Visualizing Climate Change: The Role of Construal Level, Emotional Valence, and Visual Literacy

ion level and perceived psychological distance (RQ2a), images’ abstraction level and climate change concern (RQ2b), as well as between images’ abstraction level and climate change willingness to act (RQ2c) conditional on individuals’ visual literacy?


Introduction
Visualizing climate change: The role of construal level, emotional valence, and visual literacy Climate change has become one of the most serious threats facing humankind. In 2018, the report from the Intergovernmental Panel on Climate Change emphasized the importance of limiting warming to 1.5 degrees Celsius above pre-industrial levels, and it called for immediate mitigation and adaptation actions across the globe to tackle the issue (IPCC 2018). However, currently, there is still a lack of public engagement with climate change in the United States (Pew Research Center 2017; Leiserowitz et al. 2020). When assessing the overall behaviors, only 25% of U.S. adults reported they make an effort to live in ways that help protect the environment "all the time" (Pew Research Center 2019).
Presently, signi cant psychological barriers are existent preventing people from engaging with climate change. Researchers argue that one primary barrier is the perception that the issue is psychologically distant, meaning that its uncertain impacts will affect other people, will happen in other faraway countries or sometime in the future (Brügger et al 2015). According to construal level theory (CLT), an event perceived as distant leads to a higher level of construal, that is, a more abstract mental processing of the issue (Trope and Liberman 2010). The view of climate change as abstract and psychologically distant is a major challenge to effective communication of the issue, as it can imply little personal relevance and may lead to delayed actions (Brody et al. 2012;Leiserowitz 2006). However, recent research also found evidence that decreased psychological distance or abstraction levels of climate change did not always lead to increased willingness to act (Brügger et  Nowadays, going beyond the cognitive "de cit model," scholars point out that on climate change, a polarized topic, emotions have become an important factor in shaping people's attitudes and behaviors (Nabi et al. 2018). Visuals, in particular, have the ability to arouse emotions to make the message effective. Recent climate change communication studies have started to explore emotions, visuals, and CLT, such as climate visuals' effect on emotions , the interaction between emotion and information (Nabi et al. 2018), self-conscious emotions in relation to CLT (Ejelöv et al. 2018). However, the attention was devoted to discrete emotions, textual information. How construal levels interact with affective responses (i.e., emotional valence) in the context of visuals remains unexplored.
In this study, we empirically test 1) the role of emotional valence -the positive or negative character of an emotion -in in uencing the relationship between image-induced construal level and climate change responses. In addition, 2) we examine how visual literacy in uences the effect of abstract and concrete construals of climate change as well as the potential indirect effect of emotional valence in the process.
Going beyond psychological-distance message framing (i.e., distant vs. proximal) studies and discrete emotion induction experiments, we theoretically demonstrate and empirically test the interacting effect of construal level and emotional valence in climate change visual communication. The ndings also complement prior CLT, emotion and climate change visualization literature (e.g., Ejelöv

