Climate nags: Affect and the convergence of global risk in online networks

ABSTRACT Scholars have observed the need to better understand the role of emotion in the issue of climate change, as well as to better convey the relationship between climate and other global crises. This article takes up these two positions, investigating the way social media facilitates affective connections between climate and other global risks. Using Twitter data from three global events – Covid, the 2020 U.S. presidential race, and the Russia–Ukraine war – the study examines how users connect climate change to each event. Placing these discussions in the context of online issue publics and ecocriticism, the paper examines the way users employ affect to connect these events to climate change. The paper uses a quantitatively driven qualitative approach, combining computational methods with a thematic analysis of affective expressions. Interestingly, sentiment was not universally negative, and the qualitative findings further suggest that users combine emotions in contradictory ways, expressed through the themes Weary Zealotry, The Hope–Disgust Dialectic, Climate as Proto-Disaster, Idiots and Enemies, and Global Solidarity. It is argued that a modified version of Beck’s ‘imagined communities of global risk’ provides a framework for the role of affect in people’s relationship to climate change.


Introduction
Social media are increasingly part of how citizens engage with political issues, including the issue of global climate change (Dahal, Kumar, and Zhenlong 2019;Loureiro and Alló 2020;Schünemann 2020).Transnational platforms like Twitter have been found to help like-minded climate advocates exchange solutions and solidarity across national borders, and facilitate emotional as well as policy discussions (Jacques and Connolly Knox 2016;Kirilenko and Stepchenkova 2014;Reber 2021;Schünemann 2020).
Yet climate change is only one threat in what Ulrich Beck (2011) has called the 'world risk society.'In recent years, the world has experienced events of pandemic, nuclear threat, terrorism, and rise of neo-authoritarianism (Harindranath 2011;Makhortykh and Manuel González Aguilar 2020).'There is plenty to find shocking,' (2) write Bladow and Ladino, worrying that the immediacy of other events has overshadowed climate among the public.They propose an 'affective ecocriticism' that finds more emotionally compelling ways 'to foreground connections between environmental and social justice' (3).That is, rather than presenting other global risks in competition with climate, scholars should demonstrate the way they converge.
Despite such calls, and wide scientific knowledge about the connection between a warming climate and other threats (IPCC 2022;Lawler et al. 2021), so far little social media research has captured the way users are already making topical and emotional connections between events.This paper seeks to add to the understanding of the way people connect climate to other global risks, focusing on the role of affect.Affect here is understood here to mean a practice of context-specific feeling that can be seen through relational engagement, of which emotion and sentiment are critical elements (Döveling, Harju, and Sommer 2018;Hermes 2020;Munezero et al. 2014;Papacharissi 2015).Drawing on the dual bodies of research on affect in online 'issue publics' and in risk perception, the paper asks what forms of affect help users connect climate change with other emerging forms of risk.This is important because it helps reveal how people react to and prioritize risk, necessary precursors to persuading the public of the risks of climate change.
Empirically, the paper approaches this problem using Twitter data that is not explicitly about climate, but rather focuses on three major global events: the Covid pandemic, the 2020 American presidential race, and the beginning of the 2022 Russia-Ukraine war.Tweets referring to climate change are then identified in these collections (N = 1,186,322).The networks, topics, and positive or negative sentiment are analysed using computational methods.Using these results as a basis, a qualitative thematic analysis of affective expression is then performed to provide further nuance to how users negotiate 'risk convergence' between climate change and other global events.The findings demonstrate how audiences make not just topical associations but also emotional ones.The computational sentiment analysis indicates that these can even be positive emotions (particularly in the case of the U.S. election), though the qualitative investigation further reveals modes of expression that involve seemingly conflicting emotions, including weariness/zealotry and hope/despair.I argue that the findings point to a new, affective understanding of what Beck (2011) has called 'imagined communities of global risk'.The practical implications for of these findings for climate change communication are discussed.
In the following section, I contextualize this study in the ecocriticism literature, examining why affect is especially relevant to the politics of climate change and other global risks.This is followed by an examination of research on issue publics and affect in social media.

