Image Influence on Concern about Stormwater Flooding: Exploratory Focus Groups

Abstract

Increased urbanization coupled with climate change is increasing the number and intensity of stormwater flooding events. Implementing efforts to successfully manage stormwater flooding depends on understanding how people perceive these events. While images of stormwater flooding abound, how these images influence perceptions about flooding events or management options remains understudied. Our objective is to contribute to the general understanding of how various types of images depicting stormwater runoff and stormwater related flooding influence individual and group interpretations of causes of events, major impacts of those events, and responsibility for managing stormwater related events. To this end, we convened focus groups, gave participants numerous photos of stormwater flooding, asked them to identify which images were most concerning, and to then discuss the specific aspects of the photos that prompted concern. We also tested whether a priming image implicating climate change or development as a cause of stormwater flooding influenced viewer reactions. Finally, we asked participants about preferences for who should manage stormwater. Our results revealed that photo location, the water’s appearance, and what people were doing in the photo influenced levels of concern. We also found that priming seems to affect opinions regarding urban stormwater management. Finally, there is some evidence that the absence of people in the photo may affect beliefs about who should manage stormwater.

1. Introduction

More than half of the world’s population [1] and about 80% of the US population [2] live in urban areas with rapidly expanding impervious surfaces, which means more stormwater runoff and increased pluvial flooding [3]. Climate change promises to exacerbate this situation with increased storm frequency and intensity in many places [4]. The existing stormwater infrastructure in the US is already overwhelmed and stormwater flooding causes about USD 9 billion in damages annually [5]. Hence, addressing flooding is a pressing issue and those responsible for managing stormwater need to effectively engage communities as well as individual property owners [6,7]. A challenge to actually doing this, of course, is knowing if, why, or when individuals may be motivated to take action. A review of relevant research shows that perceptions, attitudes, and behavior toward floods are quite complex and there is a need for more qualitative assessment of how and why people react to flooding [8]. Indeed, understanding perceptions of flooding is key to communicating about flood risk and such communication is necessary to encourage property owners to engage in stormwater management [9,10]. Pertinent to our study, Shultz et al. [11] conclude that employing flood-related images may be an effective means of communicating about stormwater management. Yet, despite the prevalence of photos of stormwater flooding in various media and other communication channels, there has been limited study on how these images influence viewers.

Humans are a highly visual species, hence the adages that seeing is believing and a picture is worth a thousand words. Indeed, humans can recognize the subject matter of a photograph that is visible for as little as 13 milliseconds [12]. There is significant evidence that in many contexts, people remember and recall pictures better than text [13,14,15]. The advent of a camera in every pocket along with ubiquitous media has created a culture that is seriously image rich and these images influence what people pay attention to and how they interpret information. For example, information accompanied by a photo is more likely to be deemed “true” regardless of its actual veracity and is more likely to be “liked” and shared on social media [16]. There is evidence that concrete images (e.g., color photos) are more effective at engaging viewers and motivating potential action than abstract (i.e., non-photo) images [17].

There is a growing body of work on the influence of images on perceptions and behavior related to climate change. Research on images often rely on general communication principles showing that whether a photo is likely to prompt concern or motivate individual action depends on how salient (i.e., relevant) the issue seems and the levels of perceived self-efficacy, that is, whether people think they can do something about the issue. Climate change studies focused on images typically categorize photos as showing causes, impacts, or potential actions to address climate change. In general, images of impacts seem to be the most likely to garner attention and influence views on policy or other actions [18]. Other findings reveal that photos a viewer finds salient do not necessarily prompt a sense of self-efficacy [19,20,21,22]. Showing potential ways to address climate change increases a sense of efficacy but can decrease the viewer’s sense of urgency and hence may not motivate action [18,22]. Not surprisingly, an individual’s established identity (e.g., as an environmentalist) and existing attitudes about an issue influence how photos related to that issue are interpreted [23,24]. Further, there are complex relationships among pre-existing attitudes, emotions, and behavior intent. For example, Schultz et al. [11] found that among individuals who do not have strong environmental identities, photos that evoked high levels of “disgust” (e.g., garbage in a stormwater drain) reduced respondents’ level of engagement with the image and lowered willingness to support stormwater management initiatives. Contrary to hypotheses, images that evoked “sadness” (e.g., a turtle eating plastic) did not affect viewer engagement levels compared to control conditions [11]. This aligns with the mixed findings on whether raising concern about an issue engenders action. There is evidence that levels of concern strongly correlate with support for climate change policies [25] but that raising concern to levels of despair or fear can reduce stated behavior intent [20,26].

Related to levels of concern and subsequent salience is evidence about how a photo’s location affects viewer responses. There is currently no consensus on how viewers may react to images that are local compared to images from some distant place [22]. For example, some viewers find images of climate change impacts (like local flooding) concerning, precisely because it is local. Other viewers find non-Western, global flood images more worrisome because they assume the impacts are more negative in those places compared to Western localities [18].

