Online misinformation during extreme weather emergencies: short-term information hazard or long-term influence on climate change perceptions?

Extreme weather events linked to climate change are becoming more frequent. The online public discourse on and during these events, especially on social media, attracts misinformation that can undermine short-term emergency responses, but can also be aimed at influencing long-term public perceptions of climate change. This contribution reviews existing research on online misinformation with the aim to understand the types, origins, and potential impacts of misinformation during extreme weather events like storms, floods, and wildfires. The screening of 289 publications reveals that there is scarce body of only 13 studies addressing this question. Relevant studies exploring online misinformation during extreme weather events rarely document misinformation immediately relevant for emergency responses and only recently link this to the discussion about climate change. The reviewed research provides however insights to derive a framework that can guide future research into this topic. Specifically, that misinformation in social media during environmental emergencies 1) cuts across domains and merges different areas of public interest, 2) cuts across temporal and geographical scales, and 3) needs to be studied as part of an interconnected online media landscape. Misinformation differs between emergency event types, can undermine the debate about climate change in diverse ways, appeal to completely different audiences and thus will likely require different responses and countermeasures. Structured research with comparable methodologies is urgently needed.


Introduction
Extreme weather-related events such as storms, floods, heatwaves, and wildfires have become more frequent [1].The latest IPCC report summarizes the strengthened evidence between the connection of these extreme weather-related events and climate change [2].The report also highlights the expected increase in frequency and severity of extreme weather events [2] and addresses the expected impacts on ecosystems and human livelihoods [3].These findings build on a growing body of research investigating the causal relationship of climate change and specific extreme events [e.g., 4, 5], a connection that has also garnered increasing public attention [6][7][8].

Weather extremes and perceptions of climate change
Consequently, extreme weather-related events and their emergency nature have become a key perspective in public discourse on climate change.'Emergency frames' are increasingly invoked in debates on climate change and sustainability in general [9].Wildfires, floods and storms are frequently discussed in light of climate change, both in the traditional media [8] but also online social media [10,11].The latter have become an increasingly important source of news [12,13] and a forum influencing public debates on important societal issues in general [e.g., 14,15].Environmental emergencies actually appear to intensify the discussion about climate change on social media [16], and the effect of implicit or explicit framings of climate change in social media may have a substantial effect on perceptions of the issue [10,17].
These dynamics have particular significance when viewed in light of research showing that experiencing extreme weather-related events can bear on climate perceptions and personal behavioral intentions [see for example 18,19].Recent reviews [20,21] highlight that direct personal experiences have a particularly large potential to influence perceptions on climate change, but even images of environmental disasterscommonly shared online during such eventscan trigger engagement with climate issues [22].Hence, the short-term online public discourse about climate change in the context of extreme weather events may contribute significantly to long-term public perceptions of climate change and in turn public support for policies and actions to address it as a societal challenge.Especially since social media communications play a key role during emergency events.

Disaster management, social media, and misinformation
In disaster management, social media have rapidly gained importance as communication and information tools for emergency managers [see for example [23][24][25][26].The role of platforms such as Twitter 1 and Facebook have been the primary focus of research on disaster management and crisis communications [27].These social media can be valuable sources for the early detection of crisis events [28], critical citizen-sourced information [24] and are emerging as important tools during all phases of emergency management [25,27], largely due to their affordances for rapid information dissemination [29].
At the same time misinformation has emerged as a key concern in relation to emergency responses during disaster situations [30].False information is a generic challenge during emergency management [31] independent of the information distribution channels, but has been identified as a key social challenge for the use of social media in emergency management [23].Here, misinformation carries the risk of severe consequences, especially given the ubiquity of online social media.
The same affordances for rapid information dissemination [29] that make social media a valuable element in the tool set of disaster management thus contribute to the accidental or intentional rapid spread of misinformation.Misinformation can significantly impact emergency management and in the worst case represent a direct or indirect threat to human life and health [23,30].This has been documented for a wide range of emergency situations in different domains [30], such as for example (suspected) terror attacks [32,33] or most notably the recent COVID-19 pandemic [34][35][36].