Global risk and emotion in the Anthropocene
Scholars theorize that publics are increasingly conscious of being part of not just a national entity, but a global one (Volkmer and Florian 2010).In this self-reflexive environment, Beck presented the concept of the world risk society in which publics are faced with global threats like climate change, financial crises, disease, terrorism, and nuclear proliferation (1348).These problems are at once highly documented, often manmade, and yet outside the authority of any one nation to solve.Beck argued that while one response to this situation would be a feeling of 'powerlessness and paralysis' (Beck 2000, 218), it could also compel citizens to form 'imagined communities of global risk,' in which, 'strangers in distant places are following the same events with the same fears and worries as oneself ' (2011, 1350).He suggested such transnational communities could appeal to authorities outside traditional political structures, and even outside the nationstate.
Beck does not directly address the role of affect or emotion in global risk.However, Deborah Lupton (2013) argues that emotion is closely bound up in the concept.Fear of course, is foremost, but Lupton points out that risk may also involve pleasure and control-seeking (e.g.extreme sports).In contrast to models that position risk as an objective catalyst to embodied 'gut' reactions, Lupton argues that 'emotion and risk interact with each other and in the process configure each other' (641).Thus, she argues, both must be understood within their particular contexts and understood as phenomena that are the result of people's interpretation, rather than an innate ('gut') reaction.
One of these contexts is the natural environment.In ecocriticism, the role of emotion in relation to nature and natural hazards has been a critical part of understanding the Anthropocene.Lawrence Buell (2007) argues that what he calls 'ecoglobalist affect'a whole-earth feeling about nature -predates financial and political globalization, finding it in American literature in religious writings of the colonial period.Then it was closely tied to concepts of possession and subduing the savagery of nature -precursors to what Estok (2014) calls 'ecophobia,' or fear of the agency of nature over people.In contrast, Enrique Salmón (2000), writing from an indigenous perspective, proposes a 'kincentric ecology' in which the natural world is viewed as 'not one of wonder, but of familiarity' (1329).This perspective points out the connection between sustainable practices and indigenous knowledge acquired through familiarity -a source of knowledge diminished by the Covid pandemic, among other threats (Lane 2020).
In contemporary media contexts, influenced by political psychology (Lodge and Taber 2013), affect is understood to have a critical influence on climate change perception and decision making.In a qualitative analysis of climate reporting, Höijer (2010) found framings related to hope, guilt, compassion, and nostalgia.Lachlan, Spence and Lin (2014), in a study of Hurricane Sandy in the U.S., identified 'affective expressions of risk,' whose categories included fear, dread, concern for material loss, concern for health, and anger.Loureiro and Alló (2020), applying computational methods to social media data, find differences between users in Spain and the U.K.; while fear dominated Spanish discourse around climate, British discourse was characterized more by anticipation, highlighting the complexity of affective responses to risk.Bladow and Jennifer (2018) argue that understanding emotion is a necessary precursor to persuasion and action and that 'a clearer sense of what those emotions are and how they work might reconnect environmental and social justice' (3).Climate scientists and advocates worry that climate change is more often characterized by a feeling of despair or even apathy, from which the public is easily distracted -perhaps willingly so -by other global events.Notably, in another paper, Loureiro and Alló (2021) find that the number of people tweeting about climate change dropped during the emergence of the Covid pandemic, though they did not explore how the two issues overlapped with each other.In the following section, I explore the dynamics of Twitter more closely, and the role of affect in the formation of Twitter publics.

