Growth impacts in a changing ocean: insights from two coral reef fishes in an extreme environment

Determining the life-history consequences for fishes living in extreme and variable environments will be vital in predicting the likely impacts of ongoing climate change on reef fish demography. Here, we compare size-at-age and maximum body size of two common reef fish species (Lutjanus ehrenbergii and Pomacanthus maculosus) between the environmentally extreme Arabian/Persian Gulf (‘Arabian Gulf’) and adjacent comparably benign Oman Sea. Additionally, we use otolith increment width profiles to investigate the influence of temperature, salinity and productivity on the individual growth rates. Individuals of both species showed smaller size-at-age and lower maximum size in the Arabian Gulf compared to conspecifics in the less extreme and less variable environment of the Oman Sea, suggesting a life-history trade-off between size and metabolic demands. Salinity was the best environmental predictor of interannual growth across species and regions, with low growth corresponding to more saline conditions. However, salinity had a weaker negative effect on interannual growth of fishes in the Arabian Gulf than in the Oman Sea, indicating Arabian Gulf populations may be better able to acclimate to changing environmental conditions. Temperature had a weak positive effect on the interannual growth of fishes in the Arabian Gulf, suggesting that these populations may still be living within their thermal windows. Our results highlight the potential importance of osmoregulatory cost in impacting growth, and the need to consider the effect of multiple stressors when investigating the consequences of future climate change on fish demography.


