COVID-19 Anthropause Impacts Coral Reef Ecosystems

The COVID-19 pandemic temporarily shifted the magnitude of human activity and how humans interacted with each other and with nature, providing an inadvertent ‘natural experiment’ for quantifying human impacts on various ecosystems¹. The ‘COVID-19 anthropause’ refers to the pattern of widespread alteration in human activities and their interactions with ecosystems resulting from the COVID-19 pandemic². The emergence of SARS-CoV-2 in late 2019 within human populations led to the implementation of public health measures across the world that resulted in widespread changes affecting work, travel, recreation, and other social patterns, ultimately leading to rapid and dramatic changes in how humans interact with the natural world and its inhabitants. A combined quantitative and qualitative global assessment of the COVID-19 anthropause found both positive and negative effects of human activity on species and ecosystems¹. Positive effects on wildlife and their habitats stemmed from many diverse factors, including reduced human presence and noise, whereas negative effects on wildlife originated in part from reduced conservation, management, restoration, and enforcement activities¹. Relatively few of the examples in that meta-analysis or cases found elsewhere in the literature document COVID anthropause effects on coral reef species or ecosystems, although effects have been documented on queen conch (Strombus gigas)¹, tiger sharks (Galeocerdo cuvier)¹, and abundance—but not behavior—of various reef fishes^{1,3,4,5,6,7,8}. These examples collectively highlight the intricate connections between human health and ecosystem dynamics, and in particular, how the spread of infectious diseases among humans can lead to changes in behavior, natural resource use, and resource management practices that can significantly affect the health of ecosystems. What remains unknown is if and how reduced (and subsequent increased) human presence during the COVID-19 anthropause can affect multiple reef ecosystem components simultaneously and what such changes might mean for reefs’ ecological resilience. Coastal human populations worldwide are expected to increase rapidly and disproportionately relative to inland areas in the coming decades⁹, generating increasing pressure on coastal ecosystems and natural resources. Therefore, using the COVID-19 anthropause to better understand the impact(s) of human presence on reef systems can provide important clues to how these ecosystems, and the resources they provide for coastal inhabitants, will conceivably change over time as human presence changes.

In Hawai‘i, USA, human presence and coastal marine wildlife are in frequent contact due to the high density of residents and visitors to the state. Within this context, many locations are particular ‘hotspots’ of human visitation to the coral reef ecosystem. Hanuama Bay, on the densely populated island of O‘ahu, is one such location. This bay is known for its ample wildlife, including reef fishes and endangered monk seals. Ongoing monitoring programs regularly assess these and other biological variables as well as environmental conditions, such as water clarity, to gauge reef ecosystem health. The closure of this bay to visitors in 2020 therefore provided a unique case study system in which to explore how human presence alters these diverse ecosystem facets—and therefore if and how the COVID-19 pathogen indirectly affects coral reef ecosystems.

Coral reefs are heavily human-impacted and continue to face ecological degradation worldwide due to climate change and other direct human interactions¹⁰. While system-wide evidence for COVID-19 anthropause effects is limited, prior studies provide evidence of impacts to key aspects and species of coral reef systems from human presence. In terms of environmental factors, water clarity is a commonly used metric, among others (e.g., chlorophyll), for quantifying water quality on coral reefs¹¹. Water clarity, or its converse (turbidity), can be affected by many natural and human-caused factors, primarily through changes in inputs of sediments, nutrients, and pollutants. Resuspension of existing sediments due to factors, such as wave action¹², river plumes¹², dredging ¹³, and direct human presence (e.g., diving, swimming, etc.)¹⁴ can lead to both acute and chronic changes in water clarity, depending on the spatial and temporal scale of the disturbance. Indeed, Edward et al.⁴ found that turbidity decreased on reefs after, relative to before, the COVID-19 lockdown in the Gulf of Mannar, India.

