The Extreme Climate Event Database (EXCEED): Development of a picture database composed of drought and flood stimuli
The Extreme Climate Event Database (EXCEED): Development of a picture database composed of drought and flood stimuli
Sabrina de Sousa MagalhãesID 2 3
Diana Kraiser Miranda 2 3
DeÂ bora Marques de Miranda 1 2 3
Leandro Fernandes Malloy-Diniz 0 2 3
Marco AureÂ lio Romano-Silva 0 2 3
0 Department of Mental Health, Federal University of Minas Gerais (UFMG) , Belo Horizonte , Brazil
1 Department of Pediatrics, Federal University of Minas Gerais (UFMG) , Belo Horizonte , Brazil
2 Laboratory of Molecular Medicine, Federal University of Minas Gerais (UFMG) , Belo Horizonte , Brazil
3 Editor: Francisco Leal-Soto, Universidad de Tarapaca , CHILE
The present study introduces the Extreme Climate Event Database (EXCEED), a picture database intended to induce emotionally salient stimuli reactions in the context of natural hazards associated with global climate change and related extreme events. The creation of the database was motivated by the need to better understand the impact that the increase in natural disasters worldwide has on human emotional reactions. This new database consists of 150 pictures divided into three categories: two negative categories that depict images of floods and droughts, and a neutral category composed of inanimate objects. Affective ratings were obtained using online survey software from 50 healthy Brazilian volunteers who rated the pictures according to valence and arousal, which are two fundamental dimensions used to describe emotional experiences. Valence refers to the appraisal of pleasantness conveyed by a stimulus, and arousal involves internal emotional activation induced by a stimulus. Data from picture rating, sex difference in affective ratings and psychometric properties of the database are presented here. Together, the data validate the use of EXCEED in research related to natural hazards and human reactions.
Worldwide, people are facing disasters caused by climate change. With the increasing number
of events caused by global warming (e.g., heavy precipitation events, floods, hurricanes,
wildfires and heat waves), more and more people around the world will be affected. Such
circumstances affect humans in different ways that could potentially lead to both mental health issues
and/or have an impact on a broad range of adversities to the economic, social and
environmental field [1±3]. The impact of such climate-related extreme events uncovers vulnerability
and exposure of ecosystems and human dynamics [
The doubts raised about how to cope with immediate or long-term consequences of natural
hazards motivated the development of the Extreme Climate Event Database (EXCEED). The
Minas Gerais, FAPEMIG (PPM-00183-14, <http://
www.fapemig.br/>). MARS receive both fundings.
The funders had no role in study design, data
collection and analysis, decision to publish, or
preparation of the manuscript.
database is proposed as a validated tool for use in a broad range of experimental designs
assessing emotional response, with representative stimuli of flood and drought events. Thus, this
database could could contribute to insight about how individuals respond to those stimuli.
The use of well-designed instruments that are reliable and accurate in emotion induction is
of paramount importance; still, the selection of stimuli depends on the study design and the
targeted empirical answer. EXCEED pictures could be used as emotional triggers to collect
and evaluate an extensive set of responses for both healthy and clinical samples. For instance,
experimental approaches in psychophysiological research with EXCEED stimuli could be
directed to measuring facial expression, physiological reactions (e.g., peripheral physiological
signals or electroencephalographic), behavior accuracy, or cognitive responses (e.g., attention
network or memory). Furthermore, the pictures can also stimulate investigations about beliefs
and actions related to climate change and the role of media and arts in mediating behaviors
and perceptions. Another experimental use for the database is for those who need
psychotherapeutic intervention after a natural disaster experience, including the traumatic ones. In the
latter case, the literature consistently suggests that there are immediate and long-term
consequences for mental health and human development after natural disasters [5±7]. Therefore, we
expect that these inquiries assist in the creation of important public health policies and
mitigation actions, especially in socially disadvantaged populations [
The development of sets of standardized stimuli requires a representative theory of emotion
and affect. First, emotional responses arise after the evaluation of meanings, pertinence, and
significance of the stimulus [
]. Simultaneously, the proprioceptive feedback modulates the
physiological reactions associated with the aroused feeling [
]. Two commonly used
perspectives to describe and conceptualize emotions are called dimensional theory and discrete
emotion theory [
]. The discrete emotion theory claims that there is a set of basic emotions
that triggers specific behavior and physiological responses. Proponents of this theory have
concluded that there are six basic discrete emotional states that emerge from a cognitive appraisal
of the environment: fear, disgust, surprise, anger, happiness and sadness .
Still, the model of emotion as a cognitive operation matched to internal physiological
changes is the basis for the proposition of a dimensional perspective of emotional response.
The dimensional approach to emotion postulates that the emotional response arises from
basic motivational processes, mainly organized under the influence of valence and arousal
However, initially, three-dimensional aspects were responsible for the organization of the
core affect: a) valence, a contrast between states of pleasure and displeasure; b) arousal, that
compares states of agitation and relaxation; and c) approach-avoidance state, also known as
ªmotivation directionº, which compares the tendency to approach versus the tendency to
avoid a stimulus [
]. Further development of the dimensional perspective leads to the
circumplex model of affect. This model postulates that all affective states are linear combinations of
the two primary neurophysiological systems: valence (a pleasure-displeasure continuum, also
referred to here as positive-negative continuum) and arousal (an excited-relaxed continuum,
also known as activation or alertness), along with cognitive consideration and labeling of the
experience [14±16]. In this model, valence is one of the fundamental properties of the core
], comprising the cognitive process of evaluating the environment and coding it
according to pleasantness or unpleasantness. While arousal is the other essential property of all
affective stimuli, it encompasses a continuum from relaxed, calm, sluggish, and sleepy to
stimulated, distressed, and jittered. The dimensional structure of affect can be plotted in a
schematic coordinate axis, where arousal and valence yield two polarities (high and low arousal,
and pleasure and displeasure valence, or positive and negative valence). Each emotion can be
located in this space, as a linear combination of valence and arousal. Thus, all emotions are
2 / 15
formed by the two properties of valence and arousal, and emotional responses can be properly
evaluated only through both states. Presently, the level of dominance perspective has been
eclipsed by the perspective of the circumplex model of affect.