In looking at photo content or intent, findings from previous research include that pictures of people in familiar human pursuits are most memorable [13], and “authentic’” images of people are much more positively received than “staged” images [18]. A study linking wildfire photos to concern about climate change found that images perceived as hopeful (e.g., charity efforts) did not raise concern among respondents, while images showing destruction in the built environment, smoke, or people suffering each raised concern among different groups of respondents [27]. Cross cultural studies reveal that photos of politicians, snow, or a volcano (among others) are not perceived as salient to climate change, while polar bears, landslides, and floods (among others) register as highly salient [19,20,21,28]. Most pertinent to our work, these previous studies show that images of flooding are consistently perceived as important and personally relevant to viewers. As already noted, perceived salience is not necessarily correlated with efficacy and these studies also show that flooding images can lower a sense of efficacy thereby influencing behavior. Hart and Feldman [29], however, did not find that flood images reduced perceptions of efficacy.

While research on the influence of visuals has increased in the past decade, this remains an understudied field [30,31]. In particular, in their review, Wang and colleagues [22] note that there has been minimal research on the effects of viewing specific climate-related impacts rather than the more common comparisons across photos that represent causes, impacts or potential actions. To begin to fill this gap as well as to better determine how location guides viewer reaction, our work focuses on contemporary photos of stormwater-induced flooding in four different US cities.

Extant literature also calls for employing diverse and mixed method approaches to dig more broadly and more deeply into the role that visuals play in eliciting concern, shaping attitudes or influencing behavior [18,30]. Unlike text, which has grammar and syntax rules, there is no definitive way to order images and no right or wrong way to view them. Consequently, “...different features and aspects of a picture can become cues to trigger associations, with the result that a picture allows different readings and can be subject to different or even conflicting understandings” [32]. At the same time, photos evoke a sense of objectivity and are routinely assumed to represent reality, even given contemporary knowledge of how easy it is to alter them [33]. Further, existing work on visuals often focuses on assessing how photos or photos with text are “framed” to deliver some specific message [19,24,30,34,35].

Scholarly content analysis of existing media to identify frames and their effects on viewers or experiments that intentionally test frames have offered insight into the role visuals play in shaping perceptions and attitudes about various issues. Yet, because individual identities structure how a person responds to an image, “it is a mistake to assume that concepts applied to images by researchers—such as defining a photograph of an emaciated polar bear as an ‘impact’ image—are accepted or understood as such by other viewers” [30]. Hence, allowing viewers to conduct their own qualitative content analysis and to explain, in their own words, how and why they contextualize photos in particular ways offers greater insight into viewer responses. To this end, Q-methodology, which has viewers sort and/or rank images according to researcher-defined criteria, has been effective in providing evidence on how people interpret and respond to various images. However, as Wang et al. [22] note, this approach does not typically focus on the characteristics of any specific image but assesses and/or compares categories of images (e.g., climate change causes versus impacts). Further, respondents need to engage with all provided images rather than being free to choose which images to engage with and which to ignore.

To summarize, our work responds to calls for more research that explores why particular images raise viewer concern. Our objective is to contribute to the general understanding of how various types of images influence individual and group interpretations of causes of events, major impacts of those events, and responsibility for managing stormwater related events. All of the photos in our study were of stormwater flooding, a specific climate related impact. This provided us the opportunity to consider viewer responses to a narrow and understudied topic. We employed an exploratory focus group approach based on literature reviews indicating that more qualitative research about images and public perception is warranted [8,18,30]. Focus groups are pertinent because they allow participants to express more nuanced ideas than is possible in quantitative approaches [36]. Further, our design is unique in that we sought to create a process that better aligns with lived experience, whereby individuals encounter numerous images simultaneously, and react, or do not react, to those images. Any effort to provide information or seek some action related to stormwater management should recognize that target audiences are faced with a plethora of competing images.

To build on and expand previous work, we posed these research questions:

• When a variety of photos of stormwater flooding are encountered simultaneously, what attributes of any particular photo raise concern for a viewer? Photo attribute categories include the following:
  • People or no people
  • Location (local to regional)
  • Types of infrastructure (e.g., buildings, cars)
• How does priming the focus group to consider the cause of stormwater flooding as climate change or urban development influence respondents’ reactions to photos?
• How do the priming and/or photo attributes influence general attitudes toward who is responsible for stormwater management?

2. Project Design

Figure 1 shows a broad overview of the project design. From contemporary news sources and the authors’ photo libraries, we curated 22 photos of stormwater flooding events. Our first study question determined the criteria for photo selection. We sought diverse photos from our local site and three non-local cities that included and excluded people and featured an array of infrastructure (e.g., streets, houses, cars). The selected photos were printed in full color, laminated, and labeled from A to V. Each photo had a caption that read, “Recent flooding from stormwater runoff in [the city of the photo’s location]”. Unfortunately, because most of the photos are from actual news coverage, copyright precludes sharing them in this manuscript. Figure 2 includes several author-sourced photos from the local site of Boone, North Carolina, United States to show the kinds of images used in the project, and

Table 1. Photo information. Location codes: Boone = Bo; Charleston = Ch; Detroit = De; Washington DC = DC. Select/Saw represents the number of groups that selected that photo/the number of groups that had that photo available.

