SCIENCE ADVANCES | RESEARCH ARTICLE ENVIRONMENTAL STUDIES Copyright © 2021 The Authors, some Estimating global biomass and biogeochemical cycling rights reserved; exclusive licensee of marine fish with and without fishing American Association for the Advancement of Science. No claim to Daniele Bianchi1*, David A. Carozza2, Eric D. Galbraith3,4, Jérôme Guiet1,3, Timothy DeVries5 original U.S. Government Works. Distributed The biomass and biogeochemical roles of fish in the ocean are ecologically important but poorly known. Here, we under a Creative use a data-constrained marine ecosystem model to provide a first-order estimate of the historical reduction of fish Commons Attribution biomass due to fishing and the associated change in biogeochemical cycling rates. The pre-exploitation global NonCommercial biomass of exploited fish (10 g to 100 kg) was 3.3 ± 0.5 Gt, cycling roughly 2% of global primary production License 4.0 (CC BY-NC). (9.4 ± 1.6 Gt year−1) and producing 10% of surface biological export. Particulate organic matter produced by ex- ploited fish drove roughly 10% of the oxygen consumption and biological carbon storage at depth. By the 1990s, biomass and cycling rates had been reduced by nearly half, suggesting that the biogeochemical impact of fisher- ies has been comparable to that of anthropogenic climate change. Our results highlight the importance of devel- oping a better mechanistic understanding of how fish alter ocean biogeochemistry. Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 INTRODUCTION state of the ocean, which also includes respiration by fish and Fish have not typically been considered as a dynamic component of production of sinking fecal pellets, which are integrated over the global ocean biogeochemistry. Instead, the dominant biogeochemical multicentennial time scale of ocean circulation. paradigm has focused on the interactions between phytoplankton, The animal biomass that existed before major human distur- zooplankton, and bacteria (1). Yet, in recent years, there has been a bances is a natural starting point from which the role of fish on growing appreciation for the potential role of fish, which is becom- ocean biogeochemistry can be evaluated. By providing estimates of ing increasingly important in light of marked alterations caused by carrying capacity that approximate pre-exploitation fish biomass, fishing, habitat degradation, warming, deoxygenation, and acidifi- scientific stock assessments (28) can shed light on these quantities, cation (2–5). Fish and other consumers can exert control on marine in particular in data-rich regions. Likewise, analysis of historical food webs via top-down interactions (6, 7) and affect nutrient records (24) and sedimentary proxies (29) has allowed reconstruction dynamics by storing, redistributing, and releasing carbon and of fish populations for periods that predate modern fisheries. These nutrients within and across ecosystems (8–10). But while the efforts have been complemented by numerical models, which allow biogeochemical role of fish has been documented in a variety of for extrapolation of sparse observations to entire ecosystems on environments, most notably freshwater and coral reef ecosystems large scales, coupling ecosystem dynamics to spatially varying envi- (8, 10), few studies have highlighted the role of fish on the ocean’s ronmental drivers (18, 30, 31). biological pump and seawater chemistry (11–14), and the influ- However, large uncertainties remain, as shown by the wide ence of fish has only begun to be represented in Earth system range of estimates of global fish biomass, mostly from numerical models (15, 16). models, which range from less than 1 Gt to more than 50 Gt Assessing the importance of fish in ocean biogeochemistry has (Fig. 1A). While discrepancies can partly be explained by differences primarily been hindered by a lack of knowledge regarding their in the definition of fish and the size classes considered, we find a abundance, distribution, and metabolic rates (17), properties that similar range even when comparing the same size classes and taxo- have changed markedly over the past century and remain uncertain nomic groups (Fig. 1B and section S1). These discrepancies reflect (14, 18–21). Historically, the development of fisheries was accom- variability in model formulations, ranging from application of panied by rapid and substantial biomass declines (5, 22, 23), most of ecological theory (18, 20) to intricate food web models (31), which occurred without direct scientific observation. In many compounded by the scarcity of observations available to constrain regions, industrial fisheries exploited ecosystems that had already models at the global scale. The persistent uncertainty has prompted been affected by traditional fishing, resulting in uncertain baselines a call for new, unified approaches and models that combine diverse (24). Climate change is now causing additional declines, although constraints to provide more robust estimates of the global biomass these remain much smaller than the direct fishing-related reduc- and biogeochemical cycling rates of fish (17). tions for the species targeted by fisheries (25, 26). Fishing has been Here, we take advantage of records of the intense removal of shown to have a substantial impact on ocean carbon fluxes by wild fish by fisheries to better constrain the dynamics of global fish diminishing the flux of sinking fish carcasses (27), but this does not communities and their role for biogeochemistry. We use historical capture the full impact of fish on the long-term biogeochemical reconstructions of fish catch and stock assessments (28, 32), in 1 combination with a biological-economic model (33, 34), to infer the Department of Atmospheric and Oceanic Sciences, University of California, Los Angeles, Los Angeles, CA, USA. 2Département de Mathématiques, Université du Québec à biomass, metabolic rate, and biogeochemical importance of fish Montréal, Montréal, Quebec, Canada. 3Institut de Ciència i Tecnologia Ambientals targeted by fisheries, their spatial distribution in the global ocean, (ICTA-UAB), Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, and their historical change (Materials and Methods). Because the Barcelona, Spain. 4Department of Earth and Planetary Science, McGill University, relationship between catch and biomass is complex (35) and has Montreal, Quebec, Canada. 