Despite the theoretical divergence between dimensional theory and discrete emotion
perspective, scholars who endorse either paradigm do not consider the theories as incompatible,
because human emotion seems to consist of both conscious and unconscious processes [
Ultimately, the motivational subsystems from valence/arousal underlie discrete emotional
states. However, to our purpose, the dimensional perspective provides meaningful ways to
access emotion without the need for one's inherent ability to conceptualize and verbalize
distinct emotional states. In addition, evidence indicates that the way an individual labels discrete
emotional states varies according to cultural background. Therefore, using valence and arousal
ratings enables individuals to avoid the often challenging task of discerning and applying
distinct emotional labels to differentiate the emotional experience. Due to individual differences
in emotional granularity, it is simpler to report a pleasant or unpleasant feeling and an aroused
or calm state instead of nominating emotions [
]. Thus, the dimensional perspective can be a
ªsuperordinate theoretical frameworkº [
] for studying emotional responses.
Previous studies have indicated sex differences in the processing of emotional stimuli and
emotion regulation, defined as a series of automatic and controlled responses aimed to
modulate the experience and expression of emotion [
]. Know as the female negativity bias
hypothesis, women tend to present a stronger reactivity to negative material than men [
the tendency is for the female individuals to rate negative stimuli for a high valence-high
arousal responses . On one hand, this trend may explain the vulnerability of females in the
increasing rates of anxiety disorders among them [
]. However, a major sensitivity for
negative material can contribute to empathy behaviors in the context of suffering and disasters
Currently, there are at least three pictorial databases used as methodological reference for
evaluating and classifying emotional responses: the International Affective Picture System,
], The Geneva Affective Picture Database, GAPED [
] and The Nencki Affective
Picture System, NAPS [
]. All of these databases were validated by using similar dimensional
perspective and affective ratings in the domain of valence and arousal on healthy volunteers.
One important remark considering those databases is that all of them built their negative
categories mainly upon phobic reactions, scenes of violence and injured or dead human
bodies. The lack of a full set of specific natural hazard stimulus is a gap that EXCEED aims to fill.
The choice for flood and drought stimuli was influenced by the significant prevalence of these
two events worldwide. From 1995 to 2015, floods alone accounted for 47% of all
weatherrelated disasters and, together with droughts events, affected 3.4 billion people (82% of total
affected, excluding deaths) around the world [
The media coverage of natural hazards is increasing, and the role media plays in society has
recently drawn considerable attention. At the time of this writing, the media is a central part of
everyday life. Most of the world watched and followed the repercussions of the 9/11 attack on
the World Trade Center, the 2004 tsunami in Thailand and the 2005 Hurricane Katrina in
New Orleans. The risk associated with the experience of a disaster can be manipulated,
amplified, magnified, or minimized by the media [
]. Because disasters are commonly discrete
events, the media often covers severe disasters such as floods, hurricanes and earthquakes. Less
attention is given to chronic conditions, especially droughts, because of their slow onset and
broad spatial extent, unless they become extremely critical [
]. A worldwide effort to consider
drought as a significant disaster and to protect the drought-vulnerable populations has already
begun, for example, by agencies within the World Health Organization and United Nations
3 / 15
The objective of this study was to present the development and validation of a broad set of
normative, color photographs, emotional stimuli for experimental investigations of emotion in
the context of climate associated disasters. We additionally aimed at investigating sex-related
differences in affective ratings, expecting a high valence±high arousal sex effect in participants
of female gender [19±23].
Materials and methods
All procedures of this study were approved by the appropriate local ethics committee (CAAE:
26886814.9.0000.5149). For a summary of the Method, see S1 Fig in the Supporting
Information, which describes the main methodological steps used.
EXCEED is freely accessible to the scientific community for noncommercial use, and it is
available at: <https://doi.org/10.6084/m9.figshare.4602016.v1> [
]. Each image within the
database receives an abbreviation identification letter for the category that it belongs to,
followed by a crescent number within the group, the respective attribution credit and creative
common (CC) license. The letter N represents neutral pictures (range: N01 to N50), FL
designates the flood group (range FL01 to FL50), and DR designates drought pictures (range DR01
to DR50). All pictures are in JPEG format; however, the resolution and dimensions could vary
between pictures. Because the aim was to pursue an ecological validity of the pictures, the
physical properties of the images were not controlled. This is examined in the Discussion
section as it may restrain the applicability in some methodological platforms. All EXCEED images
are under CC license, and each image is under specific license according to restrictions
modifications, and use in commercial settings (information regarding credits of the images and
respective licenses can be consulted at S1 Table).