Photo	Loc	Water	People	Infrastructure	Select/Saw	Scenario
A	DC	dark, flowing at edge of street	person with umbrella and stroller crossing street, approaching flood water	cars driving on non-flooded street; buildings in background	0/2	1
B	DC	clear, standing on street	person walking on non-flooded sidewalk	cars parked, truck driving on flooded street; buildings not obviously flooded	0/2	1
C	DC	clear, standing on street	None	buildings potentially flooded; trash bins in water	1/5	1, 2
D	DC	brown, flowing in parking lot, maybe street	None	multiple submerged cars; buildings in background; dumpster	5/5	1, 2
E	Ch	dark, standing on street/alley	two people standing in doorway looking at water at doorstep	house not yet flooded; trash bins in water	2/2	1
F	Ch	clear, standing on street	None	trash bin in flooded street; cars parked on flooded street; buildings in background	0/5	1, 2
G	Ch	clear, flowing in street/sidewalk	two people with umbrellas walking in ankle deep water	cars driving on flooded street; buildings in background, maybe flooded	0/4	1, 3
H	Ch	clear, standing on street	None	large houses, berm around focal house	2/7	1, 2, 3
I	De	dark, standing in basement	None	closeup of stairwell leading into flooded basement	0/2	1
J	De	dark, standing on street	person wading in knee deep water	several submerged cars, one semi-truck on large street	3/4	1, 3
K	De	dark, standing on sidewalk	two kids playing at edge of water covered sidewalk	parked car; trash bins in background	0/4	1, 3
L	De	dark, standing on street	None	storage bins & plastic tarp in water; buildings in background	1/4	1, 3
M	Ch	clear, bubbling from manhole cover	indistinct person in background	closeup of manhole cover; large houses in background	0/4	2, 3
N	Bo	clear, standing, settled sediment visible	None	inside a lab room	2/5	2, 3
O	Bo	brown, flowing	None	flooded outdoor stairway; building in background	1/3	2
P	Bo	brown, flowing across sidewalk	None	car driving on non-flooded street	0/3	2
Q	DC	dark, flowing in street	None	car driving on non-flooded street; buildings in background	0/3	2
R	Ch	dark, flowing in street, into drain	two sets of two people walking in ankle to knee deep water	cars parked and driving in background; flooded buildings in background	0/2	3
S	Bo	brown, flowing in hallway	person in ankle deep water giving thumbs up; indistinct people in background	interior hallway	2/2	3
T	Bo	brown, standing in courtyard	None	parked work truck; building likely flooded	1/2	3
U	Bo	dark, standing in courtyard	person wading in knee deep water	building in background	0/2	3
V	De	dark, standing in playground	None	submerged playground equipment	1/2	3

provides relevant information about all of the photos. Once photos were selected, we grouped them into three scenarios to align with our first research question. Scenario 1 included 12 photos from three non-local (i.e., not Boone) US cities, some with and some without people. Scenario 2 included 9 photos all without people, but a mix of local and non-local sites. Scenario 3 included 12 photos, both local and non-local, as well as some with and some without people. The photos in all three scenarios included various types of infrastructure. We also selected photos to ‘prime’ the focus groups about a cause of stormwater flooding as either climate change or urban development and wrote appropriate captions (Figure 3). Finally, we crafted a pre-questionnaire to gather information from respondents about their pre-existing knowledge of stormwater and what they consider their hometown. In a post-questionnaire, we gathered demographic data as well as respondents’ views on who should manage stormwater and who should pay to manage stormwater. The questionnaires are available as Supplemental Materials.

Focus groups were held in February and March 2023. We recruited students from Appalachian State University. The university research office supplied a random sample of 3000 student emails representing about 15% of the student body. Approximately one week before each scheduled focus group, we randomly selected 300 emails and sent an invitation to those students. The invitation stated they had been randomly selected to participate and they would be paid USD 20 for the one-hour time commitment. Our target was three to five people per group and five of the six originally scheduled groups included three or four participants each. Only two participants arrived for the second focus group and therefore we repeated that scenario as the seventh focus group. The number of participants for each scenario was roughly equivalent, with seven engaging with Scenarios 1 and 3 and nine engaging with Scenario 2. Further, 12 participants received the development prime, while 11 received the climate change prime.

The average participant’s age was 22 (range 18 to 29) and about 48% identified as female (

Table 2. Descriptive statistics by scenario and overall.

Scenario	1	2	3	Overall
Average age	21.9	22.7 *	22.7	22.4 *
Average years of college	3.4	3.7	3.3	3.5
Proportion female	0.29	0.67	0.43	0.48
Proportion from North Carolina, US	0.86	0.89	0.57	0.78
Proportion selected correct definition of stormwater runoff	0.71	0.56	0.71	0.65
Number of participants	7	9	7	23

Note: * These are underestimated because two graduate students did not report their age

). All of the groups met in the same conference room and the authors served as facilitators and recorders in all sessions. Participants were given consent forms to read and sign and were then asked to complete a pre-questionnaire which gauged prior understanding of stormwater runoff. Only three respondents reported that they had recently heard or read anything about stormwater, while 15 of the 23 (65%) participants selected the correct definition of stormwater runoff from a list of options. This aligns with previous survey research showing that about 59% of people can identify the correct definition [37]. Although our sample size is small and student focused, this provides some support that our respondents are not radically different from the general population in basic knowledge about stormwater.

To start the focus group, the facilitator (lead author) presented the ‘priming’ image (Figure 3) assigned to that group, read the caption and asked if participants had any questions. The facilitator next placed all of the photos for that group’s scenario on the table. Participants used the identifying letter (A to V) when referring to specific photos throughout the session.