Twitter publics and affect
If globalization has created international connections between territories, globalized digital media have provided 'the visible evidence' of these connections (Tarrow 2005, 36).Online media have upended understandings of the relationship between media and publics; rather than media 'designing social structure' (Dayan and Katz 1992, 15), online platforms allow more leeway for social structures to design their own media.Users connect to television shows, music, race, gender equality, the environment, political revolutions, and anti-war movements (Blevins et al. 2019;Castells 2015;Highfield, Harrington, and Bruns 2013;Papacharissi 2015;Pickerill, Kevin, and Frank 2011;Schünemann 2020).
In the age of social media, scholars have suggested 'issue publics' as a framework for understanding online publics (Bennett, Lang, and Segerberg 2014;Bruns and Burgess 2011;Reber 2021).The concept predates social media.Public opinion researcher Philip Converse (1964) is credited with the term, which he used to capture the idea that voters have not only varying opinions but varying strengths of opinion: 'Different controversies excite different people' he explained (53).In online digital spaces, these strengths translate into visible patterns of communication.Bruns and Burgess (2011) offer a framework specifically for identifying 'issue publics' on Twitter, through use of keywords and the communicative structures of the platform -hashtags, @mentions, retweets, and followerfollowee relationships.They argue that Twitter should not be viewed as a single space with 'separate, sealed entities' but rather as 'embedded and permeable meso-level spaces' that overlap with both macro-level networks and micro-level individual exchanges (6).In these spaces, publics may be long-standing or may form quickly -ad hoc issue publics -'to discuss breaking news and other acute events' (2).
While Converse (1964, 6) only briefly touched on the 'psychological' dimension of opinion, scholars now understand affect to be a central driver in the formation of online networks (Himelboim et al. 2016).Researchers who have tracked social media discussions on the Egyptian Revolution, Occupy Wall Street, #BLM, and other social movements argue it is not only 'rational' political interest that drive formation of these publics.As Zizi Papacharissi (2015) writes, 'All of these movements emerged out of different contexts but shared one thing in common: online and offline solidarity shaped around the public display of emotion' (5).
Coinciding with the 'affective turn' in social sciences that understands emotion as integral rather than antithetical to decision-making, affect has become a key area of study in online politics.Papacharissi goes on: Affect evolves concurrently with the flow of events technologies facilitate, adding to the ongoing movement of forces that intensify or subtract from co-occurring and interacting bodies, events, and ideas.Thus affect contributes to but also helps us to understand the set of moving forces that make any event what it is.(15) The view of affect among internet researchers is that affect involves emotion, but encompasses a broader condition of feeling and sense-making done in concert with other members of the network.Döveling, Harju, and Sommer (2018) propose that on social media, studies drawing on affect theory should move beyond 'individualized feelings' and instead consider affect as an articulation and communicative practice (3) -in other words, to not psychoanalyse the user's thoughts, but to understand the atmosphere they seek to create (3).
Even so, there are no agreed upon categories of affect as it relates to internet culture.Harindranath (2011) outlined legal, rhetorical, and emotional categories for study of 'performative public affect' related to acts of terrorism.Döveling, Harju, and Sommer's (2018) 'digital affect culture' is characterized by three forms of emotional exchange: discourse, alignment, and belonging.In their study of climate change tweets, Loureiro and Alló (2020) used Plutchik's (1982) basic psychoevolutionary emotions: fear/terror, anger/rage, joy/ecstasy, sadness/grief, acceptance/trust, disgust/loathing, expectancy/ anticipation, surprise/astonishment (see also Lachlan, Spence, and Lin 2014).
Although affect theory lends itself to qualitative analysis, studies on social media often take a quantitative approach to analysing emotion and affect (e.g.Dahal, Kumar, and Zhenlong 2019;Loureiro and Alló 2020;Williams et al. 2015).While computational methods can produce results on huge corpora of data, the outputs are often limited to positive, neutral, and negative scores.Gaspar et al. (2016), Hermes (2020) and Harindranath (2011) among others have made the case for using more qualitative approaches to studying social media audiences, specifically regarding affect theory.
Moreover, a review of the literature finds that much of the previous research has focused on single issue publics.To date, little research has been devoted to capturing the characteristic alluded to by Bruns and Burgess -that issue publics do not exist in a vacuum, but are part of wider information environments.While social media can contribute to fragmentation and polarization, they also create spaces where users can create convergence (Jenkins 2006).Therefore, in the current context, I suggest considering the potential for social media to facilitate affective forms of 'risk convergence.' With the understanding of the relationship between affect, global risk, and environment as foundation, this paper investigates the way users in the 'world risk society' make sense of the long-running risk of climate change in the context of new risk events.I pose the following questions: How do social media users understand convergence between climate and other global risks?and What are the affective drivers of these connections?In the next section, I describe the case data and the methods of analysis.