Pictures were selected mainly through an extensive online search of CC license repositories
(e.g., Wikipedia Commons, FlickrÐconstraining the search for online CCs images, Empresa
Brasil de ComunicacËãoÐa public Brazilian media agency that use CCs license for all the
images provided). Social media sites like Facebook were searched too. We only used images
from Facebook if we could contact the owner, explain our aims and receive written consent to
use the images. The vast majority of the photos were retrieved from Flickr. Some images were
free to modify, so ten pictures receive subtle alterations. CC licenses require that the user must
indicate if alterations were made, thus in compliance with CC license requirements, we list in
a supplementary file all information demanded to be in accordance with international
guidelines (S1 Table).
EXCEED database included two negative categories (flood and drought) of images that,
virtually, could culturally resemble any worldwide reality related to those events. Inclusion
criteria for these weather-related categories were that a picture must either portray a realistic event
related to the topic, such as a scenery affected by an extreme event (urban or rural), or depict
living beings in that context. Pictures were from all over the world, including Africa, Brazil,
India, and the United States, among others. For the neutral category, inanimate objects were
chosen, since finding scenes of people with neutral facial expressions was challenging [
Initially, 50 pictures were selected for each category. Consequently, the dataset at first
consisted of 150 pictures in both in landscape or portrait orientation. Our hypothesis postulated
that drought and flood pictures should exhibit valence of 1 or 2 (interpreted as negative
valence) and arousal of 4 or 5 (indication of high arousal) and that the control pictures should
be rated with 3 for both valence and arousal (neutral valence and neutral arousal). High
valence and low arousal pictures were supposed to represent positive and pleasant images,
4 / 15
which did not apply to this database. Then, we ran preliminary analysis to identify any image
that received inadequate ratings because of their unlikely or unpromising representation of a
particular category so that none of the pictures could be considered an outlier. The criterion
adopted to search for outlier pictures can be found in the Results section, although none
picture were identified as an outlier.
For the validation process, a convenience sample of 50 healthy Brazilian volunteers
participated in the study (25 women). They were recruited through online and social media
advertisements. We sought volunteers by communicating with our acquaintances, family and
A total number of 70 volunteers began to rate the pictures, but only 50 completed the
survey. Because approximately 30% of individuals dropped out, and in order to not overweight
the initial ratings versus the final ones, we decided to remove all individuals who did not reach
the end of the experiment. Thus, we do not have missing data, because responses to all the
questions were required for the development of the experiment.
The final sample of 50 individuals had a mean age of 36.94 years (ranging from 20 to 69
years; SD = 13.12). Minors were not included in the study. Most participants had at least one
academic degree (78%) and had not personally experienced flood or drought conditions
(62%), although fourteen participants (28%) reported facing flood incidents, one reported
experiencing a drought event (2.0%), and four (8.0%) reported experiencing both events. Due
to the small number of individuals who reported previous personal experience with natural
hazards and the impracticable possibility of inferential analysis within such subgroups, the
final sample consisted of all participants regardless of their experience.
The Statistical Package for the Social Sciences platform was used to perform statistical analysis.
Descriptive analysis indicated demographic profile of the sample, valence and arousal means
and standard deviation rates. We use inferential analysis to verify differences between the
EXCEED categories through ANOVA and post hoc tests, as well as examine sex-related
differences in affective ratings. If significance differences were obtained, we estimated the effect size
value through Eta Squared indices due to the possibility of bias because of the small sample. In
addition, one limitation of ANOVA is the lack of magnitude difference indicated by the test.
Effect size indicates how much variance in the dependent variable is due to the variance of the
The association between valence and arousal is an important discussion in the field. To
verify the strength and direction of the association, Pearson correlation coefficient (r) was
calculated. Generally, r values between 0.1 and 0.3 indicate a small correlation, values between 0.3
and 0.5 indicate a medium to moderate correlation, and r values more than 0.5 indicate a
strong correlation [
], although some authors argue that only r values above 0.7 should be
considered strong correlations [
The quality of the EXCEED data relies on empirical evidence for psychometric properties.
We elected two distinct methods to run the analysis. Cronbach's alpha provides information
about the homogeneity of the database. On the other hand, the split-half method enabled
conclusion about the internal consistency and we split the database, comparing the first half of
items versus the second half of items, because we want to ensure that habituation or
sensitization did not occur while experiment was in progress.
5 / 15
The validation processes comprised rating each picture for valence and arousal, because we
considered them as fundamental properties of affect. Thus, the level of analysis was the
subjective experience accessed through self-report. The chosen method (described below) was
similar to other picture databases, such as IAPS [
], GAPED [
], and NAPS [
The Survey Monkey online software was used to collect data through a link made available
via email, acquaintances, and social media. Before stimuli presentation, volunteers read a
briefing of the research and the free and clarified consent terms, then indicated with selection of
a box the agreement to participate. After this introduction and only with the authorization
box chosen, the next page presented comprehensive instructions, with examples, regarding the
task and rating scales. They were introduced to the procedures of rating and became
familiarized with the dimensions and instruments before the experiment.
EXCEED pictures were presented in a semi-randomized order, with the constraint that no
image belonging to the same category occurred consecutively [
]. Among the volunteers, the
presentation order was constant for all participants, as a limitation of the chosen software. For
each picture presented, the participant rated them for valence and arousal, and after
completing the requisite ratings for one image, a button at the end of the screen allowed the
participants to move to the following page with the subsequent stimulus (Fig 1).