Our design sought to avoid overt ‘group think’ by assessing both individual and group reactions to images. Therefore, participants were first asked to individually identify one photo that grabbed their attention and raised concern, and to use provided paper to make a brief note about what caught their eye. The group was then asked to come to consensus on which three photos were most concerning. The group discussions provide, in essence, a qualitative content analysis of specific attributes of the images that prompted concern.

Beyond initial prompts, the discussion was freeform. We recognize evidence from cognitive science showing that because the brain is always faced with competing images, any instruction to focus on something in particular influences what is actually seen [38]. Therefore, we sought to limit the amount of instruction given to the groups. Notably, the group was not asked to sort or comment on all of the photos because we wanted to see which, if any, images were ignored. Only one of the seven groups systematically walked through all the photos they were given. The project was designed for small groups to allow participants to engage as long as they wished with the images and to ensure there was plenty of time for each participant to express their own views.

All groups were recorded (video and audio), and the audio recordings were loaded into Dedoose (www.dedoose.com, accessed 1 June 2024) for analysis. We applied a combined inductive and deductive approach to coding. Our research questions, which were based on evidence from previous studies, provided a set of parent codes for references to people, location, and infrastructure. Based on image content, each parent code suggested several child codes. Once all focus groups were complete, we discussed first impressions of the results, which prompted an inductive approach to generating additional codes, including water appearance and flooding impacts. In the first coding cycle, we identified references to specific images and parent codes. In a subsequent cycle, we honed child codes. Figure 4 shows the final set of codes applied. Each author independently coded multiple transcripts, and a reliability test conducted within Dedoose showed a Kappa statistic of 0.68, which indicates strong similarity in how the transcripts were coded. Through the coding process, we captured counts of references to each image and each code as well as specific comments about why participants found an image concerning or not.

3. Results

3.1. Photo Attributes

To address our initial research question, we employed the three scenarios to learn how viewers reacted to the presence of people, local or regional photos, and various types of infrastructure. In reporting specific responses from participants, we have not corrected grammar or syntax.

3.1.1. People or No People (Scenarios 1 and 3)

Of the 10 images with people in them, five were selected by at least one individual as their most concerning photo, while groups included three distinct images with people among their most concerning. Whether a photo with people elicited concern depended on what the people were doing as well as the overall image context. If people seemed to be going about routine activities, even if walking or driving through floodwater, this allayed concern. Comments about photos A, B, G, and R, which all showed people walking through or near floodwaters, included the following:

“The momentum of A caught my eye, but people are still walking around. So that is one of the reasons why I kind of eliminated that one. And similar with B, having someone in the image, seemingly unfazed by the water level”.

“I thought G was interesting that the cars are kind of still driving, like nothing is wrong and then people are just walking around with their umbrellas, like they’re kind of used to it, which I guess if it’s Charleston, I could kind of see it being a coastal city that there might be pretty frequent flooding”.

“In picture R there are still people outside going about their day, it almost tells me that, you know, just one big flood doesn’t shut down the city. They can still go shopping, go to the stores”.

One group explicitly said that image B represented no concern at all, and another group said the same about image R.

In their individual photo selection, one participant selected photo G and noted that the ‘routine’ behavior of driving in wheel deep water was precisely what they found concerning. Likewise, two individuals in the same focus group selected photo R as most concerning and one noted “…it looks like a forceful whirl of water … I imagine it could knock a small child off their feet”. In the group discussions, however, consensus was that these situations were more inconvenient than threatening, at least compared to other images, and no group selected photos G or R as one of their top three most concerning. This finding potentially aligns with Duan et al. [27] who found that hopeful images do not raise concern about wildfire.

Images with people that unanimously raised concern in the groups that viewed them included E, J, and S. Photo S shows a flooded hallway in a university building with a person standing in the water giving a thumbs up sign (see Figure 2). One group interpreted the person’s smile and gesture as a sarcastic take on being ‘routine’ while the flooded hallway was clearly a concern. Image E showed two people in the doorway of a house looking out at floodwaters and focus group participants expressed sadness and sympathy as well as concern about this photo. Participants’ comments included that the people seemed “stuck” and “hopeless” as well as noting that the water was “coming up to the houses, like obviously causing damage”. Photo J shows a person walking in knee deep water on a road along with multiple partially submerged vehicles. While no individual selected photo J as their most concerning photo, it elicited significant concern as groups discussed it. One participant compared it with photo E because “they both show people … so it makes you feel like sympathy for them”. While photo J shows only a single person, several participants raised the idea that there could be people in the cars and one participant suggested people could be “trapped on the road and cannot get out”. Photos E and J both seemed to suggest to viewers that the people were not able to continue with normal activities. Reflecting the importance of overall photo context, images J and U both feature a person in knee deep floodwater, but photo U is in a university courtyard and there is a rainbow in the background (Figure 2). Rather than raising any concern, one group noted that the rainbow was “cool”.

For some photos, it was precisely the absence of people that raised concern. Image D, featuring muddy, turbulent water, and submerged cars in a parking lot, was overall the most concerning image for a variety of reasons, including that the absence of people in the image meant that routine behavior was not possible. As one participant said, “an interesting detail about D, there’s no humans in the photo at all...there is no sign of life whatsoever”. Another person said image D made them sad because of the “people who left their cars behind or were stranded in some way”.

