Breaking down “harassment” to characterize trends in human interaction cases in Maine's pinnipeds

Recent marine mammal recovery and growth of human populations in coastal areas has led to increased human harassment of protected pinniped populations. Yet, current monitoring approaches lack the necessary resolution to describe this emerging issue. We therefore propose a new classification scheme for monitoring of marine mammal–human interaction (HI), which defines categories based on the type of human behavior and associated risks to the animal. Among harbor, harp, and gray seal strandings (N = 3,525) from 2007 to 2019 in Maine, United States, evidence of HI was reported in 14.72% of strandings, with the majority (75.34%) of these cases involving harassment. The number of HI reports increased over time, particularly in the southern region of the state, and an analysis of geographic hotspots further identifies emerging areas of concern. Variation in the rate and type of HI among species, age classes, and seasons reflect differences in pinniped life history. The most commonly observed types of harassment included human approach, physical contact, and displacement. By breaking down harassment, we aim to inform ongoing mitigation efforts and contribute to our understanding of the impact of HI on marine mammal health, as well as offer a transferable methodological approach to monitoring human–wildlife interaction.


| INTRODUCTION
From a history of human exploitation and population decline, today many marine mammal populations in the United States (US) are recovering as a result of federal protections (Roman et al., 2013). These positive trends are particularly evident for pinniped species, including seals and sea lions (Cammen, Rasher, & Steneck, 2019). As both marine mammal and human populations grow in coastal areas, opportunities for marine mammal-human interactions (HI) increase as well. Such HI can take many forms and enact diverse impacts on marine mammal health, from individual to population level.
Marine mammal interactions with fisheries and vessels can result in severe injury or mortality (Carretta et al., 2020;Delean et al., 2020;Knowlton & Kraus, 2001), as well as chronic negative impacts on health (Rolland et al., 2012;Rolland et al., 2017). Impacts of human harassment of marine mammals are varied and less well documented, but can include both physiological and behavioral impacts and can range from innocuous to significant harm. Human approach can cause physiological stress, even in animals that do not appear visually disturbed. People poking or throwing sticks at resting animals can cause physical injuries (Boren, Gemmell, & Barton, 2002). Harassment that causes increased attentiveness or short-term displacement (e.g., flushing of hauled out pinnipeds) can also disrupt critical resting (Tyne, Christiansen, Heenehan, Johnston, & Bejder, 2018) and maternal care periods (Boren et al., 2002;Kovacs & Innes, 1990). Finally, chronic human disturbance that leads to long-term displacement from suitable habitat can have significant demographic consequences resulting in localized population declines (Gerrodette & Gilmartin, 1990;Jansen, Boveng, Ver Hoef, Dahle, & Bengston, 2015;Stevens & Boness, 2003).
To minimize such impacts to marine mammal health, many forms of HI are prohibited both in the United States and elsewhere worldwide. The US Marine Mammal Protection Act (MMPA) prohibits "takes" of marine mammals, which are defined as harassing, hunting, capturing, killing, or attempting any of these actions. Furthermore, guidelines for responsibly viewing marine mammals in the United States recommend remaining at least 50 yards away. Despite these prohibitions and guidance, HI continues to pose a risk to marine mammal species in the United States.
Much of our current understanding of the prevalence of marine mammal HI in the United States comes from a national Marine Mammal Health and Stranding Response Program (MMHSRP) that is tasked with responding to marine mammals in distress and monitoring marine mammal population health (Becker, Wilkinson, & Lillestolen, 1994). Toward these aims, authorized member organizations receive and respond to reports of stranded marine mammals, including live and dead animals found on a beach or floating in US waters. Live animals are considered stranded when they require medical attention or are unable to return to the water or their natural habitat without assistance (Wilkinson, 1991). As part of their response, organizations are mandated to collect basic information about stranded marine mammals, including species, sex, age class, stranding location, and evidence of HI. Several prior studies have described spatiotemporal trends in these datasets in an attempt to understand current threats to marine mammal health (Esquible & Atkinson, 2019;Greig, Gulland, & Kreuder, 2005;Olson, Aschoff, Goble, Larson, & Gaydos, 2020;Warlick et al., 2018).