To obtain an affective rating of the pictures, we used two strategies: a five-point graded
verbal scale (one to valence ratings and another to arousal ratings) and a pictorial scale based on
the SAM. The valence five-point scale went from very positive to very negative, with the
midpoint representing a neutral condition; participants provided ratings by completing the
sentence: "You are judging this image as. . .". The arousal scale was presented with the sentence
"Confronted with this image, you are feeling. . ." and ranged from aroused to calm, with the
midpoint indicating a neutral state.
SAM is a picture-oriented instrument that includes five cartoon-like figures for the valence
and arousal dimensions. SAM is a widely used system to measure emotional responses to a
variety of stimuli, including images and sounds [
]. For the valence dimension, SAM ranges
from a smiling and happy face to a desolate and unhappy one. When representing the arousal
dimension, SAM varies from an excited, wide-eyed figure to a sleepy figure with eyes closed.
The participants were instructed to choose from any of these five images associated with the
labels from the verbal scale above each SAM's figure. For valence ratings, we equated the
internal continuum of pleasure-displeasure with the positivity or negativity of the picture presented
and rated by individuals. SAM was selected due to its validity in determining the subjective
experience of emotion related to the appraisal of the stimuli. Its applicability covers a variety of
populations, including children and people with less education [
]. Studies have shown that
ratings of valence and arousal using this method are highly correlated with affective measures
obtained by semantic scales of emotional stimuli [
Fig 1 depicts a print-screen of a page of the experiment, in which the EXCEED picture
was presented to the upper margin and centralized, followed by the question about valence,
the five-point scale lined to SAM scales, and arousal investigation. Although the original
language of the experiment was Brazilian Portuguese, Fig 1 displayed a translated version of the
survey questions. The instructions followed those described by Dan-Glauser and Scherer
The experiment lasted about forty minutes, but time varied according to each individualÂs
rating speed, since there was no time limit. After data collection, the answers for valence and
arousal ratings were replaced with numbers (range 1 to 5). The very positive and aroused
choices were coded as 5, the subsequent choices as 4, the neutral conditions as 3, the following
6 / 15
Fig 1. An example of image presentation and respective ratings of valence and arousal for EXCEED validation.
The language portrayed is English, but the original language was Brazilian Portuguese. The Self-Assessment Manikin
(SAM) system was presented together with two five-point scales. The valence scale (the first question) went from very
positive to very negative. The midpoint represented a neutral condition. For rating, participants were instructed to
complete the sentence "You are judging this image as:º. The arousal scale (the subsequent question) was presented with
the sentence ªConfronted with this image, you are feeling:º and ranged from aroused to calm. The midpoint matched a
neutral state. Instructions paralleled those described by Dan-Glauser and Scherer [
]. (Image portrayed is DR32 by
Padre Djacy Brasileiro / CC BY-NC-ND 2.0, from EXCEED).
7 / 15
choices as 2, and very negative and calm as 1. Thus, higher numbers indicated positive valence
or greater arousal.
Results and technical validation
For each of the 150 pictures, mean value ratings were obtained for both valence and arousal
categories. The ratings ranged 1 to 5 and higher numbers indicated positive valence and
greater arousal. A score of 3, which indicates the middle point of the scales, showed a neutral
condition which could apply for both affective dimensions. First, we conducted exploratory
analyses to determine whether any picture was rated differently than was expected for their
respective groupÐfor instance, if a drought picture received neutral valence and/or low
arousal, whereas the hypothesis supported a negative valence and high arousal. Nearly of all
the drought and flood pictures should exhibit valences of 1 or 2 and arousals of 4 or 5, and the
control pictures should be rated a 3 for both valence and arousal. Therefore, to obtain
unambiguous categories and images that depicted the emotion intended, outliers were investigated.
The criteria adopted were the mean plus 1.5 standard deviation for valence ratings, the mean
minus 1.5 standard deviation for arousal ratings in flood and drought categories, and the
mean plus and minus 1.5 standard deviations for both valence and arousal in the neutral
category. Nevertheless, there was no outlier picture in our database. Therefore, the final database
contained the same 150 images.
Descriptive statistics for both valence and arousal dimensions for all EXCEED pictures
were presented as Supporting Information (S1 Table). Table 1 summarizes the main features
of each category group and shows the values for women and men separately. Group difference
was observed between the natural hazards conditions compared to the neutral, for both
dimensions (valence: F (2,147) = 313.58, p< .001; arousal: F (2,147) = 264.56, p< .001), but
not between drought and flood pictures. A Bonferroni post hoc test indicated that flood and
drought pictures were more negative than neutral and elicited more arousal in individual
participants. Fig 2 depicts the distribution of the outcome ratings in valence and arousal for each
Emotion research usually describes sex differences in emotional processing toward a
stronger physiological and cognitive response displayed by women compared to men [
the EXCEED, women and men responded slightly differently in affective ratings for the flood
and drought category stimuli (p< 0.001). Women exhibited a tendency to rate the pictures in a
more negative way and reported more feelings of disturbance when confronted with the
stimuli, although the effect sizes were small (ɳ2 = 0.22 for valence and ɳ2 = 0.27 for arousal in
flood category; ɳ2 = 0.15 for valence and ɳ2 = 0.20 for arousal in drought category).
Considering all pictures in the database, strong correlations were obtained for arousal
and valence (r = -0.82, p < .001), which was expected in order to enhance the validity of the
database. For each category, the same moderate to strong correlation pattern was observed
between affective dimensions (neutral: r = -0.69, p< .001; flood: r = -0.59, p< .001; drought:
8 / 15
Fig 2. Arousal (a) and valence (b) distribution of outcome ratings for each category. In the neutral category, the distributions of rating indicated a
neutral condition for both affective dimensions. The weather categories were rated with higher arousal and with negative valence, yet the Drought
distributions were more spread-out than Flood's. Each dot in the graphic represents a picture.
r = -0.76, p< .001). Fig 3 shows the distribution of valence and arousal for EXCEED pictures
in the affective space.