The potential for people to be at risk, even if the photo did not explicitly show that risk, increased concern. People driving cars through floodwater was generally perceived as less concerning than submerged cars where people could potentially be trapped or stranded. Similarly, image K shows children near a flooded sidewalk, and this was of no concern, prompting comments that “it just looks like they’re having fun” and it “is kind of innocent”. However, photo V of a flooded playground with no children present prompted one participant to invoke the potential for danger and that person’s group subsequently included it as one of their most concerning images.

3.1.2. Local or Not Local (Scenarios 2 and 3)

The focus groups occurred in Boone, North Carolina and the project used images from Boone to represent local events. The other images featured flooding events in Charleston, South Carolina, Washington DC, or Detroit, Michigan. Participants were asked about their familiarity with the other locations to determine whether those places could also be perceived as local. None of the participants reported any strong ties to the non-Boone locations. Among the 16 respondents who saw images of Boone, six of them selected a Boone photo as their most concerning and four of those explicitly said their familiarity with the locale was why they selected it. One respondent selected all of the local photos in Scenario 3 as most concerning, noting, “Any image that is of Boone invokes a lot of emotion for me. Seems bad to me because I’m here”.

Location does seem to have been an important influence on group photo selections. Recall that only two of the three scenarios included local images, yet four Boone photos (N, O, S, and T) were selected by at least one group as most concerning compared to three images for Detroit and two each for Charleston and Washington DC. Four of the five groups that viewed local photos selected at least one local image in their top three, and two groups included two local images in their top three. Specific comments from the groups do highlight that they were very attuned to the place in the image. For example, in one group, a participant suggested that photo O of sediment-laden water flowing down a staircase, be in their top three and another group member said, “Especially since that’s on our campus”. A participant in a different group noted, “I do walk those stairs a lot, so that one is concerning”. Both groups that viewed photo S, the flooded campus hallway, selected it, while two of five groups that viewed image N showing standing water and sediment in a campus lab room, selected it. Water being in a familiar building was raised as a concerning point and as one participant noted about photo S, “I remember my classes got canceled. I just think the fact that it went into the buildings really was shocking to me and ruined a lot of things”. Comments about photo N reflect the power of the location to raise concern, even when the image itself is not perceived as bad. One respondent noted that image N was, “not that drastic, but, it’s in Boone … You don’t really expect a lot of flooding to happen”. Conversely, photo I, which showed dirty, deep water in a building in Detroit was not discussed by either group that saw it, providing further evidence that familiarity with the location did influence levels of attention and concern.

Location also correlated with comments about preparedness. The general consensus across groups was that Boone is not prepared for flooding and that flooding is unexpected in a mountain region. As one participant noted, “Boone’s poor planning makes me so nervous … and, App [the university] is not prepared”. Discussions about coastal city Charleston, on the other hand, were split as to whether the city is prepared for flooding. One participant noted, “The pictures of Charleston are not that bad because they are more prepared for flooding, they’re more prepared for hurricanes than Boone might be”. While another participant said that Charleston photos are, “tricky cuz they’re coastal towns and the coastal flooding and the way a lot of more coastal towns have been designed in our country is not prepared for like, storm runoff or storm sea surge”.

These findings align with previous works showing that viewers often find local images more concerning because they are familiar and relatable [18,39]. Our work offers a narrower interpretation of local and non-local, as our images were all from eastern US cities, rather than contrasting local with global images. Within this narrower view, our results support previous findings that perceptions of how prepared a place is can influence viewer reaction [18].

3.1.3. Infrastructure (Scenarios 1, 2, and 3)

Cars prompted significant discussion. One aspect of this seemed to be relatability, as one discussant noted about photo D, “That could be my car in there”. In another group, someone drew a comparison between what they were seeing in image D and the Boone shopping mall, whose parking lot frequently floods. Submerged cars in photos D and J elicited high levels of concern both because of potential danger to people as well as the obvious property damage. Additionally, the flooded road in photo J raised concerns about limits to movement and access for emergency vehicles. In contrast, as already discussed, flooded streets with cars still driving evoked a sense that things must not be too bad, because people were continuing with normal behavior.

Images focused on flooding in or around buildings did prompt some discussion. In considering photo N, one participant explicitly noted a distinction between inside versus outside a building: “All of these other photos are structured from the outside so you can think about what damage might be in the building, but that really puts a powerful opinion on what it would do to the inside of someone’s home or office”.

The photos included in this study acknowledged previous research on how emotions, like disgust, affect attitudes. As part of the infrastructure attribute, six of the images included trash bins or trash. Of those, only two (photos F and L) elicited comment, and one group selected photo L as one of their most concerning. Photo L is a tight shot of storage bins and plastic sheeting sitting in the rain on a flooded street, and respondents clearly interpreted this as trash. They noted that it seemed to symbolize general negative impacts and more specifically, raised concerns about environmental contamination. As one respondent said, “the trash that’s currently being collected by the stormwater runoff is gonna spread out and it’s gonna pollute the environment more”. The group that included photo L as one of their top three most concerning compared images J and L, and eventually selected image L because it focused on the bins. As one respondent noted, “the fact that there’s not much to look at means that there’s like more impact behind the thing that you can see, you know?”