In addition to water clarity, human presence may affect behavior of reef-dwelling marine mammals. The Hawaiian monk seal (Neomonachus schauinslandi) is a pinniped found throughout the Hawaiian Islands that has been listed as Endangered under the US Endangered Species Act since 1976 and is listed as Endangered with decreasing population under the IUCN Red List¹⁵. Roughly 20 percent of the total population resides in the human-populated main Hawaiian Islands (MHI), with the remainder inhabiting the remote, un-populated Papahānaumokuākea Marine National Monument in the Northwestern Hawaiian Islands¹⁶. A 15-year long-term study examining the causes of death (COD) for monk seals in the MHI found that the largest impact on the population’s growth rate was anthropogenic factors (i.e., anthropogenic trauma; anthropogenic drowning due to entanglement) relative to natural or disease-related CODs¹⁶. While few studies, if any, have focused solely on the impacts of direct human presence on Hawaiian monk seals, the abundance of the related Mediterranean monk seal (Monachus monachus) has been shown to be negatively correlated with boat numbers and boat noise levels¹⁷. Furthermore, as the number of boats increased, seals were found to be significantly less likely to rest and sleep, and more likely to exhibit vigilance behavior¹⁷, suggesting that human presence has measurable sub-lethal impacts on this species. Studies of numerous other pinniped species worldwide have found diverse negative effects of direct human presence (both in boats and on foot), including fleeing responses, aggression, and disruption of maternal behavior (e.g., reduced lactation period duration)¹⁸.

The only group of reef organisms that has been documented by multiple studies to date to have been impacted by the COVID anthropause is coral reef fishes. Several studies of reef fishes associated with the COVID-19 anthropause observed increases in density^1,4,5,6,7, biomass³, and diversity/richness^7,8 during periods in which human visitors were absent or reduced in number. Although the time frame of these responses suggests they are due to changes in fish behavior rather than population size, no studies have attempted to isolate the mechanism(s) behind these changes, nor whether they can rebound when human presence is removed.

Here, we aim to explore the question of whether, and to what degree, the pathogen-induced global anthropause altered key biophysical ecosystem parameters in a coral reef ecosystem. We do so by conducting repeated surveys of a heavily visited reef system in Hawai‘i before, during, and after the closure of the bay to human visitors in response to the COVID-19 pandemic. Our prediction was that wildlife (specifically, monk seals and fish) would exhibit increased numbers and activity levels across a range of behaviors, and that water clarity would improve when humans were absent from the reef ecosystem. By comparing across these ‘treatments’ created by the emergence of the SARS COVID-19 pathogen, our results shed light on whether, and to what degree, human presence shapes coral reef systems on heavily visited reefs, and how the system responds when human presence is removed or reduced.

Water clarity

Water clarity in the bay (Fig. 1) was strongly influenced by human presence, as measured by both period (i.e., whether the bay was open or closed to humans; Fig. 2A; t(180.94) = 8.9934, p < 0.001) and total number of daily visitors (Fig. 2B; R² = 0.04, F(1, 478) = 24.17, p < 0.001). As the total number of daily visitors increased, water clarity decreased significantly. Specifically, when comparing the COVID-19 closure of the bay versus the post-COVID re-opening with ~750 visitors per day, water clarity was significantly lower in the latter period.

Monk seal presence

Monk seal presence was similarly affected by human presence. The percentage of days when monk seals were observed at the bay more than doubled during the COVID closure period, relative to the combined pre- and post-COVID periods when humans were present; monk seals were present ~20% of open days and ~45% of closed days (Fig. 3). Across all three periods, the number of monk seals observed per day ranged from zero to three.

Fish abundance

In an apparent contrast, fish density was lower under higher chronic human presence, but was elevated in response to immediate human presence. Specifically, diver-observed fish density differed between the three periods, with higher fish density during the COVID-19 closure and post-COVID reopening periods, both of which had lower human visitation than the pre-COVID open period (Fig. 4A). Both models (i.e., using only treatment and both treatment and transect as fixed factors) yielded qualitatively similar results (Table S2A, B). A Tukey post-hoc test revealed significant differences between fish density in the pre-COVID open period vs. the COVID closed period (p = 0.03) and the pre-COVID vs. post-COVID open periods (p = 0.005), despite the fact that the pre-COVID period surveys were conducted on the scheduled, weekly day on which human visitors were not present in the bay. In contrast, based on camera trap data, fish abundance was markedly higher in the immediate presence of humans (i.e., when a human was present in the camera field of view), whereby more fishes were observed within the camera’s FOV in the moments surrounding a snorkeler’s visit to a patch of reef than in the 25 s before or after their visit (Fig. 4B; Table S3).