Finally, an internal consistency analysis was conducted with two distinct methods and the
results are presented in Table 2. The two methods adopted indicated robust and convergent
Fig 3. The relationship between arousal and valence rating of EXCEED pictures in affective space. Each dot in the
graphic represents a picture. The dimensions exhibited a robust negative linear relationship.
9 / 15
indices. Therefore, empirical evidence for the internal validity and reliability of the database
This study aimed to develop and validate a picture database that may ecologically represent the
environmental realism associated with drought and flood. Emotion arises from a complex
interaction between cognitive appraisal and neurophysiological changes, and it is related to
valence and arousal systems, including distinct neural systems underlying each affective
]. The valuation of surrounded stimuli is made continuously and
automatically, enunciating something about its relevance and value. This internal valuation processing
produces changes in core affective states, i.e., the emotional neurophysiological state
experienced as valence (pleasant or unpleasant) and as arousal (activated or deactivated). Core affect
gives information about the relationship between the individual and its current environment
at a given point in time. Therefore, core affect provides a standard metric for comparing
different events .
For all EXCEED pictures, participants were instructed to rate according to their personal
reactions to the image. The affective rating obtained supported the hypothesis raised about the
classification of the pictures. Even though the majority of the subjects did not personally
experience extreme events, climate change awareness occurs via different media, politics and arts,
which presumably primed the individuals for the predicted emotional reactions. After all, the
induction of real affect is favored by stimuli that are pertinent to the individual. However,
higher sample variability and different affected/non-affected status of participants must be
important to further evaluate the potential use of the tool. Furthermore, the potential role of
media interference in emotional reactions must be better characterized.
One important outcome was the successful differentiation between weather categories and
the neutral group. They were rated more as displeasureable and more disturbing than neutral
pictures. Of the 150 initially chosen pictures, all achieved the inclusion criteria. Although the
drought category contained pictures with most variability within the group (see Fig 2). This
pattern could be explained by the characteristics of the sample used for the affective ratings.
The semi-arid vegetation, usually associated with dryness depicted in some pictures in the
drought category, is, in fact, a familiar and ordinary landscape commonly seen in the Brazilian
10 / 15
northeast coast or even in cities without scarce water supplies. On the other hand, any picture
that represented flood conditions seems to be more arousing and with negative valence for
Brazilian individuals. Another possible mechanism underlying these data is the common belief
that drought is not a disaster itself, an affirmation supported by the frequent absence of media
coverage about it [
]. Due to its chronic nature, people tend not to consider drought as a
natural hazard compared with acute events such as tornados, tsunamis or earthquakes. Another
important observation is that some pictures depicted disaster, but others depicted rescue
teams. Therefore, they had an intrinsic positive aspect within the harshness of the scene. The
perception of individuals about the considerations previously delineated must be better
examined to test this hypothesis. Furthermore, validation with distinct demographic conditions
would also help to elucidate the findings.
Sex differences in rating emotional stimuli were observed. Women tended to react more
and to rate pictures with more arousal and valence than men did; however, this finding only
held for the extreme conditions. There were no differences in neutral ratings. These finding
are in accordance with the female negativity bias hypothesis [
] and despite the fact that the
two-gender sample size was small, the same pattern was also observed in other studies,
including those with physiological measures that indicate a sex-distinct processing of emotional
19, 21, 22, 26
The relation between valence and arousal is an important issue in the organization of the
human affective system. It has been consistently found that increased arousal is associated
with positive and negative valence extreme values [
18, 21, 24, 26, 33, 35
]. Our findings support
this pattern of association between valence and arousal, as the negative valence pictures were
associated with increasing arousal ratings.
The negative linear relationship between the two affective dimensions considered that the
emotional experience should be regarded as a continuum, ranging from a neutral state to an
unpleasant distressed feeling. The data suggest that to analyze affective experience, both
valence and arousal should be considered, as they maintain a close linear relationship. The
high correlation observed (r = -0.82, p < .001) was expected and described in other
experiments validating pictures in databases studies. Marchewka et al. [
] found a significant
negative correlation between valence and arousal, including an index of -0.90. In a Brazilian
sample, IAPS was validated by Ribeiro et al. [
] and they also observed a correlation of -0.82
for valence and arousal.
Another empirical evidence to validate EXCEED was obtained through internal reliability
data. Both Cronbach's Alpha and split-half methods demonstrated convergence between
indices. Beside the investigation of item's consistency provide by the former, the latter method
enable the investigation about any effect resulted from the progression of the experiment. The
analyses did not suggest the occurrence of habituation or sensitizations, due to the consistency
of the indices. Hence, it revealed coherent ratings and congruent construct measures across
the categories, affective dimensions and with the entire database, providing evidence for both
internal validity and reliability.
The present study has some limitations that might affect the generalizability of the described
data. Given the complexity of the theme and some methodological limitations, we need to
interpret the results carefully. The data from EXCEED was obtained from a small, healthy,
highly educated Brazilian adult population and ratings from other distinct populations must
be obtained. After all, emotional experience also differs across age, developmental stages, and
cultural context. Children, for example, exhibit a limited capacity to describe their affective
state, having a tendency to cluster emotions under more generic labels [
]. This lack of
differentiation of affective states leads to a circumplex model of affect poorly differentiated in
children, mainly grouped to extremes of valence and less discriminative for the arousal dimension.