Photo F, on the other hand, is an overturned trash bin in the middle of a flooded street. While the bin is the focal point, the photo is a wide angle, sunny shot with buildings and cars in the background. While several participants noted that any trash present is an environmental concern, they also noted that the clear water in the photo reduced their concern. Again, reflecting the importance of context, photo K features multiple trash bins in the background, but the focal point of kids playing seems to have offset any concern. Image E also includes multiple trash bins, but no one mentioned those in expressing why the photo was concerning, instead they focused on the people looking at the floodwater.

3.1.4. Water Appearance

One result that was not explicitly anticipated in our research questions was that participants placed consistent emphasis on the water’s appearance. As Figure 4 shows, participants frequently commented on the water’s motion, depth, and clarity. Relatively shallow, still, clear water and/or clear skies suggesting that the worst of the flooding was over, generally reduced concern. For example, the few comments about image B, which includes vehicles and a person walking, noted that the floodwater was shallow and calm and therefore not concerning. In multiple discussions, participants grouped various photos together in noting that the water in all of them was “clear” or “muddy” as a variable to consider when assessing concern.

Deep, murky, moving water that seemed to be rising generated greater levels of concern. One respondent neatly captured the sentiment that multiple people expressed about photo D, “[it] was pretty shocking, you can see the water’s moving pretty fast and it’s really brown and it’s covering the cars. It gives you an idea of how high it is”. Similarly, comments about why photo S was concerning included, “it looks like a river coming down the hallway” and “it looks like it’s still coming in”. Likewise, participants noted about photo J that there “must have been a very violent current” because cars were “knocked askew”. Other participants explicitly noted that the water was “up to the windows” of the cars and that the person had water “up to his shins”. Suggesting that details about the water’s appearance and photo context are key, photo P showed sediment-laden water pooled on a sidewalk (Figure 2) and this prompted very few remarks, while photo O showed sediment-laden water flowing rapidly down a large stairway, prompting one group to include it in their top three most concerning images. Comments included, “I wouldn’t wanna walk through that water” and “I would feel threatened”.

Discussants consistently used the water’s appearance to make claims about perceived environmental impacts, and one group explicitly noted that potential effects on the environment was their primary criteria for deciding whether a photo was concerning. Across different groups, multiple people noted that image D was a concern because it was likely spreading pollution. One respondent said of photos S, T, and J, “I realize all these images are of human structures, but pollution definitely comes to mind. All this water is eventually gonna go away and it’s gonna go downstream. 100%. It’s gonna carry everything it picked up too”.

For some photos, various aspects of the water’s appearance prompted mixed reactions as to whether the situation presented a concern. Photo C, for example, was perceived by some as a pretty picture and of no concern because the water flooding the street was calm and the sky was clear, but it raised concern for others because the water seemed deep and the potential impact it was having was uncertain. In one group, a participant noted that photo C “did catch my eye just cuz it looks pretty, cuz the water’s so calm, it’s reflecting” and another participant responded, “Yeah, exactly. It just doesn’t look threatening”. Another group, however, did select C as a concern precisely because the water was not moving. One respondent noted, “I think the standing water is the worst. If it’s running, that means it’s going somewhere else” and another said, “it’s just collecting everything”.

Image N with its visible but settled sediment elicited disgust from multiple participants, “N is kind of gross. You can see the mud and stuff and it’s just kind of still in there”. In a different group, a participant said, “it’s just so dirty looking”. Others, however, suggested that the water may just be “pooling in that slight area” because “it doesn’t look as bad around the edges”, and one person said, “it’s disgusting, but it’s manageable”. Further, as already noted, some expressions of concern about photo N were less about the water itself and more about its presence inside a familiar building.

All seven groups discussed image H, which focused on a large home on a flooded street. Some participants concluded the image was not concerning because the water was clear enough to see the street, still and shallow, while others raised potential concerns about the greenish tint of the water and that the overall impact the water might have was not evident from the image. One respondent noted that photo H “is interesting ‘cuz it’s sort of the, the aftermath or it looks like it might be the aftermath. It’s not as muddy or like as movement heavy as some of the other photos are”, while another person said, “it’s just the water, you’re not seeing kind of how it’s impacting everything else”. Additionally, several people noted that their initial reaction to image H was that they were not sure what it showed. Some people initially interpreted it as a house on a waterfront while one group suggested it looked like an old Mississippi River boat. These results align with previous work showing that “image ambiguity” can prompt deeper cognitive processing about the image and can influence how people react to the image subject matter [24].

3.1.5. Image D and Selection Consistency

We do not find specific evidence of ‘group think’ occurring, and all respondents stated in the post-questionnaire that they felt that others had listened to their ideas. In five of the seven groups, at least two of the group’s most concerning images included photos that individuals in the group had independently selected. Across individuals and groups, there was consistency in which photos were deemed very concerning and which were largely ignored (

Table 3. Images selected as most concerning were consistent across groups. Images choices were not ranked. Images E, J, and S include people. Images N, O, S, and T are local. All images include infrastructure. See Table 1 for specific image descriptions.

