The effects of fish feeding by visitors on reef fish in a Marine Protected Area open to tourism

Coral reef-based tourism has risen sharply across the globe, coupled with an increase in fish feeding by visitors. Studies indicate that fish feeding is one of the leading causes of changes in distribution patterns, abundance, the structure of marine fish communities, and fish behavior. The aim of this study was to determine the effect of human presence and fish feeding on the behavior of reef fish by conducting in situ experiments in tide pools in a Marine Protected Area located at the northern limit of the Abrolhos Bank in the South Atlantic Ocean. Eight feeding sessions were conducted and filmed, resulting in a total of 160 minutes of video footage. Each filming session recorded four different experimental conditions alternating between human presence, human absence, and fish feeding. Our findings suggest that fish feeding may cause changes in fish behavior, such as habituation to human presence, conditioning to fish feeding, increased aggressiveness, attacks on humans, and short-term changes in species distribution. The continuation of fish feeding over time can cause an increase in the size of the populations of species that consume food provided during feeding and consequently trigger changes in the structure of communities.


Introduction
Coral reef ecosystems harbor rich biodiversity and are crucially important both from a socioeconomic and ecological standpoint, providing various environmental services and benefits, such as coastal protection, maintenance of fish stocks, ecotourism and species with pharmaceutical potential (Brander et al. 2007).However, coral reefs have been severely impacted by climate change (Mumby & Anthony 2015), ocean acidification (Comeau et al. 2015), overfishing (McClanahan et al. 2015), sewage and industrial waste (Wear & Thurber 2015), and tourism (Pereira et al. 2014).
Marine Protected Areas (MPAs) are areas set aside to protect marines ecosystems, including reef ecosystems, and play a critical role in conserving marine biodiversity and natural resources used by coastal populations (Giglio et al. 2015).The implementation of protective measures helps to keep reef communities healthy, thus maintaining their tourism attractiveness (Green & Donnelly 2003) and contributing to the growth of tourism in these areas (Milazzo et al. 2002).However, studies demonstrate that poorly planned or intensive tourist use can adversely affect marine ecosystems and associated marine life (Creed & Amado-Filho 1999, Eckrich & Holmquist 2000, Albuquerque et al. 2014, Giglio et al. 2016).Several researchers have investigated the impacts of tourism on reef environments, including trampling (Sarmento & Santos 2012, Giglio et al., 2017, Williamson et al. 2017), boat anchoring (Saphier & Hoffmann 2005, Beeden et al. 2014, Kininmonth et al. 2014), free and autonomous diving (Lamb & True 2014, Hein et al.2015), and fish feeding (Milazzo 2011, Feitosa et al. 2012, Bookhouse et al. 2013).
Fish feeding is a popular tourist attraction at coral reefs around the world (Giglio et al. 2015) and is used by tourism operators to lure specific species and give visitors an opportunity to observe marine life up close.However, research shows that this activity may disrupt distribution patterns, abundance and the structure of marine fish communities (Brunnschweiler & Barnett 2013, Brunnschweiler et al. 2014), as well as affecting fish health by increasing fat deposition and vulnerability to diseases caused by microorganisms and ectoparasites (Semeniuk & Rothley 2008).
One of the main impacts of this activity is changes in the natural behavior of fish.For example, by becoming used to the presence of humans, species that otherwise would not approach humans freely gather around bathers to seek supplementary feeding (Albuquerque et al. 2014).Furthermore, fish can also become dependent upon fish feeding (Ilarri et al. 2008) and, in the long term, the acquired behavior of obtaining food from human hands may reduce their ability to obtain food by themselves in the wild.Another behavioral change reported in the literature is increased aggressiveness during feeding, both in fish competing among themselves for food and towards the people who are feeding them (Milazzo 2011).Fish feeding may also lead to shifts in circadian rhythms, as observed in the naturally nocturnal southern stingray (Hypanus americanus), which has become diurnal due to daytime fish feeding activities (Corcoran et al. 2013).
Despite studies showing the negative impacts of artificial feeding, few MPAs have implemented fish feeding rules (Corcoran et al. 2013).There is also divergence of opinion among MPA managers in relation to fish feeding, given the trade-offs between financial sustainability and the negative impacts of the activity on the conservation of marine life (Hémery & McClanahan 2005, Milazzo et al. 2005).In the Great Barrier Reef Marine Park in Australia, for example, fish feeding is permitted provided that the total food used does not exceed 1 kg per day (GBRMPA 2000).In contrast, in the Booderee National Park, also in Australia, feeding marine life is prohibited.This divergence of opinion is also manifested in Kenya, where fish feeding is permitted in certain areas in the Malindi, Watamu, and Mombasa MPAs while being prohibited in the Kisite MPA.In Brazil, fish feeding was permitted in the Abrolhos Marine National Park up to 2003, when the activity was banned in the area.
Coral reef-based tourism has risen sharply across the globe (Milazzo et al. 2002), coupled with an increase in the amount of food fed to fish by visitors.Concerns over the negative impacts of fish feeding have led to an increase in studies assessing the effects of this activity on these unique ecosystems (Medeiros et al., 2007).However, only a few studies have assessed the impacts of fish feeding on coral reefs in the South Atlantic Ocean (Medeiros et al. 2007, Ilarri et al. 2008, Feitosa et al. 2012, Pereira et al. 2014, Albuquerque et al. 2014).Thus, detailed studies concentrating on this region are needed to gain a deeper understanding of the effects of fish feeding on the biology of reef fish, which range from physiological and behavioral aspects to impacts on fish communities, in order to improve the management and conservation of these areas.
The species belonging to the genus Abudefduf (Family: Pomacentridae) are omnivorous and may be considered generalists.This flexibility allows them to employ opportunistic feeding strategies, making them particularly susceptible to the effects of fish feeding.Studies have reported changes in behavior and the spatial distribution in three species -A.saxatilis (Medeiros et al. 2007, Ilarri et al. 2008, Feitosa et al. 2012, Albuquerque et al. 2014), A. sexfasciatus, and A. sparoides (Hémery & McClanahan 2005).It is believed that these changes could have damaging long-term ecological and economic impacts.
In light of the above, the aim of this study was to determine the effect of human presence and fish feeding on the behavior of reef fish by conducting in situ experiments in a MPA northern limit of the Abrolhos Bank located in the South Atlantic.