11 / 15
The majority of the sample is from Minas Gerais state (78%), which is a Brazilian state
where both natural hazards considered in the study frequently occur. So, participants have
familiarity with drought and flood. Although we tried to expand our sample to other Brazilian
states and even increase the number of individuals, the procedure of rating the images was
considered unfriendly by the participants and we did not succeed in doing so.
One can argue that sensitization occurred along the presentation of pictures for affective
evaluation. This phenomenon arises when the continuous presentation of unpleasant images
leads to an increasing aversive impact on the individuals [
] and, most commonly, it is
not observed for neutral or pleasant pictures. As a result of the lack of randomness in the
presentation of the images to participants, it may lead to more susceptibility to fatigue or over/
underestimation of the stimulus. On the other hand, some studies have found that affective
discrimination is maintained across trials [
], suggesting that emotional reactions and
emotional appraisals do not habituate or become desensitized across multiple presentations of
similar emotional material. In fact, Dan-Glauser and Scherer [
] verified that, in general, their
ratings were not influenced by the duration or sequential presentation of the stimulus. Thus,
the impact of sensitization across images could be attenuated considering those pieces of
evidence. In addition, data regarding reliability implied an important stability of the database
across picture presentation, which contrasts with the claim that sensitization or habituation
may occurred in our experiment.
As our database represents mainly disaster pictures, it was expected to impact the
participants, although a proper physiological investigation of this question has to be undertaken.
After all, the notion that arousal dimension directly represents sympathetic activation has been
]. One of the arguments is that the affective rating might be dissociated from the
actual corresponding feeling state, although this issue could be more prominent for arousal
than for valence. On the other hand, others claimed that even briefly presenting affective
pictures can activate emotional responses [
]. An association between the two methods, i.e.,
selfreport measures and physiological markers of valence and arousal, would be ideal.
Additionally, EXCEED neutral category was composed of inanimate objects, because we
were not able to find enough pictures with neutral content. In a pilot study (data not shown),
we found that even a neutral landscape will be considered positive to participants, likely due to
a spontaneous comparison between scenes portrayed in flood/drought conditions and
supposed neutral ones. Furthermore, images picturing people with neutral facial expression are
challenging to find. The deliberate decision to use inanimate objects for the neutral category
could, unfortunately, bias the emotional appraisal favoring the negative ones. However, we do
not expect that this claim could affect our data in a suggestive manner. First, the distributions
of ratings for the database (Fig 2) indicated an important internal variability. Second, some
pictures in the database represent only landscapes affected by a dry weather or a flood event,
and even those were rated with negative valence and high arousal.
Finally, the last limitation was that the EXCEED evocative stimuli pictures are inherently
heterogeneous regarding the number of elements depicted, orientation, and illumination
compared to image sets deliberately arranged (e.g., posed face photographs). Those features could
influence the appraisal of the pictures. However, the aim of the database was more to represent
a realistic scene with ecological validity than to have strict parameters of the stimulus.
Therefore, no constraint was imposed on physical properties (e.g., luminosity, contrast, quality of
the image) and no standardization rules were even set for the images. Those properties could
impose significant limitations for use in research with electroencephalography (EEG)
techniques, for example, which are important to provide external validity to the database.
However, again, the focus was to create a database with ecological validity. Furthermore, the
applicability and the purpose of the database are beyond EEG-like methodologies. For
12 / 15
example, in a recent meta-analysis and systematic review, Brown et al. [
] found that
cognitive behavioral therapy is one effective intervention recommended for traumatized children
and adolescents. EXCEED pictures can provide resources aligned to this treatment. In
addition, experimental designs aimed to investigate emotion regulation, the impact of mental
imagery, psychophysiological response in clinical samples (e.g., PTSD patients) or healthy
individuals can take advantage of the validated EXCEED database.
A standardized rated affective database supports better-controlled experiments of emotional
stimuli, as well as allows comparisons across studies, including those with different research
settings. Studies need to carefully choose the type of pictures and their value prior to testing
and controlling in order to obtain a valuable inducing material [
]. In our study, data for
valence and arousal was obtained for each picture and had the subjective experience assessed
through self-report. Valence and arousal displayed a negative linear relationship, and all
dimensions were highly correlated. Finally, EXCEED exhibited higher values for internal
validity and reliability. The psychometric properties of the data implied a consistency, stability
and validity of the entire database. Thus, we gathered empirical evidence to validate the
EXCEED database for use in psychophysiological and social research related to
climate-associated extreme events. Although these are limited findings that we intend to replicate in a
multicultural background and a more diverse sample, to our knowledge, this is the first database
specifically designed to attend a worldwide climate change demand in the field of affective
research and natural hazards.
S1 Fig. Methodological flowchart of EXCEED development and validation.
S1 Table. Characterization of EXCEED pictures, organized by category. Attribution credit,
licenses, original URLs, an indication of modifications, and ratings on valence and arousal for
each EXCEED picture.
Conceptualization: Sabrina de Sousa Magalhães, Diana Kraiser Miranda, DeÂbora Marques de
Miranda, Leandro Fernandes Malloy-Diniz, Marco AureÂlio Romano-Silva.
Formal analysis: Sabrina de Sousa Magalhães.
Funding acquisition: Marco AureÂlio Romano-Silva.