Scenario/Attributes	Prime	Image 1	Image 2	Image 3
1/people & no people	Development	D	E	L
non-local
infrastructure
1/people & no people	Climate	D	E	J
non-local
infrastructure
2/no people	Development	D	H	N
local & non-local
infrastructure
2/no people	Development	D	N	O
local & non-local
infrastructure
2/no people	Climate	D	H	C
local & non-local
infrastructure
3/people & no people	Development	J	S	V
local & non-local
infrastructure
3/people & no people	Climate	J	S	T
local & non-local
infrastructure

). Given the number of photos being viewed, this consistency suggests some common ground in what images engender concern. Photo D prompted significantly more discussion than any other photo. Seven of 16 individuals who saw it selected it as their most concerning photo, and all five of the groups that saw it included it in their top three. Photo D represents core aspects of our findings as to what raises concern, including the water’s appearance, submerged cars, and no people, suggesting that normal activity is suspended.

Other unanimous group choices included photos E and S, as both groups that saw those photos included them in their top three most concerning. In individual selections, three of seven people who saw image E selected it as their most concerning photo while three of nine who saw image S selected it as most concerning. Three of the four groups that saw photo J included it in their top three. Photo J is an anomaly because no individual selected it as most concerning. It is a relatively dark photo, and the water is deep, but not turbulent. Perhaps this precluded viewers from focusing on it during their initial scan of the photos, but as people continued to study the images, it raised concern.

At the other end of the spectrum, no individual or group selected photos A, B, I, K, M, or Q. The only comment about photo I was to describe it as “looking down into someone’s basement”. Photo Q, showing turbulent water at the edge of a street where cars were driving, was lumped into a comment noting which photos were from Washington DC. Interestingly, photo Q is the same as photo A, but with the people cropped out. These results suggest a potential hierarchy in attributes commanding attention. While photo A did elicit a few comments, the people engaged in normal activity offset any concern. In image Q, the moving cars suggest routine activity, and the floodwater not causing any obvious damage seems to have precluded participants’ interest. Discussion about photo M, which is a close-up of water bubbling from a manhole cover, included uncertainty about what the image showed and hence whether it was a concern or not. Several participants mentioned photo M only to include it in noting which images were from Charleston, while others specifically observed that the water was clear and as one person noted, therefore, “not a big deal”.

3.2. Priming: Climate Change or Development

Our second research question concerned the potential impact of priming the groups. As already noted, at the beginning of each focus group, the facilitator showed participants an image and read a caption highlighting either increased development or climate change as a cause for stormwater runoff (Figure 3). Although the priming image was left on the table throughout the focus group’s engagement, none of the participants directly referred to the image or the priming message during the discussions.

Because a seventh focus group was added, three groups received the climate change prime, and four groups received the development prime. Further, Focus Group 3 received the development prime and discussed every photo in their scenario. This contributes to more comments overall in the development groups. However, when Focus Group 3 is removed from the analysis, the development-primed groups still offer more commentary (Figure 5).

In looking at which images catalyzed more discussion, we find no obvious connection to the prime. More robust discussions in the development-primed groups may, however, reflect a greater sense of comfort and/or familiarity in talking about tangible, seemingly obvious relationships between the built environment and runoff. On the other hand, the climate change prime may be perceived as more distant and/or more controversial, potentially reducing willingness among participants to speak freely.

3.3. Preferences for Managing Stormwater

To address our third research question, the post-questionnaire included statements on who should be responsible for managing stormwater and who should pay for that management. The questions included private sector (homeowners and businesses) and public sector (city, state, or federal government) options and employed a Likert scale ranging from strongly agree to strongly disagree (

Table 4. Proportion of participants who agreed with each statement sorted by prime as to a cause of stormwater flooding.

	Prime
	Climate	Development
Homeowners have a responsibility to prevent water from running off their property	0.64	0.67
Business owners have a responsibility to prevent water from running off their property	0.64	0.83
Homeowners should pay for stormwater management on their property	0.09	0.25
Business owners should pay for stormwater management on their property	0.55	0.50
City government should have primary authority for how stormwater is managed in any community.	0.91	0.75
State government should have primary authority for how stormwater is managed in any community.	0.55	0.33
Federal government should have primary authority for how stormwater is managed in any community.	0.18	0.17
City government should pay for stormwater in their communities.	0.91	0.75
State government should pay for stormwater in their communities.	0.82	0.58
Federal government should pay for stormwater in their communities.	0.55	0.42

). While the small sample size precludes making any definitive statements, there are some noteworthy observations. Compared to those who saw the climate change prime, those who saw the development prime were more likely to distinguish between business owner and homeowner responsibilities for managing stormwater on their property. Those who saw the climate change prime, however, were much more likely to agree that city and state governments should manage stormwater compared to those who saw the development prime. These results suggest that for the private sector, discussants associate development with businesses rather than with homes. For the public sector, these results align with flood risk research showing that individuals often expect local or central government entities to address flooding issues [8,40,41]. These findings may also suggest that climate change as a cause distances the events from the private sector and nudges it toward the public sector. This interpretation has implications for how to describe or explain stormwater flooding causes when seeking support for management programs.