Data and methods
This paper uses data collected from Twitter about three global events: the Covid pandemic, the 2020 American presidential race, and the beginning of the Russia-Ukraine war.The election is perhaps not a traditional risk event.However, due to the economic, military, and diplomatic standing of the United States, the election encompasses a number of relevant issues: financial crisis, nuclear weapons, and pollution.Arguably, that the election is a different category of event provides more depth to the data; the event departs from the others in ways that will be discussed in the Findings section.
Publicly available tweets were collected using keywords related to each event from Twitter's Streaming API using the DMI-TCAT (Borra and Rieder 2014). 1 The Norwegian Centre for Research Data (ref. 892392) approved data collection without consent of the users, due to the volume of data and the public relevance of the subject matter.In order to protect the privacy of users, the Twitter handles of non-public figures have been omitted and examples of tweets have been reworded.
The subset of climate issue publics was identified using keywords in the text of tweets: climate, globalwarming, global warming, and emissions.This returned a climate subset of 1,186,322 tweets (see Table 1 for breakdown by event).The 200 most prolific usernames in the full data and all 'core' usernames in the qualitative analysis (see below) were run through the bot-detection tool Botometer and users meeting the 0.5 threshold were removed (Yang et al. 2019).

Computational analysis
Networks.A network map was constructed in the programme gephi using @mentions and retweets as ties between users.Gephi's ForceAtlas2 algorithm was applied to create a visualization of the network (Jacomy et al. 2014).This algorithm uses the tweets between users to create a kind of 'magnetic pull' that draws users together, creating a visualization that positions users according to who they tweeted with most.While network visualizations are constructed using mathematical principles, the researcher must give the network 'qualitative meaning', as suggested by Drieger (2013).Thus, the network map was then manually examined and assessed, with particular attention to the ideological viewpoints of parts of the network.
Sentiment analysis.Within computational text methods, sentiment refers to the positive-negative valence of messages, matching the text against word lists that have been assigned positive or negative connotations.In this case, I used the open-source sentiment analysis tool Valence Aware Dictionary and sEntiment Reasoner or VADER (Hutto and Gilbert 2014).VADER is specifically designed to assess sentiments in social media text; its dictionary includes emoticons, emojis, acronyms, and common slang.VADER measures the intensity of sentiment as well as the direction, returning a compound score between −1.0 and 1.0, with 0 neutral.For example, the phrase 'I love you' receives a score of 0.64 while 'I don't love you' receives −0.52.
Topic modelling.Topic modelling is an unsupervised machine learning technique to summarize topics in text.It returns probabilistic combinations of words that frequently occur together, but -as in the analysis of network maps -this method requires human interpretation and knowledge of the underlying data, as well as a decision about how many topics the data should be divided into (Dahal, Kumar, and Zhenlong 2019).Pilot tests revealed that repetition in the data due to retweeted text resulted in topics based entirely on the words in individual, highly popular tweets.Thus, the text was prepared by removing duplicate text, creating results based on repeated topics in original tweets, rather than the topics of the most shared content (although the topics of the most shared tweets are nevertheless similar to those in the original tweets).

Thematic analysis
For the manual analysis of affective expression, a subset of 'core' climate users was identified: the 846 users who appear in the climate datasets from all three events (after removing bots).The themes analysis was performed on their 752 original tweets in the data (N Covid = 313 tweets; N election = 352; N Ukraine = 87; note that not all 846 core users tweeted original content).For the analysis, the tweets were segmented by event, and within these, organized by network cluster (main or oppositional).
The qualitative thematic analysis started by considering the terms from the computational topic modelling, in tandem with categories of emotion used in previous research on climate change.These include Plutchik's (1982) categories of psychoevolutionary emotions (fear, anticipation, trust, surprise, sadness, joy, and disgust) as used by Loureiro and Alló (2020) anger, as well as Höijer's (2010) categories of emotion found in Swedish news coverage of climate (fear, hope, guilt, compassion, and nostalgia).Following Braun and Clarke's (2006) methodological framework for thematic analysis, the following codes were developed after an initial reading of the data: hope, fear, anger, preview, setback, pessimism, solution, death, urgency, disdain, mockery, failure, idiots, enemies, and one of many problems.However, this paper is not interested in named, personal emotions as such, but expressions of affect, understood as a performative and relational practice (Döveling, Harju, and Sommer 2018, 3).In addition, few of the tweets expressed what might be considered a personal emotional state, as will be discussed.Thus, the initial annotations were then condensed into five categories of affective expression that consider the discursive dimension of the tweets (Hermes 2020, 314).These categories are: Weary Zealotry, The Hope-Disgust Dialectic, Climate as Proto-Disaster, Idiots and Enemies, and Global Solidarity.

