Climate change and political polarization are two defining challenges of the 21st century. This research explores their complex interplay by examining online conversations about climate change, specifically focusing on the annual COP events. Social media platforms, like Twitter, serve as crucial spaces for public debate on climate issues, offering a mixed bag of democratizing potential and the risk of escalating polarization. A highly polarized environment can foster antagonism between ideological groups, leading to political gridlock and jeopardizing pluralistic democracies. The study of online polarization has gained significant momentum in recent years. This paper delves into Twitter data related to COP events to analyze the structure of climate change discussions, examining the diversity of viewpoints and interaction patterns among opposing groups. The researchers hypothesize that a significant opposition to the dominant pro-climate discourse has emerged, resulting in a highly polarized online climate debate. Twitter's widespread use by politicians and journalists, its broad influence, and the rich structural data it offers make it an ideal platform for this investigation. While acknowledging Twitter's limitations as a direct representation of public opinion, the authors draw upon existing research highlighting Twitter's importance in understanding climate communication, political polarization, and the spread of misinformation. The COP summits are central to this study for three reasons: first, they represent discrete, recurring events, ideal for multi-year quantitative analysis of polarization; second, their thematic focus ensures relevant tweet content and connected interaction networks suitable for robust analysis; and third, COP's prominence in international climate diplomacy makes it a key lens through which to study the intersection of climate change and political polarization.

The paper draws on existing literature that explores the role of social media in climate change communication, highlighting both its democratizing potential and its capacity to exacerbate political polarization. Studies examining online polarization trends are reviewed, emphasizing the importance of understanding the dynamics of online discussions to address issues of antagonism, political deadlock, and threats to pluralist democracies. The authors also cite research on Twitter's unique characteristics as a platform for political discourse, acknowledging both its influence and its limitations as a perfect proxy for public opinion. The literature on climate change communication across various media is considered, including research that employs computational techniques and traditional methods for analyzing polarization and politicization within the context of climate change. The paper builds upon this existing research by integrating tools from the field of infodemics, recognizing the importance of understanding the spread of information and misinformation in shaping online discourse on climate change.

The study employed a mixed-methods approach, combining quantitative analysis of Twitter data with qualitative analysis of tweet content. **Data Collection:** Twitter data, including tweets and user information, were collected using the official Twitter API, employing the search query "cop2x" (where x represents the COP number, ranging from 0 to 6). Data were gathered from June 1st of the conference year to May 31st of the following year (except for COP26, which included data up to November 14th, 2021). Supplementary Table 5 provides detailed statistics for each COP. Data were downloaded between October and November 2021. To identify political accounts, the researchers used an existing dataset of political Twitter handles from 26 countries (available at TwitterPoliticians.org and FigShare), adding any missing handles of prominent COP26 influencers manually. **Network Construction:** A retweet network was constructed for each COP, with nodes representing users and directed edges indicating retweets. Edge weights reflected the number of unique retweets between users. Only English-language retweets were included. **Polarization Measurement:** The study adopted bimodality as the key measure of polarization, defined as the presence of two distinct clusters or modes in the distribution of user opinions. Since opinions on Twitter are not directly observable, the researchers used a proxy measure—'latent ideology'—derived from retweet patterns. This method, adapted from prior research, infers ideological scores using correspondence analysis and retweet interactions, resulting in a continuous distribution of scores ranging from -1 (pro-climate majority) to +1 (climate contrarian minority). Hartigan's diptest was used to statistically assess the bimodality of the ideology distributions for each COP. The analysis of influencers was refined by excluding accounts (less than 3%) that conflated results due to factors like language, geography, or irrelevant topics, ensuring the focus remained on climate-related political ideology. **Topic Extraction and News Media Analysis:** BERT topic modeling (using BERTopic) was utilized to identify key discussion topics within each COP's Twitter data and to place them on the ideological spectrum. News media sources cited in the tweets were cross-referenced with NewsGuard's trust scores and political leanings to determine correlations between ideology and information source. **Supplementary Data:** The study also included analyses of data from YouTube and Reddit to assess the relative activity on these platforms concerning the COP events. Furthermore, an analysis of tweets containing the term "climate change" was conducted to assess the extent to which the COP discussion reflects the broader climate debate on Twitter. This involved examining monthly tweet counts, retweet activity, and cross-ideological engagement to provide a broader context for the findings.

