Media Contributions to a Chesapeake Bay Watershed Collective Identity? A Tale of Three Cities

Although collective action is needed to address many environmental challenges, it cannot proceed in the absence of collective identity, that is, evidence of group belongingness expressed in or via communicative behavior. This study looked for evidence of a collective identity in newspaper articles that referenced the Chesapeake Bay Watershed. The data were drawn from local papers published in municipalities located at the headwaters of the Susquehanna River, midway down the Susquehanna, and where the river meets the Bay. Computerized content analysis assessed the frequency with which the Chesapeake Bay and watershed were mentioned alongside a set of keywords thought to represent different facets of identity (e.g., agriculture, fishing, swimming). The results showed substantial variation in frequency across time and place but low absolute levels of coverage of the Bay and the watershed. Multidimensional scaling revealed different structures to collective identity as a function of place. These differences in content may be attributable to varying demographic and environmental characteristics along with proximity to the Bay. But, to the extent that media contribute to collective identity among residents of the watershed at all, they do so in a complex and heterogeneous manner.

When explorer John Smith arrived at the Chesapeake Bay in the 17th century, he found an estuary that teemed with blue crab, rockfish, and oysters that "lay as thick as stones" (Captain John Smith n.d.). Until recently, the Bay was one of the most productive bodies of water in the world (Cooper 1995). Primarily due to fertilizer runoff, wastewater, and stormwater from the Chesapeake Bay watershed, the oyster, crab, and rockfish populations have been decimated (Schulte 2017;Seitz 2020;Shepherd et al. 2018). By every indicator, the health of the Bay is poor (UMCES 2020).
The causes of the deterioration of the Chesapeake Bay are closely tied to human behaviors that occur upstream in the Chesapeake Bay Watershed (hereafter CBW). Consisting of 64,000 square miles, the CBW is vast in geographic terms and hugely varied with respect to economics, politics, and culture. Diversity of interests and a lack of awareness of the impact of upstream and downstream actions on the Bay are major impediments to restoring the Bay to health. However, as one ecologist noted "Nothing is going to change until people in the whole Watershed understand that they are all in this together" (C. A. Cole, personal communication, October, 2019).
Embedded in that statement is an implicit theory of change that has formal parallels in scientific theories of identity (David and Bar-Tal 2009). In brief, collective action depends on the existence of collective identity, that is, a sense of group belongingness and common purpose (Fominaya 2010). Stated differently, collective identity enables the sort of targeted, coherent collective actions that have the potential to benefit the health of the Chesapeake Bay. Once people have a sense of collective identity, they are a step closer to the sort of unified actions performed in solidarity that have the potential to lead to systemic improvements in the Bay. What is the source of collective identity? Identity theory points to messages that, at minimum, describe (a) the existence of a group and (b) one's membership status relative to it. In this study, we made use of computerized language analysis to examine news articles in three municipalities within the CBW that expressed both points. We did so for the years 2008 to 2020 in three communities located at the headwaters of the Susquehanna River, on the Bay itself, and about halfway between the two. Our inquiry sought answers to the following questions: How frequently are the Bay and the Watershed covered? What is the content and structure of that coverage? And, to what extent do coverage and structure vary by time and place?
The Nature of Identity Polletta and Jasper (2001) conceptualized identity as a "… cognitive, moral, and emotional connection with a broader community" (p. 285). Similarly, Taylor and Whittier (1995) contended that collective identity is "…the shared definition of a group that derives from members' common interests, experiences, and solidarity" (p. 105). Two key features of these definitions include (a) the existence of some group and (b) the individual's membership within it. These ideas are simple, but fundamental and necessary conditions for the existence of identity (Fominaya 2010). Simply put, individuals must have knowledge that a group exists and that they are, at least potentially, a part of it to have a sense of collective identity.