5Department of Geography, University of California, Santa Barbara, Santa Barbara, CA, USA. been subject to debate (36), we use complementary diagnostics to *Corresponding author. Email: dbianchi@atmos.ucla.edu calibrate the dynamical model, forcing an interpretation of the data Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 1 of 10 SCIENCE ADVANCES | RESEARCH ARTICLE from the relatively unambiguous characteristics of catch peaks at global, large marine ecosystem (LME) and individual stock levels (Materials and Methods and fig. S1) (37, 38). We focus on reconstructing ocean fish biomass in the absence of fisheries, which, for simplicity, we refer to as “preindustrial biomass,” analogous to the preindustrial state that is widely used in climate research. Note that several additional drivers have likely altered fish biomass over the historical period in a way that we do not assess here. These include anthropogenic climate change (3, 25, 26), pollu- tion and habitat degradation (39), and early defaunation (4, 5, 40), such as the marked early reduction in anadromous fish species (9) and marine mammals (41). Thus, the term preindustrial should be considered to reflect the ocean state in the absence of fishing rather than in the absence of human activities. Our approach necessarily makes many assumptions, and the nature of the available data constraints requires a number of simpli- fications (summarized in table S6). For example, the data used here reflect environmental conditions during years of peak harvest (1960s to early 2000s) overprinted by natural variability. Our use of Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 LME catch data to constrain the fish biomass means that the model explicitly represents only those species exploited by fisheries, a factor that we highlight by referring throughout to “targeted species.” Our high seas estimates and extensions to nontargeted species assume that the overall fish biomass, including species that are not currently exploited, can be predicted by the same environmental dependences as for targeted species in LMEs. This strategy is based on an assumption that there is no fundamental bioenergetic difference between the portion of the ecosystem that we choose to exploit and that we do not, an assumption that bears future explora- tion (42, 43). Our analysis also ignores other potential environmental Fig. 1. Estimates of the biomass of fish and other consumers in the ocean. We factors such as iron limitation (44). Last, while we estimate param- include both (A) the biomass originally reported in the papers and (B) an estimate eter uncertainty using an ensemble of simulations, we cannot of the same biomass rescaled to the range between 1 g and 1000 kg, a typical quantify structural model uncertainty, which requires using an range for marine fish and other consumers, applying a conversion factor based on ensemble of models based on different architectures (26). We hope size-spectrum theory, as described in section S1. Note that, going from (A) to (B), biomass estimates could increase or decrease, depending on whether the original that our estimates can be improved by addressing these issues size range is smaller or larger than 1 g and 1000 kg, respectively. The complete list through future work. of values and the corresponding references are listed in table S1. The values reported as “this work” (highlighted in red) reflect fish biomass in the absence of fishing. Note the logarithmic scale on the vertical axis. RESULTS Model verification that is consistent across sources and with ecological theory. The Our ensemble reproduces realistic peak catches spanning more combined use of catch reconstructions and stock assessments offers than three orders of magnitude, across LMEs encompassing the a global, long-term perspective on marine ecosystems that is not range of temperatures and productivity found over most of the available from other sources. ocean (fig. S2). Peak catches are reasonably well correlated between The model resolves a globally gridded, size-structured ecosystem the ensemble and observational reconstructions (Pearson correla- that depends on local water temperature and primary production tion coefficient R = 0.56, P < 10−4). The remaining variability (fig. (33) and is coupled to an economic model that simulates dynami- S2A) is not surprising, given uncertainties in the catch data (32) and cally interactive fish catches (34). We include targeted invertebrates the model simplifications, and suggests an important role for man- (e.g., squid) as “fish” due to their ecological similarities with true fish, agement. For example, the model overestimates fish catch in a cluster but note that the fraction of total catch from invertebrates is only around of warm-water LMEs along the Australian shelf (left side of fig. S2A), 10% (32). Although the model does simulate variable recruitment, where stringent quota-based regulations limited fishing effort (37). The our analysis considers only the sizes larger than plankton (>10 g), ensemble also produces C:B ratios in line with observations (Materials because the mortality rates of ichthyoplankton are large, variable, and Methods and fig. S2, B to D), with an average value of 0.21 ± 0.11 year−1 and poorly constrained. Thus, our estimates do not include larval as compared to 0.22 ± 0.20 year−1 in stock assessment data (28), fish, which can functionally be described as zooplankton. indicating that it simulates realistic fish production rates. We carry out the model calibration by producing a large ensemble When forced by a continuous expansion of technology, the of simulations and objectively selecting model variants with catch model reproduces a trajectory of development, peak, and decline of histories and catch-to-biomass (C:B) ratios that are in best agree- global fisheries that parallels observations of global fish catch, ment with observations (34). Our most important constraints come biomass, and effort (Fig. 2) (23, 32, 45, 46). The preindustrial fish Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 2 of 10 SCIENCE ADVANCES | RESEARCH ARTICLE biomass can then be estimated by backtracking the biomass trajec- 20% of fish remaining) occur in high-biomass, cold-water regions, tory to the beginning of the simulations, i.e., in the absence of fish and weaker depletions (approximately 60% remaining) in less catches (Fig. 2B). productive, low-biomass ecosystems (Fig. 3C). This pattern reflects the progression of model fisheries from initially profitable, biomass-­ Estimated global preindustrial biomass rich but slow-growing ecosystems, mostly located in high latitudes, and losses due to fishing to ecosystems with less initial biomass but faster growth rates, mostly In the absence of fishing, we obtain a global biomass for targeted located in the tropics and subtropics (38). While the open-access fish from 10 g to 100 kg of 3.3 ± 0.5 Gt (Table 1). About half of this dynamic omits important social aspects of fisheries, such as preindustrial biomass (1.6 ± 0.2 Gt) is found within LMEs, which, management, which played a role in the development of real-world despite covering only 20% of the ocean surface, include the most fisheries, it nonetheless captures the main progression of observed productive areas of the ocean. The largest preindustrial biomass catches (23, 38). densities are found within coastal regions and marginal seas with expansive continental shelves, extending further into the open Biogeochemical role of targeted fish in the ocean ocean in the productive mid and high latitudes (Fig. 3A). Preindustrial We estimate the biogeochemical impact of fish by considering the biomass density is much lower in the vast oligotrophic areas of the rate at which energy, in the form of biomass, passes through the fish ocean’s subtropical gyres, reflecting a combination of low primary community, and refer to it as the biomass cycling rate. The model production and dissipation of energy by small organisms (47). indicates that targeted fish were responsible for cycling biomass at a By the time of the global peak catch, the simulated biomass has globally integrated rate of 9.4 ± 1.6 Gt year−1, with about half of this decreased to 47 ± 20% of the preindustrial values averaged over cycling (4.3 ± 0.7 Gt year−1) occurring in LMEs (Table 1). The fraction Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 LMEs (Fig. 3, B and C), falling within the range suggested by recon- of primary production consumed by targeted fish was approximately structions of global stock depletions for the past century (46). The 2% when globally integrated, 1.1 ± 0.2% when globally averaged, model indicates that stronger biomass depletions (approximately and 1.0 ± 0.2% when averaged locally over LMEs, reflecting the high trophic level position of fish in the marine food web. This flow rep- resents biomass that passes through fish and is returned to the envi- ronment, as a combination of dissolved and particulate forms, and reflects gross metabolic demand plus egestion. This metabolic demand is strongly correlated with biomass: More abundant fish popula- tions require larger amounts of energy. Thus, regions hosting abundant fish stocks (Fig. 3A) also cycle large amounts of biomass (Fig. 4A). As with biomass, the fraction of primary production that is cycled through fish also shows distinctive regional patterns (Fig. 4B). In relatively cold, productive ecosystems, targeted fish processed up to 4% of the energy fixed by phytoplankton, compared to values of less than 1% in oligotrophic regions. These regional patterns can be explained by a combination of primary production and tem- perature effects (fig. S3). Where productivity is high, a larger portion of photosynthesis is cycled through fish. This dependence is further modulated by temperature; colder regions show a greater utiliza- tion of net primary production by fish (fig. S3). This dependence reflects structural assumptions of the model regarding the response of trophic webs and energy transfer under different environ- mental conditions (37, 47), as constrained by the model optimiza- tion. Nutrient and chlorophyll-rich low-temperature waters tend to host larger phytoplankton cells at the base of the food web com- pared to warm, oligotrophic waters (48). In the model, this re- sults in relatively short food webs, where primary production is more efficiently transferred to fish and other consumers before Fig. 2. Estimated temporal evolution of fish catch and biomass. (A) Fish catch being dissipated (37, 47). integrated over large marine ecosystems (LMEs) (Mt year−1). The thick red line indicates At peak catch, driven by the widespread reduction of biomass, the Sea Around Us Project (SAUP) (32) catch summed over LME, and the thin red the cycling rate of targeted fish declines to 5.5 ± 0.9 Gt year−1 globally lines indicate the 50% uncertainty on the SAUP estimates. The thick and thin solid and 2.5 ± 0.4 Gt year−1 over LMEs, corresponding to a reduction from red lines show the central estimate and the range of catch from SAUP reconstructions, the unfished level of approximately 40% when spatially averaged, respectively, and the horizontal dashed line shows the global peak catch over LME similar to the biomass decline (around 50%). The reduction of fish from SAUP (109 Mt year−1). (B) Global fish biomass (Gt) from the optimized ensem- ble of simulations. In each panel, in (A) and (B), the solid black line shows the mean cycling at the time of global peak catch is more pronounced in cold, of the ensemble of simulations, and the three blue shadings include 50% (25 to high-latitude LMEs compared to less productive low-latitude regions, 75%), 80% (10 to 90%), and 90% (5 to 95%) of the model results. All model simulations to a similar degree, or even stronger than the biomass reduction, and the SAUP catch have been aligned so that the global peak catch occurs at suggesting potential nonlinear interactions between biomass deple- time t = 0 (vertical dashed line). tion and temperature (Fig. 4C). Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 3 of 10 SCIENCE ADVANCES | RESEARCH ARTICLE Table 1. Ensemble model estimates of biomass (Gt) and cycling rates (Gt year−1) of fish targeted by fisheries. All values represent means from the optimized ensemble and have been spatially integrated (biomass and cycling rates) or averaged [cycling rates relative to net primary production (NPP)] over LME regions (i.e., excluding the high seas) and over the global ocean. Preindustrial ocean (top rows) refers to the ocean in the absence of fishing; peak catch (bottom rows) refers to the year when maximum catch is reached, considering each LME independently. LME Global Cycling rate Cycling rate Biomass (Gt) % of NPP* Biomass (Gt) % of NPP* (Gt year−1) (Gt year−1) Preindustrial ocean Commercial 1.6 ± 0.2 4.3 ± 0.7 1.0 ± 0.2 3.3 ± 0.5 9.4 ± 1.6 1.1 ± 0.2 Total 3.3 ± 1.6 8.5 ± 3.4 2.0 ± 08 6.9 ± 3.6 18.9 ± 7.8 2.2 ± 0.9 Peak catch† Commercial 0.6 ± 0.2 2.5 ± 0.4 0.7 ± 0.1 1.1 ± 0.2 5.5 ± 0.9 0.7 ± 0.1 *Ratios of cycling rates to NPP were calculated at each model grid point and then spatially averaged.   †Values determined at the time of the global peak catch integrated over LME regions in the model. Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 Roughly one-fifth of the biomass ingested by fish returns to the 600 years, making it one of the most efficient natural carbon se- environment as fecal pellets (49), an important source of large questration mechanisms in the ocean (51). organic particles from the surface ocean (11). For the unfished ocean, this implies a fecal flux of 1.9 Gt year−1, roughly one order of Extrapolation to nontargeted fish magnitude larger than the sinking carcass flux estimated in (27). The results discussed above focus on targeted fish, which are explicitly Relative to a satellite-based estimate of the total particle export (48), simulated by the model and constrained with catch and biomass the overall contribution of targeted fish remains small, on average, observations. We also extrapolate the results to the sum of targeted around 2%. However, because fish fecal pellets sink at speeds that are and nontargeted species by carrying out a set of simulations that orders of magnitude faster than small particles (up to approximately approximate the whole spectrum of animals within the model size 1000 m day−1) (11), the expected contribution of fish-produced fecal range (10 g to 100 kg; see section S4). Targeted fish are assumed to have pellets becomes increasingly significant