Fish behavior

Fish behavior—in particular, algal grazing and, to a lesser extent, aggression—varied over both chronic and acute differences in human presence. Quantile regression revealed an apparent upper bound of individual herbivorous fishes’ bite rates across all sites, resulting in a wedge-shaped distribution whereby maximum bite rate decreased as the number of humans present during the observation period increased (Fig. 5A; tau = 0.99; R² = 0.02, t = −2.18, p = 0.03). When only parrotfishes were considered, site was revealed to be a significant factor in parrotfish bites per unit time (F(3, 38) = 5.156, p = 0.004), prompting further analysis using only the most heavily-visited sector in the bay (Keyhole; see Fig. 1C map). This analysis revealed that aggregate parrotfish bites per unit time decreased dramatically at this site from full closure (4.4 bites/minute) to post-COVID re-opening (0.6 bites/minute; Fig. 5B; t(4.4858) = 3.6857, p = 0.01724). Lastly, our zero-inflated regression model via maximum likelihood revealed an apparent upper bound in the time that fishes spent attacking one another (Fig. 5C). Specifically, at lower levels of human presence, fish exhibited a wide range of time devoted to aggression, while at high human presence, little time was devoted to aggressive behaviors, resulting in a detectable negative effect of human density on the time spent by fish attacking each other (Table S4). Other behaviors quantified as time budgets, including feeding, cleaning, and burrowing, did not exhibit differences due to either chronic or acute human presence (Table S5).

By curtailing human activities, the COVID-19 pathogen indirectly affected coral reef ecosystems via changes in water clarity and the numbers and behavior of reef animals. Specifically, when human visitation to the Hanauma Bay Nature Preserve ceased due to the COVID-19 anthropause, water clarity improved, and increases were detected in the presence of endangered Hawaiian monk seals, the abundance of fish, and the rates of parrotfish herbivory (i.e., algal removal) at heavily visited sites in the bay. Following the anthropause, when humans had returned to the bay, fish were (counterintuitively) attracted to the immediate presence of snorkelers, such that fish abundance increased over scales of seconds to minutes when humans were present within a given area of reef. Overall, these results suggest that chronic human presence likely has a measurable, inhibitory effect on multiple ecosystem parameters, including the key ecosystem function of herbivory that is critical to maintaining coral reef health¹⁹.

The striking difference in water clarity between the COVID and post-COVID periods of the study demonstrates significant impacts from visitors to the bay, most likely due to sediment resuspension, which is magnified as human density increases. Water clarity was not assessed prior to the COVID-19 closure of the bay when up to 3000 visitors were utilizing the bay, but this pattern suggests that it may have been lower still under pre-COVID conditions with fourfold-plus-higher visitation intensity. This finding is in agreement with results from the COVID-19 anthropause in India, where coastal turbidity decreased during COVID-19 lockdowns⁴. Water quality is known to be an important determinant of overall coral reef health²⁰, affecting multiple key taxa through different mechanisms. For example, over a regional scale on Australia’s Great Barrier Reef, water quality was found to be strongly predictive of both coral species richness and macroalgal cover, with the former inversely related and the latter positively correlated¹¹, and high levels of suspended sediment reduced coral recruitment on Molokaʻi, Hawaiʻi²¹. Numerous other studies have found similar results at smaller scales, including examples of snorkeler- and diver-derived sedimentation effects on coral disease prevalence²², live coral cover, coral recruitment, macroalgal cover, and diversity of corals and fishes²³. Because sedimentation impacts on coral reefs are magnified with increased intensity, duration, and frequency of resuspension disturbances¹³, reducing the number of snorkelers and waders, the amount of time spent in and around the reef, and/or the number of visitor days would likely reduce the impact of snorkeler-derived effects on coral health via reduced water quality.

We found Hawaiian monk seals were roughly half as likely to visit the Hanauma Bay Nature Preserve when humans were present in the pre- and post-COVID periods versus the COVID period, during which humans were absent. This result is in accordance with previous studies of related pinniped species (e.g., the Mediterranean monk seal (Monachus monachus))¹⁷ and other pinniped species worldwide¹⁸. Our results suggest that even the dramatically reduced visitor capacity in the post-COVID period caused apparent effects on monk seal haul-out behavior. It therefore remains unknown what level of human visitor density should be maintained in order to minimize or eliminate these effects. Allowing for longer periods of ‘rest’ for the bay in which no humans are present could potentially increase the number of days monk seals could utilize this habitat. However, whether seals are likely to respond to such a change is unclear, given that with one or more days closed to visitors per week, seals would still encounter humans on the remaining days when exploring the bay as a potential haul-out site.

