coding.
At the analysis stage, the major themes and sub-themes of the coding schema remained largely consistent, reflecting the significant level of reflexivity, pre-planning, training and discussion there had been in preparing the codebook. Quantitative analyses undertaken included calculating percentage coverage for all codes/sub-codes and their valence.
The use of a data visualisation method via the application ImageJ through the macro ImagePlot (http://lab.softwarestudies.com/p/imageplot.html; see also Rose & Willis, 2019) allowed for a 'birds-eye' view of two key image properties: hue (i.e., red, orange, yellow or so on, or a combination of these) and brightness (i.e., how much light is radiated or reflected) across both the main visual themes. ImageJ has built-in functionality to measure colour hue (RGB values) and brightness (Grayscale average of RGB values) for each pixel. ImageJ calculates a median average for all pixels in the image, then the ImagePlot macro is used to plot images on a colour wheel according to these values, with brightness on the x-axis and colour hue on the y-axis.
Results from the visual content analysis and digital image visualisations were used alongside readings of individual images and image sub/theme groups, to answer the following research questions: 1. How are heatwaves visually framed in media articles about climate change and extreme heat? 2. Does the valence of textual and visual portrayals of heatwaves differ? 3. How do media portrayals of extreme heat reflect wider societal engagement on heatwaves specifically, and climate change more generally?

| RESULTS
Across all four countries, the most common codes were 'leisure activities in the sun', 'idea of heat', 'impact of heat' and 'landscape' images ( Figure 1). Where percentages are given, these refer to the percentage of coverage per country. Note that images could be multiple-coded, as visuals can depict multiple storylines or narratives (see Section 2), so the codes do not sum to 100%.

| All countries commonly visually represent heatwaves as 'fun in the sun'
All four countries, and especially the UK and the Netherlands, extensively depicted heatwaves through a visual repertoire of 'leisure activities in the sun' (Figure 2). This included leisure activities in or by water, or in outdoor spaces, as well as images of eating and drinking in the sunshine. Positively valenced images dominated, giving an overall visual depiction of 'fun in the sun'. Although there were some images in this thematic code with neither valence, there were very few images coded here with negative valencing (see also Figure 1). The ImagePlot analysis reveals how blue hues, depicting sunny skies and azure water, were common across this theme ( Figure 3). Positively valenced leisure activities in or by water were by far the most common sub-code, with images of swimming, sunbathing, splashing in fountains and so on accounting for 33% of UK, 27% of Dutch, 26% of French and 22% of German coverage. These were the sorts of images which used well-worn visual clichés and references similar to a 'holiday snap' (Rose, 2010; Figure 4). The images often depicted sociable groups of people, expressing happiness and enjoyment through their faces and body language: smiling and laughing, whilst hugging, playing, holding hands, even jumping with joy. The images were light-filled, featured blue skies and seas, and colourful leisure paraphernalia. As Paris experienced record-breaking temperatures, many people swam in the fountains in front of the Eiffel Tower -leading to repeats of this iconic visual motif in coverage across all four countries.

| Visuals depicting the 'idea of heat' were more common in German and French coverage
German and French coverage used visual tropes from the 'idea of heat' code far more than the UK and the Netherlands (German 68%, French 45%, Netherlands 32%, UK 28%; Figure 1). The 'idea of heat' included images depicting bright white sunbursts, and intense heat through 'dangerous' (Schneider & Nocke, 2017)  ( Figure 5). Thermometers depicting unusually high readings were also common (Table 1, Figure 6). A small amount of wildfire imagery was also present (0-7% across the four countries). People were far less prevalent in this theme. When people were pictured, they were mostly de-personalised; in shadow, silhouetted against the sun, with faces rarely visible.

| A dissonance between texts and images
There is a dissonance between the valence of the newspaper texts and images. The valence of accompanying text across all countries was rarely positively valenced (<1%). Yet, almost a third of images were positively valenced across the four countries (31%), with considerable variation between countries (Figure 7). The images used to fill the bars of Figure 4 are discussed further as exemplars (see Appendix S3 for uncropped images).
In the Netherlands, a headline 'Red Cross: Municipalities must prepare vulnerable people for heat' in the news website NU.nl, accompanies a holiday snap image of people paddling at the beach (Figure 4, Netherlands bar). Perhaps some of those who are vulnerable can make it to the beach to deal with extreme heat, but it is surely not a viable adaptation strategy for many others at risk. In Germany, a headline in the tabloid newspaper Bild, 'Hottest June ever-worldwide' accompanies a detailed article about the danger of climate change and heatwaves, including even an infographic depicting how climate change will threaten iconic species like polar bears and places like the Great Barrier Reef. Yet the lead image accompanying it is of a holiday snap-a man jumping from a diving board into a swimming pool (Figure 4, Germany bar). The French and UK examples present the most striking contrasts between text and image valence. On the website of French media organisation 20 minutes, the headline and sub-heading ran: 'Heatwave: What do the four levels of the national plan provide? HEAT The national heat wave plan was created after 2003 and the tragic summer during which 15,000 people died due to very high temperatures'. It accompanied a picture of children and young people playing in a city splash park (Figure 4, France bar). In the UK, an attractive, smiling young woman leans back into a city fountain, her top displaying the double entendre 'hotter than wasabi' (Figure 4, UK bar). It accompanies a Daily Mail article headline which begins: 'Hotter than hell! "Vicious" Saharan Bubble heatwave roasting Europe kills three swimmers'.