Identity Formation and the Media
To the extent that a collective identity exists, it must appear in the communal discourse of its members, which should include media. Giddens (1991) noted that mediated experiences "…have long impacted both self-identity and the basic organization of social relations" (p.4). Through their publications, local newsmakers can play especially prominent roles in making a collective identity more or less salient along with fostering senses of community, geographic identity, and social cohesion (Abernathy 2018). Kaniss (1991) argued that a central function of local news media in the U.S. has been to produce local identity as a means of linking local audiences. Local journalistic interpretative communities create meanings from events, both within their communities or those happening elsewhere, that resonate with their audiences and their regional identities (Gutsche and Shumow 2019, p. 448). Local news sources can play an instrumental role in covering environmental issues and fostering a sense of collective identity grounded in geographical or environmental issues (Pinto and Vigón 2018).
News media are considered to have the ability to disseminate messages to all citizens. They not only announce issues and events but also diffuse the values of the culture in which they are produced. Therefore, news media are one of the social institutions that construct and influence the collective identity of a group (Major 2017). By covering issues such as climate change, news media can symbolically produce and reproduce ideas of the world and the audience's place in it; therefore, its role in identity formation is essential (Olausson 2009). Media content that incorporates identity-relevant information may be used as a first step toward the development of a collective identity, and a minimum for the establishment of collective identity is the expression of groupness, which is carried out in news stories that suggest local readers live in the CBW. We see the media as enabling a representation, reflection, or even causation of collective identity. Breakwell (1986) proposed a three-step cyclical process of identity development. First, individuals are exposed to identity-relevant information. Next, they evaluate that information against information previously stored in memory. Finally, if the information is deemed acceptable, it is integrated into the individual's self-knowledge structure. The process repeats over time and with each (successful) cycle, collective identity becomes stronger and more nuanced. The notion of information exposure is vague, but people are often exposed to information via the media, and the content and frequency of exposure to information from the media can help to explain collective identity. The identity theory notion of boundary work is helpful in addressing the content of the information about the CBW to which people are exposed. As the phrase implies, the focus of boundary work is that of establishing the perimeter of the group. It involves defining who "we are" and who "we are not" (Fominaya 2010, p. 395). In many instances, boundary work is a complex process of negotiating who qualifies for group membership based on their values, history, or willingness to engage with the group's purpose (Ashforth et al. 2000). In the case of location-based identities, however, that process is simple. Geophysical boundaries based on the drainage behavior of water define where individuals resideinside versus outside of the CBW. In this context, boundary work is not metaphorical, but literal. Attribution of group membership is as straightforward as the assertion "our town is within the CBW." A second issue concerns frequency of exposure. The accessibility principle of memory holds that concepts come to mind more readily to the extent that they are practiced (Higgins 1996), and the same principle should hold for collective identity. The more frequently a person is reminded of the fact that they are a resident of the CBW, the more quickly and effortlessly that concept will occupy consciousness. In Shrum's (2008) view, this is precisely how media operate to create shared reality. That is, the more often a concept is activated, the more accessible it becomes.
In this way, local news organizations can play a prominent role in making collective identity more salient, which may, in turn, foster a sense of community, geographic identity, and social cohesion (Abernathy 2018). We note, however, that media coverage of identity-relevant issues is often reactive. Events drive coverage and events often change over time. The events of interest in one location are also often distinct from those of local relevance in another locale. Taken together, these issues prompted us to ask: RQ1a: To what extent are local newspapers providing coverage of the CB and CBW?
RQ1b: Does coverage vary by place or time?

Identity Elaboration and the Media
An essential part of building any identity involves elaborating its meaning (Breakwell 1986). For example, becoming a farmer requires more than the definition of farmers as people who manage land for the purpose of raising livestock or crops. The farmer identity can encompass a plethora of subordinate concepts including tractors, feed, fencing, crops, crop rotation, planting, harvesting, storage, purchasing, pricing, sales, and more. Naturally, some of these concepts will be more relevant to the identity of a farmer as a function of place or time. This is so because locations vary hugely with respect to the type of crops and livestock that are suited to the environment. A flood might make salient particular aspects of the farmer identity, which would be backgrounded in times of drought.
Likewise, the CBW is a complex, multifaceted, and dynamic area. We might, therefore, expect that a collective CBW identity-if there is evidence for such a thing in media coverage-would be highly variable. Coverage of events and issues that could define a collective identity could be responsive to several factors. Because David and Bar-Tal (2009) emphasized the historical/temporal and location-based aspects of collective identity, and these factors map on to measurable differences in newspaper coverage, we investigate how the nature of a CBW identity, as seen in newspaper coverage, was dependent upon location and time. As a general theoretical matter, we asked about the content and structure of the CBW identity.
RQ2a: What are the components and structure of the CBW identity?
RQ2b: Do these features vary by place or time?