Sense-making during crisis and long-term perceptions
Such instances of misinformation can emerge from the process of (collective) sense-making during an acute crisis [37], commonly observed in crisis events in general [38].Social media usage during crisis can facilitate this process of sense-making [38], aiding individual psychological mechanisms [38] to deal with for example anxiety, uncertainty and information voids [30,33].
Sense-making involves both collection and framing of data [38].This contributes to situational awareness [39] (crucial for effective responses during disaster situations [23,40]) but also the construction of mental models [39] and the interpretation of 'salient frames' [41].Hence, the use of social media and emerging instances of misinformation during an immediate crisis could also shape long-term perceptions of the crisis and its core subject.
Frames shaping the interpretation of a crisis can emerge from social media discourses but can equally be based on an individual's prior beliefs.In fact, the same research that documents the effect of experiencing weather extremes on climate attitudes also highlights that prior beliefs [21,42], political affiliation [42] and motivated reasoning [20] are key determinants for the effect on climate change attitudes.These factors may lead to strengthened awareness or belief in climate change but may also represent a vulnerability towards misinformation emerging in communications about and during environmental emergencies.Even if corrected, misinformation can have a persistent psychological effect on individuals exposed to it.This 'continued influence effect' [43] highlights the potential long-term significance of misinformation even if it occurs as a spurious phenomenon during a discrete short-term event such as a hurricane, heatwave or wildfire.

A prominent example: misinformation during wildfires
Recently, major long-lasting wildfires have emerged as particularly prominent examples of the large-scale online deliberation of weather-related extremes as evidence for a changing climate.Hashtags such as #climateemergency were widely used in Tweets commenting on the major bushfires in Australia during 2019 and 2020 [44].The broad discussion of the fires as a climate emergency [9] apparently invoked a counter-argument challenging the connection between the fires and climate change, and instead explaining the fires with an unprecedented number of arson cases [44][45][46].A similar pattern emerged during the 2020 fires in California with claims of widespread arson by so-called 'antifa' activists [47].This led to the actual impairment of emergency operations as emergency services were inundated with calls [48] and armed private citizens set up road blocks to catch alleged arsonists [47,49].
Misinformation spread during such an extreme weather-related event can thus have an immediate, shortterm effect but potentially also a lasting, long-term impactan arson rumor spread during a wildfire may impair emergency responses [48] but could also linger as an alternative explanation of the emergency's cause, feed into ongoing, deliberate attempts to undermine trust in climate science, spread climate denial [see for example [50][51][52] and undermine our collective ability to address societal challenges [53].
The referenced examples of wildfiresi.e., misinformation about arson which had the short-term impact of hindering emergency responses but also potentially influences long-term perceptions of climate changespecifically motivate this review, which aims to identify research into online misinformation during extremeweather events and identify examples with the potential of a similar combined short-term/long-term impact pattern.
To this end, research covering misinformation on social media during extreme weather-related events is reviewed, focusing on events that are specifically referenced by the latest IPCC report [2].Events that are expected to become more frequent and severe under climate change, and thus can also be expected to be a frequent backdrop of public online discussions of climate change.
The review is guided by the following research questions: • RQ1: Which events and social media platforms are covered by existing research in this area?
• RQ2: Which misinformation types and dynamics are described and explored?
• RQ3: Which misinformation short-term and long-term impacts are identified, and which countermeasures are recommended?
The paper is structured as follows: section 2 explains the methodology for the literature review, section 3 presents a summary of the results, section 4 discusses the current state of the research into this topic, and section 5 concludes with a summary of the findings and reflections on required future research.

Methods
The presented results are based on a review of peer-reviewed journal articles that combined 1) the topic of misinformation in the context of 2) social media and 3) extreme weather-related events as referenced in the IPCC AR6 WGI report [2].Relevant publications for this review were identified via searches in Scopus and Web of Science.The employed search queries relied on three groups of keyword patterns associated respectively with misinformation, social media, and extreme weather-related events (see table 1).A publication's title, abstract or keywords had to match at least one keyword in all three groups to be considered for the initial screening.
Keyword patterns for extreme weather-related events were defined on the basis of terms referencing these events in the IPCC AR6 WGI report [2].The keywords representing different social media were selected based  The initial screening of those publications focused primarily on a review of abstracts and titles but extended to partial full-text reviews where necessary to assess inclusion and exclusion criteria.The aim was to identify and include research that represented a case study covering one of the extreme event types explicitly mentioned in the IPCC AR6 WGI report [2], and relied on original data from one of the referenced social media platforms, and explored the issue of misinformation.Publications were typically excluded from further review when they were studies with a primary focus on computer science methodologies (e.g., development and testing of algorithms, verification of classification methods), when they were reviews, were not analyzing social media data or represented cases of semantic ambiguity of the search terms (e.g., 'flooded with misinformation,' 'spreading like wildfire').

Results
The review reveals a scarce knowledge base to map the types, dynamics and impacts of online misinformation during extreme weather emergencies.The 13 selected publications nevertheless point at directions for future research to improve the understanding of online misinformation emerging during the type of extreme weather emergencies that are expected to increase in connection with climate change.
In the following the reviewed studies are summarized with reference to the dimensions of interest raised by the three research questions. Figure 1 provides a high-level overview of the main characteristics of the studies included in this review (see also supplementary table S2).[10,16,44,[54][55][56][57][58][59][60][61][62][63] with regard to the main features of the case studies.The extreme weather events studied are primarily storms and wildfires, almost exclusively use Twitter data and focus on events in the USA and Australia; misinformation is mainly analyzed with a focus on emergency responses or climate perceptions, but not both.

Types of social media and extreme weather events
The reviewed studies focus almost exclusively on Twitter as a social media platform to study misinformation during extreme weather events (see supplementary table S2).Only two studies [54,59] use Facebook as a data source, one in combination with Twitter [59].
The 13 studies are all published between 2017 and 2023 covering extreme weather events from 2011 to 2021.Severe hurricanes in the United States are the primary subject of five of the considered publications; the 2017 hurricane Harvey alone features in four of the studies.Wildfires in Australia [10,16,44,63] and California [62] are covered in five studies.Two studies consider localized storm events, in Spain [59] and Canada [54] respectively.Finally, one study uses Tweets posted during the 2015 Chennai floods in India as an example [57] to explore the spread of rumors during a disaster situation.

Misinformation types and origins
Several of the reported misinformation instances seem mundane or merely entertaining without a significance on emergency responses.This includes for example the rumor of a 'shark swimming up the freeway' [55], crocodiles having escaped during the Chennai floods [57] or a video of snow allegedly being plastic purported by YouTube video shared after the 2021 Filomena snow storm in Spain [59].
In contrast, misinformation about the flooding of the New York stock exchange and a hospital fire during hurricane Harvey [56], overflow of a dam [57], misleading information about evacuation orders [60] as well as rumors of ID and immigration status checks at storm shelters during 2017 hurricanes Harvey and Irma [58,60] do have significance for immediate emergency responses.They could negatively impact public behavior and effective emergency responses and even endanger human lives.
Finally, five studies focus on misinformation about arson and a lack of sufficient preventative fire measures ('backburning') as the primary cause of bushfires in Australia and California [10,16,44,62,63].While these rumors could have had a short-term impact on emergency responses (during the 2020 wildfires in the Western United States for example similar rumors inundated emergency services [48]) these studies are also the only examples that explore misinformation during extreme events in the context of public deliberations about climate change, and thus the potential long-term impact on climate perceptions.
With few exceptions ( [10,44,63]) the studies do not address the likely origin and emergence of misinformation, i.e. whether the misinformation is emerging 'organically' (e.g., through existing communities of shared interest) or is deliberately spread, potentially with the help of automated accounts (so-called 'social bots' [64]) or 'trolls' [63,65].

Misinformation impacts and counterstrategies
Reviewed studies that attempt to measure the impact of misinformation focus primarily on methods to either quantify Twitter users' engagement with a topic, hashtag or community [e.g., [58][59][60], or measure polarization of the discourse and/or communities [e.g., 10,16,44,62].The former includes trends in propagation of Retweets, replies or sentiment, the latter uses for example network analysis focusing on user connections (following, activity) and content patterns including sentiment analysis.This may also explain the primary choice of Twitter as a study platform, which provides easy access to data relating to content, engagement, and community structures.But it could equally be argued that the choice of social media platform influences and limits the methods chosen to assess the impact of misinformation during an emergency.
The reviewed studies draw very diverse management and research recommendations from their findings: [60] and [54] for example highlight the importance of early engagement of government agencies and public officials in stopping the spread of misinformation, and in a related finding [58] suggest to publish a database of rumors to counter the spread of misinformation; [56] show that debunking early and across social media platforms appears to be effective; and in a similar vein [63] suggest the publication of preliminary analyses of emerging misinformation through mainstream media as a potential intervention to counter misinformation.
Research needs are also seen at the intersection of social and general online media [10,59], a better understanding of diffusion patterns of misinformation [44,61] as well as with regard to the identification of potential bot accounts [55] (human accounts with bot-like patterns).

Discussion
Research on the role of social media in disaster situations is extensive -Web of Science returns close to 500 results for searches on 'social media' AND 'disaster management', Scopus more than 800.The critical importance of social media as tools for disaster management in general has been demonstrated since their inception [40], and the significance of misinformation during disaster response and recovery is highlighted by several recent studies [23,30].
The turmoil surrounding the recent sale of Twitter is another point in case.It triggered warnings by emergency managers [66] that highlight both the importance of Twitter as a communication tool during natural disasters and the dangers of increasing misinformation.The presented literature review reveals however a limited body of research on misinformation in social media during extreme weather events.Events that are increasing, will further increase, and thus will provide an increasingly common frame of reference for the public discourse on climate change.This highlights a major research gap that is now seeing added urgency and significance, not least given the recent developments in the social media landscape.
Moreover, few studies included in this review explore instances of misinformation that could have an impact on emergency responses during such natural disaster events.Relevant examples include overflow of dams, flooding and fires during the emergency, or the rumor on immigration checks at storm shelters in the US.Several studies focus however on rumors which have high attention value (crocodile escape, shark swimming up freeway), but will have little significance for emergency operations, aside from introducing noise in the general flow of information relevant to an emergency.Only five studies [10,16,44,62,63] actually take up the subject of misinformation and the long-term perception and sense-making of these events, here the linkage to climate change.Considering that this was the initial motivation for the review, this indicates another major research gap.
Whether the analyzed misinformation emerges 'organically' or is spread deliberately also remains largely unexplored.Some of the reviewed studies raise [55] the issue or imply an origin, specifically around the Australia bushfires arson rumor.Here the studies see primarily an organic emergence of rumors through 'sincere activists' [44] and only a limited involvement of coordinated campaigns through social bots and trolls [10,44,63].It has to be acknowledged though that it is generally difficult to determine if misinformation is spread deliberately (i.e., can be termed 'disinformation' [67]) or emerges accidentally 4 without the intent to deceive [71].The reviewed studies are thus subject to the general challenges of mis-and disinformation research and must contend themselves with a pragmatic approach that analyses misinformation without knowledge of the intent [71].It can be argued however that knowledge about the intent is crucial information to develop effective mechanism to counter misinformation with both short-and long-term effects, and should thus be a focus for future research.

Future research framework
The reviewed studies provide however some interesting pointers for the potential structure and foci of future research into the understanding of short-term and long-term impacts of misinformation during extreme weather events, namely that 1) misinformation links across domains (e.g.touching on immigration, general politics), 2) that the impact and role of social media for disaster response and perceptions of climate change has to be studied at different geographical and temporal scales, and 3) that the interlinkage and inherent dynamics of different types of online media has to be considered in order to arrive at a comprehensive assessment that can inform communications during an environmental crisis.
Table 2 provides an overview of these three dimensions, contrasts the current research focus with future opportunities, and suggests sample case studies to explore these new directions.

Misinformation across domains
The COVID-19 pandemic has avidly demonstrated how misinformation extends beyond the immediate subject of a crisis and interacts with misinformation on other societal issues.Several studies show for example how misinformation about the safety of new vaccines, combined with political identification/self-sorting and misinformation in a polarized environment influenced the willingness to observe public health measures or to get vaccinated [34,36].
Among the studies in this review, the immigration status check rumor in the U.S. [58,60] as well as variations of the Australia and California wildfire arson rumors [10,16,44,62,63] should be interpreted from a similar perspective.These examples show how misinformation with relevance to crisis management can co-opt other domains and cross into the realm of political misinformation.This aligns with research on the perception of climate change showing that political affiliation, motivated reasoning and identity notions [43,72,73] influence views on climate change.These influential factors implicitly link the subject of climate change to other domains and contentious issues.It can thus be argued that misinformation during natural disasters is part of a larger connected misinformation ecosystem and can overlap and be influenced by other types of misinformation and political polarization.Disinformation around the August 2023 wildfires on Hawaii, connecting this environmental emergency to U.S. support for Ukraine [74], is the most recent example highlighting the relevance of cross-domain misinformation.Considering that geographical and temporal scales will vary significantly for the extreme weather-related events linked to climate change (i.e., droughts, heatwaves, storms, floods, fires) the observed focus on two event types and one social media platform highlights another urgent research gap.Short-lived local events will likely demand different misinformation management strategies than long-lasting emergencies with a global online deliberation.Long-lasting wildfires or floods allow for different discourses to develop then a hurricane lasting a few days.In contrast to wildfires, hurricanes are difficult to blame on alternative causes, but may more easily foster politicized narratives around emergency responses that could also undermine political action against climate change.
In addition, effective misinformation management and impacts will vary with different properties and user groups of platforms, i.e., regarding demography, content type, network structures and platform functions.[56] for example highlights that debunking works across media platforms.A complete understanding and effective responses of misinformation dynamics thus calls for research that is cross-platform and cuts across temporal and geographical scales.Research gaps: The interplay of different digital media in the amplification of misinformation is acknowledged by several of the reviewed studies but not fully integrated in the research design.To understand misinformation dynamics and achieve representative coverage of public audiences, studies need to focus on the parallel mapping and lifecycle of misinformation instances across multiple digital media.Potential case studies: Map a specific instance of misinformation (e.g., arson during wildfires) across the broadest possible range of digital media with the aim to fully capture the temporal and narrative dynamics of misinformation seeding, modification, and amplification.

Misinformation across media
The example of 'plastic snow' [59] as well as the debunked arson rumors during the Australia bushfires [63] highlight how the interplay of general online news production mechanisms and the incentives of the 'attention economy' [59] in turn influence the diffusion of misinformation on social media.This interplay can not only impact immediate crisis responses but also long-term perceptions and sense-making in relation to crisis situations, including extreme weather emergencies.A recent technical report about the combination of different social media platforms in the spread of arson rumors during wildfires in the U.S. underlines this point [47].Furthermore, the combination of different online media platforms and content types has emerged as a characteristic pattern of recent disinformation campaigns [e.g., 75].There is thus a need for research that extends beyond the focus on specific social media platforms and takes an integrated approach to understand (mis)information flows across platforms and media types.
Ultimately, engagement with online media will always be a limited way to assess the impact of these media and emerging misinformation [76], even when utilizing a broad coverage of interlinked media for misinformation research.Studies connecting online and offline personas are rare [see for example 65], but interviews, surveys, and longitudinal studies should feature more prominently in future research efforts, thus extending common methodological approaches using aggregate measures for misinformation engagement (e.g., likes, shares, replies), propagation (e.g., Retweet networks), or reception (e.g., text sentiment measures).
This review thus highlights that there is a need to develop research frameworks (see table 2) that consider all of these dimensions and develop a larger as well as more standardized toolset to analyze the dynamics and impact of misinformation, and inform communication and debunking strategies for both crisis managers as well as communicators interrogating these events as examples of accelerating climate change.

Conclusions
This contribution reviewed research on misinformation emerging on social media during environmental emergencies such as storms, floods, temperature extremes and wildfires, all of which are expected to become more frequent and severe with climate change.The review was specifically motivated by the question how the impact of such misinformation may extend beyond short-term effects on emergency responses and influence long-term perceptions of climate change.
High-profile emergency events generally are ripe with misinformation which can result in severe consequences.This has been avidly illustrated by misinformation during the COVID-19 pandemic.Since the emergency notion increasingly frames the deliberation (not just online) of climate change mitigation and adaptation it is crucial that we understand the emergence of misinformationboth to prevent immediate short-term negative consequences (human suffering, casualties, and environmental damage) as well as to ensure that societies can arrive collectively at effective long-term conclusions.
This review suggests that we lack a thorough understanding of the connection between short-term misinformation impacts during environmental emergencies and their long-term impact on the perceptions of these events as instances of climate change.The available studies nevertheless provided crucial observations that can help to guide further research into these questions.Namely, that future research needs to focus on misinformation emerging during all types of environmental emergencies and consider the dynamics of misinformation across different domains, media platforms and geographical scales.
Well-funded climate denial campaigns are a reality, and we can expect emergency events to be exploited for the spread of such misinformation.We must ensure that such attempts do not undermine trust in science, policies, or organizations, as this is a key prerequisite for an evidence-based discussion about these extreme weather events, their impacts, and suitable societal responses.
on a list of the most popular social media networks worldwide for 2022 Statista 2 , a partially overlapping overview by Our World in Data 3 and individual additions of other potentially relevant social media; this also included previously popular social media platforms that have declined in relevance recently.The review considers search query results available in either database up to 17. October 2023.The publication searches resulted in 268 Scopus and 153 Web of Science matches, which largely overlapped and yielded 289 unique publications.

Figure 1 .
Figure 1.Characteristics of the 13 included publications[10,16,44,[54][55][56][57][58][59][60][61][62][63] with regard to the main features of the case studies.The extreme weather events studied are primarily storms and wildfires, almost exclusively use Twitter data and focus on events in the USA and Australia; misinformation is mainly analyzed with a focus on emergency responses or climate perceptions, but not both.

Table 2 .
[10]ent and future opportunities of research into short-term and long-term impacts of online misinformation related to extreme weather emergencies and climate change.The example of self-moderated local Facebook groups during a local Tornado[54]versus the centralized countering of immigration check rumors during the 2017 hurricanes Harvey and Irma[58], as well as the arson rumors in Australia that attracted the attention of a global audience[10]demonstrate misinformation diffused at different scales.Here the mechanisms for the emergence and management of online misinformation are characterized by different geographical and time scales.Those imply different impacts of misinformation during an emergency, and, as the studies suggested, different approaches to managing this misinformation.