Literature Review
Construal level theory and climate change communication The concept psychological distance can be classi ed into four dimensions -temporal, spatial, social, and hypothetical distance, each referring to how far removed an object is from the present (Trope and Liberman, 2010). Temporal distance is de ned as the perceived distance between current time point and the time when an issue will occur. Spatial distance refers to the geographical distance between an individual and the location of the issue. Social distance refers to the extent to which a person or group of people are different or similar to oneself, and hypothetical distance explains the level of certainty that the event is occurring. According to CLT, the four dimensions are interrelated, and they all positively relate to the level of construal (Bar-Anan et al. 2006). The more distant an issue is perceived to be, the more abstractly it will be mentally processed in people's minds Liberman 2010, 2011), and such abstract mental construal further encourages people to perceive greater distance to the issue (Liberman et al. 2007).
Climate change is generally perceived as a psychologically distant issue (Brügger et al. 2015;Spence et al. 2012). It was found that reducing psychological distance has merit in promoting people's climate change-related concerns and willingness to act (Evans et al. 2014;Jones et al. 2016). However, studies also argued that other factors (e.g., individuals' cognitive style, Sacchi et al. 2016;Hart and Nisbet 2012) may affect or moderate such relationship -a reduction in psychological distance does not always lead to increased concern or behavioral intention (Brügger et al. 2016;Schuldt et al. 2018).
The effect of construal level In recent years, emotions have been examined in climate change communication studies and construal level psychological studies separately. In climate change communication, scholars explore the discrete emotions such as guilt, fear, anxiety and hope. It has been found that guilt and fear positively mediate the effect of loss frames (i.e., negative consequences of not engaging in climate change protection) on willingness to make climate change-related sacri ces (Bilandzic et al. 2017). However, fear and anxiety were found to not always increase mitigation intentions (Chadwick 2015). Persuasive messages that induce fear and anxiety can sometimes be threatening and counterproductive (O'Neill and Nicholson-Cole 2009). Speci cally, when communicating climate change for encouraging behaviors (e.g., biking, walking) rather than avoiding behaviors (e.g., not driving), hope is likely to be more effective than fear (Chadwick 2015).
One possible reason for these inconsistencies is that although discrete negative emotions like anger, fear are evaluated as negative by most people, individuals might signi cantly differ in how negatively they feel about these discrete emotions. For example, some people might feel extremely negative toward "fear," whereas others might have more negative feelings toward "anger" and not such strong negative attitudes toward "fear" (Harmon- Jones, et al. 2011). In some complicated situations (e.g., climate change), instead of discrete emotions, emotions may be more easily addressed by general psychological states such as emotional valence, as suggested by Harmon-Jones, Harmon-Jones, and Summerell (2017). In this study, we speci cally focus on overall affective responses -the emotional valence in climate change messages.
Emotional valence generally refers to the "positive" or "negative" character of an emotion, or to the character of some aspect of emotion (Colombetti 2005). As an essential component of affect, emotional valence characterizes the underlying complicated experience of feeling or mood (Barrett 2006; Frijda and Scherer 2009). According to the affect heuristic, emotional valence could skew people's judgments of a risk, bias the decision making (Keller et al. 2006), ultimately shaping judgments and risk perceptions (Slovic et al. 2005). In the communication of climate change related decisions, scholars relied on the experiential theories of risk information processing and suggested that affect and other experiential factors strongly in uenced people's risk perceptions and policy support (Leiserowitz 2006  . As aforementioned, given that "affect heuristic" is strongly associated with climate change risk perceptions (Smith and Leiserowitz 2014), in addition to cognitive deliberation, people may perceive various emotional valence of the images, which in uence their perceived risks, concern and behavioral intention. Therefore, we integrate the CLT and climate change emotions literature to explore how emotional valence might mediate the visuals' construal level effect on people's responses to climate change.
When examining visuals' effect, one factor that is worth considering is visual literacy -the set of abilities that enables people to effectively interpret, evaluate, use, and create visuals (Association of College and Research Libraries 2011). It is di cult to come up with a consensus de nition of visual literacy, as the concept is by nature interdisciplinary, encompassing education, design, and media communication contexts (Avgerinous and Pettersson 2011). Our research builds on the review of the eld (Brill and Branch 2007; Brumberger 2019), uses the communication perspectives (Messaris 1998;Trumbo 1999) to conceptualize visual literacy as the ability to think, learn and communicate visually.
Visual literacy has traditionally been explored in the educational context, with a special focus on student learning, the uses and effects of visual design (Brumberger 2019). In media communication, the public's increasing visual literacy has led to the growth of photographic advertising industry (Wilkinson 1997). As a result, with the proliferation of visually rich, screen-based materials, being visually literate has become a necessity for effective visual media communication (Riddle 2009). Media scholars also argue that visual literacy is a useful tool for comprehending visual news information, being critically aware of visual media manipulation, seeing how the camera angles are used for persuasive purposes, which all subsequently affect audience responses to the news content (Lazard et al. 2020).
In recent years, scholars have shifted to focus on the role of visual literacy in communicating scienti c issues (Brumberger 2019). In particular, the attention has been devoted to how to train science communicators in visual literacy (Rodríguez Estrada and Davis 2015), the role of visual literacy in in uencing people's environmental science message processing (Lazard and Atkinson 2015). Overall, it is found that the images of science and the underlying scienti c principles range from purely symbolic to highly representational, and they may not clearly enhance the understanding of science if there is a lack of visual literacy among the audience (Trumbo 1999(Trumbo , 2000. In terms of climate change, media scholars have called for maturing the education of literacy, including the literacy of understanding visual stories in the media, as the ability has been identi ed as a key strategy in improving public acceptance of climate science in today's age of digital information (Cooper 2011). Scholars believe that audiences have different levels of visual literacy, and more literate or sophisticated people tend to have heightened conscious awareness of how meaning is created visually (Edwards 2012) and elaborate the scienti c information more deeply (Lazard and Atkinson 2015). Thus, when researching effective science visualization, it is important to consider the audience's visual literacy levels and design visuals accordingly. In this research, we attempt to examine how visual literacy interacts with the abstract versus concrete visualizing strategies to affect a variety of climate change responses. H2 Abstract (vs. concrete) construal is associated with lower concern for climate change through positive emotion.