Our finding that overall reef fish density was lowest under the highest levels of human visitation to Hanauma Bay (i.e., the pre-COVID period) qualitatively matches patterns seen elsewhere such as Akumal, Mexico²⁴. As with that study, the mechanisms for this pattern in the present study remain untested, but are likely due to inherent threat-avoidance behaviors of fishes to humans²⁵. We conversely found that in the post-COVID period, when visitors had returned to the bay, fish were strongly attracted to the presence of snorkelers over scales of seconds to minutes, a likely sign of increased food availability surrounding snorkelers due to re-suspended organic matter from snorkelers’ fins disturbing sediment. This result is in line with Brock et al.’s²⁶ finding roughly 20 years ago that significantly more fish were observed in Hanauma Bay on days open to the public vs. closed days, a likely result of food provisioning by humans, which had been done for decades at the time of that study and ceased just prior to the study’s start. A precondition for a lack of fishing pressure and habituation to snorkelers is essential to this scenario. Our contrasting findings (i.e., Fig. 4A vs. 4B) result from different time scales of human presence, suggesting different fish behavioral responses across temporal scales and specifically that long-term risk avoidance is offset in the short term by the reward of food availability²⁷. Although untested, short-term density changes due to both chronic and acute human presence are presumed to be due to changes in fish behavior rather than population growth, suggesting that fish behavior may play a key role with regard to mediating human impact.

Furthermore, herbivore grazing behavior was dramatically reduced in the post-COVID period relative to the COVID period when humans were absent, and individual herbivore bite rates were also reduced as a function of increasing visitor numbers across these two periods. This effect is not likely due solely to differences in the density of herbivores across the different study periods because there was no significant difference in overall fish density between the COVID and post-COVID periods. Additionally, the post-COVID period had lower grazing rates than the COVID period. These findings collectively suggest that snorkeler presence can suppress the key ecosystem function of herbivore grazing.

By halting all human visitation to the study’s heavily visited coral reef ecosystem, the COVID-19 pathogen indirectly instigated a clear-cut ‘natural experiment’ that would not have been possible in its absence, and in the process revealed a number of important considerations and potential prescriptions for management strategies of the Hanauma Bay Nature Preserve and other coral reef ecosystems worldwide. First, reduced human visitation allowed multiple biophysical ecosystem parameters to shift towards levels present when unimpacted by direct human presence. Secondly, by inhibiting the grazing of benthic algae by herbivorous parrotfishes, one of the key groups responsible for maintaining balance between coral- and algal-dominance of reef benthos, human visitation to this iconic coral reef (and perhaps other reefs globally) may inadvertently promote ecosystem decline¹⁹. Our results add to the growing body of knowledge resulting from the COVID-19 anthropause and other situations that reducing human visitors to a diversity of other natural areas (both wilderness and other) often leads to rebounds in abundances and changes in behavior of wildlife¹ and subsequent restoration of key ecosystem structure, processes, and function^28,29, though studies of these indirect effects remain scarce³⁰.

To our knowledge, previous studies resulting from the COVID-19 anthropause have not shown diverse, biophysical effects from a single location as we show here. Importantly, at least two of the effects observed in this study—reduced herbivory and increased turbidity—may act synergistically to threaten coral growth and survival beyond either factor in isolation³¹. Nonetheless, our results suggest that this particular location has begun to somewhat mitigate these negative effects by re-opening, post-anthropause, with reduced human visitor density (i.e., a maximum of 3000 visitors/day pre-COVID to a maximum of 750/day during the study’s post-COVID period). This reduction may help explain why there was no detectable difference in overall fish density between the COVID closure and post-COVID periods. It may also help explain the lack of difference among fish behavior allocation, other than parrotfish herbivory, that were measured only in these latter two periods (due to lack of pre-COVID data).

Implementing visitor limits for currently unrestricted reefs elsewhere in Hawai‘i and the global tropics, or further reducing visitor limits where they currently exist, has the potential to regain lost ecological function and reduce negative human impacts on reefs. While coral reef tourism generates an estimated US$36 billion globally each year³² and nearly US$1 billion in Hawai‘i alone (Hawai‘i DLNR 2023)³³, it is important to consider the need for balance between ecological restoration goals and the economic benefits of reef tourism. These goals may not be necessarily at odds, given that past research has shown visitors are often willing to pay a higher price to visit reefs that are perceived or marketed as ‘pristine’ rather than degraded, thus potentially offsetting lost revenue due to visitor caps^34,35.