The types of HI currently documented among stranded marine mammals in the United States include vessel interaction, gunshot, fishery interaction, and other human interaction, a broad category that encompasses harassment, mutilation, ingestion of debris and several other harmful activities. In some cases, HI is detected upon physical examination of a stranded marine mammal (Hare & Mead, 1987). Vessel interactions are typified by lacerations and blunt force trauma. Fisheries interactions are characterized by the presence of fishing gear, scarring patterns, or open wounds that result from entanglement. Evidence of harassment, however, is more challenging to identify and requires observation of human behaviors. Harassment is defined by the MMPA (2007) as "any act of pursuit, torment or annoyance which has the potential to injure… or disturb a marine mammal or marine mammal stock in the wild by causing disruption of behavioral patterns, including but not limited to, migration, breathing, nursing, breeding, feeding, or sheltering." In regions of the United States with significant spatial overlap between human activity and local pinniped species, concern over the impact of HI, and particularly harassment, is growing. We argue that current data collection protocols, which categorize "harassment" broadly without further parsing, do not offer the resolution necessary to document the potential impact of these HI cases on pinniped population health. Furthermore, the current HI classification scheme makes it difficult for stranding organizations to track the types of harassment that are occurring in their region and hinders ongoing mitigation efforts.
To address gaps in our current understanding of marine mammal HI cases, we conducted a spatiotemporal analysis of harbor (Phoca vitulina), harp (Pagophilus groenlandicus), and gray (Halichoerus grypus) seal stranding records from 2007 to 2019 in the state of Maine in the United States. As part of this analysis, we propose a new classification scheme for HI that categorizes harassment based on the type of human behavior and the associated risks. This new scheme, and our analysis, will inform local stranding networks' efforts to mitigate HI and contribute to our understanding of the potential impact of HI on marine mammal health in a region experiencing growth in coastal human and pinniped populations. Ultimately, the methodological approach we describe here will also be transferable to regions elsewhere that are experiencing parallel growth in coastal development, populations of protected species, and humanwildlife interactions.

| Study system and species
The state of Maine is located in the Northeast United States abutting the Gulf of Maine with over 3,000 miles of coastline. Coastal human population density and public beach access are highest in the southern half of the state, characterized by sandy beaches. The northeastern half of the state is relatively sparsely populated, rural, and composed of largely natural resource-dependent (including fisheries) communities that abut rocky shoreline. The US Census Bureau reports increasing human populations in Maine over the past decade, particularly in coastal counties (U.S. Census Bureau, 2020), where tourism is also increasing (Maine Office of Tourism, 2020).
Three pinniped species are commonly observed and reported as stranded along the coast of Maine: harbor, gray, and harp seals. Both harbor and gray seals are found yearround in the Gulf of Maine, while harp seals are seasonal visitors (Katona, Rough, & Richardson, 1993). Harbor seal abundance peaks in Maine from spring through autumn, with pupping occurring on coastal islands and rocky shorelines from April to May. Surveys of harbor seal populations in Maine documented consistent growth during the 1980s and 1990s (Gilbert, Waring, Wynne, & Guldager, 2005), but gaps in survey effort and changes in protocols since then have led to uncertainty in more recent population trajectories (Waring, DiGiovanni Jr, Josephson, Wood, & Gilbert, 2015). Surveys of gray seal abundance in Maine clearly document increasing population trends since the mid-1990s, a result of both regional population growth and immigration (Wood, Murray, Josephson, & Gilbert, 2020). Harp seal sightings in the Gulf of Maine have also consistently increased in recent decades, likely as a result of population growth and changing oceanographic conditions (Stevick & Fernald, 1998). These seals, which breed and typically forage in Canadian waters, regularly visit the Gulf of Maine, particularly in the winter months.

| Stranding data
Records of harbor, harp, and gray seals that stranded in Maine from 2007 to 2019 were provided by three authorized stranding organizations active during this period: Maine Department of Marine Resources (MEDMR) and Marine Mammals of Maine (MMoME), which responded to stranded marine mammals in southern and mid-coast Maine from 2007 to 2011 and 2012 to 2019, respectively; and Allied Whale at the College of the Atlantic, which responded in northeastern Maine throughout the study period. From these records, we excluded any unconfirmed reports (i.e., reports from a member of the public that lack significant detail and were not confirmed by follow-up interview or corroborated by subsequent reports) and reports of animals decomposed to the point of mummification.