Introduction
Anthropogenic global climate change is having significant biological impacts on individuals, species and ecosystems (e.g. Parmesan and Yohe 2003;Bellard et al. 2012;Hughes et al. 2018;Gordon et al. 2018;Pratchett et al. 2018;França et al. 2020). Reductions in body size, coupled with reduction in lifespan (Munch and Salinas 2009), geographic shifts in species distribution (Walther et al. 2002;Parmesan and Yohe 2003;Feary et al. 2014) and changes in phenology (Walther et al. 2002;Stenseth 2002;Visser and Both 2005;Taylor 2008), are recognised as common responses to global warming in ectotherms (Daufresne et al. 2009;Gardner et al. 2011;Ohlberger 2013). Such reduction in body size is consistent with the temperaturesize rule (TSR) (i.e. body size decreases at higher temperature) (Atkinson 1994;Kingsolver and Huey 2008;Ohlberger 2013;Huss et al. 2019) and is particularly evident in aquatic environments (Forster et al. 2012;Horne et al. 2015), where fish and other aquatic organisms' average body size has already declined by 5-20% over the last two decades (Baudron et al. 2014;Audzijonyte et al. 2016;van Rijn et al. 2017). Experimental temperature-size responses, models and meta-analyses suggest that body size will further reduce by 3-5% per degree of warming in aquatic arthropods, while in fishes it may decline by 14-24% by 2050 under a higher green-house gases emission scenario (A2 scenario, IPCC 2007;Cheung et al. 2013;Pauly and Cheung 2018).
Decreasing fish sizes will impact fecundity and fisheries productivity (Baudron et al. 2014; Barneche et al. 2018), prey-predator interactions (Barnes et al. 2010) and overall ecosystem functioning (Bellwood et al. 2012;Cheung et al. 2013). Understanding what drives reductions in fishes' body size will be essential for predicting how populations are impacted by projected climate change. In this regard, despite the ubiquity of the TSR in describing the pattern of body size and temperature in a range of taxa, the underlying mechanisms of body size reduction with increasing temperature are still debated (Audzijonyte et al. 2019).
Projected increases in mean temperature and temperature variability due to climate change (i.e. 1.0 to 4.0°C increase by 2100, IPCC 2019) are likely to be particularly challenging for marine fish due to the greater energetic demands for routine metabolic activities (Gillooly et al. 2001;Pörtner and Knust 2007;Neuheimer et al. 2011;Rubalcaba et al. 2020) with, ultimately, less energy available for somatic growth. Similarly, projected intensification of the global water cycle (4-8%°C of increase in surface air temperature change) and consequent increase in water salinity in already dry regions (Durack et al. 2012;Skliris et al. 2014;Zika et al. 2018) will likely increase the osmoregulation costs for many marine species (Boeuf and Payan 2001;Ern et al. 2014), especially those living in hypersaline semi-enclosed seas. Finally, climate change is predicted to alter the rate and distribution of primary production in the world's oceans (Brown et al. 2010;Stock et al. 2017), with consequent impacts on food availability and fish growth and biomass (Martino et al. 2019). While the relationship between ocean primary productivity and fish growth is clear (Downing et al. 1990), the direction and magnitude of this relationship can vary, from increased growth due to greater food availability to growth reduction associated with excessive primary production and ocean deoxygenation, especially in highly productive and variable environments (Martino et al. 2019;Tanner et al. 2019).
One approach to understanding how marine fish demography may be impacted by future environmental variance and extremes in temperature and salinity is to study contemporary communities that exist within naturally extreme and variable environments ). Here, we compare the effect of water temperature, salinity and primary productivity on the size-at-age and growth rate of the coral reef fishes Lutjanus ehrenbergii and Pomacanthus maculosus between the environmentally extreme southern Arabian/Persian Gulf (hereafter 'Arabian Gulf') and comparatively more benign adjacent Oman Sea (Fig. 1). The Arabian Gulf has the largest thermal range ([ 20°C) and highest maximum sea surface temperature (SST) experienced by extant coral reef fishes (winter: \ 15°C, summer: [ 35°C), with fishes enduring several months of conditions considered lethal to reef fishes in other parts of the world (Riegl and Purkis 2012;Rummer et al. 2014;Vaughan et al. 2019). In addition, present summer SSTs in the Gulf are comparable to those expected for tropical oceans by 2100 (Riegl and Purkis 2012;IPCC 2019), while winter temperatures can be so low as to induce cold water coral bleaching (Coles and Fadlallah 1991). Additionally, due to restricted water exchange through the narrow Strait of Hormuz, as well as limited freshwater input and high evaporation (Reynolds 1993(Reynolds , 2002, the Arabian Gulf's waters are characterized by hypersaline conditions (annual mean salinity 42 psu) that are the highest reported for coral reefs (Bauman 2013. In contrast, the Oman Sea is subject to less extreme and variable temperature and salinity, but a higher baseline and higher interannual variation of chlorophyll-a concentration (used as a proxy for primary productivity) due to seasonal upwelling events (Coles 2003, Bauman 2013. Given differences in environmental conditions between the Arabian Gulf and the Oman Sea, we hypothesise that, in line with the TSR, (i) fishes in the Arabian Gulf would exhibit lower growth rates, as well as smaller size-at-age and maximum body size than conspecifics from the less extreme and less variable environment of the Oman Sea and (ii) that predicted differences in growth between populations (i.e. Arabian Gulf vs Oman Sea) would be driven by local environmental conditions. We investigated the role of temperature, salinity and productivity in structuring fish demography between regions by using an otolith sclerochronology approach and mixed-effect modelling and predicted that low growth rate and small body size would be the results of extreme temperature ranges in the Arabian Gulf, exacerbated by high salinity and lower primary productivity.

Materials and method
To determine whether fish growth is reduced in the Arabian Gulf compared to the Oman Sea and whether extreme environmental conditions are the drivers of somatic growth differences, we analysed the otolith structure of the blackspot snapper (Lutjanus ehrenbergii, F. Lutjanidae [Peters, 1869]) and the yellow-bar angelfish (Pomacanthus maculosus, F. Pomacanthidae [Forsskål, 1775]). Both L. ehrenbergii and P. maculosus are coral-associated species commonly found on nearshore patch and fringing reefs throughout both regions (Feary et al. 2010;Burt et al. 2011) and are known to have highly determinate growth patterns (Grandcourt et al. 2010;. Furthermore, the two species were chosen because they are phylogenetically and ecologically disparate: L. ehrenbergii is a generalist carnivore, feeding on benthic invertebrates associated with turfing algae (i.e. amphipods, isopods) and small fish (Randall 1995;D'Agostino et al. 2019), while P. maculosus is a facultative spongivore/corallivore (Shraim et al. 2017).