Despite these findings, many questions remain about the impact of human visitors on both the Hanauma Bay Nature Reserve and other reef systems. First, this study is one of the few to date that documents if and how human-induced animal behavior change on individuals, populations, and communities results in cascading effects on ecosystem functions in any ecological system³⁰, yet understanding these effects is important for wise visitor and resource management decision-making. Secondly, our study design did not allow us to answer the question of what the lower threshold is for the onset of negative effects (i.e., what density of visitors could exist without negatively affecting ecosystem structure or function), although our research does suggest that a roughly linear decline in maximum herbivory occurs as human density increases. Similarly, we did not observe negative impacts on other behaviors quantified as time budgets (e.g., feeding, cleaning, and burrowing); therefore, it remains unknown what density of human visitors would likely trigger suppression of these and other unmeasured behaviors and ecosystem metrics. Additionally, Hanauma Bay is one of the few reef systems in the main Hawaiian Islands that is fully protected from fishing and other extractive uses, and therefore maintains far higher abundance and biomass of reef fishes than most other reefs in the MHI archipelago³⁶. How these results translate to other, more impacted reefs in the main Hawaiian Islands and other regions globally remains unknown. Importantly, one interesting insight arising from these findings is that the COVID-19 anthropause affected both the abundance and behavior of reef fishes and monk seals, which suggests that to fully quantify the effect on wildlife populations, future studies should aim to measure multiple parameters among diverse marine populations.

Collectively, these results paint a picture of the various and unexpected ways in which a novel human pathogen indirectly affected a coral reef ecosystem by altering human behavior patterns. In so doing, this study highlights the interconnectedness of human and ecological systems, pointing to the need to consider potential downstream implications of seemingly unrelated human societal decisions and policies on wildlife and ecosystem management. Given the vital importance of reef tourism to local economies and the cultural, spiritual, and economic value of reefs to Indigenous and other local communities, managers must find a balance between visitor management and ecological preservation of coral reefs within Hawai’i and elsewhere in the global tropics.

Study system

Hanauma Bay, situated on Oʻahu’s southeastern coast in Hawaiʻi, is a heavily visited bay known for its unique blend of natural beauty and ecological significance. Hawaiʻi’s first no-take Marine Life Conservation District (MLCD) was designated as the Hanauma Bay Nature Preserve in 1967. This crescent-shaped bay was formed within a volcanic crater and now has markedly higher reef fish biomass than adjacent, unprotected reefs and many other Marine Protected Areas (MPAs) in the main Hawaiian Islands³⁶. While fish populations have shown recovery from previous fishing pressure and other stressors after becoming an MLCD, visitor numbers have also increased, peaking at roughly 10,000 visitors per day in the late 1980s³⁷. The beach area of the bay is approximately 10,500 m² at low tide, resulting in 1990s human densities of approximately one human per m² of beach. Recognized for its diverse marine life, the bay is still a hotspot for snorkeling, but the number of visitors was limited to 3000/day prior to the COVID-19 closure, closed entirely to visitors from April 28 to December 1, 2020, re-opened at a reduced capacity of 750 visitors/day from 2 December 2020 through 27 April 2021, and, since late April 2021, has operated at a still-reduced capacity of maximum 1400 visitors/day^37,38. All surveys conducted as part of this study were done under the original 3000 visitors/day (period pre-COVID), closed (period COVID), or 750 visitors/day (period post-COVID) capacity conditions.

Snorkeler counts

To relate our biological and environmental data to human presence, the cumulative number of snorkelers and waders in each sector (Fig. 1) was also quantified during data collection through four 10-min counts per survey day, spaced roughly evenly throughout the morning, that were later averaged. Box office counts (i.e., total daily visitors to the bay) were separately acquired from the City and County of Honolulu’s Department of Parks and Recreation to give an estimate of the daily visitor load.

Water clarity measurements

To determine water visibility across time and environmental conditions, Secchi disk water clarity measurements were taken within the four sectors of the Hanauma Bay Nature Preserve (Fig. 1C). To measure horizontal Secchi disk water clarity, a team that consisted of a pair of surveyors each conducted two measurements for a total of four per sector. One surveyor held the Secchi disk in place within the water column, while the second surveyor swam away from the disk that was connected to a transect line until the white Secchi disk was no longer visible. The distance at which they lost sight of the disk was recorded using the transect line. The same surveyor then swam further away from the disk and waited for 30 s before swimming back towards the disk, stopping to record the distance at which the secchi disk was again visible. This was repeated by switching the positions of surveyors one and two. The four measurements were then averaged to generate a single value per sector per day. During the COVID-19 closure, Secchi disk measurements were repeated on 24 dates (4 observations per day) between April 21 and December 1, 2020. Secchi disk measurements were collected following reopening on three days in December 2020 and three days in January 2021 when visitors were present in the bay (6 days total, with 4 observations per sector per day).