Computational summary: networks, sentiment, and topics
Three methods were used to quantitatively summarize the entire climate dataset, starting with mapping the networks between users (Figure 1).As found by Williams et al. (2015), the network shape suggests a strong tendency towards polarization.An investigation of major actors (labelled in Figure 1), the most popular content, and the results of the topic modelling (described below) suggests a tendency towards like-minded interaction, with climate advocates gathering in the subnetwork on the left side of the map and a smaller group of climate change doubters gathering on the right.These will be identified as the 'main network' and the 'oppositional network' respectively.It is also notable that the three events are largely intermingled in the network, particularly since one is a national election.This unity in the overall network suggests key actors in climate discourse are relatively stable, regardless of the event.
The sentiment analysis computationally detected the positive or negative valence of language of the tweets.Figure 2 shows the histograms of the compound scores in each event, indicating that the most negative sentiment appears in the Russia-Ukraine war.Sentiment during the beginning of the Covid pandemic is also negative, though to a lesser degree.Finally, the average sentiment during the election tips over to the positive side.This may be because the election provided more openings to talk about policy solutions for climate change, as will be better understood in the qualitative analysis.
Results of the topic modelling further reveal the way users created conceptual convergence between climate change and other events of global risk.In light of the likely polarization of the network, the topic modelling algorithm was run separately on the main network and the oppositional network.Table 2 shows the topics by network for each event.A few observations: 'world' or 'earth' are connected to topics in all three events in the main network.Also, the oppositional network is lexically similar to the main network, except with regard to use of terms 'hoax', 'control', and 'agenda' and the frequent reference to Swedish climate activist Greta Thunberg.

Themes of affective expression
Informed by the quantitative findings, a qualitative analysis was performed on the 752 original tweets of the users in all three events ('core' users).As previously detailed, the initial codes were consolidated into the following five themes.

Theme 1: Weary zealotry
One of the most prevalent modes of affective expression is that of a foreboding knowingan implied risk in the waiting.These suggest the unfolding event was a preview of climate risks.Users employ words like 'horrific' and 'chaos,' and refer to death.For example: Corona is just a taste of the human suffering and economic disruption that the disaster that is climate change will soon bring

All you people who think corona is horrific: what til you hear what the climate is going to do
This mode of expression is especially visible during Covid, when the quick onset of society-wide restrictions offered an example of the kind of comprehensive action In this tweet, the user pivots from Ukraine to climate, using the image of previously 'unimaginable' destruction as a rhetorical fulcrum.
In the case of the election, the event is not so much seen as a taste of climate change.Rather, the election is a crossroads between a potential way out (Biden win) and certain disaster (Trump win).For example: I don't like Biden.But the choice is between him and having your grandchildren grow up on an Earth ravaged by environmental destruction.So he gets my vote.
Weary zealotry is not necessarily characterized by fear, at least not on the part of the user.And there is little overt anger detectable in these tweets.Rather, users appear to dryly deploy mental images of death, danger, and 'horror shows' to contribute to what Anderson (2009) calls the 'affective atmosphere'.As one user in the data bluntly told their followers: 'be very afraid'.
The primacy of climate change over all other threats at times is expressed through frustration and resentment that other events were distracting from the core problem of climate change.One user sarcastically puts it this way: 'So nice of the climate change crisis to wait while we focus all our attention on corona.'Users thus express a hierarchy of risks, which they feel the media attention does not accurately reflect.

Theme 2: The hope-disgust dialectic
As suggested above, many users do not express much emotion themselves, but instead use dry and distant tones, even when referring to death.However, two nameable emotions that are more easily identifiable were hope and disgust.In terms of sentiment valence employed by VADER and other computational tools, these are opposites: positive and negative.Yet hope and disgust (or disdain) often occur together in the tweets.For example: Climate change is already here.Covid is just part of the disastrous ecological breakdown.But we have the power to rebel against our destruction and change the bad system!Here the user exhibits a certain degree of affective whiplash, moving from the idea that little progress has been made, to expressing hope that would change.Other users offer possible solutions, while also suggesting a pessimism about the feasibility.This configuration is especially pronounced during the election, when users discuss whether Biden will be able to -or even wants to -make progress on greenhouse gas emissions.Here are some examples, with this duality highlighted in bold: I have confidence that Biden-Harris will make serious efforts on climate.As they should.We are on the verge of failure.