The study reveals a substantial increase in ideological polarization on Twitter during COP26, contrasting with relatively low polarization between COP20 and COP25. This upsurge is attributed to a fourfold increase in right-wing activity compared to pro-climate groups since COP21. **COP21 vs. COP26:** Figure 1 illustrates a marked increase in both tweet creation and engagement (retweet counts) during COP21 and COP26, reflecting heightened public attention to these significant climate events. The Google Trends data supports this observation, showing an increase in public attention. The ideological spectrum analysis using 'latent ideology' shows a unimodal distribution during COP21, indicating low polarization, while COP26 displays significant multimodality confirmed by Hartigan's diptest (COP21: D = 0.0023, P = 0.003; COP26: D = 0.049, P < 2.2 × 10⁻¹⁶). This suggests a shift towards bimodal distribution and increase polarization in COP26. **Minority Influence:** Analysis of the minority group reveals an increasing influence of a broad range of climate contrarian views, which challenge the dominant pro-climate narrative. The rise of these views correlates with an increase in anti-establishment rhetoric, particularly regarding accusations of political hypocrisy. This is apparent in both the COP21 and COP26 minority influencer sets. **Political Dimension:** The involvement of elected politicians in COP26 is significant. A two-dimensional ideological spectrum (Figure 4) demonstrates a clear division between majority (pro-climate) and minority (climate-contrarian) groups, with specific political parties clustering accordingly. For example, US Republicans and former UK Brexit/UKIP politicians are prominently represented in the minority, while most other mainstream parties fall within the majority. Noteworthy is the presence of parties often seen as weak on climate action in the majority, reflecting nuanced rhetoric and a mix of pro-climate and critical statements. **Discussion Topics:** BERTopic analysis identified distinct discussion themes. The majority group focuses on climate action, COP-specific themes, and activism. Criticism of COP's effectiveness was prevalent within this group. The minority group exhibits a broader range of topics, aligning with established climate contrarian claims, including criticism of specific politicians and the promotion of broader political right-wing narratives (COVID-19, vaccines, and immigration). **Political Hypocrisy:** The theme of political hypocrisy emerged as a cross-ideological appeal, drawing in users from both majority and minority groups. This theme covers issues like the use of private jets, fossil fuel consumption, and environmental damage, showing increased prominence since December 2020. **News Media Usage:** Heatmaps correlated ideological position with news media reliability scores. The majority group tends to favour high-trust news sources, while the minority group leans towards less reliable sources. **Broader Climate Discussion:** Supplementary analyses demonstrate that the COP discussion, although keyword-based, is representative of the broader climate conversation on Twitter, showing similar trends in scepticism and polarization.

The findings address the research question by demonstrating a clear increase in ideological polarization surrounding climate change on Twitter, particularly during COP26. This heightened polarization is not simply a reflection of greater overall engagement but rather represents a significant shift in the balance of influence between pro-climate and climate-contrarian voices. The increase in right-wing activity and the prominence of 'climate contrarian' views, particularly those emphasizing political hypocrisy, are critical factors driving this polarization. The results highlight the potential risks to effective climate action stemming from this increasingly polarized landscape. The emergence of political hypocrisy as a major theme connecting otherwise disparate groups suggests a potential vulnerability in climate advocacy efforts. The correlation between ideological stance and news source reliability underscores the role of misinformation in shaping online climate debates. The study's findings have strong relevance to the field of climate communication and political science, offering valuable insights into the dynamics of online discourse around climate change and its potential implications for political action and public opinion.

This study reveals a significant rise in polarization regarding climate change on Twitter, mainly during COP26, driven by the increased influence of climate contrarian views and right-wing engagement. The prominence of political hypocrisy as a unifying theme for contrarian arguments presents a challenge for climate advocacy. Future research should monitor these trends in subsequent COP events, examining the evolving influence of minority groups, the impact of social media polarization on broader public debate, and the effectiveness of various climate communication strategies. The concept of a 'healthy plurality' of views requires further exploration and consideration.

The study's reliance on Twitter data limits its generalizability to broader public opinion. While the supplementary analysis suggests the COP discussions reflect the broader climate discussion, other platforms may exhibit different patterns. The 'latent ideology' method, while useful, is a proxy measure of opinion that may not perfectly capture the nuances of individual beliefs. The focus on English-language tweets may also exclude relevant perspectives from non-English-speaking communities. The keyword-based data collection, using 'cop2x,' may have missed certain climate-related communities on Twitter.