at depth. An idealized calculation access to a fraction of the total photosynthetic energy transferred across based on a data-constrained representation of marine particle flux (sec- food webs, estimated as 0.58 ± 0.22 (Materials and Methods). The tion S6 and table S5) (50) suggests that, at a depth of 200 m, the upper remaining energy is implicitly transferred to nontargeted consumers boundary of the ocean’s mesopelagic zone, fish fecal pellets will be in the same size range, assuming that they are identical to targeted twice as important as at the base of the euphotic zone. This ampli- fish in terms of their bioenergetics. By repeating the simulations with fication rises to a fivefold increase at 1000 m, a horizon below which this fraction set to 1, the biomass and metabolic rates are thereby extra­ exported carbon and nutrients would be sequestered for hundreds polated to the sum of targeted and nontargeted species (section S4). to thousands of years (51). At these depths, our simple calcu- The results of these additional simulations suggest that the total lation suggests that fish fecal pellets from targeted species could con- animal biomass and cycling rates in the unfished ocean were roughly tribute approximately 10% of the total global sinking particle flux twofold those of targeted species (Table 1 and Fig. 6). While this (table S5). extrapolation is weakly constrained (section S4), it suggests a pre- We use an ocean circulation inverse model (52), and a simple industrial global biomass of consumers between 10 g and 100 kg of representation of the sinking and degradation of organic particles, 3.3 ± 1.7 Gt over LME regions and 6.9 ± 3.6 Gt globally (Fig. 6). to illustrate the imprint of fish-produced fecal pellets on seawater Together, this animal biomass would have cycled organic matter at chemistry (Fig. 5 and Materials and Methods). Decomposition of rates of 8.5 ± 3.4 and 18.9 ± 7.8 Gt year−1, representing approximately fast-sinking particles preferentially occurs in the abyssal ocean, 4% of primary production when integrated globally, 2.0 ± 0.8% when where the respiration of fish-produced particles could reach up to averaged globally, and 2.2 ± 0.9% when averaged over LMEs. 20% of the total oxygen utilization (Fig. 5A). This oxygen deficit accumulates along the pathway of the deep ocean circulation, increasing from the North Atlantic to the North Pacific (Fig. 5B), DISCUSSION and is further modulated by the patterns of surface export (48). This When extrapolated to the size range between 1 g and 1000 kg respiration is accompanied by an accumulation of inorganic (section S1), our results indicate a preindustrial biomass of 5.0 Gt nutrients and carbon at depth, which, based on our model, would for targeted fish and 10.5 Gt for the sum of all fish (Fig. 1B). While contribute approximately 10% of the biological nutrient and carbon this is on the high end of previous estimates, it lies within the pub- sequestration below 1000 m depth, consistent with the ideal- lished range of 5 to 20 Gt from recent global models (19, 20) and ized calculation presented above. The total carbon sequestration meta-analyses (21) for the biomass of fish and other consumers in by fish-produced particles is approximately 100 GtC, or about 10% the ocean (Fig. 1B). of the total deep carbon sequestration by the biological pump (51). Although the biomass directly cycled by 10 g to 100 kg fish in the The sequestration time of fish-produced fecal pellets is approximately absence of fishing (18.9 Gt year−1) is only about 4% of the total Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 4 of 10 SCIENCE ADVANCES | RESEARCH ARTICLE Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 Fig. 3. Estimated biomass of fish targeted by fisheries (size range of 10 g to Fig. 4. Estimated biomass cycling rate, and its historical reduction, for fish 100 kg) and its historical reduction. (A) Preindustrial biomass of targeted fish (t km−2). targeted by fisheries (size range of 10 g to 100 kg). (A) Biomass cycling rate (B) Biomass of targeted fish (t km−2) at the time of global peak catch. (C) Fraction (%) (t km−2 year−1) by targeted fish in the preindustrial state (i.e., in the absence of fish- of biomass remaining at the time of the global peak catch, relative to the preindus- ing). (B) Biomass cycling rate of targeted fish, relative to net primary production (%). trial biomass, averaged over every LME. All maps are based on means from the (C) Reduction (%) of the biomass cycling rate by targeted fish at the time of the optimized ensemble of simulations. global peak catch, averaged over LMEs. All maps are based on means from the optimized ensemble of simulations. biomass generated by phytoplankton (53), fish have a significant potential to modify the fluxes of carbon and nutrients in the ocean 2000 m, corresponding to about 10 to 15% of the total oxygen utili- (Figs. 5 and 6), by forming fast-sinking fecal pellets (11), or via hor- zation at these depths. If the flux of fish fecal pellets had simply been izontal (9) and vertical migrations (13, 16). reduced proportionally to the fish cycling rate reduction estimated For example, our estimate suggests that fish feces are an important here, it would imply a sizable reduction of respiration in the deep and efficient component of particle export, the biogeochemical ocean. While this form of oxygen utilization is not dominant at significance of which would be expected to increase with depth. those depths, even for the unfished ocean, it is still an important Because fish fecal pellets sink much faster and further than smaller shift to consider, for example, in the deep Pacific Ocean, where the particles such as phytoplankton aggregates and zooplankton fecal accumulated effect of respiration drives oxygen concentrations close pellets (11), they could account for more than 20% (considering or even below the thresholds for hypoxia (55). In this and compara- targeted and nontargeted species) of the deep ocean respiration and ble hypoxic regions, oxygen declines of few tens of mmol m−3 are carbon sequestration driven by the ocean’s biological pump. This sufficient to limit the habitat of marine organisms, requiring specific contribution is comparable in magnitude to other secondary export adaptations to life at low oxygen (55). Consequently, any process processes such as particle subduction and downward mixing, diel increasing available oxygen would strongly influence the ecology in vertical migrations, and ontogenetic zooplankton migrations and these regions. could be important for closing the deep ocean carbon budget (51). Changes in oxygen of this magnitude are also significant when In addition, if we consider that this export may have been altered by compared to observed deep ocean deoxygenation, on the order of a more than 30% in response to fishing pressures, the resulting effect few mmol m−3 over the past century (2). These reconstructions on the biological pump would rival in magnitude