Study area
The Recife de Fora Marine Park (RFMP) is a MPA located at the northern limit of the Abrolhos Bank, approximately 4 kilometers from the coast and the city of Porto Seguro in the State of Bahia, Brazil.The park has an area of approximately 17.5 km 2 and is located between parallels 16°23"30"/16°25"06" S and meridians 38°58"30"/38°59"18" W (Porto Seguro 2016) (Figure 1).The depth in the middle of the central plateau of the park varies from 6 to 8 meters in the internal portion to a maximum of 20 meters in its western portion (Costa Jr. et al. 2002).The RFMP is one of the region's main tourist attractions and received 50,000 visitors in 2015.
Depressions in the plateau of the reef form tide pools at low tide, including the Piscina da Visitação (the visitors' pool) and Piscina do Golfinho (dolphin pool), which have historically been the park's main tourist attraction.The visitors' pool has an area of approximately 9,000 m 2 and a depth of 1.4 meters.Organized visitation to this area began over three decades ago.Although the current visitor limit is 400 visitors/ day, this number peaked at 1,200 visitors/day in the past (Porto Seguro 2016).Visitation occurs at low tide, when visitors are led to the pool by park guides, when the bathers, guides and photographers feed the fish using bread, biscuits, dry fish and dog food, and shredded shrimp (Porto Seguro 2016).The dolphin pool has an area of 4,500 m 2 and roughly the same depth as the visitors' pool, but has been closed to visitors since 2002.For the purposes of this study, the visitors' pool was called the "Feeding Area" (FA) and the dolphin pool "Control Area" (CA).
A manipulative experiment was conducted in each area under the following different experimental conditions: 1) "Pre-bather"comprising a period of five minutes before the bather entered the area, seeking to reproduce natural conditions without any disturbance; (2) "Bather presence" -comprising the first five minutes after the bather entered the area, to identify changes in fish behavior in the presence of a human; (3) "Post-bather" -comprising the 5-minute period immediately after the bather left the area, to determine how the fish reacted after the end of visual stimulation and the length of time it took for them to begin stabilizing their behavior; and (4) "Feeding" -with the presence of the bather feeding the fish for a period of 5 minutes, to assess changes caused by the presence of a person offering food (Figure 2).Fish behavior and abundance under each condition was assessed using remote filming sessions.
Sampling was conducted between March and April 2014 during the daytime at low tide before the arrival of tourists.During sampling, horizontal visibility was at least 4 meters in both the FA and CA.A GoPro Hero 3 Silver Edition (GoPro Inc.) digital camera was used to record fish behavior and quantify abundance installed at a fixed point at the edge of each tidal pool and 60 cm from the bottom.
Recording started as soon as the camera was installed.The first 8-10 minutes of the recording were discarded to allow the fish to stabilize their behavior after the researcher left the water.The Pre-bather condition comprised the 5-minute period after stabilization.At the end of this period, the bather entered the water and remained in front of the camera at a distance of exactly 1.5 meters for 5 minutes.It is important to note that, although it is probable that the changes caused by the presence of only one bather do not fully represent those caused by the 400 bathers that visit the FA each day, the identification of significant changes in the presence of one person provides an indication of the intensity of the changes caused by a larger number of people.At the end of this period, the bather left the tidal pool and the Post-bather condition began.At the end of this period, the bather entered the pool again and remained in the same position as in the Bather presence condition, this time offering food to the reef fish for a period of 5 minutes.
The above process was repeated 4 times in each area on separate occasions.For each of the four filming sessions in each area, a one-minute frame was selected from each of the 5-minute periods, giving a total of five frames per experimental condition per session and a overall total of 20 frames per condition for each area (n = 80 for both FA and CA).For each one-minute frame, the specific abundance of fish within a distance of 1.5 meters from the camera (distance between the camera and the bather) was quantified.Species were identified according to Humann & Deloach (2002) and Sampaio & Notthingham (2008).
The bather was represented by a researcher using the type of clothing and behavior used by photographers working for tour operators.The food used was the same shredded shrimp used by the tour operators, which is the most commonly used food during visitation.A total of 250 grams of food were gradually fed to the fish over the 5-minute period under the feeding conditions.This quantity corresponds to the amount fed by photographers and was based on information obtained through personal communications with the tourist agents.
To complement the experiments and to help identify possible changes in behavior, fish behavior in the two areas was recorded not only during the filming sessions, but also between experiments during normal group visits to the park, focusing on, but not limited to, the following aspects of behavior: agonistic interactions, interaction with the bather, attacks on the bather, reaction time to in-water food stimulation, and opportunistic feeding (for example, herbivores feeding on shrimp).

