Philippa Kaur

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
and JavaScript. More

In this International Year of Sustainable Mountain Development, we call for transdisciplinary research by Chilean scientists and for concerted action among all stakeholders to address the complex factors responsible for the degradation of Chile’s mountains. Mountains cover 64% of Chile’s surface and are a crucial source of water, food, energy, minerals and biodiversity.
Competing Interests
The authors declare no competing interests. More

None of the 21 targets of the United Nations’ Sustainable Development Goals (SDGs) set for 2020 was achieved. But, by our calculations, the target to protect 10% of the global ocean area (SDG14, target 5) could become a reality this year.
Competing Interests
The authors declare no competing interests. More

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain
the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in
Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles
and JavaScript. More

Data preparationHistorical dataThe primary source of data is historical parish registers, which have been transcribed under the supervision of many of the study authors over a number of decades, primarily for evolutionary demographic research. Our dataset (Supplementary Data 1) includes nine European populations, including some for which the positive relationship between maternal lifetime twinning status and maternal total births has been described previously5,6,8,10. Details for the populations used in this study are given in Table 1 and in Supplementary Table 14. The sourcing of each dataset and the socio-ecological background of each population have already been described in previous studies (see Table 1 for references). Overall, there is no reason to suspect a high level of consanguinity in these populations62, so our analyses do not account for the variable level of relatedness between individuals. The datasets cover pre-industrial periods in which the lifetime reproductive success was high and the majority of people were living and working in agrarian communities, except for the Samis (from northern Finland and Sweden) who made their living fully or in part from a combination of herding, fishing and hunting. The smooth decline of the probability of parity progression with parity (Fig. 4a) suggests that mothers did not effectively limit their reproductive success with the aim of achieving a small family size, as found in populations that have undergone a demographic transition.Data selectionWe use the term family to describe a mother and all individuals to whom she gave birth over her life. For our analyses, all families considered met the following criteria: the mother’s age was known at a monthly resolution and her life course traced until at least age 45 (approximating full reproductive life), the birth year and month of all offspring must have been recorded and consecutive births were all at least nine months apart from one another. In the case of one population (Norway) and of a few observations in the other populations, the month of birth was not available. These data were thus not considered in the results presented in main text because some of our analyses require an accurate estimation of the interbirth interval. Most analyses are thus based on data from eight populations. Nevertheless, the slope of the negative relationship between twinning and total birth remained very similar irrespectively of whether or not such data were included (Supplementary Fig. 5), which suggests that the exclusion of Norwegian data does not alter our main conclusions. Information on the populations considering also the data for which the birth months were missing is provided in Supplementary Table 14.Twin identificationIn our data, the maximum number of offspring to constitute a multiple birth was three. We use the term twin(s) to refer to offspring who were the result of the same multiple birth (including 1745 sets of twins sensu stricto and 19 sets of triplets in the filtered dataset and, respectively, 1915 and 20 in the dataset including observations that lack birth months). Although twins are sometimes explicitly indicated in the data sources, this is not always the case. Thus, for the sake of consistency across our populations, twin births were identified when at least two individuals born to the same mother appeared with similar birth dates, according to strict criteria: if the exact birth dates were available, then offspring were identified as twins if their birth dates were no more than one day apart. If the exact birth dates were not available, then an identical birth year and month were considered sufficient for positive twin identification.Analyses and simulationsCharacterisation of the relationship between twinning and fertilityWe began by characterising the relationship between lifetime twinning status and maternal total births by fitting two models. For the first, we used a Generalised Linear Mixed-effects Model (GLMM) to investigate whether the mothers of twins (twinners) had experienced a larger or smaller number of births than mothers who only had singletons (non-twinners). We fitted this GLMM on the mother-level data with the R package spaMM63 using the call:$$begin{array}{c}{{{rm{fitme}}}}left({{{{rm{births}}}}}_{{{{rm{total}}}}}sim 1+{{{rm{twinner}}}}+(1|{{{rm{pop}}}}),right.\ {{{rm{data}}}}={{{rm{mother}}}}_{{{rm{level}}}}_{{{rm{data}}}},\ left.{{{rm{family}}}}={{{rm{Tnegbin}}}}({{{rm{link}}}}={hbox{“}}{{{mathrm{log}}}}{hbox{”}})right)end{array}$$
(1)
The response variable births_total refers to all births recorded over a mother’s observed lifetime (count data). The term 1 informs the function to fit an intercept (which happens by default, but is indicated here for clarity). The predictor twinner refers to maternal lifetime twinning status (binary: twinner vs non-twinner) and is modelled as a fixed effect. The term pop refers to the population identity (qualitative variable with eight levels) and is modelled by a Gaussian random effect acting on the intercept, which allows for the modelling of the heterogeneity between populations that is not captured by the fixed effects. The argument family is used to define the error structure and the link function of the GLMM (more on this below).In a second model, we reversed which variable is used as a response and which is used as the fixed-effect predictor. This allowed us to analyse how maternal total births predicted the probability of a mother producing twins during her lifetime using the call:$$begin{array}{c}{{{{{rm{fitme}}}}}}left({{{{{rm{twinner}}}}}} sim 1+{{{{{rm{births}}}}}}_{{{{{rm{total}}}}}}+(1|{{{{{rm{pop}}}}}}),right.\ {{{{{rm{data}}}}}}={{{{{rm{mother}}}}}}_{{{{{rm{level}}}}}}_{{{{{rm{data}}}}}},\ left.{{{{{rm{family}}}}}}={{{{{rm{binomial}}}}}}({{{{{rm{link}}}}}}={hbox{“}}{{{{{rm{logit}}}}}}{hbox{”}})right)end{array}$$
(2)
The fitted models 1 and 2 are depicted in Fig. 1a, b and Supplementary Tables 1,2. While models 1 and 2 represent two sides of the same coin, the fit of both models is justified because each model formulation provides complementary information: expressing the effect of twinning on fertility relates to previous work5,6,7,8,9,10,57 and expressing the effect of fertility on twinning is a first step toward identifying what shapes twinning propensity, the focus of this paper.For the model predicting total births (model 1), we chose to use a negative binomial error structure. Using this error structure produced a fit of the data that was better than a (truncated) Poisson—the usual alternative for count data—as evidenced by much smaller marginal and conditional AIC values64. Here we specifically used a truncated negative binomial distribution because the data do not possess zeros by construction (only mothers are present in the dataset, i.e. there are no nulliparous women). For the model predicting lifetime twinning status (model 2), we chose a binomial error structure which is appropriate for binary data.Modelling the proportion of twin births among all births per mother is an effective way to avoid biases caused by differences in exposure to the risk of having twins affecting the relationship between twinning and fertility. For this, we fitted the following third model:$$begin{array}{l}{{{{{rm{fitme}}}}}}left({{{{{rm{cbind}}}}}};({{{{{rm{twin}}}}}}_{{{{{rm{total}}}}}},{{{{{rm{singleton}}}}}}_{{{{{rm{total}}}}}})right.\ sim ,1+{{{{{rm{births}}}}}}_{{{{{rm{total}}}}}}+(1|{{{{{rm{pop}}}}}}),\ ,{{{{{rm{data}}}}}}={{{{{rm{mother}}}}}}_{{{{{rm{level}}}}}}_{{{{{rm{data}}}}}},\ ,left.{{{{{rm{family}}}}}}={{{{{rm{binomial}}}}}};({{{{{rm{link}}}}}}={hbox{“}}{{{{{rm{logit}}}}}}{hbox{”}})right)end{array}$$
(3)
In this model, the variable twin_total refers to the mother’s total number of twin births (i.e. one for each twinning event), singleton_total refers to the lifetime number of singleton births, and the cbind() function serves to indicate the fitting function to model the frequency of twinning events based on these two variables, which is interpreted as number of successes and failures of a binomial experience. The fitted model 3 is depicted in Fig. 2 and Supplementary Table 3.We modified model 3 so as to test whether the effect of total births differed significantly between populations. To do so, we considered that the effect of populations on total births could either be modelled as an interaction between fixed effects or as a random slope. For the former representation, we thus compared the fit of a model with linear predictor structure defined in spaMM as 1+births_total*pop to that of a model with the structure 1+births_total+pop. For the latter representation, we compared the fit with linear predictor structure 1+births_total + (1|pop) (i.e. model 3 as introduced above) to that of 1+births_total + (1+births_total|pop). We performed this testing procedure by comparing the likelihood ratio between each pair of alternative fits to the expectation of such a ratio under the null hypothesis. The distribution of the statistics used for the test was computed using 1000 parametric bootstrap replicates, which we generated using the function anova() provided by spaMM63. The test revealed a small non-significant variation in slopes between populations (see Results). For the sake of simplicity, we thus considered the effect of births_total the same across populations in all other analyses.Modelling life-history events using GLMMsTo reveal the biological mechanisms responsible for the relationship between twinning and fertility, we first fitted statistical models describing how age, parity and twin/singleton status, as well as individual and population differences influenced three key life-history events: parity progression (PP), the duration of interbirth intervals (IBI) and the twinning outcome of births (T). These models were fitted on birth-level data by the following calls:$$begin{array}{l}{{{{{rm{fitme}}}}}}left({{{{{rm{PP}}}}}} sim 1+{{{{{rm{twin}}}}}}+{{{{{rm{poly}}}}}}({{{{{rm{cbind}}}}}}({{{{{rm{age}}}}}},,{{{{{rm{parity}}}}}}),,{{{{{rm{best}}}}}}_{{{{{rm{order}}}}}})right.\ ,+,(1|{{{{{rm{maternal}}}}}}_{{{{{rm{id}}}}}})+(1|{{{{{rm{pop}}}}}}),\ ,{{{{{rm{data}}}}}}={{{{{rm{birth}}}}}}_{{{{{rm{level}}}}}}_{{{{{rm{data}}}}}},\ ,{{{{{rm{family}}}}}}={{{{{rm{binomial}}}}}};({{{{{rm{link}}}}}}={hbox{“}}{{{{{rm{logit}}}}}}{hbox{”}})end{array}$$
(4)
$$begin{array}{l}{{{{{rm{fitme}}}}}}left({{{{{rm{IBI}}}}}} sim 1+{{{{{rm{twin}}}}}}+{{{{{rm{poly}}}}}};({{{{{rm{cbind}}}}}}({{{{{rm{age}}}}}},,{{{{{rm{parity}}}}}}),{{{{{rm{best}}}}}}_{{{{{rm{order}}}}}})right.\ ,+,(1|{{{{{rm{maternal}}}}}}_{{{{{rm{id}}}}}})+(1|{{{{{rm{pop}}}}}}),\ ,{{{{{rm{data}}}}}}={{{{{rm{birth}}}}}}_{{{{{rm{level}}}}}}_{{{{{rm{data}}}}}},\ ,left.{{{{{rm{family}}}}}}={{{{{rm{negbin}}}}}}({{{{{rm{link}}}}}}={hbox{“}}{rm log} {hbox{”}})right)end{array}$$
(5)
$$begin{array}{l}{{{{{rm{fitme}}}}}}left({{{{{rm{T}}}}}} sim 1+{{{{{rm{poly}}}}}}({{{{{rm{cbind}}}}}}({{{{{rm{age}}}}}},,{{{{{rm{parity}}}}}}),,{{{{{rm{best}}}}}}_{{{{{rm{order}}}}}})right.\ ,+,(1|{{{{{rm{maternal}}}}}}_{{{{{rm{id}}}}}})+(1|{{{{{rm{pop}}}}}}),\ ,{{{{{rm{data}}}}}}={{{{{rm{birth}}}}}}_{{{{{rm{level}}}}}}_{{{{{rm{data}}}}}},\ ,left.{{{{{rm{family}}}}}}={{{{{rm{binomial}}}}}};({{{{{rm{link}}}}}}={hbox{“}}{{{{{rm{logit}}}}}}{hbox{”}})right).end{array}$$
(6)