Data including species, sex, age class, location of stranding, and findings of HI were retrieved from the MMHSRP's National Stranding Database. These data are recorded from every reported and documented stranded animal using a standardized Marine Mammal Stranding Report Level A Data form (NOAA Form 89-864). On this form, findings of HI are indicated as "Yes," "No," or "Could Not Be Determined (CBD)." "Yes" indicates evidence of HI, but does not indicate whether or not HI is believed to be the cause of stranding or death. "No" indicates that the stranded animal was examined and there was no evidence of HI. "CBD" indicates that the assessment for HI was inconclusive or that the animal was not thoroughly examined.
The Level A form further categorizes HI cases as "vessel interaction," "shot," "fishery interaction," and "other human interaction" (in 2020, a fifth category was added: "ingestion of gear or debris"). "Other human interaction" cases are typically associated with a brief written description. These HI categories are briefly defined in Table 1, with more detail available in the Examiners Guide (National Marine Fisheries Service, 2020). We supplemented these HI data from the National Database with notes from the reporter and responders directly involved in each case, which were recorded on call logs maintained by the stranding organizations. These notes were used to recategorize HI cases based on our modified classification scheme described below.

| Developing a HI classification scheme
The development of a modified HI classification scheme for this research was motivated both by requests from stakeholders involved in marine mammal stranding response and coordination and by preliminary analyses of HI data in Maine that revealed a high skew toward "other human interaction." Stakeholders were engaged in all stages of this process, including initial design, funding acquisition, data collection and sharing, methodological development, results interpretation, and manuscript preparation (cf. "collaborative" as opposed to "cooperative" research) (Hartley & Robertson, 2007;National Research Council, 2004).
The MMHSRP had already begun the process of revising their HI categorization process (Moore & Barco, 2013), adding more detailed categories on a supplemental Marine Mammal Human Interaction Report (NOAA Form 89-864), which became required starting in 2020. This supplemental form defines the following categories of HI: vessel trauma (sharp, blunt, or both), entanglement (in gear or debris), hooking (in recreational or commercial gear), gunshot, ingestion (of gear or debris), mutilation, harassment, and other. Our aim was to design a modified HI classification scheme to further supplement the existing Level A and Human Interaction Report forms, dividing the "harassment" category into multiple, more detailed subcategories. Our goal in developing the harassment categorization system was to create a parsimonious suite of categories that balances pragmatic challenges associated with managing additional categories with the utility gained from greater specificity. Our decision-making process in defining new subcategories of harassment therefore considered the tradeoffs of defining too many overly specific categories and too few overly broad categories (i.e., splitting vs. lumping). We defined categories based on both shared aspects of the human behavior and shared potential impacts of the behavior on marine mammal health. To assess potential impacts, we relied on published literature and expert knowledge of stranding network participants (ST, RS, LJ, LD, DW), who have conducted stranding response in our study region for a cumulative total of over 100 years. We tested our modified classification scheme on subsets of the data records and revised it in an iterative process with feedback from marine mammal scientists and stranding network members. Feedback was gathered through one-on-one communication, at small research team meetings, and at regional scientific conferences.