Sample of Locations and News Outlets
Because it was not possible for us to conduct a census of news coverage of the CBW, we chose local newspapers from three municipalities on the Susquehanna River, the primary tributary of the Bay. The rationale for examining media content from these three regions includes their geographic and demographic differences, as well as some similarities. The three newspapers from this region are considered small market newspapers, daily or weekly publications with less than 50,000 in circulation (Ali et al. 2018;. They are all located within the CBW, but importantly, at different points (See Fig. 1). Cooperstown, NY sits at the headwaters of the North Branch of the Susquehanna River and is the farthest upstream to the mouth of the CBW. It is the home of the Cooperstown Crier, a weekly newspaper with a circulation of 3000. Selinsgrove, PA is located roughly halfway downstream between Cooperstown and the mouth of the CBW, along the West Branch of the Susquehanna. Its local paper, The Daily Item, has a Monday through Saturday circulation of 15,174. Finally, Havre de Grace, MD is on the Bay at the mouth of the river. The Aegis publishes on Wednesday and Friday with circulations of 14,045 to 14,144, respectively. Table 1 provides descriptive information about each of the three locales. It also highlights that the communities under study are distinct in a number of ways in addition to proximity to the Bay.

Data
We utilized the archiving service NewsBank to access articles that were published in local papers at the three sites under study. Because the earliest date of available articles varied across papers, we utilized the eight years from January 2012-May 2020. This was the longest period for which data were available for all three outlets. To emphasize the range of writing that might convey identity-relevant information, the articles themselves could be of any type, including hard or soft news coverage, editorials, letters to the editor, other opinion pieces, or death notices.

Selection of Keywords
We treated words as empirical manifestations of concepts. Put differently, we assumed a reasonably close correspondence between the utterance of the word watershed and the activation of that concept in the minds of the message recipients. But, we underscore that the unit of analysis in our research was the news article, not readers of the articles. Our goal was to assemble a representative set of words/ concepts such that we could count, then study their frequency of occurrence and co-occurrence. To gain insight into this sampling problem, all authors worked as a group to search the internet to examine verbal representations of the Bay and the Watershed, and what words were employed. For example, the Chesapeake Bay Foundation includes a tab on its website labeled The Issues (https://www.cbf.org/). The seventeen subheadings range from Agriculture to Environmental Justice to Runoff Pollution to What We Have to Lose. The search process continued until we reached saturation, that is, the point at whichin our judgmentcontinued search yielded no new information (Fusch and Ness 2015). We found the compact scheme used by the U.S. Environmental Protection Agency to be particularly useful for our purposes. It asserts that a healthy watershed benefits local communities in three ways: environmental, economic, and recreational (Environmental Protection Agency, n.d.). To represent the three ideas more concretely, we chose 13 terms (keywords) from our search for representations of the Bay: agriculture, bird, crab, fishing, flooding, hiking, nitrogen, phosphorus, photography, pollution, sediment, swimming, water quality. The terms were selected with an eye toward their overall representativeness of Bay-and Watershed-relevant concepts, not as pure indicators of a single overarching category. For instance, agriculture has implications for environmental and economic concerns, but probably not for recreation. In contrast, fishing spans all three superordinate concepts. Our group judgment was that the set of terms jointly offered a reasonable sampling of the concepts that are used to represent the Bay. Given the centrality of watershed to our inquiry, we also added that term to our list of keywords, which produced a total of 14 terms.