The response variables of models 4, 5 and 6 are thus PP, IBI and T, which refer to whether the mother went on to reproduce again or not (a boolean), the duration of the interbirth interval between the focal birth and the next (a discrete number of months) and whether the birth resulted in twins or not (a boolean), respectively. In addition to the terms that have already been defined, we now have the term poly(cbind(age, parity), best_order) to code for a polynomial describing the effect of maternal age, parity and their possible interaction. The two-variable polynomial function was applied on maternal age (with a monthly resolution) and parity (i.e. the current birth rank). Such a polynomial term allowed us to explore the influences of maternal age and parity on each response variable while encompassing the non-linearity of these predictors. We also have the predictor variable twin, which is a boolean that indicates if the previous birth event of a given mother resulted in twins or not (the variable twin and T are the same, but we used two different names to clarify when it is used as a response or as a predictor). Finally, we have the random effect “maternal identity” (maternal_id), which is used to represent intrinsic variation among mothers, that is, heterogeneity of expected response among individuals, beyond that due to the fixed effects and the population random effect. This random effect therefore measures maternal intrinsic fertility (in models 4 & 5) and twinning propensity (in model 6).To determine the best polynomial order (best_order) for the polynomial term we attempted orders from 0 to 6 and selected, for each model, the order leading to the model fit associated with the smallest marginal AIC. A polynomial of order 6 is sufficient to fit very complex shapes. Polynomial orders obtained by this procedure are given in the summary tables of the model fits given in Supplementary Tables. Importantly, maternal age and parity are highly correlated together (Spearman’s rho = 0.69), unequally correlated to response variables and exert non-linear effects. These are precisely the conditions in which collinearity issues are the most severe65. This justifies why we considered them jointly in all statistical models, as well as why we did not attempt to partition their respective biological effects in our analyses (except for the visual representation in Fig. 4).In order to study how the lifetime twinning status influenced maternal age at first birth, we also fitted the following model:$$begin{array}{l}{{{{{rm{fitme}}}}}}left({{{{{rm{AFB}}}}}} sim 1+{{{{{rm{twinner}}}}}}* {{{{{rm{births}}}}}}_{{{{{rm{total}}}}}}_{{{{{rm{fac}}}}}}+(1|{{{{{rm{pop}}}}}}),right.\ ,{{{{{rm{data}}}}}}={{{{{rm{mother}}}}}}_{{{{{rm{level}}}}}}_{{{{{rm{data}}}}}},\ ,left.{{{{{rm{family}}}}}}={{{{{rm{negbin}}}}}}({{{{{rm{link}}}}}}={hbox{“}}{rm log} {hbox{”}})right)end{array}$$
(7)
In this model, the response variable AFB corresponds to the age at first birth expressed as a number of months (discrete data) and the predictor births_total_fac corresponds to a qualitative variable referring to maternal total births (10 levels: 1, 2, …, 9, 10 + ). We here considered a possible interaction between twinner and births_total_fac. We used the negative binomial family as in model 1 but as for model 5 there is no need to consider here the truncated form of the distribution. All other terms have already been defined. The fitted model is depicted in Supplementary Fig. 1 and Supplementary Table 7.Marginal predictions for GLMMsAll predictions shown in plots or given in text represent marginal predictions. This means that the predictions for the quantities of interest (maternal lifetime births, twinning probabilities and age at first birth) are a function of coefficients of the fixed effects, and of the variance of the random effects. To be more precise, we averaged, over the fitted distribution of random effects, the predictions expressed on the scale of the response (i.e. back-transformed from the scale of the linear predictor) and conditional on the fixed and random effects. Unlike the traditional conditional predictions computed for a specific value of the random effects (often 0), such computation provides unbiased predictions and should be favoured in the context of GLMMs where random effects act non-additively on the expected response (which is the case when the link function of the model is not identity66). We estimated 95% intervals for these marginal predictions (CI95%) using parametric bootstraps with the help of the function spaMM_boot() from the R package spaMM and boot.ci() from the R package boot. More details can be found by looking at the code of the functions compute_predictions() and compare_prediction() in our supporting R package twinR (see Code availability).Simulating the life history of mothersWe produced an individual-based simulation model of human female life history to investigate the contribution of four mechanisms to the relationship, shown in Fig. 2, between per-birth twinning probability and maternal total births—an approach generally known as pattern-oriented modelling67. Each simulation proceeds in the following way: first, we initialised the simulation with representations of the exact same mothers present in the observed dataset, setting their population and maternal identities as the real ones, their starting ages at the observed values for age at first birth and their parity to one. Following this initialisation, the virtual lives of mothers proceeded as multiple iterations of a sequence of three life-history events, informed by statistical models (see below) and subject to the hypothesis being tested (Supplementary Figs. 3, 4). Specifically, for each mother, the twin/singleton status (T) of the current birth was first determined using a GLMM predicting T. Then, whether or not she will go on to reproduce at least once more was determined by simulating her parity progression status (PP) using a GLMM predicting the parity progression probability. For mothers who do continue reproducing, we finally used a third GLMM to determine the length of the interval between the current birth and the next one (IBI). At each iteration, a mother’s parity is increased by one, and age is increased by the simulated length of the interbirth interval. All predictions were performed conditionally on the value for the predictor characterised by both fixed and random effects. The process of simulating PP, IBI and T was then reiterated until all women had ceased reproduction, which happens necessarily since the probability of parity progression is lower than one. We also set this probability to zero once women reached 60 years old to save computation time in particular simulation scenarios leading to unrealistic life histories (and bad goodness of fit). Note that the maximum recorded age at which a mother gave birth was 55.1 years in our data. For the same reason, we also capped the maximum simulated duration for the interbirth interval to 30 years.Drawing life-history events from the fit of the model formulas shown above for models 4, 5 and 6 corresponds to simulating the scenario PISH (i.e. all four hypothetical mechanisms are activated). For simulating other simulation scenarios, we had to fit additional GLMMs derived from models presented above. Specifically, the term twin was dropped from model 4 to deactivate mechanism P (model 8; Supplementary Table 8); the term twin was dropped from model 5 to deactivate mechanism I (model 9; Supplementary Table 9); the term poly(cbind(age, parity), best_order) was dropped from model 6 to deactivate the mechanism S (model 10 and 11; Supplementary Tables 10, 11); and the term (1|maternal_id) was dropped from model 6 to deactivate mechanism H (model 11 and 12; Supplementary Tables 11, 12).Testing candidate mechanisms using simulationsTo test how each mechanism or association of mechanisms influenced the relationship between twinning and fertility, we ran simulations under each possible set of activated or inactivated mechanisms. We tested all possible sets and we thus built a total of 42 = 16 simulation scenarios (Supplementary Figs. 3, 4).For each simulation scenario, we ran simulation replicates (see Supplementary Notes for details and information on the numbers of replicates), then fitted model 3 on the dataset produced by each replicate and extracted the estimate for the slope associated with the term births_total in that model (β*). We then consider, in turn, that each simulation scenario may have generated the data. Each simulation scenario is thus considered as a null hypothesis which we aim at testing. Such a test is traditionally referred to as a goodness-of-fit test68. The result of such a test, a p-value, answers the question: what is the probability of obtaining a value equal to, or more extreme than, the statistic of interest, if the null hypothesis were true? The rejection of the null hypothesis by the test (i.e. a p-value ≤ 0.05) signifies a rejection of the null hypothesis, and thus, here, the rejection of a simulation scenario which represents a particular mechanism, or combination of mechanisms. In contrast, a large (i.e. non-significant) p-value would here denote support for the simulation scenario under consideration.A first candidate, as a statistic of interest to build our goodness-of-fit test, is the slope β*. However, when viewed as a goodness-of-fit test, the direct comparison of the observed and simulated slopes may be conservative when other life-history parameters are fitted to the data. This is because the data tend to be more likely given parameter values fitted to the data than given the actual (unknown) parameter values that generated the data. The goodness-of-fit test is however only guaranteed to provide uniformly distributed p-values (a feature necessary for the correctness of any null hypothesis testing) when samples are drawn under the latter parameter values. This is a general issue in statistics which has also been discussed long ago, for example, when the data-generating process is the normal distribution and a Kolmogorov–Smirnov test of goodness-of-fit is applied69. We thus designed and validated a specific procedure to correct for such bias while testing each simulation scenario (Supplementary Notes). In the text, we only report outcomes from this unbiased goodness-of-fit test (for details, see Supplementary Notes and Supplementary Table 13).Studying the effect of twinning propensity on the number of offspring using simulationsTo study how twinning propensity influences the total number of offspring that mothers produced during their lifetime, we ran two sets of simulations, each with 100 replicates. In the first set, we ran the simulation as described in the section “Simulating the life history of mothers” using the fits of the models associated with the simulation scenario PIS (i.e. fits of models 4, 5 and 12). In the second set, we did the same, except that we modified the intercept of the model predicting twinning events (fit of the model 12) by adding 2.5 to its intercept. We also tried other values, some smaller (e.g. 0.25), some larger (e.g. 5), to make sure that the magnitudes of the change of the intercept did not impact our qualitative statements. For each set of replicates, we extracted the twinning rate, the twinner rate, the mean number of offspring produced and the mean total number of births. We report the means of these metrics, as well as the 95% Central Range from simulation replicates (CR95%), which we directly computed by extracting the corresponding quantiles from the distribution generated by the replicates.Realism of the simulationsWe checked that the simulated life history closely matched that of the real mothers represented in our dataset beyond what is captured by the relationship between twinning and fertility. To do so, we compared different metrics related to fertility and twinning between the real and simulated data. We chose to perform this comparison under the simulation scenario PIS since it produces the best goodness-of-fit. The results of this quality check confirm that our simulations represent the reproductive lives of the mothers appropriately (Supplementary Fig. 6).Studying the effect of mortality on the number of offspring using simulationsTo account for the fact that not all offspring have the same expected survival, we also applied a survival weight to each simulated offspring before averaging the numbers for a given simulation set (baseline twinning propensity or enhanced, see Results). We used as weights the estimates for the probability of offspring survival between birth and adulthood provided by two publications associated with some of the data we used there. Specifically, following Helle et al.8, we used a weight of 0.603 for twins, 0.838 for singletons from twinners and 0.815 for singletons from non-twinners. Alternatively, following Haukioja et al.3, we used a weight of 0.337 for twins and 0.706 for singletons from all mothers.Implementation detailsAll statistical analyses were performed in R version 4.170. The main R packages we used were spaMM63 version 3.9.40 for the fit of all the statistical models, boot71,72 version 1.3-28 for the computation of confidence intervals based on parametric bootstraps, and R673 version 2.5.1 for defining the object used to run the simulations. The DESCRIPTION file from our package twinR (see Code availability) lists the additional R packages required for this project (e.g. those used for plotting and data manipulation).Reporting summaryFurther information on research design is available in the Nature Research Reporting Summary linked to this article. More