Monk seal surveys

The presence of Hawaiian monk seals (Monachus schauinslandi) was documented over 19 days in 2018 when the Hanauma Bay Nature Preserve was open to the public, 29 days during the COVID-19 closure to the public, and 13 days following the reopening of the Bay to the public. Observers conducted monk seal surveys for the duration of each ~4 h visit to the bay by repeatedly scanning the entire length of the beach for monk seals that were hauled out on the sand, and the presence or absence of monk seals each day was recorded. It was not possible to reliably determine the presence/absence of seals in the water; thus, only those that emerged from the water onto the beach are included. Data were analyzed using a binomial GLM (R function glm) on days with (1) and without (0) monk seals present in the two overarching treatments (i.e., when the bay was open vs. closed to visitors; Table S1).

Fish abundance surveys

A modified visual 25 m × 5 m (125 m²) belt transect³⁹ was used to survey the fish community. Eight permanent transects (two per sector, all within shallow (<2 m depth) back-reef habitat; Fig. 1) were established running perpendicular to the shoreline, with stainless steel eyebolts marking the start and end of each 25 m transect. A surveyor swam and recorded all species of fish observed in the water column, noting the number and size to the nearest cm for all species. Fish surveys were conducted prior to the COVID-19 closure on three Tuesdays when the Bay was closed to the public, ten days during the COVID-19 closure, and five days after the COVID-19 closure had ended and the bay was open to visitors. Prior to the COVID-19 closure, visitors were allowed to visit the bay on all days except Tuesdays, on which days only a small number (~10 or fewer) scientists were allowed to access the bay for research purposes. These surveys were conducted as part of an existing, pre-COVID study and subsequently became an opportunistic part of the present study following the COVID-19 closure to provide pre-closure comparison data. Data were analyzed using a linear model (R function lm) based on fish density from diver surveys (A) with only treatment (i.e., pre-, during, and post-closure of Hanauma Bay to visitors) included as predictor variable and (B) with both treatment and (fixed) transect included as predictor variables, since transects were repeated measures from the same location over time.

Fish behavioral observations

In conjunction with the abundance surveys, key fish behaviors were also monitored from April 28, 2020 to March 3, 2021, starting roughly a month after the complete shutdown of Hanauma Bay to visitors and continuing through the reopening of the bay to visitors (12 surveys total; six during the closure and six after re-opening). On each survey day, an array of GoPro cameras (aka, ‘camera traps’) was set up throughout the bay to collect video footage of fishes and invertebrates without interference from scientific diver observers. We subsequently post-processed the video using QuickTime software to extract quantitative estimates of key fish behaviors, including foraging, predator-prey interactions, and competitive interactions. For these analyses, time of entry and exit within the camera’s field of view (FOV) for individual fishes was recorded, and fish time budgets were coded into 8 different categories: (1) feeding, (2) passing, (3) guarding, (4) resting, (5) burrowing, (6) attacking, (7) cleaning, and (8) following. When feeding occurred, the number of bites by each individual was recorded. Definitions of these behavioral categories can be found in the video analysis protocol. For each category, the percentage of time the fish spent engaging in the respective behavior was calculated. Fish abundance was also measured from camera trap surveys. We also recorded the number and duration of humans (snorkelers/swimmers) within the camera’s field of view (FOV) to generate an estimate of instantaneous human presence. On either side of each instance of human presence, we delineated the period for the 25 s before and after a human was present. To examine the effect of instantaneous human presence on fish abundance, we used a generalized linear mixed model (R function glmer) with survey ID (i.e., variable rep, or the unique survey date/sector identifier) included as a random factor. For attacking data specifically, we used a zero-inflated regression model via maximum likelihood (R function zeroinfl), which allowed us to use a hurdle model to address two related questions: i.e., first, is attacking more or less likely to occur under human presence (the binomial model component); secondly, when attacking occurs, is it less frequent under human presence (via a poisson model, conditional upon the binomial component)? A Poisson distribution was selected because our response variable is the number of time points a particular fish attacked another fish while it was present in a video. These data were collectively used to understand the interaction between fish abundance and behavior in relation to human presence at the Hanauma Bay Nature Preserve.