Computerized Language Analysis
In August of 2020, we performed searches in Newsbank with the combination of Chesapeake and each of the keywords to create a dataset of news articles. This produced 544 hits. Some articles, however, were counted twice or more because they included more than one keyword. After elimination of duplicates, the database contained 312 articles.
To better understand the content and structure of these stories, we examined the frequency of occurrence of each keyword and the co-occurrence among word pairs. It was typical for the co-occurrence of word pairs to be unequal. For example, in a given article agriculture might appear five times when bird appeared only twice. In such cases, we recorded the higher of the two values as the cooccurrence score.
To generate these data, we employed the corpus analysis tool AntConc (Anthony 2005). AntConc is a corpus analysis tool that is widely used in media, journalism, and communication studies (Bednarek and Carr 2021). Previous studies have used it to analyze newspaper articles and social media posts (Fuster-Márquez and Gregori-Signes 2018; Walsh 2020). Because AntConc only supports the concordance search of one word or word group, we cross-searched keywords in different folders, where a folder contained all of the articles in which the keyword appeared. For example, we searched agriculture in the remaining 13 keyword folders to determine how often this term co-occurred with others. To look at the frequencies of keywords in each pair of concurrences, we used the concordance plot search function of AntConc to calculate how many times every keyword was mentioned in each location. Multidimensional Scaling (MDS) To explore the structure among the keywords and the focal concept (i.e., watershed), we used multidimensional scaling. This technique is a means of visually representing a set of concepts with regard to their similarity with one another. Technically, it is a mathematical method for displaying a matrix of proximities in N-dimensional space. For MDS users, the first problem to resolve is the number of dimensions (N) that best captures variation among the concepts. Two statistics aid in this decision. First, stress values of .00 are perfect, whereas values of .20 are considered poor according to the Kruskal and Wish (1978) rule of thumb. But, regardless of sample size, stress values decrease as dimensionality increases. Second, the R 2 for each N-dimensional solution represents the proportion of variance accounted for by that solution. Larger values are preferred and higher dimension solutions yield greater R 2 values. Because both stress and R 2 are influenced by the size of the matrix and the distribution of data within it, the overriding criteria for choosing among the solutions are subjective: parsimony balanced against interpretability.
The ALSCAL routine in SPSS version 25 was used to conduct a nonmetric multidimensional scaling of the word count data. Because location was of interest, the first step to creating input matrices was to partition the data by town. Next, we created a within-story co-occurrence matrix for each location. This involved identifying the number of times each of our 14 keywords appeared within each story. The result was a similarity (i.e., co-occurrence) matrix for each location. Because ALSCAL requires dissimilarity matrices as input, we first inserted values of 0 in the diagonal to indicate that each keyword was identical with itself (i.e., not at all dissimilar). Then, off-diagonal entries were computed by subtracting each element in the co-occurrence matrix from the highest frequency value in that matrix plus 1. As a result, all off-diagonal elements were greater than 0, and higher values represented higher levels of dissimilarity.

Coverage of the Chesapeake Bay
RQ1a asked to what extent local newspapers provide coverage of the Chesapeake Bay. The data suggest that coverage is quite limited. Our search returned a total of 312 articles over an 8-year period. This number may, at first, appear substantial. But it is an average of 39 articles per year in total, or 13 per year per paper. Given a total number of stories equal to 13,809 for Cooperstown, 108,979 for Selinsgrove, and 13,444 for Havre de Grace, the overall proportion of stories focused on the Bay was tiny (312/136,232 = )0.229%.
RQ1b focused on whether coverage varied by place and time. Figure 2 gives an affirmative answer to both. Over the 8-year period under study, The Cooperstown Crier published 18 articles out of a total of 13,809. This can be compared to 110 for Selinsgrove's Daily Item out of 108,979, and 184 for the Havre de Grace paper, The Aegis out of 13,444. Thus, total coverage (i.e., raw frequency) is inversely related to distance from the Bay. The relative frequencies, that is 1.36%, 0.100%, and 0.130%, showed proportionally greater coverage in Havre de Grace with Selinsgrove and Cooperstown being roughly on par with one another. Figure 2 also makes it clear that coverage varies substantially over time. Variability is greatest in Havre de Grace and somewhat more modest, but still notable in Selinsgrove. There is far less variability in the Cooperstown data, but the variance is constrained by the low frequency of articles. In an attempt to understand what might account for the variations over time, members of the research team grouped the articles by location. This effort did not reveal any event or set of events that explained the peaks and valleys in Fig. 2. However, the patterns in Fig. 2 are sufficient to answer RQ1b: Frequency of coverage of the Chesapeake Bay did vary as a function of place and time. In fact, variability in the frequency of coverage appears to increase as locations are closer to the Bay.