| Data analysis
Using our new classification scheme, we re-assigned a HI type to all stranding records that indicated evidence of HI on the Level A form. All records were initially classified by one member of the research team (EN) and then independently verified by a second member (KC), using the definitions for HI types provided in Table 1. Records were re-classified using information included in the reporter and responder's comments appended to each stranding report. As in prior schemes, it was possible to assign multiple HI types to a single stranding record. When insufficient information was included in the comments (N = 10 cases), the case was classified based on its Level A categorization. Correspondence between Level A categories and our new classification scheme is shown in  However, we acknowledge that this approach for cases without further details may have resulted in some incorrect assignments, as the Other Human Interaction category can also include ingestion, mutilation, and other types of HI such as vehicular trauma. We used our new classification scheme to analyze spatial and temporal trends in the proportion of strandings which involved HI in Maine from 2007 to 2019. All statistical tests were implemented in R. For temporal analyses, stranding data were aggregated at an annual and seasonal scale, where seasons were defined as follows: winter = December, January, and February; spring = March, April, and May; summer = June, July, and August; and fall = September, October, and November. Seasonal averages were compared using a Pearson's chi-square contingency table test. Annual trends were described using linear regression models, and the model slopes were compared between regions and among species using the lstrends function in the lsmeans R package, which evaluates leastsquares means predicted from linear models (Lenth, 2016). The annual proportions of HI cases reported per species were compared using a Kruskal-Wallis test, after Shapiro-Wilk's tests for normality revealed a significant deviation from the normal distribution for the annual proportion of HI cases reported for harp seals. Finally, change in the number of HI types experienced per animal over time was evaluated using linear regression.
For spatial analyses, stranding data were aggregated by stranding response region and into grid cells measuring 0.1 Â 0.1 . The annual proportions of strandings with HI for each region were checked for normality using the Shapiro-Wilk's test, and then compared using a two-tailed student's t-test as no deviations from normality were detected. Hotspots of HI were identified using Openshaw's geographical analysis machine (GAM) scanning approach (Openshaw, Charlton, Wymer, & Craft, 1987), implemented in the DCluster R package with a 1 Â 1 km grid, 15 km radius circle, and alpha of .002 (Bivand, 2006). This approach identifies locales where HI cases are higher than expected based on the number of strandings reported in that locale and the state-wide frequency of HI cases. This hotspot analysis was conducted with stranding data from all years, and with stranding data from a subset of years so as to compare hotspots over time. For the latter analysis, the study period was subdivided into three approximately equal-length periods that represent low (2007-2011), medium (2012-2015), and high (2016-2019) proportions of HI cases reported.

| Spatiotemporal trends in pinniped-HI cases
Following the data filtering steps described above, there were a total of 3,525 confirmed strandings of harbor, harp, and gray seals in Maine from 2007 to 2019. Of these cases, 519 (14.72%) were reported to involve HI, and the number of HI reports increased over time throughout our study period (Table 2). We observed a significant difference in rates of HI among seasons across the study period (Pearson's Χ 2 = 102.65, df = 6, p < .0001), with the highest proportion of HI cases observed in the spring (Table 2 and Figure S1, Supporting Information).
The number of reported strandings and HI cases was higher in southern Maine than northeastern Maine (Figure 1a,b), although the two regions do not report a significantly different average annual proportion of strandings with HI (t = À0.26, p = .80; Table 2). The number of reported HI cases has increased significantly faster in southern and mid-coast Maine than northeastern Maine (t = À2.66, p < .05; Figures 1c and S2). This increase is particularly noticeable in southern and midcoast Maine after 2012, when MMoME began responding to stranded animals in this region.
Four geographic hotspots of HI, areas of higher than expected proportions of HI cases relative to the number of strandings and state-wide HI frequency, were identified when data from all years were aggregated (Figure 1b). Although the exact location and number of these hotspots vary over time throughout our study period ( Figure S3), they are consistently distributed throughout the coastline, in areas of relatively high and low stranding density. Thus, HI hotspots were not all colocated with areas that reported high numbers of HI cases, but rather higher than expected proportions of HI.
Harbor seals were the most frequently reported stranded pinniped, representing over 70% of the total strandings each year (Table 2). There was no significant difference among species in either the average annual proportion of HI cases reported per species (Kruskal-Wallis Χ 2 = 1.25, df = 2, p = .54) or the rates at which the frequency of HI cases increased over time (all pairwise tests p > .05) (Table 2 and Figure S4).

| Types of HI and a modified classification scheme
The most commonly observed type of HI was Other Human Interaction (79.00%), as categorized on the Marine Mammal Stranding Report Level A Data form, followed by an equal number of vessel (9.83%) and fishery (9.83%) interaction cases (Figure 2a). Few gunshot cases (1.73%) were observed in the dataset. When using the Marine Mammal Human Interaction Report form categorization scheme, the most commonly observed type of HI was harassment (75.34%), followed by entanglement (10.98%, includes entanglement in fisheries gear and marine debris) and vessel trauma (9.83%) (Figure 2b). All three top types of HI were most commonly reported in southern Maine (Figure 1d-f). Hotspots for entanglement were distributed from mid-coast to southern Maine, while hotspots of harassment and vessel trauma were identified in midcoast Maine and northeastern Maine.