Investigation: Sabrina de Sousa Magalhães, Diana Kraiser Miranda.
Methodology: Sabrina de Sousa Magalhães, Diana Kraiser Miranda, DeÂbora Marques de
Miranda, Leandro Fernandes Malloy-Diniz, Marco AureÂlio Romano-Silva.
Project administration: DeÂbora Marques de Miranda, Marco AureÂlio Romano-Silva.
Resources: Marco AureÂlio Romano-Silva.
Supervision: DeÂbora Marques de Miranda, Leandro Fernandes Malloy-Diniz, Marco AureÂlio
13 / 15
Validation: DeÂbora Marques de Miranda, Leandro Fernandes Malloy-Diniz, Marco AureÂlio
Visualization: DeÂbora Marques de Miranda, Marco AureÂlio Romano-Silva.
Writing ± original draft: Sabrina de Sousa Magalhães, Diana Kraiser Miranda.
Writing ± review & editing: Sabrina de Sousa Magalhães, DeÂbora Marques de Miranda,
Leandro Fernandes Malloy-Diniz, Marco AureÂlio Romano-Silva.
14 / 15
1. Coyle KJ , Van Susteren L. The psychological effects of global warming on the United States: and why the U.S. Mental Health Care System is not adequately prepared . National Wildlife Federation . 2012 . http://www.climateaccess.org/sites/default/files/NWF_Psychological%20Effects.pdf.
2. Ebi KL , Bowen K. Extreme events as sources of health vulnerability: drought as an example . Weather Clim Extrem . 2015 ; 11 : 95 ± 102 .
3. Furr JM , Corner JS , Edmunds JM , Kendall PC . Disasters and youth: A meta-analytic examination of posttraumatic stress . J Consult Clin Psych . 2010 ; 78 : 765 ± 780 .
4. Pachauri RK , Allen MR , Barros VR , Broome J , Cramer W , Christ R et al. Climate change 2014 : synthesis report. Contribution of Working Groups I, II and III to the fifth assessment report of the Intergovernmental Panel on Climate Change . 2014 . https://www.ipcc.ch/pdf/assessment-report/ar5/syr/SYR_ AR5_ FINAL_full_wcover .pdf
5. Lamond JE , Joseph RD , Proverbs DG . An exploration of factors affecting the long term psychological impact and deterioration of mental health in flooded households . Environ Res . 2015 ; 140 : 325 ± 334 . https://doi.org/10.1016/j.envres. 2015 . 04 .008 PMID: 25909883
6. Maclean JC , Popovici I , French MT . Are natural disasters in early childhood associated with mental health and substance use disorders as an adult? Soc Sci Med . 2016 ; 151 : 78 ± 91 . https://doi.org/10. 1016/j.socscimed. 2016 . 01 .006 PMID: 26789078
7. Rataj E , Kunzweiler K , Garthus-Niegel S . Extreme weather events in developing countries and related injuries and mental health disorders-a systematic review . BMC public health . 2016 ; 16 ( 1 ): 1020 . https:// doi.org/10.1186/s12889-016 -3692-7 PMID: 27682833
8. Bennett CM , Friel S. Impacts of climate change on inequities in child health . Children . 2014 ; 1 ( 3 ): 461 ± 473 . https://doi.org/10.3390/children1030461 PMID: 27417491
9. Scherer SR , Dan ES , Flykt A . What determines a feeling's position in affective space? A case for appraisal . Cognition Emotion . 2006 ; 20 : 92 ± 113 .
10. Lindquist KA , Wager TD , Kober H , Bliss-Moreau E , Barrett LF . The brain basis of emotion: a meta-analytic review . Behav Brain Sci . 2012 ; 35 : 121 ± 202 . https://doi.org/10.1017/S0140525X11000446 PMID: 22617651
11. Bolls PD . Understanding emotion from a superordinate dimensional perspective: a productive way forward for communication processes and effects studies . Commun Monogr . 2010 ; 77 : 146 ± 152 .
12. Mauss IB , Robinson MD . Measures of emotion: a review . Cognition Emotion . 2009 ; 23 : 209 ± 237 . https://doi.org/10.1080/02699930802204677 PMID: 19809584
13. Ekman P. An argument for basic emotions . Cognition Emotion . 1992 ; 6 : 169 ± 200 .
14. Posner J , Russell JA , Peterson BS . The circumplex model of affect: an integrative approach to affective neuroscience, cognitive development, and psychopathology . Dev Psychopathol . 2005 ; 17 : 715 ± 734 . https://doi.org/10.1017/S0954579405050340 PMID: 16262989
15. Russell JA. A circumplex model of affect . J Per Soc Psychol . 1980 ; 39 : 1161 ± 1178 .
16. Russell JA . Core affect and the psychological construction of emotion . Psychol Rev . 2003 ; 110 ( 1 ): 145 ± 172 . PMID: 12529060
17. Barrett LF . Valence is a basic building block of emotional life . J Res Pers . 2006 ; 40 : 35 ± 55 .
18. Kuppens P , Tuerlinckx F , Russell JA , Barrett LF . The relation between valence and arousal in subjective experience . Psychol Bull . 2013 ; 139 : 917 ± 940 . https://doi.org/10.1037/a0030811 PMID: 23231533
19. Kinner VL , Het S , Wolf OT . Emotion regulation: exploring the impact of stress and sex . Front Behav Neurosci . 2014 ; 8 : 1 ± 8 .