Video analysis protocol

Below we outline the detailed video analysis protocol adapted from Madin et al. (2019). This methodology ensures a standardized and efficient process for collecting high-resolution data on fish abundance and behavior that is applicable across various reef environments.

Camera parameters include the following:

• Cameras used: GoPro Hero (specifically Hero 4 Session, Hero 5 Black, and Hero 7 Black).

• Field-of-view boundary: 3 m in front of the camera (except where noted in datasheets).
• Resolution: 1080p for GoPro Hero 4 Session; 720p for GoPro Hero 7 and Hero 5.
• Frame rate: 60 fps.
• View setting: “Wide”.

Cameras were set up in the following way:

1. 1.

   Locate the first marker (bolt) of the permanent transect at each site. Attach float or flagging tape temporarily. This serves as the video’s focal point and field of view (FOV) marker.

2. 2.

   Choose an area with a suitable substrate for recording. Ensure that the view towards the transect bolt/marker is not obstructed by large objects (e.g., reef structures).

3. 3.

   Lock the transect tape at a 3-m distance. Find a location that is clear and has a suitable attachment substrate at exactly 3 m from the transect bolt.

4. 4.

   Attach the camera to the substrate using zip ties, ensuring it faces the FOV marker.

5. 5.

   Turn the camera on to start recording.

6. 6.

   Indicate which transect you are filming by placing a specific number of fingers in front of the lens. The number corresponds to the sequence of the transect (e.g., one finger for the first transect).

7. 7.

   Point towards the FOV marker for the video analyst’s reference.

8. 8.

   Roll up the transect tape and detach it.

9. 9.

   Swim away and, once back on land, record metadata on waterproof paper. Metadata includes location, site, transect, date, start time, camera number, maximum depth, minimum depth, FOV distance, and hand signal number.

10. 10.

    Return after 35 min to collect the camera and FOV marker. The 35 min includes a 5-min buffer to allow for fish behavior normalization.

When analyzing video, the first 5 min of each video were not used to allow fish to return to normal behavior. The remaining video was divided among analysts who received standard-length segments from each site and date, ensuring uniform distribution of inter-observer variation. Analysts used an Excel spreadsheet template for data extraction, focusing on the time of entry and exit of each fish within the FOV, species identification, and behavior quantification in the ten categories above.

Analysts completed the following steps:

1. 1.

   Complete the Excel Data Template and fill in the required columns with information such as observer name, video time, fish species, behavior, etc.

2. 2.

   Record the time when each fish enters the FOV.

3. 3.

   Identify the fish species and record the common name, genus, and species.

4. 4.

   Observe and record the fish behavior, including feeding activity (if applicable).

5. 5.

   Record the time when the fish leaves the FOV and estimate the percentage of time spent in each activity over the duration of time the fish was visible in the FOV.

6. 6.

   Rewind the video to the entry time of the previous fish and repeat the process for each fish in the FOV.

Behavior was quantified by allocating fish behavior amongst 10 categories, defined below. Each behavior is quantified by estimating the percentage of time a fish spends in each category while visible.

• Feeding: when the mouth contacts the substrate. It often (but not always) is accompanied by obvious body jerks that look like it’s ‘bouncing off’ the substrate.

• Passing: swimming, when not engaged in any other apparent activity (e.g., feeding, attacking, cleaning, etc.), even if hovering with little forward motion.
• Guarding: mostly stationary and hovering around the same spot, but clearly watching other fish that pass by; primarily displayed by territorial damselfishes who aim to defend their home algal patch from intruders.
• Resting: completely stationary on the seafloor; mostly only displayed by hawkfishes, rays, flatfish, etc.
• Burrowing: digging its body into the sand and/or remaining partially covered by sand.
• Attacking: chasing or biting another animal; aggressive displays that do not result in two animals physically connecting (e.g., via a bite) are still considered attacking.
• Attacked: the recipient of an attack, as described above.
• Cleaning: feeding on the outside of an organism’s body or inside its mouth; usually only displayed by a few species of wrasses and possibly other fishes.
• Cleaned: the recipient of cleaning, as described above.
• Following: staying continuously under, immediately next to, or actually attached to another, larger fish; only displayed by a small number of species, such as remoras.