F I G U R E 1
Major thematic codes by valence, by country. Positively valenced counts are dark coloured, negatively valenced are light coloured, mixed valence are stippled and neither valence are hashed. Y-axis is percentage (%) of images by code and valence. Note that images could be multiple-coded.

| Nascent climate solutions journalism
Historically, 'climate solutions' (broadly defined as positive stories of opportunities, adaptation or solutions) have received much less attention than those focusing on negative impacts, despite the potential of solutions journalism for increasing climate engagement (Doulton & Brown, 2009;Hulme, 2009;O'Neill et al., 2015;Painter, 2013). One study showed for example that positive adaptation stories only represented 2% of climate reporting (Moser & Boykoff, 2013). In the present study, climate solutions imagery was absent throughout almost the entire dataset, except for one outlet-the Netherlands local and regional news organisation Algermeen Dagblad (AD). AD used visuals coded under 'health and wellbeing' or 'other' far more than other outlets (see Appendix S4 for links to images). For example, although AD used the visual cue for hot weather of people eating ice-creams, the people depicted were elderly residents inside a care home, rather than young families queuing at an ice-cream van on a benign sunny day. When AD depicted a young family, they were not escaping the heatwave with a holiday to the beach, but were pictured in an everyday situation at home-sitting in front of a fan, looking visibly uncomfortable and fed up with the heat. Further examples included an air-conditioned town hall which had been made available for vulnerable people to visit in order to escape the heat; and an urban greening project designed to reduce the urban heat island effect.

| CONCLUSION
The results of this research have important implications for visual climate communication. First, the media visual discourse of the 2019 heatwave in France, Germany, the Netherlands and the UK was problematic; and especially as we F I G U R E 2 All images across the four countries in the 'leisure activities in the sun' code, clustered by sub-code ('leisure activities in or by water', 'leisure activities in an outdoor, public space', 'eating and drinking in the sunshine'). Note how sub-code 'activities in or by water' (top group of images) dominates this code. only included heatwave articles which made mention of climate change. Across all countries, it was common to visually represent heatwaves as 'fun in the sun'. Much coverage featured images of people having fun in the sun at beaches, pools or city fountains. Although heatwaves may offer enjoyment for some, when heatwaves are imagined as high days and holidays, as events to be welcomed and eagerly anticipated, they are displaced from uncomfortable everyday realities of how heatwaves manifest at home, at work, for travel and so on. Another common set of visual cues depicted the 'idea F I G U R E 4 Percentage of media images depicting the percentage of positively valenced 'leisure activities in or by water', by country. The figure's bars use part of a typical image from each country's dataset (see 'a dissonance between texts and images' discussion).

F I G U R E 3
Median colour hue (y-axis) and brightness (x-axis). Blue hues, depicting sunny skies and azure water were common in the 'leisure activities in the sun' theme. The green hues towards the top of the wheel (12 o'clock) were where the few images coded in 'leisure activities in an outdoor, public space' and 'eating and drinking in the sunshine' were located. Plotted using ImagePlot (http://lab. softwarestudies.com/p/imageplot.html). of heat', through bright sunbursts, intense red heat maps, and high readings on thermometers. At least in these images, the negative aspects of extreme heat are visually cued. But in this visual theme, people are largely excluded. Whilst danger might be implicit, people are still marginalised-not to mention impacts of extreme heat on animals, plants, and other non-humans. Yet heatwaves are understood not only as physical processes of energy transfer, as portrayed in the 'idea of heat' theme, but also in bodily sensations, memory, language and imagination (Abrahamson et al., 2008). This 'distancing' visual framing is common in the news media (O'Neill, 2020); and is also reflected in public perceptions of climate change as an issue which can be abstract and distant (Wang et al., 2019).
Both of these depictions are concerning in media coverage about heatwaves and climate change. Those most vulnerable to the impacts of extreme heat are excluded and voiceless, the real concerns for threats to their health and wellbeing during extreme heat events marginalised. It is also a lost opportunity for dialogue around who is vulnerable and how to reduce vulnerability. A study with chronically ill elderly people (i.e., those at higher risk of extreme heat) found that they do not perceive themselves as at risk from heat extremes (Wolf et al., 2010). Indeed, that study's participants suggested media storylines could usefully centre and humanise information about heat risk, instead of relying on traditional weather warnings (Abrahamson et al., 2008).
Second, our results show that there is a dissonance between images and text for media heatwave coverage. Almost a third of all images were positively valenced, yet hardly any texts were. Images and text advanced contrasting narratives of heat extremes. Media outlets are increasingly working towards more responsible and accurate coverage of climate change-for example, there are far fewer occasions now where the journalistic norm of balanced reporting contributes to informationally biased coverage, unlike in the early 2000 s when 'balance as bias' was far more prevalent (Boykoff & Boykoff, 2004;McAllister et al., 2021). However, there is clearly a problem with the visual imagery accompanying media narratives of heatwaves and climate, if holiday snaps are considered appropriate visual representations alongside news reports of heatwave vulnerability and even deaths.
The image-text dissonance echoes the findings of DiFrancesco and Young (2011) who found images and text pulling in different narrative directions in their study of climate change news in Canadian print media. They hypothesised that this could be due to editors, rather than article journalists, attaching the images post-facto to articles; and editors choosing images for their edgy qualities rather than their match to the text. Indeed, news media outputs are a result of complex interactions between macro trends, the organisational meso-level, and the micro-level of individual journalists (Shoemaker & Reese, 1995). An interview study with journalists highlights the complexity of climate news production and the everyday structures and routines (including the covering of climate news stories by general news reporters rather F I G U R E 7 Percentage (%) of images coded by valence, per country.