Research questions and hypotheses
H3 Negative emotion will mediate the relationship between construal level and willingness to act in that more concrete construal will lead to greater negative emotions, which will lead to greater willingness to act on climate change.
As previous studies have started to consider individuals' literacy levels when examining proenvironmental message effect (Lazard and Atkinson 2015), we also took visual literacy into consideration to contribute to the methodology and empirical evidence on climate change visual communication. We explore the following research questions. RQ1 To what extent does individuals' visual literacy in uence the relationship between images' abstraction level and their perceived psychological distance (RQ1a), abstraction and concern (RQ1b), abstraction and willingness to act on climate change (RQ1c)? RQ2 To what extent is the indirect relationship via emotional valence between images' abstraction level and perceived psychological distance (RQ2a), images' abstraction level and climate change concern (RQ2b), as well as between images' abstraction level and climate change willingness to act (RQ2c) conditional on individuals' visual literacy?

Method
Materials and procedure Pretest procedure Pretest served as a manipulation check. In it, we selected nine abstract and nine concrete climate change images and tested if they successfully encouraged people's abstract and concrete construal, respectively. First, participants were randomly assigned to either the abstract or concrete condition in an online experiment. Then, after viewing the nine images, participants were given the Behavioral Identi cation Form (BIF; Wegner 1987, 1989) which measured their difference in construal level.

Main study procedure
In the main study, participants were randomly assigned to one of two experimental conditions (abstract, concrete). In both conditions, participants rstly viewed nine climate change-related images and captions. Then, they were asked to complete a questionnaire asking about their emotional reactions, perceived psychological distance of climate change, concern for climate change, behavioral intention, visual literacy. Participants' demographic characteristics were also measured.

Experimental stimuli
We selected 18 climate change images for this study, including nine with abstract features and nine with concrete features. These abstract and concrete features have all been identi ed and veri ed in prior experimental psychology research (Burgoon et al. 2013) and in the previous content analysis research which clari ed how levels of abstraction should be measured. For instance, according to Lee et al. (2014), colorful images are more concrete than their black-white image counterparts. Thus, we coded the presence of color as a concrete feature and the absence of color (i.e., black-white display) as an abstract feature. In general, images used in our abstract condition were black-white non-photographs with no people, focused primarily on the causes of climate change and portrayed climate change as a gradual, relatively invariant, and stable process (e.g. gradual temperature change, greenhouse gas emissions). Images in the concrete condition were colorful photographs, featuring human victims in the United States, detailed information about the present temporally and socially urgent situation) and emphasizing the certain consequences as well as the incidental disastrous aspects of climate change (e.g. wild res, hurricanes). All images were modi ed slightly to display in similar size and style.

Participants
Participants were adults recruited from around the United States via Amazon Mechanical Turk (MTurk). MTurk is an e cient crowdsourcing platform for recruiting online research participants, and the platform is especially reliable for experimental data collection (Paolacci et al. 2010). To ensure data quality, we concealed the research purpose, prescreened workers to include only those with excellent survey completion records and used attention-check questions to avoid inattentive responses.

Emotional valence
Overall perceived emotional valence was measured with the question "Overall, how would you rate your subjective feelings to the climate change images you just viewed?" Responses ranged on a Likert-type scale from 1 (completely unhappy) to 9 (completely happy) (M = 2.71, SD = 1.76). The question was adapted from a study by Marian and Kaushanskaya (2004).