estimates of indicate that the contribution from global warming can only explain climate change impacts (54). half of the observed oxygen loss (3), requiring a significant compo- Furthermore, the magnitude of fish-mediated fluxes is sufficient nent due to unresolved biological processes (2). While a causal link to alter the sensitive balance of oxygen in the deep sea. We estimate between fisheries and deoxygenation cannot be firmly established that respiration of fecal pellets produced by targeted fish is respon- yet (15), our results suggest that the magnitude of fish-mediated sible for an average of about 20 mmol m−3 oxygen utilization below export makes it an integral part of oceanic oxygen regulation so that Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 5 of 10 SCIENCE ADVANCES | RESEARCH ARTICLE Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 Fig. 6. Summary of simulated fish biomass, metabolism, and biogeochemical impacts. (A) Preindustrial ocean (i.e., in the absence of fishing). (B) Ocean at the Fig. 5. Illustration of the impact of fish cycling on ocean biogeochemistry. The time of the global peak catch. In both panels, numbers in black show fish biomass two panels show estimates of the dissolved oxygen utilization caused by the in Pg; in red biomass cycling processes in Pg year−1, for all fish in bold, and for decomposition of sinking organic particles produced by targeted fish and other con- targeted fish in brackets (size range of 10 g to 100 kg for both groups). Processes sumers as a fraction (%) of the dissolved oxygen utilization caused by remineralization affecting ocean biogeochemistry consist of respiration, production of sinking of all sinking particles. The estimate reflects fish cycling rate in the unfished ocean particles, and diel vertical migrations (DVMs), all included in the metabolism (Fig. 4A). Details of the calculation are provided in section S7. (A) Zonal average. term. Total metabolism and particle production are explicitly quantified in this (B) Average for the deep ocean (all depths are greater than 1000 m). paper, but not DVM. See also Table 1 for numerical values. changes in fish populations would be expected to drive change in ocean oxygen concentrations of a similar magnitude to those among the consequences of rebuilding marine fisheries and ecosystems recently observed. That is, deoxygenation changes could be larger (56, 57). Evaluating the signature of fisheries on ocean biogeochem- than appreciated, as the drop in fish biomass (and thus respiration) istry is a complex problem that requires a better characterization of may have masked a substantial fraction of the effect. However, we the impacts of fish on productivity and particle export (15–17), and caution that the possibility for trophic cascades and shifts between of the ecosystem changes that follow the depletion, for example, communities (including phytoplankton and zooplankton) could changes in size structure and trophic cascades (7). Nonetheless, it is lead to nontrivial patterns of change (15). likely that changes in biomass and cycling rate caused by fisheries Our results provide a first-order quantification of global fish would reverberate through the food web to alter carbon, nutrient, biomass and metabolism and their decline caused by fisheries. and oxygen cycles relative to the unfished ocean (15), given the tight Generating this quantification required making several simplifying stoichiometric relationships between these elements. While a number assumptions regarding the marine ecosystem, leading to a number of of challenges remain (table S6), global ecosystem models will be essen- important caveats and uncertainties (table S6). While acknowledging tial to quantifying interactions between fisheries and biogeochemistry these necessary simplifications, we estimate that, at the time of and should aim at representing fish and other consumers as integral global peak catch, the simulated biomass of targeted fish was reduced components of Earth system cycles. Recent marine ecosystem model to less than half of its unfished state, and the cycling rate by about intercomparison and ensemble analyses are promising steps in this 40% over LMEs, with marked reductions in productive, cold-water direction (26). Given the large change in global fish communities ecosystems. These estimates are likely to underestimate the total that has resulted from industrial fishing and early defaunation (40), change, because they do not include the detrimental effects of habitat and the emerging effect of additional stressors such as ocean warm- degradation (4) and climate change (3), the latter of which has ing (25), the development of a better understanding of these bio- probably contributed an additional 4 to 5% fish biomass decline geochemical implications is an urgent undertaking. over the past century (25, 26). While classical fisheries science would argue that this level of biomass change is close to desired for achieving maximum sustainable MATERIALS AND METHODS yields (56), this overlooks the broader ecosystem implications of Model description and experimental design that degree of biomass change. Biogeochemistry is just one ecosystem We use the BiOeconomic mArine Trophic Size-spectrum (BOATS) process potentially affected by such change but should be considered model (33, 34), which consists of a size-based representation of fish Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 6 of 10 SCIENCE ADVANCES | RESEARCH ARTICLE ecology and life history, coupled to an “open-access” economic Biomass estimates are only available for scientifically assessed model of fishing effort. BOATS represents all targeted species with stocks and represent only a fraction of the targeted fish biomass in individual mass between 10 g and 100 kg, as three continuous, inde- the ocean. To overcome this limitation, we use the stock assessments pendent size spectra with different asymptotic sizes, which provide in the RAM database to define the C:B ratio at the scale of LMEs. a rough representation of fish diversity and are directly comparable The C:B ratio is computed from the general total catch (the to size-dependent catch reconstructions (32). While these species variable “TCbest”) and general total biomass (the variable “TBbest”) include invertebrates (e.g., mollusks and crustaceans), they are reported by the RAM Legacy Database (28). We combine these dominated by fish, and we simply refer to them as fish in the paper. ratios with the global SAUP catch data to constrain the model fish The model is implemented on a global, two-dimensional grid at 1° biomass. That is, a model with a realistic representation of fish spatial resolution and is forced with monthly climatological tempera- catches and C:B ratios would also produce a realistic representation tures and satellite-derived primary productivity from observations of fish biomass. We recognize that C:B ratios vary by species, representative of the 1990s to 2000s period (34). We use the same environmental conditions, and level of exploitation. To minimize forcings for simulations with and without fishing. Thus, our estimate the effects of these sources of variability, we select C:B ratios for of preindustrial fish biomass (i.e., in the absence of fishing) does not available stocks at the time of maximum catch, which tend to reflect include the impacts of anthropogenic climate change, which has been the inherent productivity of stocks [e.g., (37)], and aggregate them estimated to be relatively minor for the historical period (25, 26). over LMEs, where environmental conditions tend to be more uniform Biological and ecological processes (e.g., growth, mortality, than at the global scale. Accordingly, we determine the C:B ratio at reproduction, and recruitment) are parameterized on the basis of peak catch for 25 LMEs from 292 stocks by calculating the total relationships that draw from macroecological theory and observa- catch and dividing it by the total biomass in each LME, and taking Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 tions (33). The ecological model is coupled to a representation of the average over the highest 5 years of catch. Using the total catch fishing effort and catches based on the open-access Gordon-Schaefer and biomass from each LME ensures that the resulting C:B ratios fishery economics model (section S2) (34), which reasonably ap- are representative of the overall LME productivity, rather than proximates to the dynamics of fisheries under weak regulation, up reflecting individual stocks. The list of LME used, the number of to the time of the global peak catch in the 1990s (23, 38). Because stocks, and the corresponding average C:B ratios are summarized our focus is on the difference between fished and unfished biomasses, in table S4. rather than details of the historical transient, we simply use a single, high rate of technological progress of 7% year−1 throughout the Monte Carlo approach and optimized ensemble simulations, on the higher range of observational estimates (23). Without posterior constraints, the range of model parameters pro- Note that the hindcasts likely underestimate catches in the early duces model solutions that span several orders of magnitude for the decades of the simulations (Fig. 2A), primarily because preindustrial globally integrated fish biomass and catch (34). Following previous artisanal and subsistence fisheries are not explicitly represented. work (23, 33), we adopt a Bayesian Monte Carlo approach to select Among the model parameters, we define the fraction of primary a set of model parameters (“optimized ensemble”) producing solu- production that is potentially available to targeted fish, , as the tions that best agree with observational estimates of fish catch and energy input to the base of the food web that can ultimately support biomass at the scale of LMEs. The approach is schematically shown targeted species, and the vast majority of which is implicitly respired in fig. S1 and is summarized by the following steps: by zooplankton and bacteria (section S2.2). As default, we set the 1) We start from an initial set of N = 10,000 simulations (“Monte value of  equal to 1, but we reassess it as part of the model optimi- Carlo ensemble”), with parameters randomly assigned based on zation procedure. ranges drawn from the literature. Details of this Monte Carlo A summary of the model rationale, equations, and parameters ensemble are presented in (34, 23) and are summarized in section is presented in section S2. A summary of model assumptions and S3. Although we use the same Monte Carlo ensemble, the selection limitations is presented in table S6. of optimized members (described below) differs from the two prior works, leading to a new, larger ensemble that spans a broader range Observational constraints of parameters. We use two observational datasets to constrain model catch and 2) For each simulation in the Monte Carlo ensemble, we calculate biomass at the global scale. The first is the Sea Around Us Project the global peak catch as the average of the five largest annual catches database (SAUP) 2010 release (32), which contains reconstruction integrated over LMEs. On the basis of these peak catches, we esti- of global annual catches based on Food and Agriculture Organization mate the fraction of primary production potentially available to fish (FAO) data, encompassing industrial, artisanal, and recreational () that would produce in each simulation a global peak catch fisheries, as well as bycatches. The SAUP database spans the ocean equal to the SAUP estimate, 110 Mt year−1, summed over LMEs exclusive economic zones from 1950s to 2010, providing a global, (32). We then exclude all the runs that would require more energy long-term perspective on fish catches. The second dataset is the than potentially available from net primary production to sustain RAM Legacy Database version 4.3 (28), which contains a synthesis the global fish catch (i.e., all the runs for which  > 1). This leaves of more than a thousand stock assessments from the global ocean, us with N = 962 simulations. including biomass and catch data. 3) To preserve a realistic partitioning of fish captures by size, we We aggregate and analyze the SAUP data at the scale of 66 LMEs, constrain the catch for fish in the medium and large size groups to which provide ecologically consistent, regional-scale spatial units be in the observed range relative to fish in the small group (34). This that can be directly compared with spatially integrated model output. reduces the number of runs to N = 618. Following previous studies (37), we further consider the peak catch- 4) To ensure that global catches are supported by realistic rates of es as the average of the largest 5 years of catches in each LME. fish biomass production, we compare the model C:B ratios averaged Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 7 of 10 SCIENCE ADVANCES | RESEARCH ARTICLE over 25 LMEs to the observational estimate from the RAM database layer, and close to the global average) but acknowledge that this (28). We do so by only retaining runs for which a Kolmogorov-­ neglects a portion of remineralization that is driven by other Smirnoff test indicates a probability greater than 50% that the model mechanisms such as dissolved organic matter and vertically migrating C:B ratios follow the same distribution as observations. This selec- organisms (51). Below the euphotic zone, particles sink and decompose, tion leaves N = 66 runs. following a remineralization profile described by a power-law 5) To ensure that the selected runs reflect geographical patterns equation (section S6) (50). We specify the total surface particle of observed catches across a range of temperatures and productivi- export flux from a satellite-based observational synthesis (48) and ties, we select model solutions that produce similar peak catch at assume that these particles remineralize with a power-law exponent