Sample collection of focal species for life-history comparison
Sample collection encompassed six sites within the Arabian Gulf (three reefs and four local fish landing sites, spanning from Dhabiya to Umm Al Quwain) and seven sites within the Oman Sea (two reefs and five local fish landing sites, spanning from Dibba to As Seeb) (Fig. 1, Table S2). As both species are relatively site attached and the nearest sites between the two regions were separated by more than 300 km (Grizzle et al. 2016;Burt et al. 2016), negligible movement between regions was predicted (Buchanan et al. 2019). Fish sourced from landing sites were purchased and sampled only upon confirmation of their local origin from the fisherman. Individuals collected in situ (i.e. from the reefs) were collected using the fish anaesthetic clove oil (juveniles) or spear guns (sub-adults and adults) and euthanized using an ice slurry. Upon collection, all individuals were measured (standard length  [SL], nearest mm), weighed (total weight, nearest g) and sagittal otoliths removed, cleaned in ethanol and stored dry.

Age determination
To determine age, 1012 and 210 otoliths were sectioned from L. ehrenbergii and P. maculosus specimens, respectively (Table 1). Each otolith was ground to the nucleus to produce a thin transverse section (* 300 lm) and mounted on a microscope slide using established procedures (i.e. Taylor and McIlwain 2010). Otoliths were examined under transmitted light with a low power microscope and individual ages estimated by counting the number of annual increments, or annuli, along the dorsal antisulcus axis of each otolith (Fig. 2). Previous studies have verified the deposition of annual increments for both species (Grandcourt et al. 2010;. Blind reads (with respect to size and collection location) of annual increments were performed by the first author (DD) on three separate occasions for each specimen. Final age was determined when two or more counts agreed, and the precision of age estimation was calculated using the index of average percentage error (IAPE) (following Wakefield et al. 2017). If agreement was not achieved after three counts, the sample was excluded from the analysis. For individuals \ 1 year, otolith sections were further ground with lapping film, polished by hand with 0.3 lm alumina powder and viewed through a compound microscope by one of the co-authors (BMT). For these individuals, age was then estimated by counting the number of daily increments, with three blind reads performed for each individual. Final age for juvenile fishes was taken as the median of three counts when all counts were within 10% of the median. Samples with counts [ 10% of the median were excluded from the analysis.

Population-level growth, body size and life span
To model and compare the population growth rate of P. maculosus and L. ehrenbergii between the Arabian Gulf and the Oman Sea, the re-parameterised version of the von Bertalanffy growth function (rVBGF, Francis 1988) was used (Visconti et al 2018). The rVBGF describes growth based on modelled body size of individuals at three ages, L s , L x and L l , allowing a direct comparison of mean sizeat-age data between populations. Age s was selected to reflect the ascending part of the growth trajectory (i.e. the period of fast, early growth), x represents the onset of a reduction in growth rate, while l represents the asymptote of growth (i.e. when growth reaches a plateau) (Trip et al. 2008(Trip et al. , 2014. To model species growth based on the trajectory of the region-specific growth curves, ages 1 and 5 years (based on annuli counts) were taken as L s (hereafter 'L 1 ') and L l (hereafter 'L 5 '), respectively, while L x was calculated at 3 years (hereafter 'L 3 '). The rVBGF model was fitted using the age (years) and length data (mm SL) of each sample and by constraining the curve to a length of settlement of 10 mm SL (Grandcourt et al. 2011).
The best-fit model parameters, L 1 , L 3 and L 5 , describing each dataset were determined by minimising the negative log of the likelihood, given a probability density function with a Poisson distribution (Haddon 2001). Mean maximum age (hereafter 'Age max ') and mean maximum body size (hereafter 'L max ') were estimated for both species in each region. Age max and L max were taken, respectively, as the average age (years) and body size (SL) of the 25% oldest or largest adult (i.e. C 3 years old) individuals, respectively (Trip et al. 2008(Trip et al. , 2014. Differences in mean Age max and L max between populations were analysed using linear models. n, sample size; L 1 , mean size-at-age one; L 3 , age three; L 5 , age five; -k, negative log-likelihood; r, standard deviation. All sizes are in mm SL, Ages are in years unless indicated otherwise