Statistical analysis
A two-way nested ANOVA was used to compare fish abundance between different experimental conditions in both areas, followed by Tukey's post hoc test when a statistically significant result was obtained.Abundance data was log (x + 1) transformed to meet ANOVA assumptions.The analyses were performed using the software package Statistica 8.0 (StatSoft 2007), adopting a 5% significance level.

Results
A total of 2,816 individuals comprising nine families and 17 species were identified from the 160 frames, which is equivalent to 32.6% of the 43 species registered in the RFMP.Fourteen of these species consumed food in the CA, compared to only four in the FA (Table 1).
The most abundant species under all conditions in the Feeding Area and Control Area were Abudefduf saxatilis and Haemulon aurolineatum, respectively (Figure 3).
In the FA, the abundance of A. saxatilis differed significantly between different experimental conditions (F = 80.89, df = 3, p <0.05) and areas (F = 76.62,df = 1, p <0,05).Species abundance was significantly greater under feeding conditions in both the FA and CA.The density of A. saxatilis was significantly greater in the presence of the bather than in the absence of a bather in the FA.Densities did not differ significantly between the other conditions (Figure 4, Table 2).
The abundance of A. saxatilis in the FA was shown to significantly greater under experimental conditions involving human presence (Figure 4, Table 2).Abundance in the FA was over twice that in the CA in presence of feeding.
Opportunistic behavior was observed in herbivorous species such as Stegastes fuscus, Scarus trispinosus, Sparisoma axillare, Acanthurus chirurgus, A. coeruleus, and A. bahianus, which consumed shrimp.In the FA, in the presence of a bather (without feeding) and in the presence of feeding, large numbers of A. saxatilis approached the bather, migrating from different areas of the pool to the experiment site.In contrast, in the CA, besides not attracting this species, the presence of the bather scared away other species.Aggressive behavior was observed in A. saxatilis during feeding in the FA, consisting of agonistic interactions (biting and chasing other fish) with conspecific individuals and other species, such as Stegastes fuscus, Acanthurus chirurgus, and Scarus trispinosus.A. saxatilis also directed attacks at the bather, biting his hands during feeding.This type of aggressive behavior did not occur in the CA.
The grouping of A. saxatilis promoted by feeding in the FA may lead to "thefts" from S. fuscus' algae farms.It is possible that the abundance of A. saxatilis leads to increased agonistic behavior from individuals of the species S. fuscus, which leave their algae farms unprotected when they chase intruders.In such situations, conspecific individuals and other species, such as A. saxatilis, Acanthurus bahianus, and A. chirurgus, feed on the cultivated algae.In some cases, individuals of the territorial species S. fuscus stopped defending their territories to consume the food provided by the bather, thus enabling other fish to "steal" from their farms.