Content of CBW Coverage
RQ2a inquired as to the content of the CBW identity. Table 2, which presents the frequency of keywords in total and broken by municipality, provides an answer. It is important to bear in mind that all of the articles in this table were screened such that they contained one or more mentions of watershed. After looking first at the column labeled Row Total, it can be seen that watershed is the most frequently mentioned keyword (92 times) when the data are collapsed across location. This is followed by pollution (73) and water quality (66). Fishing, agriculture, and flooding come next with 51, 49, and 37 occurrences respectively. Thus, these five terms comprise the main contents of the watershed identity, at least as it is conveyed in local newspapers across municipalities.
Turning attention to the frequencies by municipality, it appears that there are large variations, but they are difficult to judge in the frequency data. The proportions, however, allow for direct comparison and produce a surprising finding. After adjusting for base rate, frequency of occurrence of the keywords is remarkably similar across locations. That is, the proportions in each row do not differ much from one another. This suggests that the overall content of the news articles in the three locales is homogeneous. There are two notable exceptions. Water quality is higher in Selinsgrove than either Cooperstown or Havre de Grace. This may be the result of frequent flooding in Selinsgrove, where onequarter of the borough's total land area is within FEMA designated 100-or 500-year floodplains (see also The Selinsgrove Historical Association). Second, watershed is mentioned comparatively more often in Cooperstown than in Selinsgrove, where it is mentioned more often than in Havre de Grace. This may follow from Cooperstown being located at the headwaters of the Susquehanna River, or it might reflect the differing local knowledge or identity about being located within the watershed.

Structure of CBW Coverage
Although the frequency counts for keywords provided valuable information about media coverage, they do reveal how often the keywords occurred relative to one another or how those co-occurrences were patterned. Consequently, we conducted MDS analyses having two, three, or four dimensions for each location 1 In each case, we determined that the two-dimensional solution was most useful for our purposes. The stress and R 2 values were 0.06 and 0.96, respectively, for Cooperstown, 0.19 and 0.77 for Selinsgrove and 0.20 and 0.73 for Havre de Grace. Although the stress values for Selinsgrove and Havre de Grace were poor by Kruskal and Wish's (1978) standard, we considered them adequate given the small number of objects in the matrix and the relatively modest number of articles in each location. The two-dimensional solutions were strongly favored in terms of ease of interpretation and the ability to compare results across locations. Thus, we retained the twodimensional solutions.  Given that the overall results were not meaningful, a pooled analysis hides true and meaningful variation across locations. Figure 3 displays the results for Cooperstown. To highlight the findings relevant to a watershed collective identity, we have drawn an oval around the focal concept watershed and its closest keywords. From this grouping of keywords, we can conclude that most articles on the CBW discuss water quality, pollution, and nitrogen. If we were to slightly expand the size of the oval, it would include agriculture and phosphorus. Hence, in Cooperstown, media discourse about the CBW is focused mainly on degraded water quality and the chemical pollutants that cause it.
It is informative to note the presence of another theme in The Crier's coverage, one which is defined by the cluster of terms at the far right of the figure. Their proximity to one another suggests a distinct set of articles that emphasize crab, swimming, and photography as well as sediment and flooding. A similar, though smaller cluster of keywords concerns fishing and hiking. Although these concepts relate to the Bay and are perhaps used in news stories about recreation near Cooperstown, they are less likely to co-occur with watershed and, therefore, less likely to contribute to the collective identity that was the focus of our study.
In Selinsgrove (Fig. 4), watershed is tightly clustered with hiking, pollution, photography, water quality, and sediment. This is an indication that coverage of the CBW consists of articles that consistently intermingle recreational concepts (hiking and photography) with water quality concepts (pollution and sediment). Curiously, other Bay-related recreational activities (i.e., swimming and fishing) are quite distant from watershed.
A second theme is defined by the cluster at the top of the figure, which contains crab, bird, and swimming. Another relevant theme on the left side of the figure documents the co-occurrence of articles that emphasize agriculture and nitrogen, perhaps based on nitrogen's use as a fertilizer. Selinsgrove is surrounded by a substantial amount of agricultural land, and research documents the environmental degradation associated with the use of fertilizers (Shaviv and Mikkelsen 1993;Zhang et al. 2015). As with Cooperstown, we see a set of articles that deals with Bay-related concepts, but do so without reference to the watershed. Figure 5 shows that the Havre de Grace data present clusters that differ markedly from those in the previous two analyses. The terms crab, sediment, and flooding cluster tightly around watershed, thereby defining this set of terms as central to a collective watershed identity. A second cluster, which appears at the bottom of the figure, is composed of nitrogen, phosphorus, fishing, and photography. These terms appear in articles that are less likely to include the term watershed.