Individual annual growth estimation
To determine whether yearly growth rate differs within species between regions, annual increment widths were measured and compared between individuals. A subset of the total samples (148 and 144 otoliths of L. ehrenbergii and P. maculosus, respectively; Table 2) that showed clear edges between annual increments and with at least one completed annual increment (i.e. individuals C 2 years old) were photographed and examined. As otolith growth is an appropriate representation of fish somatic growth in our species (i.e. significant positive relationship between otolith radius and individual body length, see Fig. S1a, b), increment widths were measured (mm) along the dorsal antisulcus axis (from core to edge) using MorphoJ (v1.06). Distances between the outer edges of each annual increment were taken to indicate the width of each consecutive growth increment (Fig. 2a, b). Through back-calculation from age at capture (hereafter 'AAC') and year of capture, age ('Age') and calendar growth year ('Year') were, respectively, assigned to each increment (Table 3). The first two annual increments were not used in the analysis due to poor visualisation in the inner region of the otolith, while the last (marginal) increments were excluded because they did not represent a full year of growth (Martino et al. 2019). For each region, the sample size was at least five increment measurements per calendar year. In total, increment widths were measured from 148 L. ehrenbergii (Arabian Gulf = 58; Oman Sea = 90) and 144 P. maculosus (Arabian Gulf = 107; Oman Sea = 37) individuals (Table 2). Increment widths encapsulated growth of individuals between 2002 and 2015 in L. ehrenbergii and between 1993 and 2016 in P. maculosus (Table 2).

Growth mixed modelling
Sources of annual growth variation in P. maculosus and L. ehrenbergii were investigated using a series of hierarchical mixed-effects models predicting variation in otolith increment widths (following Morrongiello and Thresher 2015). Growth was modelled separately for the two regions to avoid the co-variation of multiple environmental factors with regions. All mixed modelling was performed in R (R core development team 3.5.1, 2019), using the packages lme4 (Bates et al. 2018), effects (Fox 2003), and AICcmodavg (Mazerolle 2019). Modelling could not be done for P. maculosus in the Oman Sea due to insufficient sample sizes.  To explore potential drivers of inter-annual growth variation within the model, we used both intrinsic and extrinsic predictors. Intrinsic predictors (i.e. intra-population drivers) were the fixed effects of Age and AAC, which account for expected age-related trends, bias, and selectivity (Morrongiello and Thresher 2015), while the random effects of FishID and Year were utilized (Table 3 and Supplementary materials-'intrinsic predictors' for details). Extrinsic predictors (i.e. environmental predictors) were mean, maximum and minimum annual SST (°C) ('mean, max and min SST', respectively), annual mean salinity ('salinity', psu) and annual mean chlorophyll-a ('chl-a', mg m -3 ) ( Table 3 and Supplementary materials-'extrinsic predictors' for details). Due to the lack of sex data, sex was not used as growth predictor; nevertheless, patterns of sexual dimorphism are known for our species (i.e. female individuals grow larger in L. ehrenbergii, while males grow larger in P. maculosus) (Grandcourt et al. 2010(Grandcourt et al. , 2011, and were taken into account during the interpretation of the results.

Model selection
Growth, Age, and AAC were log-transformed to meet model assumptions. All variables (intrinsic and extrinsic) were mean centred to facilitate the convergence of the model and the interpretation of the random slopes (Morrongiello and Thresher 2015). Analyses were run with a two-stage approach: in the first stage, we built a base set of linear mixed models that encompassed a series of random effect and fixed intrinsic effect structures to identify the optimal intrinsic model (Tables S3-S6 and Supplementary materials-'models selection' for details of models selection, ranking and fitting). In the second stage of the analysis, we investigated whether environmental factors influenced growth. We extended the optimal intrinsic model for each region to include single or combinations between two environmental variables, which were fitted separately and ranked using Akaike's information criterion corrected for small sample sizes (AIC C ) (Burnham and Anderson 2004) and DAIC c values (Table S7, S8). Interactions between SST effects and between SST and salinity were not included in the model selection as a high level of correlation was expected. Additionally, collinearity between all the remaining combinations was explored using variance inflation factor (VIF) and models with VIF above 3 were removed from comparison (Hair et al. 1998). Finally, linear temporal growth trends were investigated by adding Year as a fixed effect to the optimal intrinsic model.