Discussion
The number of species observed consuming food shows that fish feeding directly affects one third of the species of reef fish found in the RFMP.In a study conducted in MPAs in the southeast of Kenya, the number of species that consumed bread fed by tourists (14 species) was identical to that found by the present study (Hémery & McClanahan 2005).The fish identified by this study also included individuals from the families Pomacentridae, Acanthuridae and Labridae, suggesting that fish feeding can affect ichthyofauna in different geographic regions in similar ways.However, the lack of experimental studies does not allow comparisons to be made with other regions.
The low species richness among individuals that consumed food provided by the bather in the FA is probably due to the high abundance and aggressive nature of A. saxatilis in this area.It is also interesting to note that four of the species that consumed food in the CA did not consume food in the FA, despite being present in both areas.This finding corroborates the findings of other studies that suggest aggressive species tend to benefit more from fish feeding by excluding nonaggressive species (Perrine 1989, Orams 2002).
The results also suggest that feeding promoted an increase in the abundance of A. saxatilis in the FA.Although studies conducted prior to the introduction of tourism to this tidal pool do not exist, this hypothesis is supported by the behavioral changes observed in this study and by the findings of other studies (Medeiros et al. 2007, Ilarri et al. 2008, Feitosa et al. 2012).A. saxatilis is an abundant species in tropical reefs in the Atlantic Ocean (Humann & Deloach 2002) and is considered to be a generalist and opportunistic due to its omnivorous diet (Deloach 1999).A. saxatilis has also been reported to be the most abundant species in feeding areas in other reefs in the northeast of Brazil, including Picãozinho (Medeiros et al. 2007, Ilarri et al. 2008) and Maragogi (Feitosa et al. 2012), where feeding was pointed out by the authors to be the primary cause of the increased abundance of this species.Two species of the genus Abudefduf (A. sexfasciatus and A. sparoides) were also found to be the most abundant species in feeding areas in the Malindi, Watamu, and Mombasa MPAs in southeastern Kenya (Hémery & McClanahan 2007).
Our findings indicate that the intensification of fish feeding activities in MPAs can lead to an increase in the size of populations of Abudefduf and other generalist fish species, leading to a corresponding increase in the number of omnivores in areas where feeding activity is frequent.It is known that fish feeding is associated with higher fish survival and reproduction rates (Sweatman 1996), favoring an increase in the size of populations of fish that consume the food provided during feeding.However, the impact of these increases in population size on the structure of reef communities remains unclear.Therefore, monitoring these fish populations is essential to ensuring the effective management of MPAs.Our results support the theory proposed by other authors that suggests that fish feeding is the leading cause of the increased abundance of omnivores associated with increases in the size of the population of A. saxatilis (Medeiros et al. 2007, Ilarri et al. 2008, Feitosa et al. 2012).The results of this study should therefore be used to inform tourism management planning in the RFMP and in other MPAs where fish feeding is permitted.
The behavioral findings of this study suggest that fish feeding also causes changes in the natural behavior of certain species.A. saxatilis showed a change in behavior in the FA, approaching the bather even in the absence of feeding.This suggests a change in behavior conditioned by human presence (visual stimulus) and the presence of food (chemical stimulus).According to Bond (1979), behavioral changes can be caused by both visual and olfactory stimuli and may lead to differences in fish distribution patterns.The behavior of A. saxatilis towards people has been studied in feeding areas in reefs in the states of Paraiba (Medeiros et al. 2007, Ilarri et al. 2008, Feitosa et al. 2012) and Pernambuco (Feitosa et al., 2012) in the Northeast Region of Brazil.Furthermore, A. sexfasciatus showed the same pattern of habituation to the presence of humans in feeding areas in MPAs in Kenya (Hémery & McClanahan 2007).The conditioning of A. saxatilis may be explained by the learning ability of fish, as reported by Shettleworth (1984).In extreme cases, conditioning of the behavior of fish to the presence of humans can lead to dependence on fish feeding (Harriot 2002).
We also observed aggressiveness among individuals of the species A. saxatilis in the FA, evident in frequent agonistic interactions in competition for food.It is believed that fish feeding may cause increased aggression and changes in the genetic structure of populations due to the possible natural selection of more aggressive fish (Moribe 2000, Semeniuk & Rothley 2008, Hammerschlag et al. 2012).Furthermore, biting during agonistic interactions may cause skin lesions, making the injured fish more susceptible to infections (Brookhouse et al. 2013).