CONTINUUM: JOURNAL OF MEDIA & CULTURAL STUDIES
Trump has been apocalyptically bad.He won't get the Christian version of the apocalypse, but he's sure happy to destroy the planet.And there are a lot of #climate voters ready to get rid of him Many of the more hopeful tweets discuss Biden's promised re-entrance into the Paris Climate Agreement, which President Trump had withdrawn from.Similarly, during Covid, users describe a permanent paradigm shift.'We can't go back to business as usual,' says one.Although users are cynical, they see that a major event is changing the menu of possibilities.Although the Russia-Ukraine war is overwhelmingly negative in the sentiment analysis, some users suggest the convergence of the war with climate change could have a positive outcome: 'When the Ukraine war leads to an energy crisis,' writes one user, 'we can finally make a transition to renewable energy.'Theme 3: Climate as proto-disaster Some tweets used climate not just as comparison or metaphor, but as root cause that helps make sense of the unfolding event.For example, this user depicts the tragedy of Covid deaths as a direct result of failure to act on climate change: America is at risk from corona because of our terrible air quality.Asthma and other lung diseases by the millions are underlying conditions.Wouldn't be such a problem if we had acted sooner on #climate Climate change becomes the origin of the newly emerging mass tragedy of a pandemic, proving the urgency of the climate crisis.This mode of expression was even more pronounced in the case of Ukraine, when many users are grappling with the complex geopolitics.One user suggests that the war can be explained by Europe's efforts to shift from coal as a result of climate change and Ukraine's natural gas reserves: 'it's finally all clear' they write.
The use of climate as proto-disaster is also visible in the election, when climate becomes indicative of a certain worldview.Users frequently add climate to a list of issues that includes race, guns, abortion, and healthcare.Users describe a 'way of thinking' that is responsible for not only failure to act on climate change, but also other political evils.One user suggests that Trump's inability to accept climate change reflects the same intellectual failing as his inability to accept his loss in the election and 'the fact that he's broke'.Other times climate is the issue that helps users clarify the race.One user argues (in a way that overlaps with Theme 1): Every issue the candidates debated tonight can be understood through climate change.It's part of every threat we're facing, and more people are paying attention.It's not political.It's moral.
Users following the war in Ukraine at times incorporate irony and dark or inappropriate humour that prioritizes climate -the original disaster -ahead of the new conflict.One user suggests the U.S. could send nuclear bombs to Ukraine to spark a nuclear winter and counteract global warming.During Covid lockdowns, many users point out that at least greenhouse gas emissions have declined.
As for the oppositional network, the reductionism of these tweets become a source of almost joyful mockery.The prioritization of climate over people, for the oppositional users, is seen as typical of the other side.For example: Based on the tweets I'm seeing I'm beginning to think corona was made by a group of climate zealots to kill off the world's population and 'save' planet Earth, LOL However, the oppositional network engaged in its own version of affective heuristics, as will be discussed in the fourth theme.

Theme 4: Idiots and enemies
Although many of the tweets addressed climate change directly, it was also common to see users express disdain for groups of people -people responsible for climate change, but also possibly other problems.In the main network these people were often identified as right-wing or anti-science idiots.Some examples: If you want to know who's responsible for Ukraine, just look toward those claiming climate is a hoax flat earthers, creationists, moon landing nuts, and of course climate change hoaxers.The stupidity never ends.
These tweets take on an accusatory tone that says the problem is not really climate change: it's particular people.
However, this form of affective expression was more prominent in the oppositional network than in the main network.Here, after all, climate change was never the issue.Rather, climate change is one of many longer running problems, often stemming from a corrupt elite: At least Trump isn't afraid of the Climate nags.That's why I don't like @BorisJohnson.He joined the sheep.@BBCNews Since the invasion of Ukraine, corona and global warming have disappeared . . .gee how odd 'Climate crisis' 'pandemic' 'threat of nuclear war in Ukraine'.These are headlines meant to make you afraid so they can divide us.Don't listen to their agenda.
While users in the oppositional network criticize 'climate nags' for making everything about climate, in the above tweets, they also connect multiple global risks to each otheralthough in this case, the cause is global elite, corrupt politicians, or the mainstream media.Notably, each of the above tweets also connects the emotion of fear to climate, though not in the way the main network does; here, the thing to be afraid of is the global elites.Interestingly, the implication of this sentiment is a sense of global interconnectedness and solidarity -through resistance to this common enemy.The final theme further takes up the concept of global community.