Population-level growth, body size and life span
The age range of L. ehrenbergii individuals was similar between regions, ranging between 38 days and 15 years in the Arabian Gulf and 27 days to 16 years in the Oman Sea; whereas P. maculosus individuals' age range differed between regions comprising of both younger and older Calendar year in which the increment was formed, quantifies inter-annual growth variability Age|FishID Random age slope for FishID random intercept, allows for individual-specific differences in the growth -age relationship Age|Year Random age slope for Year random intercept, allows for year-dependent differences in age-specific growth individuals in the Arabian Gulf compared to the Oman Sea (9 days to 31 years and 81 days to 22 years, respectively) ( Table 1). Although there is no generally applicable target value for precision given variation among species (Campana et al. 2001), the IAPE for L. ehrenbergii and P. maculosus was, respectively, 3.9 and 3.5% which is below the expected value of 5% in high-volume ageing facilities (Morison et al. 1998). Both L. ehrenbergii and P. maculosus had consistently smaller asymptotic size and maximum length (L max ) in the Arabian Gulf compared with the Oman Sea (Fig. 3a, b; Table 1). Specifically, L max of L. ehrenbergii and P. maculosus individuals was 19% (linear regression, F = 153.3, df = 1,49, p \ 0.001) and 13% smaller (F = 9.968, df = 1,34, p = 0.003), respectively. Maximum age (Age max ) was significantly lower in the Arabian Gulf compared to the Oman Sea for L. ehrenbergii (F = 4.362, df = 1,49, p = 0.042), but did not significantly differ between P. maculosus populations (F = 1.859, df = 1,34, p = 0.182) ( Table 1).

Extrinsic conditions
The magnitude and temporal trends of extrinsic parameters varied significantly between the Arabian Gulf and the Oman Sea (Fig. 4, Table S1). Both regions recorded relatively high inter-annual variation in min, mean and max SST, while inter-annual variation in salinity was high in the Arabian Gulf but minimal in the Oman Sea. In addition, both regions recorded relatively high inter-annual variation in chlorophyll-a concentrations, which was stronger in the Oman Sea than the Arabian Gulf (Fig. 4a). Despite no difference in annual mean SST between regions, the Arabian Gulf had higher intra-annual SST variation, with temperature significantly colder in winter and warmer in summer compared to the Oman Sea, as well as higher annual mean salinities, and lower annual mean chlorophylla concentrations (Fig. 4b, Table S1).

Growth predictors
Across species and regions, growth declined with age and salinity (Tables 4 and 5, see Table 4 for random effect structure details). Negative effects of salinity on growth rate were most apparent within the Oman Sea populations of L. ehrenbergii (-107.2% annual growth per unit increase in annual mean salinity [psu -1 ]), while Arabian Gulf populations of L. ehrenbergii and P. maculosus also showed lower growth with higher levels of salinity (-30.3% and -5.1% annual growth per unit increase in annual mean salinity [psu -1 ], respectively) ( Fig. 5a and Table 5).
There was little consistent effect of increasing temperature on growth rate between regions, though regional patterns were apparent ( Fig. 5b and Table 5). Water temperature within the Arabian Gulf was positively correlated to growth in both species, with increases in max SST having a predicted effect on L. ehrenbergii's growth rate of ? 12.7% per°C, and increases in mean SST an effect of ? 6.5% per°C in P. maculosus. There was no effect of temperature on the Oman Sea populations of L. ehrenbergii.

