Protecting nursery areas without fisheries management is not enough to conserve the most endangered parrotfish of the Atlantic Ocean

Marine protected areas (MPAs) are a primary strategy for marine conservation worldwide, having as a common goal the protection of essential habitats to enhance fish population recovery. However, MPAs alone may not be effective because species are not isolated from critical impacts occurring outside their boundaries. We evaluated how protecting critical nursery habitats affect the population of an important fishing target, using a 6-year database to predict juvenile hotspots and estimate population trends of the endemic and endangered parrotfish Scarus trispinosus within a mosaic of MPAs at the Abrolhos Bank, NE Brazil. We found that important nursery habitats are within no-take areas, but both juvenile and adult populations still show a declining trend over time. MPAs failed to ensure population maintenance and recovery likely due to overfishing in adjacent areas and the lack of compliance to management rules within multiple-use and within no-take MPAs. MPAs alone are not enough to protect ecologically important endangered species, but is still one of the only conservation strategies, particularly in developing countries. Our results shed light on the need for a wider adoption of more effective conservation policies in addition to MPAs, both in Brazil and in countries with similar governance contexts.


Material and methods
Study area. We sampled 28 reefs including no-take and multiple-use reserves with depths varying from 0.5 to 27 m between 2003 and 2008 in the Abrolhos Bank, eastern Brazil (16° 40′-19° 40′ S, 39° 10′-37° 20′ W; Fig. 1). The region consists of a wide enlargement of the continental shelf (46,000 km 2 ) that shelters the largest and richest reef environment in the South Atlantic 27,28 . Four different types of Marine Protected Areas (MPAs) have been established in the region, covering an area of about 6250 km 2 : Corumbau Marine Extractive Reserve, a co-managed MPA with multiple-uses created in 2000, which comprises Itacolomis reefs (1); Cassurubá Extractive Reserve, a co-managed MPA with multiple-uses created in 2009; Ponta da Baleia Reserve, a large multiple-use MPA created in 1993, which comprises Parcel das Paredes (3) and Sebastião Gomes reefs (4), and is considered as a "Paper Park" due its lack of proper management 28 ; and the Abrolhos Marine National Park, a no-take MPA crated in 1983, which comprises two distinct portions, one inshore and poorly enforced (Timbebas reefs [2]) and another offshore and more intensively enforced (Abrolhos Archipelago [5] and Parcel dos Abrolhos reefs [6]; Fig. 1).
Fish and benthic surveys. In each of the 28 sites, juveniles (i.e. individuals ≤ 10 cm of total length, TL, around 4 months old 22 ) and adults (i.e. individuals > 10 cm of TL) of S. trispinosus were counted through a nested stationary visual census technique 29 . The density of S. trispinosus in each visual census was calculated by the equation N/π × r 2 , where N is the number of individuals counted and π × r 2 is the sampling area, which r = 2 (12.6 m 2 ) for juveniles and r = 4 (50.2 m 2 ) for adults. Field protocols were approved by the Brazilian legislation (ICMBio-MMA-Brazilian Ministry of Environment) and sampling was carried out using observational and non-destructive techniques, under the permit SISBIO-11709-1. The fish surveys were conducted in the inshore reefs between 2003 and 2008, and in the offshore reefs between 2005 and 2008, totalling 3987 visual censuses (details in Supplementary Table S1). All surveys were conducted during the austral summer (January-April). Benthic cover data was obtained from Francini-Filho et al. 27 and used as predictors for juveniles' distribution. Seven morpho-functional groups of benthic organisms were used: turf algae (i.e. epilithic algal matrix), crustose calcareous algae (CCA), fire corals (milleporids), fleshy macroalgae, sponges, stony corals (scleractinians) and zoanthids.