In this study, we aimed to determine if tropical cyclones in the northwestern GOM differentially affected mesozooplankton community structure. We found that multivariate community structure varied between storm and non-storm years. However, among the three hurricanes, only the post-Harvey mesozooplankton communities were distinct from years where no storms occurred. Multivariate dispersion, a measure of variance within community structure, did not differ between storm and non-storm years. This result refutes our hypothesis that variability in zooplankton community structure would be higher during storm years opposed to non-storm years. Our prediction that post-storm mesozooplankton communities would differ from non-storm communities was supported, as was our expectation that mesozooplankton community structure varied among storms. We hypothesized that due to the major flooding and rainfall of Harvey, reduced salinity would likely be the main driver of mesozooplankton community differences relative to non-storm years. This expectation was partially met; differences in mesozooplankton community structure between Harvey and years with no storms were driven by salinity, stratification, and ‘distance-from-shore’ rather than solely salinity. This indicates that NWGOM zooplankton community structure varies holistically with biophysical conditions rather than being primarily driven by one or two dominating factors24. Moreover, we found that the presence of Hurricane Harvey, rather than temperature, explained the second greatest amount of variance in zooplankton community structure reflecting the importance of considering how complex disturbance mechanisms might compound and result in a unique ecological responses following a tropical cyclone. The overall PSEM showed that high salinity was directly associated with reduced fluorescence and depressed zooplankton abundance. Higher abundances were found to result in higher community dominance (i.e., lower evenness). Conversely, high salinities indirectly reduced zooplankton evenness via greater water column stratification. Lower fluorescence at higher salinities supports the spatial patterns in NWGOM phytoplankton identified by Kurtay et al.20 following Hurricane Harvey. Those authors observed declines in overall phytoplankton abundance and communities increasingly dominated by cells  More