Discussion
We compared the growth of two common coral reef fish L. ehrenbergii and P. maculosus between the environmentally extreme Arabian Gulf and comparably benign Oman Sea, and examined the impact of temperature, salinity and primary productivity on the somatic growth of populations within each region. We found that both species were significantly smaller at age and attained a smaller maximum body size within the Arabian Gulf than within the Oman Sea. However, contrary to the expectation that extremes in temperature would be the main driver of somatic size reduction (i.e. consistent with the TSR), reductions in body size were mainly related to variation in salinity. Such results indicate that salinity may be a vital determinant of both species' growth trajectories. In comparison, temperature had a slightly positive effect on growth rates of Arabian Gulf populations of both species, with no measurable effect on Oman Sea fish populations.
In our study, growth declined with higher mean annual salinity across species and regions, indicating the impacts on demography of a highly saline environment are a reduction and, ultimately, truncation of life history. Indeed, together with temperature, food availability and photoperiod, salinity is known to be a major factor in determining fish development and growth (Boeuf and Payan 2001). Moreover, as osmoregulation is energetically costly (i.e. between 2 and 30% of daily energy expenditure in marine fish) (Boeuf and Payan 2001;Ern et al. 2014;Kultz 2015), changes in salinity may result in significant energy diversion from growth (Boeuf and Payan 2001). Here, we did not observe the expected crossing of growth curve trajectories between populations living in different thermal regime as predicted by the TSR (i.e. faster juvenile growth but earlier asymptotic growth for fish living in warmer environments) (Trip et al 2014), suggesting that salinity and perhaps even productivity are exhibiting a strong influence on differences in growth between regions. Interestingly, growth rates of populations within the Oman Sea, where salinity conditions are predominantly stable, were impacted more strongly by salinity fluctuations compared to populations within the Arabian Gulf, suggesting Arabian Gulf's populations may have higher capability of acclimation to seasonal and interannual changes in environmental conditions (Rummer and Munday 2017;D'Agostino et al. 2019). While numerous studies have examined the implications of increasing temperature and temperature variability on coral reef fish growth (e.g. Munday et al. 2008;Messmer et al. 2017;Taylor et al. 2019), to our knowledge there has been no investigation of the combined effect of increasing temperature and salinity on coral reef fish demography. This is despite the evidence of synergism between the two stressors (Claireaux and Lagarde 1999;Jian et al. 2003) and the likely occurrence of both stressors in already dry regions and hyper-saline (sensu saltier than ocean salinity) semi-enclosed seas in the near future (such as the Arabian Gulf, Red Sea and Mediterranean Sea) (Durack et al. 2012;Skliris et al. 2014;Zika et al. 2018).
Positive effects of high SST on growth rates of both species within the Arabian Gulf may show that high summer water temperatures do not exceed the population's typical temperature range in this region. For example, there    Brandl et al. (2020) showed that fish's thermal maximum tolerance limits largely surpass (i.e. between 1 and 2.5°C) the maximum summer temperature recorded in the region (i.e. 36°C). Finally, the positive effect of mean SST on Arabian Gulf populations of P. maculosus may suggest that intermediate temperatures experienced during milder winters or slightly warmer springs or autumns may benefit the growth rate of this species (Johansen et al. 2015;Djurichkovic et al. 2019). The smaller body sizes observed in Arabian Gulf fishes of both species may represent a life-history trade-off between metabolic demands (i.e. increased osmoregulatory cost) and size, with likely flow-on effects to population structure. Although not universally accepted, the gill-oxygen limitation theory (GOLT) states that as gills function as a two-dimensional surface with growth limited by geometrical constraints, any three-dimensional increases in fish body size, and consequent increases in oxygen demand, may not be met by adequate oxygen supply (Pauly and Cheung 2018, but see Audzijonyte et al. 2019). Consequently, larger-bodied individuals may be unable to compensate for increased metabolic demands associated with high salinity and temperature, due to the incapability of the respiratory system to supply enough oxygen. While we did not observe a strong negative effect of max SST (when dissolved oxygen is at its lowest and the oxygen demand at its highest) on fish growth (Shapiro Goldberg et al. 2019), large-bodied individuals are expected to have limited capacity to increase mass-specific maximum metabolic rate in warmer conditions (Messmer et al. 2017;Rubalcaba et al. 2020) and to be more likely to approach their maximum physiological capacity (Pauly and Cheung 2018). Hence, reduced individual body size in the Arabian Gulf may represent a life-history trade-off, whereby survival is enhanced through smaller body size and, therefore, reduced metabolic demands. Additionally, reduced body size was recently observed in cryptobenthic fish species in the Arabian Gulf compared to conspecific in the Oman Sea (Brandl et al. 2020). Here the authors, based on gut content analyses, suggested fish body size reduction was due to a mismatch between individuals elevated energetic costs of living in the extreme environment of the Arabian Gulf and reduced food quality and diversity within this habitat.