Theme 5: Global solidarity
The fifth and final theme focuses on affective expressions of solidarity -often expressed as a form of global solidarity.Although this might denote a certain level of compassion, at times, these users employ a pessimistic, even ecophobic, affect, more in line with Theme 1: The world must realize that power of nature can't be ignored.Disease and climate change will destroy humans if we can't come together.
Yet users are also positive towards the possibility of a kincentric global cooperation (Salmón 2000), particularly after seeing the response to the Covid pandemic.Some examples: It's every country's choice now: will they learn from corona and make it a defining moment?The level of cooperation on this pandemic can be a model for the #ClimateCrisis After January 20 President Biden will rejoin the Paris Agreement.Climate denial is over.Thank you to everyone who shared the truth and made it happen.We have a chance now to save our planet The use of the globe emoji in the first tweet calls to mind Lawrence Buell's (2007) reflection on the Earth photo taken from space, which he identifies as a symbol of the selfconscious global citizen (227).Parallel to universalized calls for cooperation, users speak specifically to their own contexts, using other countries as examples.For example, after the U.S. election, an Australian user tweets at the prime minister: @ScottMorrisonMP Now that Trump is on his way out, you have no excuses.It's time to build a green economy, or face the world's criticism.This identification of nation-state responsibility shows an awareness of the interconnections between decisions made in one country and the future of another country (Beck 2011;Tarrow 2005) -thus, a global interconnection, but one that largely relies on discrete national policies.I will return to this idea in the concluding discussion.

Concluding discussion: Risk convergence in the Anthropocene
This study aims to capture the role of affect in transmitting messages on climate change, using a more holistic view of online publics -namely, that users do not exist in singleissue isolation, but experience overlappings and convergences with other issues and major events (Bruns and Burgess 2011).Previous literature has established the role of sentiment, emotion, and affect in online issue publics (Anderson 2009;Blevins et al. 2019;Papacharissi 2015) and its role in environmental issues (Bladow and Jennifer 2018;Buell 2007;Estok 2014, Höijer 2010;Loureiro and Alló 2020).The computational approaches suggest, in line with scholars' context-specific view of affect (Anderson 2009, Hermes 2020;Lupton 2013), that the feelings raised by climate varies across the different events.Moreover, it is not always negative.While the Russia-Ukraine War is characterized by negative sentiment, Covid was less so, and the 2020 U.S. election had a cumulatively positive score.Yet as noted by Gaspar et al. (2016), this computational approach is not able to take in the 'richness of social media expressions' (181).For example, the following tweet received one of the most negative scores (−0.93): Climate change is already here.Covid is just part of the disastrous ecological breakdown.But we have the power to rebel against our destruction and change the bad system!Yet this tweet (an example of Theme 2: The Hope-Disgust Dialectic) arguably has a positive point of view.In other words, when dealing with crises like climate change, it may not be immediately clear what 'positive' and 'negative' entail.
Therefore, jumping off from the concepts identified in the computational topic modelling, a qualitative thematic analysis was performed.This finds that first, publics connect risks both through the spectre of disastrous outcomes (Theme 1), and through the expectations (or scepticism) of institutional structures to address the risk (Theme 2).Yet while the sentiment analysis implies a feeling on the part of the user, the qualitative findings suggest a knowing detachment.That is, users are not so much engaging in a 'public display of emotion' (Papacharissi 2015, 6), but an expression aimed at inciting emotions in others.Moreover, we see evidence of what Harindranath (2011) calls recontextualization, or the 're-presentation of social events through the contextual rationale of another social event' (148).In this case, the affective templates for climate change are applied to the new and lesser understood threats (Theme 3), including who's to blame (Theme 4).Finally, these risks trigger feelings, positive and negative, of a global one-ness, connected through the need for action (Theme 5).It is this last point I wish to discuss further.
Beck theorized, 'It is the reflexivity of world risk society that produces the reciprocal relation between the public sphere and globality' (1350).Indeed, we see a kind of self-awareness in the reference to not only global climate change, but the users' recognition of their own global communication about climate change in transnational issue publics.However, I argue we do not see in the data support for Beck's assertion that these publics are interested in 'bursting the container of the nation-state'.For example, in the network map (Figure 1), we see that users direct their tweets at elected officials within nation-states, rather than supranational bodies like the E.U. or the U.N. or even the IPCC.(Arguably the oppositional network attribute more potency to these entities.)Instead, users presented the global public sphere as a sentimental rather than institutional connection.As Twitter and other transnational 'digital affect cultures' (Döveling, Harju, and Sommer 2018) become more integrated into public discussions of climate change (Jacques and Connolly Knox 2016;Kirilenko and Stepchenkova 2014;Reber 2021), understanding such perceptions of institutions becomes more relevant to researchers, policymakers, and organizations involved in public communication.
To that point, the findings suggest three main takeaways.First, rather than viewing online attention as either/or, a more fruitful approach may be to consider the opportunities in the convergence of attention -that is, the potential for overlapping rather than overshadowing.Of course, the way climate change overlaps with other crises is scientifically well documented (IPCC 2022;Lawler et al. 2021), but often this is put in terms of rational connections.The second takeaway is that risk convergence also entails emotional and affective connections.As we see in the analysis, these connections may at times involve emotions we view as opposites (e.g.Theme 2).Thus, persuasion campaigns that focus on simplified versions of affective response to climate change -positive/negative or fear/indifference -may be out of step with online conversations.Finally, it is worth considering the tendency of users in the data to position themselves as emotionally removed, while employing affective expressions aimed at inciting emotions in others.This evokes Döveling, Harju, and Sommer's contention that affect is 'something people do instead of have' and points to the need for strategies that enlist users as doers rather than recipient havers of emotions.