To further describe the large number of harassment cases, we developed and implemented a new classification scheme of harassment subtypes (Table 1). Nine subtypes were defined: human approach, physical contact, displacement from land and water, unauthorized collection, covered, feeding, canine interaction, and other or unspecified harassment. It is important to note that multiple subtypes can be selected when multiple human behaviors are experienced by a stranded animal, and some of the harassment subtypes are inevitably linked. For example, human approach is required in all cases of physical contact, displacement from land or water, unauthorized collection, covered, and feeding. Physical contact is required in all cases of unauthorized collection.
Human approach was the most commonly observed type of harassment (N = 346; 88.49% of harassment cases) (Figure 2c). Of these cases, 57.80 and 54.05% also involved physical contact and displacement or unauthorized collection of the animal, respectively. Displacement occurred both with and without physical contact. There were some cases of animals being picked up or pushed by a member of the public into or out of the water. There were also cases of flushing, when a seal retreats from land to the water because of a disturbance. We observed a higher proportion of cases than expected (i.e., a hotspot) involving displacement from land in southern Maine (Figure 1i). Hotpots of human approach and physical contact overlapped with those for harassment cases identified in mid-coast and northeastern Maine (Figure 1g,h).
Up to five HI categories were associated with a single stranding (N = 12; 2.32% of HI cases) in our dataset, although almost half of HI cases experienced only one HI behavior (N = 232; 44.79% of HI cases) ( Table S1). The most common combinations of HI types were: human approach, physical contact and displacement from land (i.e., pushing animals back into the water) (9.85%); human approach and physical contact (e.g., petting, picking up) (8.30%); human approach, physical contact, and unauthorized collection (6.95%); and human approach and displacement from land (i.e., flushing) (6.18%). All cases that experienced five distinct HI types included human approach, physical contact, and either displacement from land or water and/or unauthorized collection; in addition, these cases often included covered or feeding. All of these cases were harbor seals that stranded between March and September; these cases included eight pups, one yearling, one subadult, and two animals of unknown age class. Three of these cases occurred in northeastern Maine.
The average number of HI types experienced per animal increased over the study period (r 2 = .61, p < .01), which likely reflects a shift in how harassment has been recorded over time with greater detail in the notes and photodocumentation in recent years that enable T A B L E 2 Average annual and seasonal proportion of strandings with evidence of human interaction (HI) and the rate at which the frequency of HI cases has increased over time among stranded harbor, harp, and gray seals and two regions of Maine distinction between multiple HI behaviors. Eight of the 12 cases that experienced five distinct HI types were reported in the last 2 years of the study, although the earliest of these cases was reported in 2010. We also noticed differences in HI type frequency over time for some subcategories of harassment; for example, flushing was primarily reported in 2018 and 2019, while pushing animals back into the water has been reported consistently since 2013. Because of these potential changes in how harassment behaviors were recorded over time, we do not attempt to further interpret temporal trends in HI types in this study. Finally, we observed differences in HI type experienced among pinniped species and age classes (Figure 3). Human approach was the most common type of HI for harbor (69.54%) and harp seals (75.00%), but for gray seals, reports of entanglement (54.76%) exceeded human approach (26.19%). Pups (62.43% of all human approach cases) and yearlings (25.43%) were the most common age class to be approached by humans across all three species; adults (2.02%) were rarely approached. Entangled and gunshot animals were more likely to be adults (entanglement: 52.63%; gunshot: 88.89%), and a range of age classes experienced vessel trauma.

| DISCUSSION
Human interaction with marine mammals has known risks of negative impacts for both the marine mammals and humans involved. Yet, in Maine, reports of HI have continued to increase in recent years, primarily attributed to harassment. To address current gaps in our understanding of the potential impact on pinniped populations of the different types of human activities associated with harassment, we undertook a retrospective analysis of stranding records from Maine from 2007 to 2019. As a part of this analysis we proposed an updated classification scheme for harassment, which aims to better enable assessment of the potential impacts of human behavior and associated HI on local pinniped populations and provide stranding networks with more detailed information that can be used when designing response and HI mitigation strategies.