Psychological distance of climate change
All four dimensions of psychological distance were included. Perceived spatial distance was measured by four questions such as "I feel the estimated distance between the geographical areas(s) that are being negatively affected by climate change and my location could be __ ." Perceived temporal distance was measured by three questions such as "climate change is an urgent threat or risk for my generation." Perceived social distance was measured by questions such as "People generally similar to me can feel/experience the negative impacts of climate change". Perceived hypothetical distance was measured by Likert scale questions such as "It is certain that climate change will have a negative impact on me." The responses ranged from 1 (strongly disagree/extremely close) to 7 (strongly agree/extremely distant). All four dimensions of distance were collapsed into one composite score of perceived psychological distance (M = 3.26, SD = 1.15, Cronbach's alpha = .869), with higher values indicating greater perceived distance.

Willingness to act
We divided 25 "willingness to act" questions into two types -mitigation intention and adaptation intention, each with ve-point Likert-type responses ranging from 1 (very unlikely) to 5 (very likely). For each, questions were split between generic and speci c intentions. Generic mitigation intention was measured with questions such as "I plan to take some actions to stop climate change." Speci c adaptation intention was measured with questions including "I intend to buy ood insurance or other climate-related disaster insurance for my (future) house." Then we aggregated the overall mitigation intention and overall adaptation intention questions to form a scale on "willingness to act," which was found to be internally consistent (M = 3.51, SD = .75, Cronbach's alpha = .859). All questions were adapted from previous studies (Broomell et

Concern for climate change
We used three questions to measure participants' concern for climate change. All questions utilized Likert-type responses ranging from 1 (not at all concerned) to 5 (a great deal concerned). For example, one question asked, "Considering any potential effects of climate change which might be on you personally, how concerned, if at all, are you about climate change?" (M = 3.39, SD = 1.27, Cronbach's alpha = 0.91).

Visual literacy
Visual literacy was measured using four visual thinking and visual creation items such as "How often do you draft mind mapping, diagramming (see examples below) in your day-to-day thinking?" "Do you do PowerPoint slides presentations on a regular basis?" "How skillful are you with photo software such as Adobe Photoshop?" "How skillful are you with illustration software such as Adobe Illustrator?" Responses ranged from 1 (entry-level amateur/Never) to 5 (Expert professional/Always) (M = 2.03, SD = 0.77, Cronbach's alpha = 0.78). All questions were adapted from previous study (Brumberger 2011). In this study, other demographic characteristics were measured as well, including age, gender, race, ethnicity, geographic region, political orientation, and education level.

Pretest
Results from Pretest showed that respondents who were exposed to abstract images were more likely to prefer abstract behavioral descriptions (n = 93, M = 16.53) in the BIF test than those who viewed concrete images (n = 99, M = 14.35) (p = .024, Cohen's d = .33). These results suggest images selected for this study successfully encouraged participants in the abstract group to have higher level of construal than those in the concrete condition. Thus, the stimuli manipulation was successful.
A series of independent sample t tests were conducted to examine the main effect of construal level experimental manipulation. Participants who viewed abstract images were found to perceive signi cantly greater psychological distance to climate change than those who viewed concrete images (M abstract = 3.45, SD abstract = 1.13, M concrete = 3.08, SD concrete = 1.15, t(446) = 3.35, p < .01, Cohen's d = .32).

The mediating role of emotion
To examine H1, we conducted a mediation analysis using the PROCESS Macro Model 4 for SPSS (Hayes, 2018), with 5,000 bootstrap resamples and 95% con dence intervals. We included images' abstractionlevel conditions as independent variable, emotional valence as mediator, and perceived psychological distance as dependent variable. Social demographic variables were also entered into the model as covariates to control for their potential impacts. Findings on the indirect effects partially support H1, as emotional valence partially mediates the relationship between perceived psychological distance and images' abstraction level ( = .14, SE = .04, 95%CI [.07, .22]; Fig. 1). Table 1 reports the estimates of the mediation model.
Next, we performed the same analysis with climate change concern as dependent variable. Results showed that images' abstraction level did not have a direct impact on climate change concern. However, it had an indirect effect on the concern ( = − .16, SE = .04, 95%CI [-.24, − .07]; Fig. 2) via emotional valence: Abstract (vs. concrete) construal led to increased emotional valence, which then reduced the concern for climate change. Therefore, H2 was supported.
To test H3, we ran the same model with willingness to act as dependent variable. It was shown that emotional valence also negatively mediated the in uence of images' abstraction level on willingness to act on climate change ( = − .07, SE = .02, 95%CI [-.11, − .03]; Fig. 3). Thus, H3 was supported.