Bowen, B. W., Rocha, L. A., Toonen, R. J. & Karl, S. A. The origins of tropical marine biodiversity. Trends Ecol. Evol. 28, 359–366 (2013).PubMed 
Article 

Google Scholar 
Lam, V. W. et al. Climate change, tropical fisheries and prospects for sustainable development. Nat. Rev. Earth Environ. 1, 440–454 (2020).ADS 
Article 

Google Scholar 
Halpern, B. S. et al. Recent pace of change in human impact on the world’s ocean. Sci. Rep. 9, 1–8 (2019).CAS 
Article 

Google Scholar 
Capitani, L., de Araujo, J. N., Vieira, E. A., Angelini, R. & Longo, G. O. Ocean warming will reduce standing biomass in a tropical western atlantic reef ecosystem. Ecosystems https://doi.org/10.1007/s10021-021-00691-z (2021).Article 

Google Scholar 
Lima, L. S. et al. Potential changes in the connectivity of marine protected areas driven by extreme ocean warming. Sci. Rep. 11, 1–12 (2021).Article 
CAS 

Google Scholar 
Sale, P. F. et al. Transforming management of tropical coastal seas to cope with challenges of the 21st century. Mar. Pollut. Bull. 85, 8–23 (2014).CAS 
PubMed 
Article 

Google Scholar 
Dunstan, P. K. et al. How can climate predictions improve sustainability of coastal fisheries in Pacific Small-Island Developing States?. Mar. Policy 88, 295–302 (2018).Article 

Google Scholar 
Martins, I. M. & Gasalla, M. A. Perceptions of climate and ocean change impacting the resources and livelihood of small-scale fishers in the South Brazil Bight. Clim. Change 147, 441–456 (2018).ADS 
Article 

Google Scholar 
Moura, R. L. et al. Spatial patterns of benthic megahabitats and conservation planning in the Abrolhos Bank. Cont. Shelf Res. 70, 109–117 (2013).ADS 
Article 

Google Scholar 
Lesser, M. P., Slattery, M. & Leichter, J. J. Ecology of mesophotic coral reefs. J. Exp. Mar. Biol. Ecol. 375, 1–8 (2009).Article 

Google Scholar 
Bryan, D. R., Kilfoyle, K., Gilmore, R. G. Jr. & Spieler, R. E. Characterization of the mesophotic reef fish community in south Florida, USA. J. Appl. Ichthyol. 29, 108–117 (2013).Article 

Google Scholar 
Fukunaga, A., Kosaki, R. K., Wagner, D. & Kane, C. Structure of Mesophotic Reef Fish Assemblages in the Northwestern Hawaiian Islands. PLoS One 11, e0157861 (2016).PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
Kahng, S., Copus, J. M. & Wagner, D. Mesophotic coral ecosystems. In Marine Animal Forests (eds Rossi, S. et al.) 1–22 (Springer International Publishing, Paris, 2016). https://doi.org/10.1007/978-3-319-17001-5_4-1.Chapter 

Google Scholar 
Rocha, L. A. et al. Mesophotic coral ecosystems are threatened and ecologically distinct from shallow water reefs. Science 361, 281–284 (2018).ADS 
CAS 
PubMed 
Article 

Google Scholar 
Bongaerts, P. et al. Deep reefs are not universal refuges: Reseeding potential varies among coral species. Sci. Adv. 3, e1602373 (2017).ADS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Rosa, M. R. et al. Mesophotic reef fish assemblages of the remote St. Peter and St. Paul’s Archipelago, Mid-Atlantic Ridge, Brazil. Coral Reefs 35, 113–123 (2016).ADS 
Article 

Google Scholar 
Medeiros, A. P. et al. Deep reefs are not refugium for shallow-water fish communities in the southwestern Atlantic. Ecol. Evol. 11, 4413–4427 (2021).PubMed 
PubMed Central 
Article 

Google Scholar 
Reid, D. G. SEFOS—Shelf edge fisheries and oceanography studies: An overview. Fish. Res. 50, 1–15 (2001).Article 

Google Scholar 
Heyman, W. D. & Kjerfve, B. Characterization of transient multi-species reef fish spawning aggregations at Gladden Spit, Belize. Bull. Mar. Sci. 83, 531–551 (2008).
Google Scholar 
Paxton, A. B. et al. Four decades of reef observations illuminate deep-water grouper hotspots. Fish Fish. 22, 749–761. https://doi.org/10.1111/faf.12548 (2021).Article 

Google Scholar 
Frédou, T. & Ferreira, B. P. Bathymetric trends of northeastern Brazilian snappers (Pisces, Lutjanidae): Implications for the reef fishery dynamic. Braz. Arch. Biol. Technol. 48, 787–800 (2005).Article 

Google Scholar 
Longhurst, A. R. & Pauly, D. Ecologia dos oceanos tropicais (Edusp, 2007).
Google Scholar 
Olavo, G., Costa, P. A., Martins, A. S. & Ferreira, B. P. Shelf-edge reefs as priority areas for conservation of reef fish diversity in the tropical Atlantic. Aquat. Conserv. Mar. Freshw. Ecosyst. 21, 199–209 (2011).Article 

Google Scholar 
Eduardo, L. N. et al. Identifying key habitat and spatial patterns of fish biodiversity in the tropical Brazilian continental shelf. Cont. Shelf Res. 166, 108–118 (2018).ADS 
Article 

Google Scholar 
Silva, M. B., Rosa, R. S., Menezes, R. & Francini-Filho, R. B. Changes in reef fish assemblages in a cross-shelf euphotic-mesophotic gradient in tropical SW Atlantic. Estuar. Coast. Shelf Sci. 259, 107465 (2021).Article 

Google Scholar 
Doty, M. S. & Oguri, M. The island mass effect. ICES J. Mar. Sci. 22, 33–37 (1956).Article 

Google Scholar 
Gove, J. M. et al. Near-island biological hotspots in barren ocean basins. Nat. Commun. 7, 1–8 (2016).Article 
CAS 

Google Scholar 
Letessier, T. B. et al. Remote reefs and seamounts are the last refuges for marine predators across the Indo-Pacific. PLoS Biol. 17, e3000366 (2019).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Heywood, K. J., Barton, E. D. & Simpson, J. H. The effects of flow disturbance by an oceanic island. J. Mar. Res. 48, 55–73 (1990).Article 

Google Scholar 
Signorini, S. R., McClain, C. R. & Dandonneau, Y. Mixing and phytoplankton bloom in the wake of the Marquesas Islands. Geophys. Res. Lett. 26, 3121–3124 (1999).ADS 
Article 

Google Scholar 
Henry, G. W. & Lyle, J. M. National recreational and indigenous fishing survey (2003).Coutis, P. F. & Middleton, J. H. Flow-topography interaction in the vicinity of an isolated, deep ocean island. Deep Sea Res. Part Oceanogr. Res. Pap. 46, 1633–1652 (1999).ADS 
Article 

Google Scholar 
Cardoso, C., Caldeira, R. M. A., Relvas, P. & Stegner, A. Islands as eddy transformation and generation hotspots: Cabo Verde case study. Prog. Oceanogr. 184, 102271 (2020).Article 

Google Scholar 
Tchamabi, C. C., Araujo, M., Silva, M. & Bourlès, B. A study of the Brazilian Fernando de Noronha island and Rocas atoll wakes in the tropical Atlantic. Ocean Model 111, 9–18 (2017).ADS 
Article 

Google Scholar 
Motta, F. S. et al. Effects of marine protected areas under different management regimes in a hot spot of biodiversity and cumulative impacts from SW Atlantic. Reg. Stud. Mar. Sci. 47, 101951 (2021).Article 

Google Scholar 
Agardy, T., di Sciara, G. N. & Christie, P. Mind the gap: Addressing the shortcomings of marine protected areas through large scale marine spatial planning. Mar. Policy 35, 226–232 (2011).Article 

Google Scholar 
Shucksmith, R. J. & Kelly, C. Data collection and mapping—Principles, processes and application in marine spatial planning. Mar. Policy 50, 27–33 (2014).Article 

Google Scholar 
Queffelec, B. et al. Marine spatial planning and the risk of ocean grabbing in the tropical Atlantic. ICES J. Mar. Sci. 78, 1196–1208 (2021).Article 

Google Scholar 
Rubio-Cisneros, N. T. et al. Poor fisheries data, many fishers, and increasing tourism development: Interdisciplinary views on past and current small-scale fisheries exploitation on Holbox Island. Mar. Policy 100, 8–20 (2019).Article 

Google Scholar 
Samhouri, J. F., Haupt, A. J., Levin, P. S., Link, J. S. & Shuford, R. Lessons learned from developing integrated ecosystem assessments to inform marine ecosystem-based management in the USA. ICES J. Mar. Sci. 71, 1205–1215 (2014).Article 

Google Scholar 
Long, R. D., Charles, A. & Stephenson, R. L. Key principles of marine ecosystem-based management. Mar. Policy 57, 53–60 (2015).Article 

Google Scholar 
Hewitt, J. E., Anderson, M. J. & Thrush, S. F. Assessing and monitoring ecological community health in marine systems. Ecol. Appl. 15, 942–953 (2005).Article 

Google Scholar 
Caselle, J. E., Rassweiler, A., Hamilton, S. L. & Warner, R. R. Recovery trajectories of kelp forest animals are rapid yet spatially variable across a network of temperate marine protected areas. Sci. Rep. 5, 1–14 (2015).Article 
CAS 

Google Scholar 
Díaz-Pérez, L. et al. Coral Reef Health Indices versus the Biological, Ecological and Functional Diversity of Fish and Coral Assemblages in the Caribbean Sea. PLoS One 11, e0161812 (2016).PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
Topor, Z. M., Rasher, D. B., Duffy, J. E. & Brandl, S. J. Marine protected areas enhance coral reef functioning by promoting fish biodiversity. Conserv. Lett. 12, e12638 (2019).Article 

Google Scholar 
Pennino, M. G. et al. Fishery-dependent and -independent data lead to consistent estimations of essential habitats. ICES J. Mar. Sci. 73, 2302–2310 (2016).Article 

Google Scholar 
Hilborn, R. & Walters, C. J. Quantitative Fisheries Stock Assessment: Choice, Dynamics and Uncertainty (Springer Science and Business Media, 2013).
Google Scholar 
Bohnsack, J. A. & Bannerot, S. P. A stationary visual census technique for quantitatively assessing community structure of coral reef fishes (1986).Jones, R. S. & Thompson, M. J. Comparison of Florida reef fish assemblages using a rapid visual technique. Bull. Mar. Sci. 28, 159–172 (1978).
Google Scholar 
Kimmel, J. J. A new species-time method for visual assessment of fishes and its comparison with established methods. Environ. Biol. Fishes 12, 23–32 (1985).Article 

Google Scholar 
Michalopoulos, C., Auster, P. J. & Malatesta, R. J. A comparison of transect and species-time counts for assessing faunal abundance from video surveys. Mar. Technol. Soc. J. 26, 27–31 (1992).
Google Scholar 
Gray, J. S., Ugland, K. I. & Lambshead, J. Species accumulation and species area curves: A comment on Scheiner (2003). Glob. Ecol. Biogeogr. 13, 473–476 (2004).Article 

Google Scholar 
Mallet, D. & Pelletier, D. Underwater video techniques for observing coastal marine biodiversity: A review of sixty years of publications (1952–2012). Fish. Res. 154, 44–62 (2014).Article 

Google Scholar 
Langlois, T. J. et al. Cost-efficient sampling of fish assemblages: Comparison of baited video stations and diver video transects. Aquat. Biol. 9, 155–168 (2010).Article 

Google Scholar 
Logan, J. M., Young, M. A., Harvey, E. S., Schimel, A. C. G. & Ierodiaconou, D. Combining underwater video methods improves effectiveness of demersal fish assemblage surveys across habitats. Mar. Ecol. Prog. Ser. 582, 181–200 (2017).ADS 
Article 

Google Scholar 
Koslow, J. A. The role of acoustics in ecosystem-based fishery management. ICES J. Mar. Sci. 66, 966–973 (2009).Article 

Google Scholar 
Bertrand, A. et al. Broad impacts of fine-scale dynamics on seascape structure from zooplankton to seabirds. Nat. Commun. 5, 1–9 (2014).ADS 
Article 
CAS 

Google Scholar 
Benoit-Bird, K. J. & Lawson, G. L. Ecological insights from pelagic habitats acquired using active acoustic techniques. Annu. Rev. Mar. Sci. 8, 463–490 (2016).ADS 
Article 

Google Scholar 
Sutton, T. T. Vertical ecology of the pelagic ocean: Classical patterns and new perspectives. J. Fish Biol. 83, 1508–1527 (2013).CAS 
PubMed 
Article 

Google Scholar 
McClatchie, S., Thorne, R. E., Grimes, P. & Hanchet, S. Ground truth and target identification for fisheries acoustics. Fish. Res. 47, 173–191 (2000).Article 

Google Scholar 
Cappo, M., Speare, P. & De’ath, G. Comparison of baited remote underwater video stations (BRUVS) and prawn (shrimp) trawls for assessments of fish biodiversity in inter-reefal areas of the Great Barrier Reef Marine Park. J. Exp. Mar. Biol. Ecol. 302, 123–152 (2004).Article 

Google Scholar 
Harvey, E. S., Cappo, M., Butler, J. J., Hall, N. & Kendrick, G. A. Bait attraction affects the performance of remote underwater video stations in assessment of demersal fish community structure. Mar. Ecol. Prog. Ser. 350, 245–254 (2007).ADS 
Article 

Google Scholar 
Fitzpatrick, B. M., Harvey, E. S., Heyward, A. J., Twiggs, E. J. & Colquhoun, J. Habitat specialization in tropical continental shelf demersal fish assemblages. PLoS One 7, e39634 (2012).ADS 
CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Rooper, C. N., Hoff, G. R. & De Robertis, A. Assessing habitat utilization and rockfish (Sebastes spp.) biomass on an isolated rocky ridge using acoustics and stereo image analysis. Can. J. Fish. Aquat. Sci. 67, 1658–1670 (2010).Article 

Google Scholar 
Jones, D. et al. Evaluation of rockfish abundance in untrawlable habitat: Combining acoustic and complementary sampling tools (2012).O’Driscoll, R. L. et al. Species identification in seamount fish aggregations using moored underwater video. ICES J. Mar. Sci. 69, 648–659 (2012).Article 

Google Scholar 
Fernandes, P. G., Copland, P., Garcia, R., Nicosevici, T. & Scoulding, B. Additional evidence for fisheries acoustics: Small cameras and angling gear provide tilt angle distributions and other relevant data for mackerel surveys. ICES J. Mar. Sci. 73, 2009–2019 (2016).Article 

Google Scholar 
Gastauer, S., Scoulding, B. & Parsons, M. An unsupervised acoustic description of fish schools and the seabed in three fishing regions within the Northern Demersal Scalefish Fishery (NDSF, Western Australia). Acoust. Aust. 45, 363–380 (2017).Article 

Google Scholar 
Blanluet, A. et al. Characterization of sound scattering layers in the Bay of Biscay using broadband acoustics, nets and video. PLoS One 14, e0223618 (2019).CAS 
PubMed 
PubMed Central 
Article 

Google Scholar 
Campanella, F. & Taylor, J. C. Investigating acoustic diversity of fish aggregations in coral reef ecosystems from multifrequency fishery sonar surveys. Fish. Res. 181, 63–76 (2016).Article 

Google Scholar 
Domokos, R. On the development of acoustic descriptors for semi-demersal fish identification to support monitoring stocks. ICES J. Mar. Sci. 78, 1117–1130 (2021).Article 

Google Scholar 
Villalobos, H. et al. A practical approach to monitoring marine protected areas: An application to El Bajo Espíritu Santo Seamount near La Paz, Mexico. Oceanography 34, 32–43 (2021).Article 

Google Scholar 
Hazin, F. H., Zagaglia, J. R., Broadhurst, M. K., Travassos, P. E. P. & Bezerra, T. R. Q. Review of a small-scale pelagic longline fishery off northeastern Brazil. Mar. Fish. Rev. 60, 1–8 (1998).
Google Scholar 
Lessa, R. P. et al. Distribution and abundance of ichthyoneuston at seamounts and islands off north-eastern Brazil. Arch. Fish. Mar. Res. 47, 239–252 (1999).
Google Scholar 
Dominguez, P. S., Zeineddine, G. C., Rotundo, M. M., Barrella, W. & Ramires, M. A pesca artesanal no arquipélago de Fernando de Noronha (PE). Bol. Inst. Pesca 42, 241–251 (2014).Article 

Google Scholar 
Lopes, P. F. M., Mendes, L., Fonseca, V. & Villasante, S. Tourism as a driver of conflicts and changes in fisheries value chains in Marine Protected Areas. J. Environ. Manag. 200, 123–134 (2017).CAS 
Article 

Google Scholar 
Outeiro, L., Rodrigues, J. G., Damásio, L. M. A. & Lopes, P. F. M. Is it just about the money? A spatial-economic approach to assess ecosystem service tradeoffs in a marine protected area in Brazil. Ecosyst. Serv. 38, 100959 (2019).Article 

Google Scholar 
Garla, R. C., Chapman, D. D., Wetherbee, B. M. & Shivji, M. Movement patterns of young Caribbean reef sharks, Carcharhinus perezi, at Fernando de Noronha Archipelago, Brazil: The potential of marine protected areas for conservation of a nursery ground. Mar. Biol. 149, 189–199 (2006).Article 

Google Scholar 
Bertrand, A. FAROFA 1 cruise. RV TUBARAO Tigre. https://doi.org/10.17600/18001399 (2017).Article 

Google Scholar 
Bertrand, A. FAROFA 2 cruise. RV TUBARAO Tigre. https://doi.org/10.17600/18001411 (2018).Article 

Google Scholar 
Bertrand, A. FAROFA 3 cruise. RV TUBARAO Tigre. https://doi.org/10.17600/18001381 (2019).Article 

Google Scholar 
Bertrand, A. et al. Acoustic data from FAROFA surveys, 2017-09-15 to 2019-04-22. https://doi.org/10.17882/71024 (2020).Salvetat, J. et al. Underwater video observations from FAROFA surveys, 2017-09-15 to 2019-04-22. https://doi.org/10.17882/76019 (2020).Pawlowicz, R. M_Map: A mapping package for MATLAB, version 1.4 m (computer software) (2020).Péter, A. Solomon Coder: The Concept of Behavioral Elements, Categories and the Representation of Data in Solomon Coder (2019).Priede, I. G., Bagley, P. M., Smith, A., Creasey, S. & Merrett, N. R. Scavenging deep demersal fishes of the Porcupine Seabight, north-east Atlantic: Observations by baited camera, trap and trawl. J. Mar. Biol. Assoc. U. K. 74, 481–498 (1994).Article 

Google Scholar 
McQuinn, I. H. et al. Description of the ICES HAC standard data exchange format, version 1.60 (Conseil international pour l’exploration de la mer, 2005).
Google Scholar 
Trenkel, V. M. et al. Overview of recent progress in fisheries acoustics made by Ifremer with examples from the Bay of Biscay. Aquat. Living Resour. 22, 433–445 (2009).Article 

Google Scholar 
Perrot, Y. et al. Matecho: An open-source tool for processing fisheries acoustics data. Acoust. Aust. 46, 241–248 (2018).Article 

Google Scholar 
Salvetat, J. et al. In situ target strength measurement of the black triggerfish Melichthys niger and the ocean triggerfish Canthidermis sufflamen. Mar. Freshw. Res. 71, 1118–1127 (2020).Article 

Google Scholar 
Lavery, A. C. et al. Determining dominant scatterers of sound in mixed zooplankton populations. J. Acoust. Soc. Am. 122, 3304–3326 (2007).ADS 
PubMed 
Article 

Google Scholar 
MacLennan, D. N., Fernandes, P. G. & Dalen, J. A consistent approach to definitions and symbols in fisheries acoustics. ICES J. Mar. Sci. 59, 365–369 (2002).Article 

Google Scholar 
Barros, M. J. G. Analises da Ictiofauna marinha e habitats associados atraves de videos subaquatica. (Universidade Federal de Pernambuco, 2020).
Google Scholar 
Sazima, C., Bonaldo, R. M., Krajewski, J. P. & Sazima, I. The Noronha wrasse: A jack-of-all-trades follower. Aqua J. Ichthyol. Aquat. Biol. 9, 97–108 (2005).
Google Scholar 
Soto, J. M. R. Peixes do arquipélago Fernando de Noronha. Mare Magnum 1, 147–169 (2001).
Google Scholar 
Krajewski, J. P. & Floeter, S. R. Reef fish community structure of the Fernando de Noronha Archipelago (Equatorial Western Atlantic): The influence of exposure and benthic composition. Environ. Biol. Fishes 92, 25 (2011).Article 

Google Scholar 
Sazima, I., Sazima, C. & da Silva-Jr, J. M. Fishes associated with spinner dolphins at Fernando de Noronha Archipelago, tropical Western Atlantic: An update and overview. Neotropical Ichthyol. 4, 451–455 (2006).Article 

Google Scholar 
Petitgas, P. Use of a disjunctive kriging to model areas of high pelagic fish density in acoustic fisheries surveys. Aquat. Living Resour. 6, 201–209 (1993).Article 

Google Scholar 
Chiles, J.-P. & Delfiner, P. Geostatistics: Modeling Spatial Uncertainty Vol. 497 (Wiley, 2009).MATH 

Google Scholar 
Bez, N. & Braham, C.-B. Indicator variables for a robust estimation of an acoustic index of abundance. Can. J. Fish. Aquat. Sci. 71, 709–718 (2014).Article 

Google Scholar 
Switzer, P. Min/max autocorrelation factors for multivariate spatial imagery. Comput. Sci. Stat. (1985).Bez, N. Global estimation based on indicators factorization (2021).Assunção, R. V., Silva, A. C., Martins, J. & Montes, M. F. Spatial-temporal variability of the thermohaline properties in the coastal region of Fernando de Noronha Archipelago, Brazil. J. Coast. Res. 75, 512–517 (2016).Article 

Google Scholar 
da Silva, A. C. et al. Surface circulation and vertical structure of upper ocean variability around Fernando de Noronha archipelago and Rocas atoll during spring 2015 and fall 2017. Front. Mar. Sci. 8, 598101 (2021).Article 

Google Scholar 
Breiman, L., Friedman, J., Olshen, R. & Stone, C. Classification and regression trees. Wadsworth Int. Group 37, 237–251 (1984).MATH 

Google Scholar 
Therneau, T., Atkinson, B., Ripley, B. & Ripley, M. B. Package ‘rpart’. Available Online Cran Ma Ic Ac Ukwebpackagesrpartrpart Pdf Accessed 20 April 2016 (2015).Kuhnert, P. M., Duffy, L. M., Young, J. W. & Olson, R. J. Predicting fish diet composition using a bagged classification tree approach: A case study using yellowfin tuna (Thunnus albacares). Mar. Biol. 159, 87–100 (2012).CAS 
Article 

Google Scholar 
Breiman, L. Bagging predictors. Mach. Learn. 24, 123–140 (1996).MATH 

Google Scholar 
Kuhnert, P. M., Henderson, A.-K., Bartley, R. & Herr, A. Incorporating uncertainty in gully erosion calculations using the random forests modelling approach. Environmetrics 21, 493–509 (2010).MathSciNet 

Google Scholar 
Kuhnert, P. M. & Mengersen, K. Reliability measures for local nodes assessment in classification trees. J. Comput. Graph. Stat. 12, 398–416 (2003).Article 

Google Scholar 
R Core Team. R: A language and environment for statistical computing (2020).ParisTech, M. ARMINES: RGeostats: The Geostatistical R Package (2020).Kahle, D. J. & Wickham, H. ggmap: Spatial visualization with ggplot2. R J 5, 144 (2013).Article 

Google Scholar 
Pimentel, C. R. et al. Mesophotic ecosystems at Fernando de Noronha Archipelago, Brazil (South-western Atlantic), reveal unique ichthyofauna and need for conservation. Neotropical Ichthyol. 18 (2020).Ilarri, M. I., Souza, A. T. & Rosa, R. S. Community structure of reef fishes in shallow waters of the Fernando de Noronha archipelago: Effects of different levels of environmental protection. Mar. Freshw. Res. 68, 1303–1316 (2017).Article 

Google Scholar 
Schmid, K. et al. First fish fauna assessment in the Fernando de Noronha Archipelago with BRUVS: Species catalog with underwater imagery. Biota Neotropica 20 (2020).de Araújo, M. E. et al. Diversity patterns of reef fish along the Brazilian tropical coast. Mar. Environ. Res. 160, 105038 (2020).PubMed 
Article 
CAS 

Google Scholar 
Krajewski, J. P., Floeter, S. R., Jones, G. P. & Leite, F. P. Patterns of variation in behaviour within and among reef fish species on an isolated tropical island: Influence of exposure and substratum. J. Mar. Biol. Assoc. U. K. 91, 1359–1368 (2011).Article 

Google Scholar 
Mendes, T. C., Quimbayo, J. P., Bouth, H. F., Silva, L. P. & Ferreira, C. E. The omnivorous triggerfish Melichthys niger is a functional herbivore on an isolated Atlantic oceanic island. J. Fish Biol. 95, 812–819 (2019).PubMed 

Google Scholar 
Petitgas, P. & Levenez, J. J. Spatial organization of pelagic fish: Echogram structure, spatio-temporal condition, and biomass in Senegalese waters. ICES J. Mar. Sci. 53, 147–153 (1996).Article 

Google Scholar 
Burgos, J. M. & Horne, J. K. Characterization and classification of acoustically detected fish spatial distributions. ICES J. Mar. Sci. 65, 1235–1247 (2008).Article 

Google Scholar 
Russ, G. R. Grazer biomass correlates more strongly with production than with biomass of algal turfs on a coral reef. Coral Reefs 22, 63–67 (2003).Article 

Google Scholar 
Friedlander, A. M. & Parrish, J. D. Habitat characteristics affecting fish assemblages on a Hawaiian coral reef. J. Exp. Mar. Biol. Ecol. 224, 1–30 (1998).Article 

Google Scholar 
Munday, P. L. Does habitat availability determine geographical-scale abundances of coral-dwelling fishes?. Coral Reefs 21, 105–116 (2002).ADS 
Article 

Google Scholar 
Martins, K. et al. Assessing trophic interactions between pelagic predatory fish by gut content and stable isotopes analysis around Fernando de Noronha Archipelago (Brazil), Equatorial West Atlantic. J. Fish Biol. 99, 1576–1590 (2021).CAS 
PubMed 
Article 

Google Scholar 
Costa, B., Taylor, J. C., Kracker, L., Battista, T. & Pittman, S. Mapping reef fish and the seascape: Using acoustics and spatial modeling to guide coastal management. PLoS One 9, e85555 (2014).ADS 
PubMed 
PubMed Central 
Article 
CAS 

Google Scholar 
Kavanagh, K. D. & Olney, J. E. Ecological correlates of population density and behavior in the circumtropical black triggerfish Melichthys niger (Balistidae). Environ. Biol. Fishes 76, 387–398 (2006).Article 

Google Scholar 
Lubbock, R. The shore fishes of Ascension Island. J. Fish Biol. 17, 283–303 (1980).Article 

Google Scholar 
Price, J. H. & John, D. M. Ascension Island, South Atlantic: A survey of inshore benthic macroorganisms, communities and interactions. Aquat. Bot. 9, 251–278 (1980).Article 

Google Scholar 
Robertson, D. R. & Allen, G. R. Zoogeography of the shorefish fauna of Clipperton Atoll. Coral Reefs 15, 121–131 (1996).ADS 
Article 

Google Scholar 
Gasparini, J. L. & Floeter, S. R. The shore fishes of Trindade Island, western south Atlantic. J. Nat. Hist. 35, 1639–1656 (2001).Article 

Google Scholar 
Lubbock, R. & Edwards, A. The fishes of Saint Paul’s rocks. J. Fish Biol. 18, 135–157 (1981).Article 

Google Scholar 
Feitoza, B. M., Rocha, L. A., Luiz-Júnior, O. J., Floeter, S. R. & Gasparini, J. L. Reef fishes of St. Paul’s Rocks: New records and notes on biology and zoogeography. Aqua 7, 61–82 (2003).
Google Scholar 
Ferreira, C. E. L., Floeter, S. R., Gasparini, J. L., Ferreira, B. P. & Joyeux, J. C. Trophic structure patterns of Brazilian reef fishes: A latitudinal comparison. J. Biogeogr. 31, 1093–1106 (2004).Article 

Google Scholar 
Floeter, S. R. et al. Atlantic reef fish biogeography and evolution. J. Biogeogr. 35, 22–47 (2008).
Google Scholar 
Morais, R. A., Ferreira, C. E. L. & Floeter, S. R. Spatial patterns of fish standing biomass across Brazilian reefs. J. Fish Biol. 91, 1642–1667 (2017).CAS 
PubMed 
Article 

Google Scholar 
Walsh, W. J. Patterns of recruitment and spawning in Hawaiian reef fishes. Environ. Biol. Fishes 18, 257–276 (1987).Article 

Google Scholar 
Walsh, W. J. Aspects of Nocturnal Shelter, Habitat Space, and Juvenile Recruitment in Hawaiian Coral Reef Fishes (University of Hawaii, 1984).
Google Scholar 
Caldeira, R. M. A., Groom, S., Miller, P., Pilgrim, D. & Nezlin, N. P. Sea-surface signatures of the island mass effect phenomena around Madeira Island, Northeast Atlantic. Remote Sens. Environ. 80, 336–360 (2002).ADS 
Article 

Google Scholar 
Martinez, E. & Maamaatuaiahutapu, K. Island mass effect in the Marquesas Islands: Time variation. Geophys. Res. Lett. 31, 18 (2004).Article 

Google Scholar 
Messié, M. et al. The delayed island mass effect: How islands can remotely trigger blooms in the oligotrophic ocean. Geophys. Res. Lett. 47, e2019GL085282 (2020).ADS 
Article 

Google Scholar 
de Souza, C. S., da Luz, J. A. G., Macedo, S., de Montes, M. J. F. & Mafalda, P. Chlorophyll a and nutrient distribution around seamounts and islands of the tropical south-western Atlantic. Mar. Freshw. Res. 64, 168–184 (2013).Article 
CAS 

Google Scholar 
Travassos, P., Hazin, F. H., Zagaglia, J. R., Advíncula, R. & Schober, J. Thermohaline structure around seamounts and islands off North-Eastern Brazil. Arch. Fish. Mar. Res. 47, 211–222 (1999).
Google Scholar 
Bakun, A. Ocean triads and radical interdecadal variation: Bane and boon to scientific fisheries management. in Reinventing fisheries management 331–358 (Springer, 1998).Agostini, V. N. & Bakun, A. ‘Ocean triads’ in the Mediterranean Sea: Physical mechanisms potentially structuring reproductive habitat suitability (with example application to European anchovy, Engraulis encrasicolus). Fish. Oceanogr. 11, 129–142 (2002).Article 

Google Scholar 
Hamner, W. M., Jones, M. S., Carleton, J. H., Hauri, I. R. & Williams, D. M. Zooplankton, planktivorous fish, and water currents on a windward reef face: Great Barrier Reef, Australia. Bull. Mar. Sci. 42, 459–479 (1988).
Google Scholar 
Valenzuela, J., Bellwood, D. & Morais, R. Ontogenetic habitat shifts in fusiliers (Lutjanidae): Evidence from Caesio cuning at Lizard Island, Great Barrier Reef. Coral Reefs 40, 1687–1696 (2021).Article 

Google Scholar 
Curley, B. G., Kingsford, M. J. & Gillanders, B. M. Spatial and habitat-related patterns of temperate reef fish assemblages: Implications for the design of Marine Protected Areas. Mar. Freshw. Res. 53, 1197–1210 (2002).Article 

Google Scholar 
Ferrari, R. et al. Habitat structural complexity metrics improve predictions of fish abundance and distribution. Ecography 41, 1077–1091 (2018).Article 

Google Scholar 
Maida, M. & Ferreira, B. P. Coral reefs of Brazil: An overview. in Proceedings of the 8th International Coral Reef Symposium Vol. 1 74 (Smithsonian Tropical Research Institute Panamá, 1997).Pittman, S. J., Costa, B. M. & Battista, T. A. Using lidar bathymetry and boosted regression trees to predict the diversity and abundance of fish and corals. J. Coast. Res. 2009, 27–38 (2009).Article 

Google Scholar 
Costa, T. Análise comportamental e distribuição da atividade pesqueira no Arquipelágo de Fernando de Noronha (Nordeste, BR) baseada em dados de GPS. (Universidade Federal Rural de Pernambuco, 2019).
Google Scholar 
Spalding, M. D. et al. Marine ecoregions of the world: A bioregionalization of coastal and shelf areas. Bioscience 57, 573–583 (2007).Article 

Google Scholar 
Claudet, J., Pelletier, D., Jouvenel, J.-Y., Bachet, F. & Galzin, R. Assessing the effects of marine protected area (MPA) on a reef fish assemblage in a northwestern Mediterranean marine reserve: Identifying community-based indicators. Biol. Conserv. 130, 349–369 (2006).Article 

Google Scholar 
Caveen, A. J., Gray, T. S., Stead, S. M. & Polunin, N. V. C. MPA policy: What lies behind the science?. Mar. Policy 37, 3–10 (2013).Article 

Google Scholar 
Hernández, C. M. et al. Evidence and patterns of tuna spawning inside a large no-take Marine Protected Area. Sci. Rep. 9, 1–11 (2019).
Google Scholar  More

Stringed instruments can be worth millions of dollars, particularly those made in northern Italy in the seventeenth and eighteenth centuries. There are also many copies and forgeries.As a forest ecologist, I use dendrochronology — or tree-ring dating — to understand how trees grow, as well as to investigate historical environmental conditions. The widths of tree rings vary according to meteorological conditions, so samples can be dated by cross-referencing against databases of ring-width series.In 2010, I was an expert witness in a legal case about a viola supposedly made in the sixteenth century. I agreed with two laboratories that had independently concluded that it could not have been made before the late eighteenth century.Dendrochronology cannot precisely date when an instrument was made, but it can identify the most recent year that the wood it was made from was part of a growing tree. Tree rings give probabilities and levels of confidence in a date according to the availability of appropriate reference series.Dendrochronological analyses can produce conflicting dates, creating confusion about the method’s reliability. The most famous example is the ‘Messiah’ violin made by Antonio Stradivari in 1716. In 1999, doubts were raised about its authenticity when, among other reasons, tree-ring analysis found that the instrument could not have been made before 1738, a year after Stradivari’s death. The finding was based on examining a photograph of the instrument, and was later retracted. The controversy highlighted the limitations of the technique for verifying the authenticity of musical instruments.I also use dendrochronology to understand the impacts of droughts, and I’m fascinated by how trees grow, how water is transported in them and whether species differences are evolutionary adaptations. More