Discussion
The motivation for this project evolved from a simple set of premises. Despite efforts to improve the health of the Chesapeake Bay, it remains in precarious condition. This is largely the product of human behavior that occurs not in the Bay, but throughout the vast watershed that feeds it, particularly from agricultural operations upstream. Hence, any possible solution to this degradation is likely to depend on residents' understanding that their collective actions will determine the fate of the Bay. Media, especially local media, are more than a common source of information about a community. They often are critical providers of information about local issues that would otherwise not be found elsewhere, as well as institutions with resonance throughout a community in terms of community building and democratic function. The presence or absence of local newspapers can work to build or diminish a sense of community, including fostering a sense of cohesion, inclusion, or common senses of community values and norms (Ali et al. 2018;Lowrey et al. 2008). Local news has been linked to the degree and quality of local civic participation; conversely, when local newspapers shutter, levels of political engagement have been found to decrease (Hayes and Lawless 2018). Or, as others have found, they can function as a form of social control, excluding particular social groups from inclusion in community spaces and stories (Gutsche 2015).
We sought to understand the extent to which local newspapers discussed key terms relevant to the CBW and therefore may or may not be contributing to the construction of a collective watershed identity. Our research revealed that the general answer was: "Not much." But, there is nuance to this conclusion that deserves closer attention. We turn to those issues next.

Coverage of the Chesapeake Bay
A place-based identity is established when individuals experience a sense of connection with some location (Kavaratzis and Hatch 2013). These locations might include a country, a park, a family home, or in our case, a watershed. However, identity is necessarily premised on awareness of the place as a place. In other words, knowledge of the location and its boundaries are key to the development of an identity.
Given the centrality of local news to a community's sense of itself, we looked first at media coverage of the Chesapeake Bay. We found clear evidence of Bay-related news stories, though by almost any standard, the frequency of that coverage was sparse. Across location, fewer than .50% of the articles published from 2012-2020 in local newspapers mentioned the Bay. Thus, while it was possible that media coverage encourages awareness of the Bay, that possibility was dishearteningly small. Nonetheless, there were interesting variations by time and place.
In an explicit reference to temporal dynamics, Proshansky et al. (1983) asserted that "…place-identity will be modified over the course of the individual's life cycle" (p. 60). By plotting frequency of coverage over the time period of our investigation, we observed substantial year-toyear variation. But, because our efforts to identify exogenous predictors of this variation were unsuccessful, we leave that task to future research.
With respect to location, the results were clearer. Proximity to the Bay emerged as a strong predictor of coverage with Havre de Grace showing roughly twice the coverage of Selinsgrove, and Selinsgrove showing five times the coverage of Cooperstown. To the extent that "… identities are talked into existence" (Hunt and Bedford 2004, p. 445), the quantity of that talk appears to be tied to location. This inference is, however, qualified by the presence of many differences between the three municipalities (Table 1), any or all of which could contribute to variations in the frequency of coverage and increases as distance from the Bay decreases. The prevailing logic of our study also suggests that collective identity motivates collective action to benefit the Bay. It should also be noted that not everyone within the CBW has the same interests and motives. Conflicting tensions among residents of the CBW might hinder their participation in actions that benefit the Bay, and future research can determine whether and what types of collective identity generate different actions. Media coverage and responses to it are likely contextualized by the community it reaches, and future research can examine how different forms of media coverage shapes (dis)identification with the CBW in different communities.