Visual literacy
The rst set of research questions (RQ1a-c) asked to what extent visual literacy could moderate the effects of images' abstraction level on people's responses to climate change. To address these questions, we conducted a moderated mediation analysis using PROCESS Macro Model 8 to test both the direct and indirect paths from images' abstraction level to the climate change response variables. Similar to the mediation model, we entered images' abstraction level (abstract = 1, concrete = 0) as independent variable, emotional valence as mediator, perceived psychological distance as dependent variable and social demographic variables as covariates. The remaining analyses were identical to the rst except that concern, willingness to act were entered as dependent variables, respectively.
The analyses revealed that our image manipulation interacted with visual literacy to affect people's perceived distance to climate change (β = − .26, p < .05; Fig. 4), concern for the issue (β = .31, p < .01; Fig. 5), and their willingness to act (β = .16, p < .05; Fig. 6), after controlling for social demographic variables. Speci cally, on the one hand, for people with low visual literacy (1 SD below the mean), abstract (vs. concrete) images increased their perceived psychological distance to climate change (β = .40, SE = .13, p = .00, 95%CI [.15, .65]). Such construal level manipulation did not affect the perceived distance among people with high visual literacy. On the other hand, for people with high visual literacy (1 SD above the mean), abstract (vs. concrete) images signi cantly increased their concern (β = . 34 Bridging these areas of study, we made an initial attempt to explore the mediating role of emotional valence and the moderating role of visual literacy in the relationship between (abstract, concrete) construal level and climate change engagement. We found evidence that concrete (vs. abstract) images brought people negative (vs. positive) feelings, which subsequently led people to perceive less psychological distance to climate change, generate greater concern and willingness to act. In addition, people with low (vs. high) visual literacy were more in uenced by the construal level visual manipulation.
And unlike highly visually literate people who were motivated by abstract images to engage with climate change, people with low level of visual literacy were positively in uenced by concrete images. Overall, we The analyses show that participants who viewed abstract images perceived greater distance between climate change and themselves. This is aligned with environmental communication literature which explained that local frames of climate change could encourage people's worry and concern about the issue (Bloodhart et al. 2015). It also con rms CLT which suggests that people's perceived psychological distance to an issue is positively in uenced by the level of abstraction at which the issue is mentally represented (Trope and Liberman 2010). Future visual communication scholars and advocates could follow this line of thinking to explore how to use concretizing strategies to make climate change become psychologically close to the public.
Supporting the hypotheses, we found that abstract images led to increased perceived psychological distance, less concern and willingness to act via positive emotions. This nding is in close accord with past CLT literature which explains that compared to concrete construals, abstract construals encourage people to generate more pros (reasons in favor of a course of action) than cons (reasons against a course of action), consequently having more positive perceptions of an issue (Eyal et al. 2004;Williams et al. 2014). Also, the negativity elicited by concrete thinking of climate change could further encourage climate change engagement. It is likely that when portraying climate change negatively, a low level, concrete construal could help motivate people with urgency and action-oriented thinking. Whereas under a high level, abstract construal, the negativity might become vague, involve less intense emotions, and thus raise less concern or behavioral intentions (Van Lent et al. 2017). Although fear was sometimes overwhelming and discouraging (O'Neill and Nicholson-Cole 2009), future scholars could build on our nding and continue to explore whether or not concretizing certain negative emotions (e.g., using anger instead of sadness) might be able to effectively promote climate change engagement.
Despite concrete thinking encouraging concern and willingness to act via negative emotions, we found that construal level mindsets did not directly in uence people's concern for climate change, nor did they have a direct impact on willingness to act. These ndings are aligned with CLT which originally explains only the positive relationship between construal level of and psychological distance of an issue, not how such construal level cognitive style is related to environmentally friendly attitudinal and behavioral Furthermore, ndings from our analysis show that individuals' visual literacy played an important moderating role between construal level and climate change outcomes, though it did not in uence the indirect effect of emotional valence on the relationship. This is consistent with previous work by McDonald et al. (2015), which claimed that the effect of abstract versus concrete messages on people's climate change decision-making is dependent upon various individual-level factors. Speci cally, we found that less visually literate people were more in uenced by the construal level visual manipulation, and concrete images made them perceive signi cantly less psychological distance to climate change. On the contrary, the perceived psychological distance among people with higher visual literacy was not in uenced by the construal level visual effect. One possible explanation is that highly visually literate people were able to critically consume visual information. As a result, they were less subject to visual abstraction or concreteness persuasion tactics. This con rms the study by Messaris (1994), Messaris and Moriarty (2005), unveiling that audience critical visual thinking and design skills can greatly impact the effectiveness of visual communication.