Figure 1 .Figure 2 .
Figure 1.Network map.Notes.The network map shows the tweets collected from each event that also refer to climate change.Lines ('ties') represent a retweet or mention another user.The most retweeted and mentioned users ('nodes') have been labelled, and are sized according to the relative number of other users that retweeted or mentioned them.Map created in gephi using the Force Atlas2 algorithm.n nodes = 679,696; n ties = 1,022,016

Table 2 .
Topic modelling.healthneedworldactionresponse future threat crises Topic2 recovery green energy crisis economic post economy new plan uk Topic3 join today new webinar models climates live current impact pandemic Topic4 pollution air pandemic co2 lockdown latest environment impact drop long Topic5 china world news spread die war deaths year 2020 new Topic6 science trump real like hoax people just believe scientists racism Election Topic1 science policy real said planet future issue man forest earth Topic2 covid care health presidential climateaction america american 19 white pandemic Topic3 president paris vote donald china agreement taxes pence accord news Topic4 plan new deal energy fracking green oil support fossil zero Ukraine Topic1 crisis nuclear energy protest power video world climateemergency security stop Topic2 oil invasion gas biden energy fuel europe action pandemic prices Topic3 people just world like need help think right support nato Topic4 fossil fuels human dependence peace roots induced world future hope Oppositional NetworkCovidTopic1 models science like people just don wrong think years experts Topic2 hoax agenda trump world just fear china control people new Topic3 crisis pandemic greta expert news world thunberg cnn economic says Election Topic1 covid oil fracking energy policy gas news american economy fossil Topic2 new deal plan green vote harris support believes campaign greta Topic3 china paris world obama did accord president taxes tax health Ukraine Topic1 energy covid crisis just like policy oil agenda west europe Topic2 nuclear nato right know military think years russians good want Topic3 biden gas oil news help stop hoax new going saysNotes.Iterations = 100; words occurring in less than 1% and more than 98% of tweets excluded; stop words include the search terms used to collect data.necessary to address climate change.However, in the case of the war in Ukraine, this expression of affect compares the harms of the Russian attack and future harms of climate change: