Introduction
According to Professor Petteri Taalas, the Secretary General of the World Meteorological Association, ‘The number of weather, climate and water extremes are increasing and will become more frequent and severe in many parts of the world as a result of climate change’ (McGrath, Reference McGrath2021). Between 1970 and 2019, there has been an increase in the number of extreme weather events with a threefold increase in the number of recorded disasters and an increase in economic losses, and although the death rate remains very high, it has reduced, mainly due to better early warning systems and responses (World Meteorological Association, 2021). Given the increasing evidence that extreme weather and anthropogenic (i.e. caused by humans) climate change are linked by meteorological science (e.g. Perkins-Kirkpatrick et al., Reference Perkins-Kirkpatrick, Stone, Mitchell, Rosier, King, Lo, Pastor-Paz, Frame and Wehner2022) and in the public eye (e.g. Wong-Parodi and Rubin, Reference Wong-Parodi and Rubin2022), this study considers psychological responses to both extreme weather and climate change. The next two sections introduce key concepts in the climate change and extreme weather literatures, respectively, followed by an outline of the uncertainty distress model (Freeston et al., Reference Freeston, Tiplady, Mawn, Bottesi and Thwaites2020) and how it may apply in this context.
Climate change, psychological distress and climate change action
The links between climate change and psychological distress are now well established. For example, a study of adults from 25 countries across all continents (n=10,143) reported robust weak relationships between stronger negative emotions in response to climate change, increased insomnia, and poorer mental health (Ogunbode et al., Reference Ogunbode, Pallesen, Böhm, Doran, Bhullar, Aquino, Marot, Schermer, Wlodarczyk, Jiang, Salmela-Aro, Hanss, Maran, Ardi, Chegeni, Tahir, Ghanbarian, Park, Tsubakita, Tan, van der Broek, Chukwuorji, Ojewumi, Reyes, Lins, Enea, Volkodav, Sollar, Navarro-Carrillo, Torres-Marín, Mbungu, Onyutha and Lomas2023). Likewise, a scoping review of 26 studies found consistent evidence that young people are worried about climate change and that negative emotions in response to climate change are positively associated with poor mental health (Ramadan et al., Reference Ramadan, Randell, Lavoie, Gao, Manrique, Anderson, McDowell and Zbukvic2023). They also consider a number of differently named constructs such as climate anxiety, climate change anxiety, climate change worry, and eco-anxiety, but also ecological grief and distress, eco-paralysis, environmental concerns, distress, and grief and solastalgia (loss of place). Literature about climate change concerns sometimes uses a language of worry (e.g. Whitmarsh et al., Reference Whitmarsh, Player, Jiongco, James, Williams, Marks and Kennedy-Williams2022) but sometimes uses the more neutral term of concern (e.g. Semenza et al., Reference Semenza, Hall, Wilson, Bontempo, Sailor and George2008; Spence et al., Reference Spence, Poortinga and Pidgeon2012). The point at which emotional reactions or distress becomes problematic in that it impacts on people’s functioning remains to be determined, but it is likely that some people report what would be considered clinical levels of distress (see Whitmarsh et al., Reference Whitmarsh, Player, Jiongco, James, Williams, Marks and Kennedy-Williams2022).
Based on UK survey data (n=1338), Whitmarsh et al. (Reference Whitmarsh, Player, Jiongco, James, Williams, Marks and Kennedy-Williams2022) suggest that climate anxiety may be a motivating factor for some higher effort pro-environmental behaviours (e.g. buying second hand or repurposing) but not for other lower effort behaviours (e.g. buying products with less packaging). A study in Finland (n=2070) found significant medium-strong positive associations between measures of both climate anxiety and climate hope with engagement in climate action and the number of climate mitigation actions (Sangervo et al., Reference Sangervo, Jylhä and Pihkala2022).
Extreme weather and psychological distance of climate change
Psychological responses to extreme weather can be considered at two levels, first as a response to the actual event as a natural or anthropogenic disaster, and second as a proxy response to climate change when extreme weather is considered as evidence for climate change. At the first level, a systematic review of 35 studies on the impacts of extreme weather events in Europe (Weilnhammer et al., Reference Weilnhammer, Schmid, Mittermeier, Schreiber, Jiang, Pastuhovic, Herr and Heinze2021) found risk of cardiovascular and respiratory mortality increases due to extreme heat events and droughts, increased risk of mortality due to extreme cold events, and that mental health morbidity might increase due to wildfires. As a specific example of the impact of extreme weather on mental health, a meta-analysis of studies following up exposure to flooding in the UK in the previous year, found 30.6% of participants reported significant levels of PTSD symptoms, k=4, N=1359; 95% CI 11.68–49.05 (Cruz et al., Reference Cruz, White, Bell and Coventry2020).
The second level, that people’s emotional response to extreme weather is a proxy response for climate change, is predicated on the notion that people draw on their experience of extreme weather as evidence that climate change is happening. The science around the attribution of extreme weather events to anthropogenic climate change has developed rapidly (for a summary, see Clarke et al., Reference Clarke, Otto, Stuart-Smith and Harrington2022), and this discourse can be found in many readily accessible sources, including the UK Government’s Met Office (Attributing extreme weather to climate change - Met Office). The link can also be seen in the ‘local warming’ literature, where a meta-analysis (k=31, n=82,952) found a small but significant effect equivalent to a 1°C increase in local temperature increased worry about climate change by 1.2%, regardless of whether studies had meteorological evidence of the increase in temperature or asked about people’s perceptions (Sugerman et al., Reference Sugerman, Li and Johnson2021). Consequently, there are plausible links between personal experience of weather, extreme weather, broader concerns about climate change, and so distress. The position taken in this study is that the experience of extreme weather and concerns about climate change do not have a one-to-one correspondence, but they are frequently related, and either or both may be related to psychological distress. Consequently, from a clinical perspective where people may present with distress that has multiple determinants, both extreme weather and concerns about climate change should be considered.
The psychological distance of climate change typically consists of three components (see Keller et al., Reference Keller, Marsh, Richardson and Ball2022): geographical and/or spatial distance (i.e. here vs elsewhere), temporal (i.e. now vs in the future), and social (i.e. to people like me vs not like me). Less distance is linked to stronger negative emotional reactions (e.g. Chu and Yang, Reference Chu and Yang2019), and less distance has been linked to more behaviour (including behavioural intentions) towards climate change mitigation (for a review, see Maiella et al., Reference Maiella, La Malva, Marchetti, Pomarico, Di Crosta, Palumbo, Cetara, Di Domenico and Verrocchio2020). Brügger (Reference Brügger2020) argues that distance may be seen as a set of beliefs that may change and are being updated as more information from different sources becomes available.
Extreme weather, climate change and the uncertainty distress model
Based on treatment development work and conceptual work in the years before the pandemic, Freeston et al. (Reference Freeston, Tiplady, Mawn, Bottesi and Thwaites2020) proposed a trans-situational model of uncertainty distress and considered how it may apply to the pandemic (see Fig. 1).
Figure 1. The trans-situational model of uncertainty distress (Freeston et al., Reference Freeston, Tiplady, Mawn, Bottesi and Thwaites2020; p. 9).
The model is trans-situational and can be applied to a range of settings. In the context of extreme climate change, each of the elements takes on a specific meaning. Each element is addressed in turn below.
A real-world situation
In this case the situation is both anthropogenic climate change in a narrow sense as the warming of the climate, but in the broader sense it is the implications for individuals, their choices, society, humankind, the biosphere, and the planet. In a more immediate sense, extreme weather is an actual event which people may experience acutely, or may serve as a proxy for evidence of climate change. Like many real-world situations there will be levels of threat that may easily be recognised. However, even when the situation happens there is likely to be uncertainty about the degree of those threats, the exact cause or timing of the event, how the event will unfold, whether it will have long lasting impacts, and/or the probability of re-occurrence.
Actual and perceived threat
Extreme weather may represent actual threat with the possibly of loss, damage, injury, etc. Actual threats by climate change to all levels of human existence can be demonstrated in the general sense, but indicators may also be carbon taxes, restrictions on emissions, and other limitations on people’s actions. Within climate change psychology, psychological distance is one element of perceived threat, but it may be further operationalised as the likelihood of bad things happening, how bad they will be, and how imminently they will happen in line with models of anxiety (Riskind et al., Reference Riskind, Williams, Gessner, Chrosniak and Cortina2000; Salkovskis, Reference Salkovskis1991).
Life disruption
People’s lives may be disrupted subjectively, by the sense that the world they live in is broken or breaking, that the world is no longer safe, or that they no longer have the freedom of action that they previously enjoyed. For flooding, there is likely to be some disruption to transport and deliveries, and often power, water, drainage/sanitation, even when an individual’s accommodation is not directly affected. In a cohort of people exposed to flooding in the UK in 2013–2014, greater depression, anxiety and PTSD were reported 12 months later with actual experience of flooded accommodation, disruption to domestic utilities, deeper floodwater and disruption to health/social care and work/education (Waite et al., Reference Waite, Chaintarli, Beck, Bone, Amlôt, Kovats, Reacher, Armstrong, Leonardi, Rubin and Oliver2017). Among those households that were actually flooded at the time, significantly higher rates of anxiety, depression and PTSD were detected among those who were evacuated (Munro et al., Reference Munro, Kovats, Rubin, Waite, Bone, Armstrong, Beck, Amlôt, Leonardi and Oliver2017). Finally, at two-year follow up, significantly greater psychological morbidity remained among those who reported persistent damage to properties (Jermacane et al., Reference Jermacane, Waite, Beck, Bone, Amlôt, Reacher, Kovats, Armstrong, Leonardi, James and Oliver2018). Thus, both the degree and duration of disruption are predictive of greater and prolonged distress.
Actual and perceived uncertainty
High degrees of uncertainty often exist around extreme weather in terms of occurrence, duration, severity, and whether it occurs as predicted. For example, significantly higher rates of depression and PTSD were detected one year later among those who had not received a warning compared with those who had at least 12 hours warning; worse outcomes were reported when the initial event had greater uncertainty (Munro et al., Reference Munro, Kovats, Rubin, Waite, Bone, Armstrong, Beck, Amlôt, Leonardi and Oliver2017). Uncertainty about climate change is harder to operationalise, because at the limit these are the ‘unknown unknowns’, i.e. we do not know what we do not know. We have a series of ‘facts’ that are based on past data to model the present and the future. The degree to which people reject that information or do not integrate it into their beliefs about the world can be said to reflect their actual uncertainty about climate change.
On the other hand, perceived uncertainty is defined within the model by their endorsement of statements about what is not known about climate change, its causes or whether it is happening. The perception (after McCloskey, Reference McCloskey1996) is that vital information is missing, or available information is conflicting, noisy, or confusing. The final element of uncertainty, also identified by McCloskey (Reference McCloskey1996) is mistrust or distrust and questions the sources or information and/or the motivations behind the information.
Dispositional and situational intolerance of uncertainty (IU)
Dispositional IU is defined as ‘a tendency to be bothered or upset by the (as yet) unknown elements of a situation, whether the possible outcome is negative or not’ (Freeston et al., Reference Freeston, Tiplady, Mawn, Bottesi and Thwaites2020; p. 6, original emphasis). It is a construct with both time-invariant and time-varying components (Knowles et al., Reference Knowles, Cole, Cox and Olatunji2022). It is conceptualised as relatively stable, but from the perspective of the somatic error theory of uncertainty (Freeston and Komes, Reference Freeston and Komes2023), disruption and so unsafety will lead to elevations in the current level of IU, whereas re-establishing safety should lead to a return to longer term or typical levels for the individual. The greater the degree and duration of disruption, the greater the experience of unsafety, and so the longer the elevation of IU. Situational IU is the extent to which people are bothered by the uncertainty they experience as aversive in a given situation. It is psychometrically distinct from perceived threat in that situation (see Pepperdine et al., Reference Pepperdine, Lomax and Freeston2018), and both situational IU and threat contribute to anxiety and other forms of distress.
Uncertainty reducing behaviours
These behaviours have the primary function of reducing uncertainty rather than reducing threat. Although they may take many forms specific to the situation, information seeking is especially relevant in the context of climate change and extreme weather (e.g. Hmielowski et al., Reference Hmielowski, Donaway and Wang2019; Whitmarsh et al., Reference Whitmarsh, Player, Jiongco, James, Williams, Marks and Kennedy-Williams2022).
Uncertainty distress
In this model, distress is assessed with reference to a situation rather than ‘in general’. Distress is defined as heightened negative emotional reactions to the situation (as well as reduced positive reactions) and can be measured as one of the several measures of climate, eco- or environmental anxiety or distress (Ramadan et al., Reference Ramadan, Randell, Lavoie, Gao, Manrique, Anderson, McDowell and Zbukvic2023). Alternatively, adjustment disorder (O’Donnell et al., Reference O’Donnell, Agathos, Metcalf, Gibson and Lau2019) has some conceptual advantages as it links the response to the stressor and includes the possibility of adaptation in the resolution (Chen et al., Reference Chen, Bagrodia, Pfeffer, Meli and Bonanno2020). For extreme weather, the event will pass and its consequences may resolve, but for climate change if understood as continuing to happen, adaptation may be the only response possible in the longer term. Finally, if exposure to extreme weather is relevant, then post-traumatic-like symptoms may also be relevant as an indicator of more extreme stress response to a potentially threatening situation (for a meta-analysis of PTSD following extreme weather, see Chique et al., Reference Chique, Hynds, Nyhan, Lambert, Boudou and O’Dwyer2021).
Aim and approach
The aim of this study is to integrate constructs from two literatures, namely the broader climate change distress literature and the uncertainty distress model to better understand people’s responses to extreme weather and climate change. Network analysis has gained popularity in clinical psychology as it approaches psychological disorders as functionally related symptoms, i.e. the symptoms constitute the disorder and are not caused by some underlying latent cause (for a full discussion, see McNally, Reference McNally2021). The uncertainty distress model proposes that distress and associated behaviours arise from the interaction between uncertainty and threat in response to a disruptive stressor, in this case either extreme weather and/or climate change. Therefore, a network analysis approach has been chosen to (1) consider extreme weather and climate change simultaneously, (2) integrate two sets of constructs from an emerging climate change distress/climate change action model and a more explicit UD model, (3) conduct an exploratory study of these variables, and (4) capitalise on the visual nature of networks that have some similarities with psychological formulations or case conceptualisations by describing or hypothesising the potential links between different observable and measurable factors.
Method
Participants
Links to the survey were distributed on social media, student research participant schemes, and list holders/moderators of community groups were asked to post the link, including those addressing weather events and climate change. The sole inclusion criterion was adults, aged 18 and over. However, potential participants were informed that there would be questions about extreme weather events that could be upsetting for some people and so they may wish to not participate if this were likely. Participants not seeking credits were offered the opportunity to be entered into a draw to win a voucher (£20 per 50 participants). Overall, 395 people clicked on the survey link. Of these, 25 clicked only without consenting (6.3%), 12 consented only (3.0%), eight provided socio-demographic data only (2.0%), 23 provided data about extreme weather and climate change only (5.8%), the remainder (n=327, 82.8%) completed significant amounts of the survey.
There were 48 men (14.7%), 276 women (84.4%), two did not identify with either and self-described their gender in the text box available (0.6%), and one (0.3%) preferred not to say. Participants were aged 18 to 75, the mean age was 22.1 (SD=8.8), and 89.5% were aged 18 to 25. English was first langague for 93.5%, and 98.2% had received education in English to high school level; all but one partcipant rated their proficiency as moderate to excellent. The large majority were currently residing in the UK (n=310; 96.9%), with five from Asia, three from Europe, and one each from Africa and the Americas. There were 20 (6.1%) who described themselves as Asian or Asian British, three (0.9%) as Black, African, Black British or Caribbean, 17 (5.2%) as mixed or multiple ethnic groups, 285 (87.2%) as White, and two as Arab (0.6%). Four (1.2%) described themselves as high school graduates, 176 (53.8%) reported A-levels as their highest qualification, 121 (37.0%) reported some or completed undergraduate study, two (0.6%) reported vocational qualifications, and 24 (7.3%) held postgraduate qualifications. Most (n=257, 79.3%) were students.
Measures
Standardised measures of psychological distress
The following five measures assess symptoms of worry, anxiety, depression, post-traumatic stress symptoms, and adjustment disorder.
Penn State Worry Questionnaire-3 (PSWQ)
Berle et al. (Reference Berle, Starcevic, Moses, Hannan, Milicevic and Sammut2011) developed an ultra-brief version of the PSWQ from the full 16-item PSWQ (Meyer et al., Reference Meyer, Miller, Metzger and Borkovec1990). The three items measure worry on a 5-point Likert scale, from 1 (not at all typical of me) to 5 (very typical of me). Kertz et al. (Reference Kertz, Lee and Björgvinsson2014) established a cut-score of 11 with sensitivity = 0.71 and specificity = 0.73.
Generalized Anxiety Disorder Questionnaire 2 (GAD-2; Kroenke et al., Reference Kroenke, Spitzer, Williams, Monahan and Löwe2007, Reference Kroenke, Spitzer, Williams and Löwe2010)
The GAD-2 consists of the first two items of the GAD-7 rated on a 0 (not at all) to 3 (nearly every day) scale over the last two weeks. Scores range from 0 to 6, and a cut-off score ≥3 indicates clinically significant anxiety.
Patient Health Questionnaire-2 (PHQ-2; Kroenke et al., Reference Kroenke, Spitzer and Williams2003)
The PHQ-2 consists of the first two items of the PHQ-9 rated on a 0 (not at all) to 3 (nearly every day) scale over the last two weeks. Scores range from 0 to 6, and a cut-off score ≥3 indicates clinically significant depression.
International Adjustment Disorder Questionnaire (IADQ; Shevlin et al., Reference Shevlin, Hyland, Ben-Ezra, Karatzias, Cloitre, Vallières, Bachem and Maercker2020)
The IADQ assesses the core features of adjustment disorder (AdjD) outlined in the International Classification of Disease (ICD-11; World Health Organization, 2022). The questionnaire consists of three main sections; the first asks about nine stressful life events, but replaced here with ‘climate change’. The second section assesses six core symptoms of AdjD, related to the index event covering both pre-occupation and failure-to-adapt in the last month. The third section has three items measuring functional impairment. The measure is scored algorithmically for the five criteria for AdjD; if all five are met then the person meets criteria for AdjD. The items in the last two sections were also summed as an indicator of psychological distress.
Primary Care PTSD Screen for DSM-5 (PC-PTSD-5; Prins et al., Reference Prins, Bovin, Smolenski, Mark, Kimerling, Jenkins-Guarnieri, Kaloupek, Schnurr, Pless Kaiser, Leyva and Tiet2016)
This measure assesses five core symptoms of PTSD in reference to exposure to a traumatic event rated yes or no. Scores range from 0 to 5. A cut-off score of 4 indicates likely PTSD. Participants were presented with the questionnaire only if they had experienced extreme weather in their area in the last five years.
Psychology of climate change measures
Three core concepts from the psychology of climate change literature were assessed, namely, distress, concern and action.
Climate change distress
Eleven items were rated on a 9-point Likert scale from 0 (not at all) to 9 (extremely) in response to ‘how you feel in response to extreme weather and climate change’. In line with broad definitions of climate change and ecological distress, five were in the anxious register (anxious, concerned, panicked, scared, worried), three may be experienced in the transgressor role of moral injury (ashamed, regretful, guilty) while three were in the register of the transgressee role of moral injury (angry, frustrated and betrayed).
Climate change concerns
Ten common concerns from the literature were assessed on a 0 (not at all) to 8 (extremely) 9-point Likert scale reflecting a range of egoistic, altruistic and biospheric concerns (e.g. Helm et al., Reference Helm, Pollitt, Barnett, Curran and Craig2018). The instructions were framed neutrally rather than as worry or fear; one can express concern without necessarily engaging emotionally.
Climate change action
Positive climate change actions were assessed with 19 items covering both simple behaviours where many will engage (e.g. walked or taken public transport instead of driving) to those that environmental activists might do (e.g. took part in a protest/rally about a climate change issue). Items were drawn from the Environmental Action Scale (Alisat and Riemer, Reference Alisat and Riemer2015), from Brody et al. (Reference Brody, Kang and Bernhardt2010), and the Pro-Environmental Behaviour Scale (Dono et al., Reference Dono, Webb and Richardson2010) and were rated on a 5-point Likert scale from 0 (never) to 4 (frequently) for the last six months.
Disruption measures
Disruption was assessed with a composite measure of exposure to extreme weather events, and personal experience of extreme weather and mitigation (see Supplementary material for construction of the composite), and two face-valid items measuring subjective disruption.
Extreme weather
A questionnaire was adapted from the Household Natural Hazards Preparedness Questionnaire (Williamsburg Emergency Committee, 2004) and measures exposure to eight different types of extreme weather, the number of exposures to extreme weather in the last five years, and 12 items about the impacts of extreme weather on themselves or family and friends.
Personal experience of extreme weather
A single item measures personal experience of extreme weather on a 7-point Likert scale from strongly disagree to strongly agree (Fownes and Allred, Reference Fownes and Allred2019).
Mitigation measures
Mitigation measures were assessed for each type of extreme weather they had experienced based on a questionnaire developed for flood mitigation (Brody et al., Reference Brody, Kang and Bernhardt2010). The number of mitigation methods varied according to weather type (see Supplementary material).
Subjective disruption
The first item assessed how much the participant’s life had been disrupted by extreme weather and its impacts; the second item was about climate change more generally and its impacts. Each was rated on a 9-point Likert scale from 0 (not at all) to 8 (extremely).
Threat measures
The concept of threat was operationalised with two measures.
Perceived proximity of climate change
Perceived proximity of climate change was assessed with 19 items drawn from Jones et al. (Reference Jones, Hine and Marks2017) covering construal of climate change on geographic (6 items), temporal (8), and social (5) dimensions. Participants indicated agreement with each statement on a 7-point Likert scale from 0 (strongly disagree) to 6 (strongly agree). Each subscale scale was scored with higher scores in the direction of close proximity, i.e. climate change is happening here, now, and to people like me.
Perceived threat
This consisted of two items measuring likelihood and severity of something bad happening, with two others assessing imminence and how dangerous extreme weather or climate change would generally be for the participant. Participants rate each question on a 9-point Likert scale from 0 (not at all) to 8 (extremely).
Uncertainty measures
The three uncertainty variables consisted of rating climate change facts as lacking or having a factual basis (actual uncertainty), agreement with statements expressing uncertainty about climate change (perceived uncertainty), and actual trust/mistrust in information sources (mistrust).
Factual basis of climate change
Fourteen statements derived from climate change science were stated in an affirmative but neutral manner (where predictions were concerned, words like ‘could’ were used) and were rated on a 7-point Likert scale from 1 (completely disagree) to 7 (completely agree).
Perceived uncertainties in climate change science
Five items measured agreement with statements about the degree to which climate change science is not agreed and two items (reversed scored) that it is agreed/happening (see Supplementary material). The items were rated on a 7-point Likert scale from 1 (completely disagree) to 7 (completely agree).
Mistrust in the communication of climate change science
Four items measured trust in different sources of information in climate change (governments/international agencies, scientists, traditional/mainstream media, social media), one item about trust in one’s own preferential media, one item about mistrusting because of scientific disagreement, and one item about only trusting officially rejected or marginalised sources. Participants rated each question on a 9-point Likert scale from 0 (not at all) to 8 (extremely), and the total scale was scored in the direction of mistrust.
Intolerance of uncertainty measures
Four measures cover dispositional IU, situational IU, information seeking, and uncertainty reducing behaviours.
Dispositional intolerance of uncertainty
This is measured with a brief version of the Intolerance of Uncertainty Scale (IUS-5; Bottesi et al., Reference Bottesi, Mawn, Nogueira-Arjona, Romero Sanchiz, Simou, Simos, Tiplady and Freeston2020), a 5-item version of the Intolerance of Uncertainty Scale (IUS; Freeston et al., Reference Freeston, Rhéaume, Letarte, Dugas and Ladouceur1994), the gold standard measure for IU. Participants rate each item on a 5-point Likert scale from 1 (not at all characteristic of me) to 5 (extremely characteristic of me).
Situational intolerance of uncertainty
This scale was based on Pepperdine et al. (Reference Pepperdine, Lomax and Freeston2018) and consists of four pairs of items; the first of each pair refers to the uncertainty of a target item and the second is how bothered or upset the participant is by the uncertainty. The four target items are (the uncertainty of) not knowing enough about climate change, what is being done in response to climate change, what will happen in the future, and the longer-term impact of climate change. Participants rate each question on a 9-point Likert scale from 0 (not at all) to 8 (extremely), and a total score is calculated.
Information seeking
Four questions asked about time per day engaging with the information sources about climate change concerns within a typical day, namely, traditional media, social media, other sources (e.g. friends, colleagues, etc.), and alternative media sources (e.g. media sources which challenge or oppose the mainstream viewpoint). These items were rated on a 1 to 10 scale from 1 (not at all), through 2 (less than 15 min) and 3 (15–30 min), up to 10 (>7 h).
Uncertainty reducing behaviours
The short version of the Intolerance of Uncertainty Behaviours in Everyday Life (IUBEL; Clifford et al., Reference Clifford, Hardcastle, Lambert, Beckwith, Bottesi, Wilkins, Mclean and Freeston2015) consists of the original six items with one additional item developed within the context of the pandemic (I am acting as if the worst is happening right now). The items are rated on a 9-point Likert scale from 0 (not at all) to 8 (extremely).
Data analysis
All preparatory analyses were conducted with SPSS (version 28.0) and the network analyses with JASP version 0.17.03. Data missing at the item level was replaced with the mean of the participant’s items if less than 30% of the items were missing on that scale. At the sample level, there was 0.80% missing data overall, with no more than 2.80% for a given variable; 91.19% of participants provided complete data. Due to the fact that the majority of the missing data was towards the end of the survey, the data were not missing completely at random (Little’s MCAR test, chi-square = 249.306, d.f.=200, p=.010). Multiple imputation and full information maximum likelihood were not available with the software used, so expectation maximisation was used.
Network analysis is evolving and determining the required sample size is not yet well established; there is no equivalent of power analysis. There are established post hoc methods for considering the adequacy of the sample by estimating the stability of the network on increasingly smaller bootstrapped subsamples of the original sample. If the average correlation between the estimates from the full sample and those from the reduced sample remain sufficiently high, then the network is said to be stable and interpretable with caution. The benchmark with 50% of the sample removed is that the average correlation should >=.70, indicated by a correlation stability coefficient CS (.70) >= .50 (Epskamp et al., Reference Epskamp, Borsboom and Fried2018).
Given that EBIC Glasso is sensitive to non-normality (e.g. Slipetz, Reference Slipetz2021), four variables with skew > |1| were transformed using the Box-Cox transformation (Wessa, Reference Wessa2021). Following transformation, skew on three was reduced to <0.20, but for the fourth (post-traumatic stress symptoms) skew reduced from 2.22 to 1.41, but remained skewed.
Procedure
The study was a single group online survey on a secure dedicated survey platform (QualtricsTM). An information sheet was followed by a consent form, socio-demographic variables, extreme weather variables followed by the climate change and uncertainty distress measures used in the network analysis. The standardised measures of distress followed in randomised order and the survey was completed with a debrief sheet. On the information sheet and debrief, participants who so desired were signposted to public sector, professional association, or charitable sources for generic mental health advice as well as specific sources for climate change anxiety. Recruitment took place between 19 November 2020 and 5 February 2021.
Results
Extreme weather
As can be seen in Table 1, the most frequent experiences of extreme weather were flooding (55.9%) and strong winds (29.1%); approximately one-third had no exposure to extreme weather in their area, one-third exposure to one type and one-third to two or more types. In terms of episodes, just over a quarter (27.1%) reported no exposure in the last five years, 44.6% reported 1–2 episodes, with 28.1% reported 3 or more episodes. In terms of being personally affected by extreme weather, 44.3% agreed they had been affected, but to different extents
Table 1. Exposure to extreme weather
As can be seen in Table 2, of those who had experienced one or more episodes of extreme weather (229), over half (54.4%) had received a warning. Three-quarters (75.1%) reported no impacts, but the rest reported one or more of taking precautionary measures, material damage, financial loss, insurance problems or personal injury. Just over half (53.2%) knew of family or friends who had to take one or more of precautionary measures, suffered damage, injury or even death.
Table 2. Impact on self and on family and friends
Reliability and descriptive statistics
As can be seen in Table 3, the majority of scales had adequate reliability (< .70) apart from some of the short scales (e.g., disruption with two items).
Table 3. Reliability and descriptive statistics for all variables
a Derived from principal component score from Supplementary material Table S2 based on six indices of extreme weather (mean = 0.00 and SD = 1.00). The component score has been arithmetically rescaled by multiplying by 10 and adding a constant (14) and rounded to result in a value 0–47. The transformed score correlates at 1.00 with the raw principal component score.
b These variables exhibit excessive skew, > |1.00|, and scores were transformed by Box-Cox transformation for network analyses.
c These variables represent adjustment disorder symptoms to climate change generally and post-traumatic stress symptoms in response to exposure to extreme weather events specifically.
Psychological distress
Based on the established cut-offs for the standardised measures, 20.2% of the sample met criteria for excessive worry, 30.9% scored above the GAD-2 cut-off for anxiety, and 34.7% on the PHQ-2 for depression. In relation to climate change specifically, 8.5% met criteria for adjustment disorder. Although the pre-occupation criterion was met by many (65.3%), fewer met failure to adapt (23.0%), and only 10.4% met impairment criterion. With respect to extreme weather events, 9.8% met criteria for PTSD. In summary, while there were high levels of general distress at around 20–35%, specific distress responses to climate change and extreme weather potentially in the clinical range were around 10% or below.
Network analysis
The 18 variables described above were entered into the network analysis and are represented as circles or nodes in Fig. 2. The nodes are colour coded to reflect the conceptual grouping of the variables. Positive associations or edges are the lines in blue and negative associations or edges are the lines in red; the thicker the edge, the stronger the association. There were 72 out of 153 non-zero edges (sparsity = .529). The full matrix may be found in the Supplementary material (Table S2). In line with articles reporting network analysis, a narrative description of the network is followed by reporting estimates of accuracy and stability
Figure 2. Network analysis of climate change and uncertainty distress variables.
As can be seen starting on the right-hand side of Fig. 2, starting with perceived distance or proximity of climate change as ‘happening to me’ and ‘happening here’ have strong edges, along with a slightly weaker edge to the perception of ‘happening now’. Unsurprisingly, there is a strong edge between ‘happening to me’ to perceived threat (likely, severe and imminent danger) which in turn has a strong edge to emotional distress (e.g. anxious, guilty, angry) and a somewhat weaker edge to subjective disruption and situational IU, and an even weaker edge to adjustment disorder symptom severity.
Environmental concerns (egoistic, altruistic and biospheric concerns) also have strong edges to distress and subjective disruption. The composite variable of exposure to extreme weather has an edge to subjective disruption and importantly to post-traumatic stress symptoms. In turn, post-traumatic stress symptoms edges have moderate edges with information seeking and environmental action (which also have a strong edge between them), and also with dispositional IU.
Dispositional IU has moderate to weak edges with the other clinical variable, AdjD symptom severity, as well as with perceived uncertainty (the perceptions of what is not known), situational IU and IU behaviour. Dispositional IU also has a weak negative edge with environmental action, that is, the greater one is intolerant of uncertainty, the less likely one is to take action. While dispositional IU is on the periphery of the network, situational IU is at the centre. Situational IU predictably has edges with dispositional IU and uncertainty behaviours, but also with temporal proximity, perceived threat, and symptom severity.
The top group of uncertainty variables, actual uncertainty about what is known, perceived uncertainty about what is not known, and mistrust have strong edges between them and strong but negative edges with the perception that climate change is ‘happening now’: low temporal proximity is associated with higher uncertainty about climate change. Temporal distance effectively links the uncertainty variables to the rest of the network. Perceived uncertainty has weak, positive edges with information seeking but a negative edge with climate action: scoring high on what is not known about climate change is associated with information seeking inaction. Moreover, uncertainty about what is known is associated with lower distress. Finally, there is a weak positive edge between mistrust and PT symptoms.
Finally, AdjD symptom severity has positive edges with perceived threat, climate change distress, situational and dispositional IU and a weak edge with PTS symptoms. It also has positive edges with the behavioural variables of pro-environmental action, information seeking and IU behaviour.
Centrality
While the visual representation is the first level of analysis, it is also relevant to consider the centrality or relative importance of the nodes: strength is the sum of the absolute edge weights and expected influence is the sum of raw edge weights, so effectively allows positive and negative weights to cancel each other out. As can be seen in Fig. 3, the normalised centrality indices have been plotted (relative indices are in the Supplementary material). The nodes with the highest strength and expected influence are adjustment disorder, perceived threat and environmental distress. In contrast, temporal proximity (‘happening now’) has high strength, and low influence because of strong positive and negative edges. Other nodes with higher expected influence are social proximity (‘happening to me’), subjective disruption, situational IU, and uncertainty behaviour.
Figure 3. Centrality plot of normalised strength and expected influence indices.
Accuracy and stability
While the edge weights have been estimated to construct Fig. 2, their accuracy needs to be considered. One approach is bootstrapping where in this case 1000 random subsamples have been taken from the dataset allowing estimation of the confidence intervals (Epskamp et al., Reference Epskamp, Borsboom and Fried2018). Inspection of the plot of the edges (Supplementary material, Fig. S1) suggests that positive edges greater than .100 and negative edges less than –.150 have confidence intervals that do not include zero. Thus 27 positive and two negative edges appear estimable; these are the thicker edges in Fig. 2 and should be given the most consideration. Only these 29 edges will be significantly different from edges constrained by the lasso model to zero weight. The 27 significant positive edges are not significantly greater than the other 37 positive edges, nor are the two significant negative edges significantly different from the other six negative edges.
Inspection of node centrality accuracy indicates that the six nodes with greatest strength indices have significantly greater strength than the six nodes with least strength (see Supplementary material, Fig. S2); the remaining six nodes have strength indices that fall in between and do not differ from either the stronger or the weaker. This simply means we cannot be confident about the other weaker edges or nodes in the model, and should base our cautious interpretation on the stronger edges and nodes
The average correlation between the estimates from the full sample and those from the reduced sample remained at .70 down to 30% of the sample, although lower CIs fell below .70 at about 32% of the sample. The correlation stability coefficient CS(.70)=.68, greater than the benchmark figure of 50%, CS(.70)=.50, indicating that the order of edge strength is interpretable with caution (Epskamp et al., Reference Epskamp, Borsboom and Fried2018). A similar approach is used for strength indices. Again, CS(.70)=.68, indicating that node strength reported above is interpretable with caution.
Discussion
The aim of this study was to integrate constructs from two literatures, namely the broader and emergent climate change distress literature and the uncertainty distress model, in order to better understand people’s responses to both extreme weather and climate change. A significant number of participants reported exposure to extreme weather and a range of consequences were reported. While clinical levels of distress in terms of worry, anxiety and depression were reported by 20–30% of the sample (unsurprising in the first year of the pandemic), less than 10% reported either clinical levels of adjustment disorder symptoms specifically in response to extreme weather/climate change and/or post-traumatic stress symptoms in response to an extreme weather event.
Climate change and psychological distress
In line with much of the literature, pro-environmental behaviours were associated with distress (e.g. Sangervo et al., Reference Sangervo, Jylhä and Pihkala2022; Whitmarsh et al., Reference Whitmarsh, Player, Jiongco, James, Williams, Marks and Kennedy-Williams2022) as measured in this study by the severity of AdjD symptoms and environmental distress. While climate change concerns were linked to emotional responses, they were not linked directly to actions. A more granular level of analysis of different types of behaviour may be required to unpick whether, for example, concerns may link to actions that are lower effort and part of general societal responses to climate change (e.g. informing day-to-day consumer decisions), while distress may link to higher effort behaviours, activism, or fundamental lifestyle changes. Again, consistent with the literature, the geographical, temporal and social proximity variables were related to each other but were not redundant (e.g. Keller et al., Reference Keller, Marsh, Richardson and Ball2022), with temporal proximity at greater distance from the others. Moreover, temporal proximity was linked to action, but not to the distress variables suggesting that believing climate change is happening now may be a ‘cooler’ driver of pro-environmental behaviours. Finally, in line with the literature cited earlier (Cruz et al., Reference Cruz, White, Bell and Coventry2020), there was a link between degree of exposure to extreme weather and PTSD symptoms. In conclusion, the relationships between the four psychology of climate change variables were broadly consistent with previous literature. The next section discusses the integration of these into the UD model.
Extreme weather, climate change and the uncertainty distress model
The most familiar part of the UD model is the link between perception of threat drawn directly from cognitive models of anxiety (e.g. Salkovsksis, Reference Salkovskis1991) with perceptions that the consequences of an event will be likely, bad and imminent. The social aspect of psychological distance expressed in neutral terms, namely that climate change is ‘happening to me’ was linked to perceived threat (something likely, bad and imminent). In turn, threat was associated both directly to AdjD symptom severity and indirectly through environmental distress, thus supporting one the key parts within the UD model.
Temporal proximity as well as its link to action noted above, appears to act as the link to the three uncertainty variables (actual, perceived, and mistrust) which formed an almost separate part of the network. As noted earlier, temporal proximity (‘happening now’) has high strength (the sum of absolute weights of the edges) but lower expected influence (where positive and negative edges cancel each other out). It has been argued that in some psychopathology networks such nodes are not important as they may act as deactivators (e.g. Robinaugh et al., Reference Robinaugh, Millner and McNally2016). In this case deactivating one part and activating another may be important: if the perception that climate change is happening ‘now’ increases (perhaps through extreme weather), then uncertainty about climate change may decrease and engagement in pro-environmental action may increase.
Against the predictions of the UD model developed in the context of the pandemic, perceived uncertainty in the case of climate change was not linked to distress. It is consistent with the climate change literature where not accepting the reality of climate change is a less distressing position (Wullenkord and Reese, Reference Wullenkord and Reese2021). However, as a person becomes less uncertain and so sees climate change as ‘happening now’ and if climate change is also seen happening ‘to me’, then the network suggests that there will be a greater perception of threat, and so increased likelihood of distress. Thus, the psychological distance or proximity variables, especially temporal proximity, may play an important modulating role in the network and should not be neglected, despite low expected influence (but high strength).
In the uncertainty distress model, dispositional IU is seen as having a modulating role and this is partially reflected in the network in that it is associated both with perceived uncertainty as well as to situational IU. It also plays its familiar transdiagnostic vulnerability role with edges to both post-traumatic stress symptoms and AdjD symptom severity. In contrast to the peripheral position of dispositional IU, situational IU is quite central in the model with edges to environmental distress and AdjD severity as well as to perceived threat, once again consistent with the model. There is also a link to generic uncertainty-reducing behaviours which are in turn are linked to environmental action and AdjD severity, consistent with the UD model that these various behaviours may be part of a maintenance cycle. The edge between information seeking and AdjD symptom severity is also consistent with reciprocal roles in a maintenance cycle.
Exposure to extreme weather (part of the real-world disruption) was linked to subjective disruption, as predicted by the UD model, but against predictions from more recent theoretical work (Freeston and Komes, Reference Freeston and Komes2023), there was no link between subjective disruption and IU. However, disruption was not well operationalised in this study as it did not consider what or how had been disrupted in terms of routine, sense of belonging, familiar signals of safety, etc.
In summary, the observed network is consistent with several parts of the UD model, but the role of disruption and the contribution of uncertainty to distress are not. In the first case this could be due to the limited way in which disruption was measured before more recent theoretical developments (Freeston and Komes, Reference Freeston and Komes2023). Whereas being uncertain about what is happening may be distressing in some situations (e.g. the pandemic), being uncertain about whether climate change is happening may be a more comfortable position than the certainty that climate change is happening here, now and to me. For climate change at least, uncertainty may be linked to denial or used to limit responsibility, and so reduces distress and need for action. Consequently, uncertainty about climate change may need to be reduced in order to lead to greater action, perhaps by reducing psychological distance and making climate change seen as happening ‘now’ and ‘here’. However, reducing uncertainty may also lead to increased distress, in part through increases in perceived threat, especially if climate change becomes more personally proximal, that is, happening ‘to me’.
Networks based on cross-sectional data have much in common with disorder specific models in that they examine and describe group level relationships between variables, usually at one point in time. Individual formulation draws on both nomothetic knowledge (i.e. the models and the supporting evidence) and idiographic knowledge (i.e. this specific person, their history, current situation, values, strengths and difficulties) to develop a shared understanding (e.g. Dudley et al., Reference Dudley, Kuyken and Padesky2011). Networks differ from individual formulations in that any or all of the relationships identified in the network may not apply to a specific individual, nor do they capture the changing relationships over time as the person reacts to a situation or event as it unfolds or due to therapy. Networks, like empirically supported models of anxiety disorders, OCD, and eating disorders identify factors that may be relevant for a specific person reporting concerns or distress in a particular area.
Clinical implications
From an uncertainty perspective, some level of climate distress may be a normative and indeed adaptive reaction to climate change, as both the uncertainty and indeed threat are real (Whitmarsh et al., Reference Whitmarsh, Player, Jiongco, James, Williams, Marks and Kennedy-Williams2022). Unlike anxiety disorders, a sole focus on increasing tolerance to uncertainty as one might consider with GAD may not be first line of intervention. For example, it may be helpful to first use interventions re-establishing a sense of safety (Freeston and Komes, Reference Freeston and Komes2023), not just in the sense of reducing danger (which may not ultimately be possible in the case of climate change), but that a felt sense of safety may be portable, adaptable or re-constructable through maintaining or (re-)establishing low level routines, and perhaps in this context through social connectedness. Consistent with the sparse climate anxiety intervention literature, using an evidence-grading approach, Bingley et al. (Reference Bingley, Tran, Boyd, Gibson, Kalokerinos, Koval, Kashima, McDonald and Greenaway2022) recommend 15 interventions under the headings of problem-focused action, emotion management, and enhancing social connections which all have evidence of beneficial effects on individual, social, and environmental outcomes (https://psychologicalsciences.unimelb.edu.au/research/climate-change-anxiety#climate-change-anxiety-overview). Of the 15 interventions listed, 11 have interpersonal/peer, social, or community dimensions, emphasising that to date much of the evidence for individual outcomes is through increasing some degree of social connectedness. Similarly, a first trial of a modular internet CBT intervention specifically designed for psychological distress associated with climate change (Lindhe et al., Reference Lindhe, Bengtsson, Byggeth, Engström, Lundin, Ludvigsson, Aminoff, Berg and Andersson2023) reported positive results among completers. It was based mainly on generic CBT strategies for different aspects of distress but interestingly included a module addressing loneliness.
While connecting with nature or place, especially ‘nearby nature’ can be an effective way of building a sense of safety and wellbeing as many people found in the pandemic (e.g. Phillips et al., Reference Phillips, Wells, Brown, Tralins and Bonter2023) there is a note of caution. Bingley et al. (Reference Bingley, Tran, Boyd, Gibson, Kalokerinos, Koval, Kashima, McDonald and Greenaway2022) identified that nature-based interventions may have mixed effects on individual outcomes, with evidence that they may be beneficial for some, but detrimental for others, perhaps because for some people, connection with nature may highlight losses.
Next, for those with ongoing high levels of anxiety and who may be showing excessive monitoring and checking (e.g. water levels, weather forecasts, etc.), the types of interventions developed for IU could be considered. These could include group-based interventions, especially within a community affected by climate change or its effects as extreme weather, aimed to increase tolerance of uncertainty such as Making Friends with Uncertainty (Mofrad et al., Reference Mofrad, Tiplady, Payne and Freeston2020).
Likewise, given the links between information seeking and the distress measures, information management strategies may be helpful. These include distinguishing between (1) helpful/necessary information, (2) other or greater information that may be accurate but that comes at a cost of greater uncertainty or threat, and (3) unhelpful information where the intention is to attract attention. A structured approach to separating out these types, reducing some, and rebalancing with a greater focus on local, community, and value-driven information can lead to a more balanced ‘information diet’ (Ashrafi-Rizi and Kazempour, Reference Ashrafi-Rizi and Kazempour2020; see also https://www.uncertaintyincontexts.com/materials-downloads/ for worksheets and a worked example).
Finally, uncertainty-informed interventions would consider engaging with an uncertain future, and in this case with pro-environmental actions. The challenge is helping people who are experiencing the loss of the future to find a realistic way to engage with an uncertain future while not denying uncertainty and threat, not engaging in wishful thinking, and not giving up or feeling hopeless. Engagement with the future requires hope (Ojala, Reference Ojala2023) and is likely to involve some combination of meaningful and values driven action, seeking social connectedness at a level appropriate for the individual, while practising judicious self-care to address the emotional demands of confronting a problem of such scale (see also Pihkala, Reference Pihkala2022). Finally, for self-care, Gerber (Reference Gerber2023) proposes that self-compassion may play a key role for activists, and that interventions could be adapted from those developed and evaluated for health care professionals (Neff et al., Reference Neff, Knox, Long and Gregory2020). Lindhe and colleagues (Reference Lindhe, Bengtsson, Byggeth, Engström, Lundin, Ludvigsson, Aminoff, Berg and Andersson2023) also included a self-compassion module in their modular treatment.
Limitations
Network analysis is still evolving as an approach and determining the required sample size is not yet well established. In this study, the sample may be limited for the number of nodes and so the number of parameters to estimate, but the bootstrapped accuracy and stability data suggest that cautious interpretation is possible. The sample has limited diversity and is mostly young, White, female and educated, although exposure to extreme weather and its consequences varied from no exposure to quite high levels of exposure and consequences. The sample and the data are conditioned by recruitment methods through social media and attempts by the co-authors to access different groups. There are a number of novel measures, however reliabilities are generally adequate to excellent and most have their origins in previous literature or data. Ultimately the data is cross-sectional and causality cannot be determined. However, most CBT models are recursive in nature, suggesting that these are mutually reinforcing processes, and that any potential arrow pointing in one direction may have an accompanying arrow pointing in the other direction.
Future research
Some potentially important aspects of several constructs have been overlooked, namely, heatwaves among the types of extreme weather (perhaps because the study was planned in the UK in November 2020), the role of warnings about extreme weather which may contribute to actual and perceived threat and uncertainty, different types of disruption, and a fuller range of emotional responses. The next steps would be to shorten some of the existing measures, address constructs that have been overlooked and then repeat the study on a larger sample and recruited to increase the diversity and representativeness of the sample, the geographical distribution and so the range of experiences of climate change and extreme weather.
Conclusion
While recognising that climate change distress may be a normative reaction, this study demonstrates that some people may show clinical levels of difficulty in adjusting to climate change and that some people will show post-traumatic-like reactions to extreme weather events. The study has integrated key psychological variables from the climate change literature (concern, proximity, distress, and action) and variables from the uncertainty distress model using network analysis. It has largely replicated the findings from the former and integrated them into the latter. The challenge, increasingly documented in the literature, is how to establish and maintain a suitable balance between sustained engagement in pro-environmental action while addressing, for at least some people, clinical levels of what may be realistic, ongoing and complex emotional distress. This distress may lead to functional impairment as people try to adjust to a specific but multi-faceted stressor which will not go away and may become more acute in the future. How people adjust may vary greatly from one person to another, and so when intervention is required, strategies may need to be carefully tailored to the individual.
Key practice points
-
(1) Peoples’ reactions to climate change will be complex and may involve pre-existing beliefs, their experience of climate change, and their own responses to uncertainty.
-
(2) While many people will have degrees of concern and emotional distress in response to extreme weather and climate change, a small but not insignificant proportion may show significant post-traumatic stress symptoms and/or symptoms consistent with adjustment disorder.
-
(3) The underlying models and network analysis reported here suggest a range of factors that could be considered when assessing and formulating someone reporting experience of extreme weather and/or climate change distress.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1754470X24000205
Data availability statement
The data and metadata are available in the Materials and Data at https://doi.org/10.25405/data.ncl.25681455
Acknowledgements
While actively writing this article I (M.F.) have also been digging a wildlife pond in my garden. Both have been three years in the planning and conducting the preparatory steps. The contrast has been interesting, the balance essential.
Author contributions
Mark Freeston: Conceptualization (lead), Data curation (lead), Formal analysis (lead), Investigation (lead), Methodology (lead), Project administration (lead), Supervision (lead), Validation (lead), Visualization (lead), Writing – original draft (lead), Writing – review & editing (lead); Letitia Sermin-Reed: Conceptualization (supporting), Formal analysis (supporting), Investigation (supporting), Methodology (supporting), Writing – review & editing (supporting); Saskia Whittaker: Conceptualization (supporting), Formal analysis (supporting), Investigation (supporting), Methodology (supporting), Writing – review & editing (supporting); Joanne Worbey: Conceptualization (supporting), Formal analysis (supporting), Investigation (supporting), Methodology (supporting), Writing – review & editing (supporting); Chloe Jopling: Conceptualization (supporting), Formal analysis (supporting), Investigation (supporting), Methodology (supporting), Writing – review & editing (supporting).
Financial support
None.
Competing interests
Mark Freeston receives royalties for books and honoraria for training on anxiety-related topics, including the understanding of uncertainty in mental health.
Ethical standards
This research has been conducted in line with the ethical principles of the British Association of Behavioural and Cognitive Psychotherapy (BABCP) and the British Psychological Society (BPS). The study received approval from the Faculty of Medical Sciences Ethics Committee, at Newcastle University, approval number: 6737/2020. Participants provided informed consent.
Open access
Comments
No Comments have been published for this article.