Additionally, we found that abstract images drove highly visually literate people to engage with climate change. Whereas less visually literate people were more motivated to engage by concrete images. We speculate that in abstract images, there is a lack of detailed information, and meanings are symbolic, indirect, and are constructed via black-white, data-intensive graphs, cartoons. All of these features require higher visual thinking and design ability. This way, people with high visual literacy are able to understand the abstract images more deeply, process them more effectively, and thus being able to get motivated by them. For less visually literate audiences, concrete images communicate the direct, representational rather than symbolic meanings to them with detailed information, human elements, colorful design, and these facilitate better, easier visual message processing and thus more favorability among them. This result also explains one of our earlier ndings that abstract and concrete construals did not directly in uence people. We speculate that abstract and concrete visual portrayals of climate change activate not only different construal levels but also people's visual skills. And for viewers with different visual literacy levels, the processing style and the level of understanding of the visual information might be different. Future research should continue to explore visual literacy as an in uential personal factor and take other audience characteristics (e.g., cognitive style, visual preference) into account when examining visual communication effectiveness. These ndings shed light on the importance of incorporating visual literacy into visual communication efforts.
Overall, our study is a rst step toward an integration of construal level and emotions in climate change visual communication. We demonstrate that concrete visuals of climate change elicit negative emotions toward the issue, which further leads to greater concern, willingness to act. There are also individual differences in terms of visual information processing. For people with high visual literacy, visual construal level effect on perceived distance to the risk is minimal, and abstract images are more effective in encouraging climate change engagement. However, for less visually literate people, we found the opposite patterns.
This study is not without limitations. First, this study measured visual literacy by self-reported questions that focused on visual creation skills, visual use and thinking habits. The questions on how well people interpret visuals did not reach internal consistency and were thus eliminated from further analyses.
Second, in this study, we did not differentiate between low-level (concrete) and high-level (abstract) behaviors, or near future (short-term) and distant future (long-term) behaviors, which were found in prior research to possibly associate with different concrete and abstract construals, respectively. Last, emotional valence in our study captured only the overall affective responses people had towards the images, and it is unknown how intense people felt towards the images and what exact discrete emotion contributed to the attitudinal or behavioral change. Future research should continue to investigate the interactions between construal level and emotions, with more variables being considered such as emotional intensity, the abstraction and concreteness of the emotion.
It is expected that this study will encourage future research on the intersections among construal level of

Declarations Funding
The authors received nancial support for the research from the Center for Advanced Media Studies at the University of Nevada Reno.

Con icts of interest/Competing interests
The authors declared no potential con icts of interest with respect to the research, authorship, and/or publication of this article.

Availability of data and material
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request. Note: a. Abstract=1, Concrete=0; N = 448; Adjusted for age, gender, race, region, political orientation, education, and income. Bootstrap con dence intervals were constructed using 5000 resamples and 95% bias-corrected con dence intervals; *p < .05, **p < .01, *** p < .001 (two-tailed); β: unstandardized coe cient, SE= Standard error.   we performed the same analysis with climate change concern as dependent variable. Results showed that images' abstraction level did not have a direct impact on climate change concern. Note: Social demographic variables of age, gender, race, political orientation, region, education, and income were  The analyses revealed that our image manipulation interacted with visual literacy to affect people's perceived distance to climate change Page 25/26 Figure 5 The analyses revealed that our image manipulation interacted with visual literacy to affect people's perceived distance to concern for the issue.