Although reductions in body size may play an important role in coping with variable and extreme temperature and salinity, other mechanisms may be involved in facilitating population stability. For example, recent work has highlighted the importance of behavioural and feeding plasticity in coping with the Arabian Gulf's extreme environmental conditions. The pale-tail damselfish (Pomacentrus trichrourus) appears able to mitigate bioenergetic inefficiency associated with the fluctuating and extreme water temperature within the Arabian Gulf by downregulating costly activities during winter and summer, while upregulating activity and increasing energy stores in spring (D'Agostino et al. 2019). In this respect, P. trichrourus, P. maculosus and Pomacentrus aquilis all show a degree of feeding plasticity in the Arabian Gulf (Shraim et al. 2017;D'Agostino et al. 2019), suggesting that such plasticity in feeding may be an important factor in Table 5 Fixed effect parameter estimates (± SE) and test statistics for optimal models describing intrinsic, temporal and environmental effects on Lutjanus ehrenbergii and Pomacanthus maculosus' growth (see Tables S1-8 for model selection and base  understanding growth rate and overall body size of all three species. Reductions in fish body size, within the magnitude of lowered size reported in the present work, are expected to have substantial consequences for trophic interactions, ecosystem function, fisheries and global protein supply (Shackell et al. 2010;Cheung et al. 2013). Fish body sizes are already reducing due to intensive fishing pressure (Stergiou 2002), while there is also evidence to show the potential additive effect of climate change and oxygen limitation on individual (and therefore population) body size spectrum (Cheung et al. 2013;van Rijn et al 2017;Rubalcaba et al. 2020). Notably, theoretical models predict that small reductions in individual body size (i.e. 4% over 50 years) may lead to a 50% increase in mortality as well as a 5-35% reduction in biomass and catch (Audzijonyte et al. 2013). However, most importantly, as fish fecundity exponentially increase with body size, lowered body size is expected to have substantial negative consequences on population reproductive output, replenishment and longterm persistence (Baudron 2014;Barneche et al. 2018), which will have cascading effects on the wider ecosystem, potentially impacting coastal economies and threatening food security.
Understanding the mechanisms by which fishes endure the Arabian Gulf's extreme and variable environment will be vital to understand how low latitude coral reef fish populations may cope with predicted changes in their environment. The present work shows that sub-tropical fish communities can persist within extreme environmental conditions, albeit with substantial trade-offs to their life history and demographic structure; individuals of both L. ehrenbergii and P. maculosus showed smaller size-at-age and lower maximum size in the Arabian Gulf compared with conspecifics in the Oman Sea. Our results suggest that to predict the effect of climate change on fish demography, Black lines indicate 95% CI. Note different scales on the y and x-axis. Predicted effects refer to the model outputs described in Table 4 and 5 Coral Reefs (2021) 40:433-446 443 the combined effect of osmoregulatory and thermal stress needs to be considered, especially in regions with limited oceanic exchange and predicted increase in SST and evaporation. Ultimately, to tease apart the effect of extreme temperature and salinity on growth of Arabian Gulf fish populations and establish the role of other potential coping mechanisms (i.e. feeding and behavioural plasticity) in mitigating environmental stressors, laboratory-based physiological-behavioural experiments combining multiple stressors and conditions, while assessing detailed energy budget, are needed.
Acknowledgements This work was funded by a PhD scholarship from the University of Nottingham to DD and supported by NYU Abu Dhabi Marine Biology Core Technology Platform. JB's funding was supported by the NYUAD Water Research Center, funded by Tamkeen under the NYUAD Research Institute Award (project CG007), while GC's funding were supported by the American University of Sharjah (research Grant FRG19-M-G74). Field fish collection was carried with permission of the Environment Agency, Abu Dhabi (protocol No. EAD-TMBS-RP-0), and according to the NYU-Abu Dhabi animal ethics guidelines. Finally, many thanks to Andrew MacColl, Dan Exton and the anonymous reviewers for useful advice and constructive suggestions.

Compliance with ethical standards
Conflict of interest On behalf of all authors, the corresponding author states that there is no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.