20. Gardener EKT , Carr AR , MacGregor A , Felmingham KL . Sex differences and emotion regulation: an event-related potential study . Plos One . 2013 ; 8 : 1 ± 9 .
21. Lithari C , Frantzidis CA , Papadelis C , Vivas AB , Klados MA , Kourtidou-Papadeli C , Pappas C , Ioannides AA , Bamidis PD . Are females more responsive to emotional stimuli? A neurophysiological study across arousal and valence dimensions . Brain Topogr . 2010 ; 23 : 27 ± 40 . https://doi.org/10.1007/ s10548-009 -0130-5 PMID: 20043199
22. Aluja A , Rossier J , Blanch AÂ , Blanco E , MartÂõ-Guiu M , Balada F. Personality effects and sex differences on the International Affective Picture System (IAPS): a Spanish and Swiss study . Pers Indiv Differ . 2015 ; 77 : 143 ± 148 .
23. Luo P , Zheng X , Chen X , Li Y , Wang J , Deng L , Zheng X . Sex differences in affective response to different intensity of emotionally negative stimuli: An event-related potentials study . Neurosci Lett . 2014 ; 578 : 85 ± 89 . https://doi.org/10.1016/j.neulet. 2014 . 06 .032 PMID: 24970750
24. Bradley MM , Lang PJ . The international affective picture system (IAPS) in the study of emotion and attention . In: Coan JA , Allen JJB , editors. Handbook of emotion elicitation and assessment . New York: Oxford University Press; 2007 . p. 29 ± 46 .
25. Dan- Glauser ES , Scherer RK . The Geneva affective picture database (GAPED): a new 730- picture database focusing on valence and normative significance . Behav Res Methods . 2011 ; 43 : 468 ± 477 . https://doi.org/10.3758/s13428-011 -0064-1 PMID: 21431997
26. Marchewka A , Żurawski è , JednoroÂg K , Grabowska A . The Nencki Affective Picture System (NAPS): Introduction to a novel, standardized, wide-range, high-quality, realistic picture database . Behav Res Methods . 2014 ; 46 : 596 ± 610 . https://doi.org/10.3758/s13428-013 -0379-1 PMID: 23996831
27. CRED, UNISDR. The human cost of weather related disasters 1995 ± 2015 . 2015 . United Nations, Geneva. http://cred.be/HCWRD.
28. Vasterman P , Yzermans CJ , Dirkzwager AJE . The role of the media and media hypes in the aftermath of disasters . Epidemiol Rev . 2005 ; 27 ( 1 ): 107 ± 114 .
29. Howe PD , Boudet H , Leiserowitz A , Maibach EW . Mapping the shadow of experience of extreme weather events . Climatic change . 2014 ; 127 ( 2 ): 381 ± 389 .
30. Magalhães SS , Kraiser DM , Miranda DM , Malloy-Diniz LF , Romano-Silva MA . EXCEEDÐThe Extreme Climate Event Database; 2017 [cited 2017 April 25 ]. Database: figshare [Internet]. https://doi.org/10. 6084/m9.figshare. 4602016 .v1
31. Cohen J . ( 1988 ). Statistical Power Analysis for the Behavioral Sciences . New York, NY: Routledge Academic.
32. Hinkle DE , Wiersma W , Jurs SG . Applied Statistics for the Behavioral Sciences. 5th ed . Boston: Houghton Mifflin; 2003 .
33. Bradley MM , Lang PJ . Measuring emotion: the Self- Assessment Manikin and The Semantic Differential . J Behav Ther Exp Psy . 1994 ; 25 : 49 ± 59 .
34. Morris JD . Observations SAM: the Self- Assessment Manikin; an efficient cross- cultural measurement of emotional response . J Advertising . 1995 ; 35 : 63 ± 68 .
35. Ribeiro RL , PompeÂia S , Bueno OFA . Comparison of Brazilian and American norms for the International Affective Picture System (IAPS) . Rev Bras Psiquiatr . 2005 ; 27 : 208 ± 215 . PMID: 16224608
36. Cahill L. Why sex matters for neuroscience . Nat Rev Neurosci . 2006 ; 7 : 477 ± 484 . https://doi.org/10. 1038/nrn1909 PMID: 16688123
37. Colibazzi T , Posner J , Wang Z , Gorman D , Gerber A , Yu S , Zhu H , Kangarlu A , Duan Y , Russell JA , Peterson BS . Neural systems subserving valence and arousal during the experience of induced emotions . Emotion . 2010 ; 10 ( 3 ): 377 ± 389 . https://doi.org/10.1037/a0018484 PMID: 20515226
38. Barrett LF . Are emotions natural kinds? Perspect Psychol Sci . 2006 ; 1 : 28 ± 58 . https://doi.org/10.1111/j. 1745- 6916 . 2006 . 00003 . x PMID : 26151184
39. Codispoti M , Bradley MM , Lang PJ . Affective reactions to briefly presented pictures . Psychophysiology . 2001 ; 38 : 474 ± 478 . PMID: 11352135
40. Bradley MM , Cuthbert BN , Lang PJ . Picture media and emotion: effects of a sustained affective context . Psychophysiology . 1996 ; 33 : 662 ± 670 . PMID: 8961788
41. Brown RC , Witt A , Fegert JM , Keller F , Rassenhofer M , Plener PL . Psychosocial interventions for children and adolescents after man-made and natural disasters: a meta-analysis and systematic review . Psychol Med . 2017 ; 47 : 1893 ± 1905 . https://doi.org/10.1017/S0033291717000496 PMID: 28397633