Human interaction was reported in approximately 15% of reported harbor, gray, and harp seal strandings in Maine each year from 2007 to 2019 (Table 2), similar to rates previously reported among stranded pinnipeds on the west coast of the US (11-16%) (Keledjian & Mesnick, 2013;Warlick et al., 2018). The highest number of strandings and HI cases in our study region occurred in southern Maine (Figure 1), where the coastline is more accessible, more densely populated, and more frequently visited by tourists, and where stranded seals are typically more visible against a backdrop of sandy beach as compared to rocky coastline. In Maine and elsewhere, this association of strandings and human population density has been attributed to both a reporting effect and human impacts on coastal marine mammal populations (Esquible & Table 1 for definitions of each type of HI et al., 2020). The reporting effect refers to the increased likelihood that a seal will be observed and reported in areas of higher human density. Because we do not see a difference in the proportion of HI cases between the more accessible and developed southern and mid-coast Maine coastline and the more rural and rocky northeastern Maine coastline, we can attribute the variation in number of HI cases reported across the state primarily to the reporting effect.
The proportion of Maine pinniped strandings involving HI has increased over our study period, similar to trends observed on the west coast of the United States over the past several decades (Goldstein et al., 1999;Keledjian & Mesnick, 2013;Warlick et al., 2018). In Maine, the increase in HI prevalence may be related, in part, to rehabilitation capacity. The University of New England Marine Animal Rehabilitation Center offered rehabilitation capacity for stranded pinnipeds in the state from 2001 to 2014. Following its closure, rehabilitation capacity was limited to partnering out-of-state centers until 2016 when MMoME opened its own rehabilitation center. From 2014 to 2016, MMoME reports that with less rehabilitation capacity in the state, fewer stranded animals could be collected and stranded animals spent longer on the beaches, thus increasing their opportunities for HI. This southern region also experienced change in stranding response effort and capacity over the study period; marine mammal stranding response in this region was conducted by MEDMR up to 2011 and by MMoME thereafter. Although the Level A form has remained constant throughout the study period, it is possible that these other changes in personnel and protocols have affected reporting rates for HI.
We find that the dominant type of HI that affects pinnipeds in our study region is harassment (Figure 2). From prior research, we know that the impacts of harassment on pinniped populations can range from little to no impact, particularly in cases of habituation (Boren et al., 2002;Engelhard, Baarspul, Broekman, Creuwels, & Reijnders, 2002;Holcomb, Young, & Gerber, 2009), to population-scale impacts with demographic consequences (Gerrodette & Gilmartin, 1990). Yet, even habituation of individual animals can have broader reaching consequences, for example, when habituated animals interact with humans in detrimental ways, including aggression (Walsh, Kovaka, Vaca, Weisberg, & Weisberg, 2020) and fisheries depredation (Cook, James, & Bearzi, 2015). Such interactions can lead to negative public perception of pinnipeds and motivate demands for population control (Cook et al., 2015;Jackman et al., 2018).
Using a modified classification scheme (Table 1), our analyses reveal the diversity of types of harassment experienced by single animals and across stranded animals in our dataset. The majority of the harassment cases observed in Maine from 2007 to 2019 involved close approach by a human, and more than half of these approach cases further involved physical contact. Human approach, with or without physical contact, often resulted in displacement of the stranded animal. Although this study did not attempt to measure the impact of HI on affected marine mammals, MMoME and Allied Whale have both documented significant impacts of harassment, including abandonment, injury, and death. An analysis by species and age class further reveals that despite finding no difference in the proportion of HI cases per species or the rates at which the frequency of HI cases changed over time (Table 2), there were clear differences in the types of HI experienced by each species and age class (Figure 3). Human approach was the most common type of HI for harbor and harp seals, which were predominantly pups and yearlings, while  Table 1 for definitions of each type of HI entanglement was most common among gray seals, which included a larger proportion of adults. Similar species-specific differences have been reported in other studies (Goldstein et al., 1999;Warlick et al., 2018).