Data analysis. Density of juveniles and adults. We assessed differences in the density of juveniles and adults among areas using permutation-based analysis of variance (Permutation-based ANOVA), which does not require normality or homogeneity of variances, using the package "lmperm" 30 (aovp function) in R software 31 . We ran a post hoc pairwise permutational test to assess the significant contrasts, using the package "rcompanion" 32 (pairwise PermutationTest function). In order to calculate mean densities per location, we aggregated the abundance data from all visual censuses with all years combined (n = 3987; mean densities = N individuals/N visual censuses per location). Mean densities were plotted using the package "yarrr" 33 (pirateplot function).
Modelling juveniles' distribution. In order to identify the hotspots of S. trispinosus juveniles, we used hierarchical Bayesian hurdle spatio-temporal models. These types of models are implemented to deal with high numbers of zero in the response variable (S. trispinosus' juveniles), in two stages: (1) modelling presence/absence in order to obtain the envelope of the predicted probability of presence of the species studied (binomial distribution) and (2) modelling the juveniles' density (Gamma distribution; Shapiro and Kolmogorov-Smirnov normality tests, p-value ≤ 0.001) of the studied species only in areas where species were predicted to be present 34 . For both stages, the explanatory variables included all environmental and benthic variables, the observer random effect, the year factor and a spatially structured random effect that account for the spatial autocorrelation. The models were per-  31 . For fixed parameters, vague priors were assigned with zero mean and a variance of 100. Variable selection was performed beginning with all possible interaction terms, but only the best combination of variables was chosen. Such choice was based on three different measures: (1) the Watanabe-Akaike information criterion (WAIC), (2) the Root Mean Square Error (RMSE), and (3) the adjusted coefficient of determination (R 2 ). The best (and most parsimonious) model was chosen based on the compromise between low WAIC values, low RMSE values, and high R 2 values, containing only relevant predictors, i.e., those predictors with 95% confidence intervals not covering zero. Functional responses were plotted using "ggplot2" package 36 in R software. In addition to the benthic cover, five environmental variables were also considered as potential predictors of juveniles' distribution 37 : Sea Surface Temperature (SST in °C) and Sea Surface Salinity (SSS in PSU), depth (in meters), rugosity (low = 0 vs high = 1) and distance to land (in meters; details in Supplementary Methods: Environmental variables).
Predictions were obtained for each year from 2003 to 2008 using the annual mean of the selected environmental and benthic variables. In addition, an average map of the entire period was also generated using the environmental means of the entire time series as predictors (see details of the temporal variation of SST and SSS in Supplementary Figs. S1, S2). Prediction validation was performed using two separated approaches. Firstly, the predicted and observed values using the full dataset were compared. Secondly, a tenfold cross validation using a random half of the dataset was performed to build the model and the remaining data to test the prediction. In both cases two statistics were calculated: Pearson's correlation coefficient r and the average error (AVEerror 38 ).
Modelling abundance trends. Abundance trends of both juveniles and adults were modelled using Bayesian time series models. Sites were aggregated within each location, and adults' and juveniles' abundance trends were modelled separately for each location, as well as for the entire region.
Three different models were used in each case: (1) an autoregressive model of order 1 (AR1), (2) a random walk model of order 1 (RW1) and (3) a random walk model of order 2 (RW2) 39 . For all models a Poisson distribution was implemented and the linear predictor was linked to the mean using the natural logarithm. Comparison among models were performed using the Watanabe-Akaike information criterion (WAIC) and Conditional , with significant differences between Timbebas reefs (2) and Parcel dos Abrolhos (6; permutation test, p < 0.05). The lowest densities were recorded in the Itacolomis (1) and Sebastião Gomes reefs (4), with significant differences between these two locations (permutation test, p < 0.05, Fig. 2B).  Supplementary  Fig. S3). Based on juveniles' density, the best-fitted model predicted their preferred habitat requirements, which were: higher rugosity, CCA, turf, zoanthids, fleshy macroalgae and sponges cover, and a SSS optimum value around 37.05 PSU (Fig. 4). Similarly to the SSS, fire corals and stony corals also presented an optimum range of correlation with the juveniles, with optimum values around 6% of fire corals cover and 30% of stony corals cover (Fig. 4). The best-fitted model also included the year factor that account for the temporal variability and the random spatial effect that account for the spatial intrinsic variability of the data (Supplementary Table S2).  (1); TIMB Timbebas reefs (2); PPAR Parcel das Paredes reefs (3); SEBG Sebastião Gomes reefs (4); ARCH Abrolhos Archipelago (5) Supplementary Table S1).

Discussion
Many studies show the benefits of Marine Protected Areas (MPAs) in conserving and restoring marine biodiversity 1,13,43 . Despite their critical importance, in some cases MPAs alone are not an effective strategy because species are not protected from critical impacts outside the MPAs boundaries 44 . We found that despite two out of the three most important nursery habitats for Scarus trispinosus being within no-take areas of a MPA (i.e. Timbebas [2] and Parcel dos Abrolhos reefs [6]; Fig. 3), both juvenile and adult populations showed declining trends over time within such areas. The declining trends suggest that the excessive removal of adult individuals from the population outside no-take MPAs and inside multiple-use reserves may be decreasing the generation of new recruits. Both results combined reinforce that, despite being essential, MPAs cannot be the only conservation tool to protect an endangered species targeted by fisheries. In such cases, specific conservation policies in addition to MPAs must be adopted. The density of S. trispinosus' juveniles was considerably higher in reefs closer to the coast (i.e. Timbebas [2] and Parcel das Paredes reefs [3]; Fig. 2A), which consist of shallow (1-18 m deep) and structurally complex reefs with a characteristic form of mushroom-shaped pinnacles 28 . Inshore shallower reefs and other coastal marine habitats are important nursery grounds for many reef fishes, because these habitats have more food resources and refuges for juveniles, and lower predation risk compared to adult habitats 2,45,46 . Due to the wide expanse of the continental shelf in the Abrolhos region, some shallow reefs occur even far from the coast 28 . This is the case of the reefs at Parcel dos Abrolhos (6), located 60 km from the coast, which are among the habitat hotspots for juveniles. Despite being slightly deeper, Parcel dos Abrolhos (6) harbours similar habitat conditions compared to the hotspots in the inshore reefs, but with a lower juvenile density, likely due to the greater distance from the coast and slightly greater depths ( Fig. 2A; Fig. 3).
All benthic cover variables were important to determine the juvenile hotspots, likely because the benthic morpho-functional groups used in the analysis are abundant and well distributed in all study sites 27 (Fig. 4). Reef attributes and the availability of preferred food sources are commonly within the main drivers of parrotfish distribution 47 . Some of the substrates selected by the best-fitted model are recognized as the main grazing targets of S. trispinosus, especially crustose coralline algae (CCA), turf and fleshy macroalgae 48 . In addition, fire and stony corals were important to sustain high abundance of S. trispinosus juveniles, indicating that the maintenance of healthy reefs and corals is critical in nursery habitats for this species. Although S. trispinosus may also feed on  www.nature.com/scientificreports/ stony corals, the species allocate only a small fraction of its bites to this type of substrate 49 . Therefore, it is more likely that the associations of juveniles with fire and stony corals relate to the structural complexity provided by these invertebrates, rather than the species feeding activity. Despite the physical and biological similarities among the three hotspots of juvenile abundance (Timbebas [2], Parcel das Paredes [3] and Parcel dos Abrolhos [6]), the enforcement is critically different among these areas. Although there was some temporal variation in the importance of the hotspots (Supplementary Figure S3), the location and protection level within each area were the same throughout the years. The offshore portion of the Abrolhos Marine National Park, that includes Parcel dos Abrolhos reefs (6), have a stronger enforcement due to the presence of the Federal Environmental Agency and the Brazilian Navy in the Abrolhos Archipelago. On the contrary, Timbebas reefs (2) are weakly enforced and poaching occurs frequently 12 . Even so, Timbebas reefs are one of the most important areas in the Abrolhos Bank in terms of abundance and biomass of small carnivores and large herbivores, including S. trispinosus 12 , indicating that the higher density of juveniles mainly emerged from habitat requirement and not necessarily from the MPA protection alone, since small-sized juveniles are not fishing targets. On the other hand, Parcel das Paredes reefs (3), which is the largest complex of inshore shallow reefs in the region 50 and one of the main nurseries areas, had a significantly lower juvenile' density compared to Timbebas (2), despite having habitat similarities ( Fig. 2A). Differently from Timbebas (2), Parcel das Paredes reefs (3) are located in a multiple-use reserve subjected to constant fishing pressure due to its lack of proper management. Therefore, the low juvenile density in Parcel das Paredes (3) can also result from high fishing pressure on adults, reducing the supply of juveniles in this area.
Almost two tons of S. trispinosus is fished per month by artisanal fisheries in the Abrolhos Bank 51 . Moreover, the species is one of the main targets of recreational spearfishers in the region 52,53 . Overfishing may affect the demographic structure of parrotfish populations by modifying their vital rates of mortality and inducing species to change sex at smaller age and sizes 54,55 . This is especially problematic for a relatively larger, late maturing and longer-lived hermaphrodite parrotfish such as S. trispinosus 21,22 . MPAs may prevent some of these effects by exporting larger older individuals to surrounding areas, and/or by receiving recruits that can fully develop to www.nature.com/scientificreports/ adult sizes within MPA boundaries 6,56 . Well-designed networks of MPAs can maintain source-sink population dynamics and its temporal variability if important habitats such as nurseries and breeding areas are effectively protected 5 . For S. trispinosus, which is known to undergo an ontogenetic habitat shift from inshore to offshore reefs, with juveniles being more abundant near the coast and mature adult individuals in deeper offshore reefs 21,22 , protecting inshore and offshore habitats is essential to maintain the species source-sink population dynamics. The Abrolhos Marine National Park (no-take) seems to be a good model of that, since it comprises both inshore and offshore portions. However, when fishing pressure outside the MPA is strong enough, the input of larvae decreases as the adult population declines 57 . Intense fishing pressure on S. trispinosus outside the MPA and the occasional poaching occurring inside the MPAs, related to lack of compliance and proper enforcement, may explain the declining pattern of adult and juvenile populations. The protection of the main nurseries habitats is not preventing the species from declining, probably because of the stronger fishing pressure on adults, both inside multiple-use reserves and unprotected reefs. In other words, the lack of protection for adults, which could exist through fisheries management in addition to MPAs, may be compromising the species' life cycle. Surprisingly, the offshore Parcel dos Abrolhos (6) was the only location that experienced population increases between 2006 and 2008 (Fig. 5). In addition to being a better enforced area, the population from Parcel dos Abrolhos (6) may be less susceptive to other coastal disturbances due to its larger distance from the coast, may acting as a refugium 58,59 for S. trispinosus. The structural complexity of Parcel dos Abrolhos reefs (6), which is similar to Timbebas (2) (5); PABR Parcel dos Abrolhos reefs (6). Different colors indicate different Marine Protected Areas. The shaded regions represent 95% credibility interval (CI). Please note that graphs within this plate are in different scales to accommodate and represent the within site variation. Graphs were plotted using the package "ggplot2" 36 in the R software 31 .
Scientific Reports | (2020) 10:19143 | https://doi.org/10.1038/s41598-020-76207-x www.nature.com/scientificreports/ of protection are deeper reefs in the east side of the Abrolhos Marine National Park. These areas aggregate high biomass of commercially important fishes, and consequently, fishing activities intensively occur in the area 60 .
Although the predominant occurrence in shallower waters, evidences indicate that larger mature individuals of S. trispinosus also occurs in offshore reefs between 50 and 60 m deep 21,60,61 , suggesting that the expansion of the Abrolhos Marine National Park toward deeper reefs may benefit the species, specially by protecting adult individuals. Besides that, proper management of Ponta da Baleia Reserve and the inclusion of no-take zones in this area could benefit inshore populations of Parcel das Paredes reefs (3), which is one of the juveniles' hotspots and remain poorly protected. Despite the encouraging expectations derived from expanding protection areas, our results suggest that expanding protected areas without setting clear rules for managing the areas, pursuing compliance among stakeholders and intensifying the enforcement inside no-take areas, may be insufficient to protect fishing-target endangered species such as S. trispinosus. This is the case of many regions worldwide where fishing-target species declined despite MPAs creation efforts 44,62 , including other iconic species, the bumphead parrotfish (Bolbometopon muricatum) in the Solomon Islands 63 . Even though MPAs may enhance the persistence of exploited parrotfishes 13 , we showed that the protection of the main nursery areas of S. trispinosus is not enough to prevent the declining trend in the juvenile's abundance, likely because of intensive fishing of adults in adjacent areas. In 2014, S. trispinosus fishing was nationally banned after the species was listed as endangered on the Brazilian Red List of Endangered Species (Decree No. 445), but with no significant enforcement. Most recently, in 2018, the Brazilian National Recovery Plan for endangered species (Decree No. 59-B) regulated the species fishing under restrict rules, which included the ban on recreational fishing and fishing nets, determined spearguns as the only fishing gear allowed and a slot-size limit for catches between 39 and 63 cm total length. The proposed slot-size limit aims to protect both immature and older mature individuals (including most males) which are those with a greater reproductive capacity, according to the species' demographic traits 21,22 . The plan also proposes that fishing would only be allowed within multiple-use marine protected areas by authorized artisanal fishers, with continuous monitoring. The Abrolhos bank seems to be a suitable region to enforce these measures, due to the presence of multiple-use reserves and the fact that most of the artisanal catches are within the proposed slot size limit 21 . Until the present moment, however, the government has not enforced the plan and the species keep being fished indiscriminately. Given the critical situation, efforts to implement the management measures should be taken as soon as possible, otherwise, the complete fishing ban will be the only way to aid the species recovery.
We are aware that our estimates would benefit from a longer timeseries, but given the significant declines and increasing threats reported to this species 19,[21][22][23] , evaluating population trends and the effectiveness of conservation measures is imperative. Therefore, in the absence of such longer term data we modelled the six-year time series using a robust method that provided reliable estimates accounting for this potential limitation. Also, one can argue that the temporal trends we observed could be due to other anthropogenic impacts rather than fisheries, such as climate change or invasive species. However, the Abrolhos bank did not experience any significant effect related to climate change or thermal anomalies 64 and had minimal variability in sea temperature and salinity within the time span of our study (see Supplementary Fig. S1 and Fig. S2). Also, the region did not experience problems related to invasive species back in the early 2000's. Recently, the expansion of the invasive coral species of the genus Tubastraea in the southern Abrolhos Bank 65 (more than 100 km from our study sites) raised concerns, but no invasive fish species was ever recorded in the region. Therefore, the strongest and most consistent impact on S. trispinosus populations is fishing pressure, which removes about two tons of the species from the Abrolhos Bank monthly 51 . Conservation actions targeting this species should focus on fisheries management, enforcement and compliance to guarantee its long-term survivorship.
The creation process of new MPAs and fishery management plans must involve fishers and other stakeholders in order to reach compliance and avoid the dissemination of paper parks 66 . Parrotfish populations, including S. tripisnosus at the Abrolhos Bank, and other important fishing-target and endangered species elsewhere, will only benefit from MPA networks 1,43 if it comes associated to other conservation strategies, including a wellimplemented demographic-based fisheries management plan that considerably reduce fishing pressure.

Data availability
The data and codes are available at https ://zenod o.org/recor d/39643 27#.XyF1W Z5Kgd W.