Figure 6a,b show results segregated by private versus public actions and by scenario. Regarding private actions, the agreement that homeowners and businesses should be responsible for managing stormwater on their property is stronger than the agreement that they should pay to manage that stormwater. For government actions, the responses are mostly reversed. Respondents are somewhat more likely to say the government should pay than to say they should have authority. Response consistency is highest related to city government. In Scenarios 1 and 3, all participants agree that city government should have primary authority for managing stormwater in their community. Agreement that the city government should pay is also strong. Providing support for the validity of focus group findings, these responses are similar to a survey which asked the same questions of more than 700 people in 13 states [37].

One interesting result from our study is that agreement is relatively weak on all questions in Scenario 2. Especially noteworthy is that although more respondents engaged with Scenario 2 than with Scenarios 1 or 3, none of those respondents agreed that homeowners should pay to manage stormwater on their property. Recall that none of the photos in Scenario 2 included people. This suggests that the photos may have influenced views about managing stormwater, and this has implications for what kinds of images planners and managers may want to use in stormwater-related informational materials targeting individual homeowners.

3.4. Study Limitations

Our focus groups were student-based, and therefore do not reflect the demographics of the general population. However, as noted, we do have evidence that our respondents do not differ from the general population in terms of knowledge about stormwater or options for managing stormwater. The criticisms that arise when student samples are used are familiar and show up often in the experimental economics literature where it is common to rely heavily on student subjects. As this has been a topic for decades, it has been discussed often in the literature (see [42] for an overview). Whereas students are younger and better educated than the general population, any convenience sample (including professionals) will bring their own set of qualities which may be unrepresentative of the larger population. With large samples, one can control for subject characteristics using regression analysis. With qualitative research, however, that is not usually an option.

Like many focus groups, our sample size is small. Although the study was designed for small groups, we had a lower response rate during recruitment than we would have liked. There may be some lingering post-COVID effect as other participatory laboratories on campus are having similar experiences with student recruitment.

Of course, individual participant’s values, experiences, and biases influence how they interpret images, and this study was not intended to decipher that type of detail from the focus groups. Further, while we designed the study to avoid ‘group think’, there are always going to be group dynamics within any focus group that may influence what participants choose to share. Researchers can minimize ‘group think’ by encouraging individuals to share their views and using post-questionnaires to determine whether some members of the group felt their views were not considered. In our study, we collected individual responses before the group began their discussion. We found consistency in most cases between individual responses and group decisions.

4. Conclusions

Our study is unique as it offers a more focused understanding of how viewers react to stormwater flooding photos and what, specifically, may evoke a threat. This offers insight into how viewers may perceive images included in communication about flooding or actions to mitigate flooding. As communities face increased risk from stormwater flooding, there is a concomitant need to garner support for management options. Modern informational materials and communication channels are highly visual, and our results highlight some distinctions among images that are likely to be ignored, attract attention, and/or catalyze concern. Our most interesting findings relate to the presence or absence of people and submerged cars, and what the water looks like in photos. While other research has highlighted that showing people in photos can draw viewers’ attention [13,43], our work suggests that in flood-specific contexts, how people are portrayed is important. If the people in the image appear to be engaged in routine activity, like walking or driving through floodwater, this may reduce any perceived threat from the flooding event, which subsequently may lower support for managing stormwater. Several photos with no people but showing deep, dirty, or turbulent water were concerning precisely because people were not engaged in routine activity. Further, photos of submerged cars prompted concerns because of the potential danger to people as well as the property damage. Yet, our study also found that the lack of people in photos seemed to reduce the level of reported support for either the private or the public sector to manage stormwater.

Another conclusion we draw is that what the floodwater looks like in photos matters. Our focus group participants used water depth, clarity, and motion as cues to assess danger, property damage, and environmental concern. In general, photos showing deep, turbulent, and/or sediment-laden water raised more concern than those showing shallower, still, and/or clear water. Our results also suggest, however, that there are complex relationships among the water’s appearance, the location of the flooding, what people are (or are not) doing, and how viewers react. Location can affect viewers sentimentally but may also raise or lower concern depending on the viewer’s belief about how prepared the region is to handle flooding from stormwater.

Previous work suggests that municipalities need to develop effective education and outreach programs about stormwater management [7] and our work provides some insight into the importance of careful image selection when developing such materials. Further, as an exploratory project, our work lays a foundation for thinking about future research, including a deeper examination of viewer response to specific photo attributes. For example, the Protection Motivation Theory tries to explain when/why people are likely to engage in actions to address potential hazards, including flooding. Work within this frame includes understanding how people perceive any potential threat, perceptions about possible actions to take as well as the importance of communicating about risks and possible responses [9,40]. Yet, despite the prevalence of photos of stormwater flooding, there has been limited study on how viewing such images may align with the Protection Motivation Theory, especially how images affect perceived threat.

More specifically, based on our findings, continuing to assess viewer reactions to people in flood-related photos seems germane to creating effective communication strategies related to managing stormwater. Likewise, our study suggests there is reason to elicit more detailed perceptions of the relationships among people, location, and water appearance. Another promising avenue of study could be to explore in more detail how focused photos like these influence perceptions of stormwater management options. Additionally, our findings suggest that the priming condition establishing climate change or development as the cause for stormwater flooding did catalyze some differences in how participants engaged with the photos and subsequent management preferences. Delving more deeply into whether this effect holds with larger or different respondent groups is warranted.