Components and Structure of a CBW Identity
Identity theory also asserts that collective identities are composed of structured sets of concepts that jointly give meaning to the identity. Thus, our second set of research questions focused more tightly on the components and structure of a Chesapeake Bay Watershed identity. Following a review of Bay-related websites, we selected a sample of keywords to represent the universe of possible related concepts. After limiting the database to articles that contained watershed, we observed that, across locations, pollution, water quality, fishing, agriculture, and flooding were the most frequently mentioned terms. In our view, linking pollution and water quality to watershed is beneficial insofar as it reminds readers of the problems the Bay is facing. Importantly, it also frames the problem in collective terms. Use of the term watershed references the entire 64,000 square miles that comprise the CBW, not merely the residents of the Bay. Similarly, the frequent occurrence of fishing, agriculture, and flooding apply to the watershed as a whole. As with coverage of the Bay, we suggest these findings might be seen as both a glass half full and a glass half empty. We found evidence of the potential for media coverage to encourage a sense of connection with the watershed, but the frequency with which these terms co-occurred was quite low. Moreover, even if coverage does exist, it is unclear whether people know that they live within the bounds of the CBW. The answer to this question would require research oriented toward individuals that would complement the messagefocused work reported in this paper. For instance, a cultural consensus analysis can investigate agreement and disagreement that exists among individuals living within the CBW (Paolisso 2008), offering insight into the components and structure of a CBW identity at an individual level. Such research could establish baseline knowledge of the CBW and consider what messages, when communicated socially or transmitted via the media, might enhance knowledge and a collective CBW identity.
Also, as with coverage of the Bay, we looked for variations in Watershed coverage due to time and place. In this case, there was insufficient data to carry out MDS analyses partitioned on both variables. Hence, we broke the dataset by place alone and conducted three separate analyses. These proved illuminating in two respects. First, we uncovered interesting differences in the components of a watershed identity as a function of place. For example, pollution and water quality were part of the focal cluster with watershed in Cooperstown and Selinsgrove, but not in Havre de Grace. Selinsgrove also encompassed hiking, photography, and sediment, whereas Havre de Grace was composed of only sediment, flooding, and crab. In short, the collective watershed identities that emerged from media coverage in different locations overlapped somewhat, but were unique to each place. Our impression was that issue coverage became more serious (vs. recreational) as proximity to the Bay increased.
The possible causes of these variations are numerous. Table 1, for example, shows differences across locations on a variety of sociodemographic indices. There is also reason to suspect that local environmental conditions can partially explain the manifestation of keyword groups. The "grades" issued by the University of Maryland Center for Environmental Science (n.d.) on a 100-point scale are relevant to this point (https://ecoreportcard.org/report-cards/chesapea ke-bay/bay-health/). For instance, the North Branch of the Susquehanna, where Cooperstown is located, scores much better than the West Branch of the Susquehanna (Selinsgrove) or the mouth (Havre de Grace) on nitrogen (North = 87 vs. South = 37) and on phosphorus (North = 69 vs. South 42). It seems plausible that environmental degradation receives more media coverage in the areas where it is most severe. However, as the MDS analyses show, these particular problemsnitrogen and phosphorus are not covered in conjunction with the watershed in the lower Susquehanna municipalities. Regrettably then, they cannot contribute to a collective watershed identity in Selinsgrove and Havre de Grace despite the fact that these pollutants enter the waterway primarily from upstream sources (which might include Cooperstown).