F I G U R E 6
All images across the four countries in the 'idea of heat' code, clustered by sub-code (sunbeams, thermometers, weather map, fire, thunderstorm). than environment beat specialists), which might contribute to such effects (Strauβ et al., 2021). We hypothesise that the dominance of global image libraries (e.g., see Aiello, 2016); the fast-paced and competitive nature of online news environment, the way images can drive engagement with a website via thumbnail images, and the news media organisation's social media engagement strategy (e.g., photographs requested from readers to illustrate a story) may all be key drivers of visual content. These hypotheses of climate text-visual news production deserve further research. Additionally, researchers could usefully investigate audience engagement with the different types of heatwave imagery.
Third, the results of this study raise the question: how can extreme heat be visualised in ways which are sensitive to risks and injustices, but without resorting to a visual discourse of fear (O'Neill & Nicholson-Cole, 2009)? The dominance of the current visual narrative around extreme heat disregards the unevenly distributed harms on vulnerable communities and misses an opportunity to imagine different climate-changed futures. It is an opportunity lost to build resilience to heat extremes in everyday life.
There are historical-cultural reasons why particularly northern Europeans might welcome news of an impending period of sunny summer weather (Fox, 2005); and at least in the UK, public perceptions of heatwaves were previously associated with positive as well as negative emotions (Taylor et al., 2014). However, extreme heat has now begun to be identified by UK publics as a serious climate risk (Steentjes et al., 2020) and solutions journalism is gaining ground industry-wide (https://coveringclimatenow.org/resource/climate-solutions/) and within individual media companies (https://www.theguardian.com/world/series/the-upside). So, how could a broader understanding of what heatwaves entail be visualised? How might a space for more hopeful climate futures be imagined? A starting point for this comes from within our own dataset. AD visualised everyday living-with-heat, which will be part of climate adaptation-much more so than being able to escape to the beach or splash in a city fountain during every heatwave. The AD photographs effectively brought climate change, and heat extremes, closer to home; in ways which opened up the visual discourse to concerns of justice and equity. It may be that as a local and regional outlet, AD focused on local stories of how people were coping with the heat, and commissioned freelance photographers for specific images to accompany these stories. As with research showing the important role local TV weathercasters can play in shaping peoples' climate perceptions (Bloodhart et al., 2015), future research could usefully investigate the role of local and regional news organisations as potential change-makers in the climate journalism space, and how their audiences respond to these very different types of solutions-orientated images.
Finally, it would also be illuminating to investigate how heatwave imagery is visualised beyond the European context, and beyond this largely manual analysis of online news. Is 'fun in the sun' a visual trope used globally, or is it limited to the global North? How can digital analyses of large numbers of visuals circulating online (what Rose and Willis (2019) define as 'hyper-seeing'), for example, add to understanding? What do other visualisations of heat extremes tell us about how people engage with weather and climate change?
To conclude, current visual portrayals of heatwaves in the European media are problematic for how they displace and marginalise vulnerability, and in how they exclude opportunities for imagining a more resilient future. Media portrayals both shape and reflect society. It is therefore upon us all-not just image collection curators, photojournalists or editorsto critically question visual portrayals of extreme heat and facilitate dialogue about the unequal impacts of these weather events. Co-produced research with such actors could seek ways in which to open up a more inclusive and responsible visual discourse.

ACKNOWLEDGEMENTS
SO was funded through a Leverhulme Research Fellowship (RF-2021-599). SH acknowledges ESRC SWDTP PhD scholarship funding. Thanks to Angeliki Balayannis for discussions about heatwaves and the imagination; to Andreas Schuck for contributing to discussions around the coding protocol development; to Travis Coan for assistance with inter-coder reliability; and to Anke Wonneberger for insights into the Dutch news media ecosystem.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.