In Maine and elsewhere, these trends may reflect differences in the life history and ecology of pinniped species. For example, adult gray seals spend little time hauled out on Maine's coastal beaches, preferring rocky islands instead, and are large enough to deter close human approach in most cases. Young harbor seals, on the other hand, which pup in coastal Maine during the late spring and early summer months, are more likely to interact with people on the beaches during the coinciding season of peak tourist visitation and beach use. This observation is also consistent with seasonal differences in HI frequency, which is relatively high during the spring and summer months (Table 2 and Figure S1). In Maine, harbor seal mothers commonly leave their pups alone on beaches for short periods of time while foraging offshore (Wilson, 1978). The assumption by the general public that these solo pups require their assistance can be particularly problematic as HI at this susceptible life stage can disrupt mother-pup bonding and lead to abandonment and reduced survival (Osinga, Nussbaum, Brakefield, & Udo de Haes, 2012).
By breaking down the harassment category of HI, we provide stranding networks with more detailed information that can be used when designing their response and outreach strategies. Descriptive spatial analyses provide information about where the number of HI cases reported each year is high, which may be one way to identify regions where HI is likely to impact marine mammal population health and where mitigation efforts (e.g., public outreach campaigns) should be targeted. These areas in Maine and elsewhere are often associated with high-density human populations (Goldstein et al., 1999;Keledjian & Mesnick, 2013;Warlick et al., 2018). The complementary hotspot analyses identify regions with higher than expected proportions of HI based on the number of strandings. Interestingly, the areas with the highest density of HI cases did not frequently overlap with areas identified by the hotspot analysis, which seem more ephemeral in nature (Figures 1  and S3). While the total number of HI cases can be low in these hotspots, this approach serves as a useful indicator for stranding networks to identify potential emerging areas of concern that warrant close monitoring and dedicated outreach.
Our new understanding of common HI types and occurrence holds value in its potential to inform more targeted future public outreach campaigns and assess the outcomes of any associated policy change. While this research did not directly assess the human populations engaged in HI behaviors, describing HI type and distribution offers insights into which human groups to target for effective mitigation and what strategies to employ. For example, efforts to reduce entanglement among gray seals in mid-coast Maine will require a different strategy (e.g., fishermen outreach, marine debris beach clean-ups) than campaigns to protect harbor seal pups from human approach in southern Maine (e.g., beach signage, public presentations to community groups). The design and implementation of such public outreach campaigns would benefit from additional research into what triggers different human harassment behaviors. Similar to the described spatial variation in HI type, human motivators for pinniped interaction may vary geographically, and our research could inform the design of subsequent studies of human attitudes and perceptions toward pinnipeds across the study region.
Beyond its relevance to local public outreach programs, our modified HI classification scheme further enhances the existing national system for assessing HI and could thereby serve as a framework to monitor the effects of national-level policy change. As many pinniped populations in the United States recover, there are increasing calls to revisit and revise current MMPA protections for these species (Cammen et al., 2019). Should such policy change be enacted, it is critical that stranding organizations and researchers have methods in place to track potential downstream effects on human behaviors toward marine mammals and associated pinniped health outcomes.
The subcategories we added for harassment were informed by the types of human behaviors that were observed over a 13-year period in Maine, and thus are likely robust for the pinniped species in this region. A similar scheme may be valuable in other regions, such as the Pacific Northwest coast of the United States, which also reports the greatest proportion of their pinniped HI cases as falling within the "Other Human Interaction" category (Warlick et al., 2018). However, we acknowledge that other regions and other species may experience different common types of HI, and in such cases, the current HI classification schemes may be sufficient or benefit from other modifications. For these other regions, we believe that our methodological approach of redesigning a targeted classification scheme will be transferable and valuable in future analyses.
National Stranding Database and reviewing earlier versions of this manuscript; and Shannon Brown, Sarah Burton, Elizabeth Piotrowski, and Kai LaSpina for their work on data entry. Funding for this research was provided by a NOAA John H. Prescott Marine Mammal Rescue Assistance Grant (NA18NMF4390041).