Changes in fish behavior due to intense fish feeding activities have been reported by other studies.At the Shark Reef Marine Reserve in Fiji, fish feeding has been shown to cause shifts in the movement patterns of the bull shark (Carcharhinus leucas), which attended the feeding area with increasing frequency over time, leading to changes in habitat loyalty (Brunnschweiler & Barnett 2013).In the Cayman Islands, daytime fish feeding activities led to a reversal of the feeding patterns of the naturally nocturnal southern stingray (Hypanus americanus) and affected their spatial distribution, leading to unnatural grouping (Corcoran et al. 2013).Thus, our findings are consistent with the behavioral changes observed in other studies conducted with other species in other regions around the world.
Our findings also show that species that are endemic to Brazil, such as S. trispinosus, S. axillare, H. brasiliensis, and S. fuscus, consumed the food provided.Besides being endemic, S. trispinosus is an endangered species from the family Labridae found in the South Atlantic (Padovani-Ferreira et al. 2012) and is considered extinct in other parts of Brazil, such as Arraial do Cabo in the State of Rio de Janeiro (Floeter et al. 2007).Labridae also plays an important role in controlling macroalgae populations in reef environments (Francini-Filho & de Moura 2008, Bonaldo et al., 2014) and thus affects the balance between coral reefs and macroalgae, which is highly important to the maintenance of healthy coral reefs (Ainsworth & Mumby 2015).Given the conservation status and role played by this family of fish, PMA management and conservation strategies should take into consideration the possible effects of fish feeding on the biology of these fish.
The change in the feeding behavior of the herbivores S. fuscus, Sc. trispinosus, S. axillare, A. chirurgus, A. coeruleus and A. bahianus, indicates that fish feeding may cause changes in food preferences.Herbivorous fish are essential for maintaining the resilience of reef environments (Bellwood et al. 2004, Mumby et al. 2006).Therefore, considering the importance of the ecosystem services provided by these herbivores, further research should be conducted into the effects of fish feeding on the food preferences of these species.
Biota Neotrop., 18(3): e20170339, 2018 http://www.scielo.br/bnhttp://dx.doi.org/10.1590/1676-0611-BN-2017-0339 Fish feeding in RFMP may also be affecting the health of these animals by increasing fat deposition around vital organs (Moribe 2000, Orams 2002), susceptibility to microbial infections, stomach ulcers and ectodermal parasites, and skin lesions (Brookhouse et al. 2013;Semeniuk & Rothley 2008).Furthermore, the grouping of animals caused by fish feeding may favor the spread of diseases among conspecific individuals and other species (Orams 2002).These and other effects of fish feeding on fish health should be taken into account to ensure the effective management of MPAs.
Our findings suggest that changes have occurred to the population of A. saxatilis, which are likely to resonate throughout the entire structure of the ichthyofauna.The present study highlights behavioral changes among reef fish resulting from fish feeding activities in the RFMP.These changes include habituation to human presence, conditioning to fish feeding, increased aggressiveness, attacks on humans, and shortterm changes in species distribution.To prevent the intensification of changes caused by fish feeding, the competent authorities should take appropriate steps to control the quantity and frequency of feeding and monitor the dynamics of affected fish populations.

Figure 1 .
Figure 1.Location of the of Recife de Fora Marine Park (Porto Seguro, Bahia, Brazil).The letters indicate the tide pools: "A" = Dolphin pool, "B" = Visitors' pool, "C" = Panam pool, not included in this study.

Figure 2 .
Figure 2. Experimental conditions in the Recife de Fora Marine Park.

Figure 4 .
Figure 4. Average abundance of Abudefduf saxatilis (± SD) under different experimental conditions in the Feeding and Control Areas of Recife de Fora Marine Park.Different letters indicate a significant difference between treatments according to the results of the Tukey post hoc test.

Table 1 .
Fish species registered under feeding conditions in the Feeding Area and Control Area.Recife de Fora Marine Park, Bahia, Brazil."Consumed" indicates species that consumed the food provided.

Table 2 .
Tukey's post hoc test results for Abudefduf saxatilis showing the association between abundance and the factors experimental condition and area.Recife de Fora Marine Park, Bahia, Brazil.
FA = Feeding Area; CA = Control Area; NS = not statistically significant