Collective Identity and Place Branding
Although our analysis of media coverage cannot rule out the possibility that a collective watershed identity is alive and well in other communication channels, that possibility seems unlikely. For that reason, we should consider more proactive means by which a watershed identity might come into existence. The literature on place branding, that is, the conscious effort to promote awareness of and attachment to a place, provides some helpful suggestions (cf., Kavaratzis and Hatch 2013). A recurring theme in this body of inquiry is an emphasis on human interaction. For example, work on sense of place among Chatham Islanders (a part of New Zealand) finds four components to the brand: Rights, roles, responsibilities, and relationshipseach of which highlights one aspect of human connectedness (Aitken and Campelo 2011). Similarly, Insch and Florek (2010) underscore the notion that social interaction is the binding agent between people and places. So, in addition to making CBW residents aware of their shared geographic membershipa precondition of collective identityhow might social bonds be developed?
Research on the contact hypothesis provides several answers (Harwood et al. 2016). For one, the contact hypothesis itself suggests that communication between group members leads to enhanced cooperation. This effect is seen most powerfully when certain conditions are met: participants are of equal status, the situation calls for cooperation, development of close relationships is possible, and institutional support is present (Allport 1954). These specifics aside, there is compelling evidence that contact alone can create a superordinate identity (Harwood et al. 2016), which is precisely what is needed to encourage collective action in the CBW. The contact hypothesis has been supported using several different channels of computermediated communication and in discussions about fairly severe or contentious issues (Cao and Lin 2017;Walther et al. 2015). These findings provide hope that modern communication channels can help provide a means of connection through which careful and focused communication can establish a common, superordinate identity.
Our study also intersects with research on ecocultural identification within the CBW. Ecocultural identification is a process of delineating relations among human and non-human beings (Trombley 2020). Scientific modeling, such as multidimensional scaling, attempts to integrate data on human and non-human beings to develop ecocultural identifications. Developing models and establishing identifications makes it possible to manage the CBW by identifying agents affecting the landscape and water quality. The present study identifies news media, specifically newspapers, as agents capable of creating or disseminating a collective identity and promoting collective action within the CBW ecocultural identification. Future attempts to identify how a CBW identity manifests collectively or individually serves to further refine ecocultural identification within the CBW. Moreover, manipulating messaging to modify a CBW collective identity or promote social interaction in service of collective action is a mobilization of an ecocultural identification.
Efforts to understand the development and activation of a CBW identity would benefit from expanding the spatial and temporal dimensions considered. Trombley (2018) argued perceptions, behaviors, and management efforts all exist within historical contexts of trauma and violence (e.g., colonization, capitalism, ecological exploitation, etc.), which demand recuperation rather than restoration. The key difference being recuperation requires confrontation with past trauma and violence that allowed degradation to occur and be perpetuated in the first place. The broad scope of news media leaves hope for the possibility that these traumas can be addressed in collective discourse, perhaps prompted by newspaper coverage.
Continued research into this topic can extend our exploratory results in several ways. We observed the prevalence of content related to the CBW in three local newspapers along with how the structure of a CBW identity might vary across those locations. The collective identity that is transmitted via newspapers is likely only one sort of collective identity. Future research can work to more fully understand the nature of that collective identity along with what collective actions in support of the Bay it motivates. Then, scholars can consider whether different forms of communication, including social or interpersonal interactions, produce similar or different forms of collective identity and the extent to which those forms of identity compel people to enact different forms of collective action. Ultimately, researchers can come to an understanding of what forms of collective identity promote the actions that best aid the health of the Chesapeake Bay along with how to establish those identities.

Conclusion
There appear to be meaningful divergences in the collective identities of the three towns that are reflective of their differing demographic and environmental characteristics. Proximity to the Bay may explain some of these differences, but as a whole the CBW is a complex heterogeneous landscape of fragmented people and places. The news media can function to influence the public and policy agendas and are at the same time influenced by them. Media accounts that include language that creates awareness of issues, challenges, policies, or opportunities for communities within the CBW can serve to provide insight toward how communities might perceive themselves, or be unaware, of their geographic location within a watershed. As the social scientist Bernard Cohen (1960 p. 165) wrote: "[The press] may not be successful much of the time in telling people what to think, but it is stunningly successful in telling most people what to think about".
Our results indicated that the potential for formation of a collective watershed identity has been unevenly aided by local media coverage and only to a small degree in an absolute sense. However, our data, and therefore our ability to generalize, were limited along at least fourdimensions. First, our sample of media messages was based on only 8 years of data. Because frequencies were low, our ability to discern patterns over time was inhibited. Second, we sampled from only three newspapers. Although they were chosen to capture geographic variation across the watershed, a larger number of locations would have been desirable. It should also be noted that both the municipalities included in this study and their associated newspapers differed in several meaningful ways (see Table 1). In part because of these differences, we do not report significant differences between locales. Rather, we attempt to explain why these locations might experience differences in collective identity related to the CBW in their newspapers. Third, we used "watershed" as the focal keyword in our analyses because it is the most common way the area feeding into the Chesapeake Bay is described, though some people refer to that area using terms like "water catchment" or "river basin." Finally, we gathered data only from local newspapers. While these traditional media have the potential to influence collective identity, a more comprehensive understanding of watershed identity would require data drawn from social media and face-to-face interpersonal interactions to contextualize the ways in which media coverage reflects or influences the perceptions of local citizenry, as well as a more granular examination into how that may lead to systemic change. We hope that future research is able to address these qualifications.
In sum, to the extent that the well-being of the Chesapeake Bay hinges on a collective watershed identity, our results suggest that it is unlikely to develop organically from media coverage. The creation of a watershed identity will need to be planful and proactive.