the scale of LMEs as the SAUP reconstructions. Specifically, we cal- equal to −0.7, a representative value for the ocean (50). The export culate the Pearson correlation coefficient R between modeled and of fish-produced particles is estimated based on the mean fish observed peak catches for each LME and take all solutions with egestion from the optimized ensemble. We assume that fish-produced R > 0.60 as our optimized ensemble. This leaves us with N = 31 solu- particles are formed above the euphotic depth (a conservative tions, with average  of 0.58 ± 0.22 and average C:B ratio over LME assumption), and sink with a velocity 10 times larger than the bulk of 0.21 ± 0.11 year−1. sinking speed of all particles, consistent with observations (11). We We finally rerun the optimized ensemble for 200 years without represent this rapid sinking by imposing a power-law exponent for catches to reach an equilibrium representative of the biomass of fish particle flux equal to −0.07 (section S6). We then use the ocean targeted fish in the absence of fishing, followed by a continuous circulation inverse model to solve separately for the three-dimensional 220-year increase of catchability at a rate of 7% year−1, mimicking steady-state concentration of dissolved inorganic carbon resulting the historical progression of global fisheries (23). In each of these from (i) the remineralization of all particles and (ii) the remineral- Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 runs, the fraction of primary production potentially available to ization of fish-produced particles (details on the model equations fish, , is set to the value required to produce the global catch of and solution are discussed in section S7). The results (Fig. 5) are 110 Mt year−1 from SAUP (section S4). To obtain an approximate presented as the oxygen utilization resulting from this remineraliza- estimate of the biomass of all consumers (that is, both targeted and tion, which is calculated by using a typical stoichiometric ratio of nontargeted species), we also run a second instance of this optimized oxygen to carbon of −150:106 (58). ensemble, where we set  = 1 in each run. For this second set of simulations, we assume that targeted and nontargeted consumers SUPPLEMENTARY MATERIALS follow the same dynamics and together represent all consumers Supplementary material for this article is available at https://science.org/doi/10.1126/ within the model size range of 10 g to 100 kg. sciadv.abd7554 Model diagnostics REFERENCES AND NOTES We focus our analysis on the biomass of fish, both in the unfished 1. C. Le Quere, S. P. Harrison, I. Colin Prentice, E. T. Buitenhuis, O. Aumont, L. Bopp, state and at the time of the global peak catch. We further calculate H. Claustre, L. Cotrim Da Cunha, R. Geider, X. Giraud, Ecosystem dynamics based the metabolic demand of fish and compare it to the net primary on plankton functional types for global ocean biogeochemistry models. Glob. Chang. Biol. 11, 2016–2040 (2005). production that ultimately sustains it. We use the new biomass 2. S. Schmidtko, L. Stramma, M. Visbeck, Decline in global oceanic oxygen content during production rate and the trophic efficiency from the model (see the past five decades. Nature 542, 335–339 (2017). section S5) and determine the biomass ingested by fish, and returned 3. L. Bopp, L. Resplandy, J. C. Orr, S. C. Doney, J. P. Dunne, M. Gehlen, P. Halloran, C. Heinze, directly to the environment as a combination of organic (e.g., fecal T. Ilyina, R. Séférian, J. Tjiputra, M. Vichi, Multiple stressors of ocean ecosystems in the 21st pellets and excretion) and inorganic (respiration) fluxes, but not century: Projections with CMIP5 models. Biogeosciences 10, 6225–6245 (2013). 4. J. B. C. Jackson, M. X. Kirby, W. H. Berger, K. A. Bjorndal, L. W. Botsford, B. J. Bourque, retained within the fish size spectrum. We express this cycling rate R. H. Bradbury, R. Cooke, J. Erlandson, J. A. Estes, Historical overfishing and the recent in units of wet biomass processed per unit area per unit time (i.e., collapse of coastal ecosystems. Science 293, 629–637 (2001). t km−2 year−1). As a proxy of the particle production by fish, we 5. J. A. Estes, J. Terborgh, J. S. Brashares, M. E. Power, J. Berger, W. J. Bond, S. R. Carpenter, assume that about 20% of the biomass cycling rate is returned to the T. E. Essington, R. D. Holt, J. B. C. Jackson, R. J. Marquis, L. Oksanen, T. Oksanen, R. T. Paine, E. K. Pikitch, W. J. Ripple, S. A. Sandin, M. Scheffer, T. W. Schoener, J. B. Shurin, environment as fecal pellets (49). Assuming a constant carbon A. R. E. Sinclair, M. E. Soulé, R. Virtanen, D. A. Wardle, Trophic downgrading of planet content of organic matter of 10 gC/cWB (18), we compare these earth. Science 333, 301–306 (2011). quantities to biogeochemical rates expressed in carbon units, such 6. R. T. Paine, Food webs: Linkage, interaction strength and community infrastructure. as generation of new biomass from photosynthesis and export of J. Anim. Ecol. 49, 666 (1980). organic matter from the upper ocean as particles, which we take 7. J. K. Baum, B. Worm, Cascading top-down effects of changing oceanic predator abundances. J. Anim. Ecol. 78, 699–714 (2009). from existing satellite-based estimates (48). 8. M. J. Vanni, Nutrient cycling by animals in freshwater ecosystems. Annu. Rev. Ecol. Syst. 33, 341–370 (2002). Oxygen utilization due to fish-produced particles 9. C. E. Doughty, J. Roman, S. Faurby, A. Wolf, A. Haque, E. S. Bakker, Y. Malhi, J. B. Dunning, We use an ocean inverse circulation model (52) to estimate the J.-C. Svenning, Global nutrient transport in a world of giants. Proc. Natl. Acad. Sci. U.S.A. 113, 868–873 (2015). impact of remineralization of fish-produced fecal pellets in the 10. J. E. Allgeier, D. E. Burkepile, C. A. Layman, Animal pee in the sea: Consumer-mediated ocean interior. The ocean inverse circulation model provides a nutrient dynamics in the world’s changing oceans. Glob. Chang. Biol. 23, 2166–2178 (2017). data-based estimate of ocean transport, which we use to solve for 11. G. K. Saba, D. K. Steinberg, Abundance, composition, and sinking rates of fish fecal pellets the steady-state distribution of dissolved biogeochemical tracers in the Santa Barbara Channel. Sci. Rep. 2, 716 (2012). (i.e., oxygen and carbon) resulting from remineralization of organic 12. P. C. Davison, D. M. Checkley Jr., J. A. Koslow, J. Barlow, Carbon export mediated by mesopelagic fishes in the northeast Pacific Ocean. Prog. Oceanogr. 116, 14–30 (2013). matter in the ocean interior. We assume that remineralization is 13. D. Bianchi, E. D. Galbraith, D. A. Carozza, K. A. S. Mislan, C. A. Stock, Intensification mainly driven by the flux of organic particles produced in the of open-ocean oxygen depletion by vertically migrating animals. Nat. Geosci. 6, euphotic zone (here set equal to 74 m, the depth of the second model 545–548 (2013). Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 8 of 10 SCIENCE ADVANCES | RESEARCH ARTICLE 14. R. W. Wilson, F. J. Millero, J. R. Taylor, P. J. Walsh, V. Christensen, S. Jennings, M. Grosell, 40. D. J. McCauley, M. L. Pinsky, S. R. Palumbi, J. A. Estes, F. H. Joyce, R. R. Warner, Marine Contribution of fish to the marine inorganic carbon cycle. Science 323, 359–362 (2009). defaunation: Animal loss in the global ocean. Science 347, 1255641 (2015). 15. J. Getzlaff, A. Oschlies, Pilot study on potential impacts of fisheries-induced changes 41. H. Whitehead, Estimates of the current global population size and historical trajectory in zooplankton mortality on marine biogeochemistry. Global Biogeochem. Cycles 31, for sperm whales. Mar. Ecol. Prog. Ser. 242, 295–304 (2002). 1656–1673 (2017). 42. T. R. Anderson, A. P. Martin, R. S. Lampitt, C. N. Trueman, S. A. Henson, D. J. Mayor, J. Link, 16. O. Aumont, O. Maury, S. Lefort, L. Bopp, Evaluating the potential impacts of the diurnal Quantifying carbon fluxes from primary production to mesopelagic fish using a simple vertical migration by marine organisms on marine biogeochemistry. Global Biogeochem. food web model. ICES J. Mar. Sci. 76, 690–701 (2019). Cycles 32, 1622–1643 (2018). 43. X. Irigoien, T. A. Klevjer, A. Røstad, U. Martinez, G. Boyra, J. L. Acuña, A. Bode, F. Echevarria, 17. G. K. Saba, A. B. Burd, J. P. Dunne, S. Hernández-León, A. H. Martin, K. A. Rose, J. Salisbury, J. I. Gonzalez-Gordillo, S. Hernandez-Leon, S. Agusti, D. L. Aksnes, C. M. Duarte, D. K. Steinberg, C. N. Trueman, R. W. Wilson, Toward a better understanding of fish-based S. Kaartvedt, Large mesopelagic fishes biomass and trophic efficiency in the open ocean. contribution to ocean carbon flux. Limnol. Oceanogr. 66, 1639–1664 (2021). Nat. Commun. 5, 3271 (2014). 18. S. Jennings, F. Melin, J. L. Blanchard, R. M. Forster, N. K. Dulvy, R. W. Wilson, Global-scale 44. E. D. Galbraith, P. Le Mézo, G. S. Hernandez, D. Bianchi, D. Kroodsma, Growth limitation predictions of community and ecosystem properties from simple ecological theory. of marine fish by low iron availability in the open ocean. Front. Mar. Sci. 6, (2019). Proc. R. Soc. B Biol. Sci. 275, 1375–1383 (2008). 45. R. A. Watson, W. W. L. Cheung, J. A. Anticamara, R. U. Sumaila, D. Zeller, D. Pauly, 19. S. Jennings, K. Collingridge, Predicting consumer biomass, size-structure, production, Global marine yield halved as fishing intensity redoubles. Fish. 14, 493–503 (2013). catch potential, responses to fishing and associated uncertainties in the world’s marine 46. Y. Rousseau, R. A. Watson, J. L. Blanchard, E. A. Fulton, Evolution of global marine fishing ecosystems. PLOS ONE 10, e0133794 (2015). fleets and the response of fished resources. Proc. Natl. Acad. Sci. U.S.A. 116, 12238–12243 20. M. B. J. Harfoot, T. Newbold, D. P. Tittensor, S. Emmott, J. Hutton, V. Lyutsarev, M. J. Smith, (2019). J. P. W. Scharlemann, D. W. Purves, Emergent global patterns of ecosystem structure and 47. J. H. Ryther, Photosynthesis and fish production in the sea. The production of organic function from a mechanistic general ecosystem model. PLOS Biol. 12, e1001841 (2014). matter and its conversion to higher forms of life vary throughout the world ocean. 21. Y. M. Bar-On, R. Phillips, R. Milo, The biomass distribution on Earth. Proc. Natl. Acad. Sci. U.S.A. Science 166, 72–76 (1969). 115, 6506–6511 (2018). 48. J. P. Dunne, R. A. Armstrong, A. Gnanadesikan, J. L. Sarmiento, Empirical and mechanistic 22. R. A. Myers, B. Worm, Rapid worldwide depletion of predatory fish communities. Nature models for the particle export ratio. Global Biogeochem. Cycles. 19, (2005). Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 423, 280–283 (2003). 49. B. Kooijman, Dynamic Energy Budget Theory for Metabolic Organisation (Cambridge Univ. 23. E. D. Galbraith, D. A. Carozza, D. Bianchi, A coupled human-Earth model perspective Press, ed. 3, 2009). on long-term trends in the global marine fishery. Nat. Commun. 8, 14884 (2017). 50. F. Primeau, On the variability of the exponent in the power law depth dependence 24. H. K. Lotze, B. Worm, Historical baselines for large marine animals. Trends Ecol. Evol. 24, of POC flux estimated from sediment traps. Deep. Res. Part I Oceanogr. Res. Pap. 53, 254–262 (2009). 1335–1343 (2006). 25. C. M. Free, J. T. Thorson, M. L. Pinsky, K. L. Oken, J. Wiedenmann, O. P. Jensen, Impacts 51. P. W. Boyd, H. Claustre, M. Levy, D. A. Siegel, T. Weber, Multi-faceted particle pumps drive of historical warming on marine fisheries production. Science 363, 979–983 (2019). carbon sequestration in the ocean. Nature 568, 327–335 (2019). 26. H. K. Lotze, D. P. Tittensor, A. Bryndum-Buchholz, T. D. Eddy, W. W. L. Cheung, 52. T. DeVries, The oceanic anthropogenic CO2sink: Storage, air-sea fluxes, and transports E. D. Galbraith, M. Barange, N. Barrier, D. Bianchi, J. L. Blanchard, L. Bopp, M. Büchner, over the industrial era. Global Biogeochem. Cycles 28, 631–647 (2014). C. M. Bulman, D. A. Carozza, V. Christensen, M. Coll, J. P. Dunne, E. A. Fulton, S. Jennings, 53. P. G. Falkowski, R. T. Barber, V. Smetacek, Biogeochemical controls and feedbacks M. C. Jones, S. Mackinson, O. Maury, S. Niiranen, R. Oliveros-Ramos, T. Roy, on ocean primary production. Science 281, 200–206 (1998). J. A. Fernandes, J. Schewe, Y. J. Shin, T. A. M. Silva, J. Steenbeek, C. A. Stock, P. Verley, 54. C. Laufkötter, M. Vogt, N. Gruber, Long-term trends in ocean plankton production J. Volkholz, N. D. Walker, B. Worm, Global ensemble projections reveal trophic and particle export between 1960-2006. Biogeosciences 10, 7373–7393 (2013). amplification of ocean biomass declines with climate change. Proc. Natl. Acad. Sci. U.S.A. 55. L. A. Levin, Manifestation, drivers, and emergence of open ocean deoxygenation. 116, 12907–12912 (2019). Annu. Rev. Mar. Sci. 10, 229–260 (2018). 27. G. Mariani, W. W. L. Cheung, A. Lyet, E. Sala, J. Mayorga, L. Velez, S. D. Gaines, T. Dejean, 56. R. Hilborn, R. O. Amoroso, C. M. Anderson, J. K. Baum, T. A. Branch, C. Costello, M. Troussellier, D. Mouillot, Let more big fish sink: Fisheries prevent blue carbon C. L. De Moor, A. Faraj, D. Hively, O. P. Jensen, Effective fisheries management instrumental sequestration—Half in unprofitable areas. Sci. Adv. 6, eabb4848 (2020). in improving fish stock status. Proc. Natl. Acad. Sci. U.S.A. 117, 2218–2224 (2020). 28. D. Ricard, C. Minto, O. P. Jensen, J. K. Baum, Examining the knowledge base and status 57. C. M. Duarte, S. Agusti, E. Barbier, G. L. Britten, J. C. Castilla, J.-P. Gattuso, R. W. Fulweiler, of commercially exploited marine species with the RAM legacy stock assessment T. P. Hughes, N. Knowlton, C. E. Lovelock, Rebuilding marine life. Nature 580, 39–51 database. Fish. 13, 380–398 (2011). (2020). 29. L. Kavanagh, E. Galbraith, Links between fish abundance and ocean biogeochemistry 58. L. A. Anderson, On the hydrogen and oxygen content of marine phytoplankton. Deep Sea as recorded in marine sediments. PLOS ONE 13, e0199420 (2018). Res. Part I Oceanogr. Res. Pap. 42, 1675–1680 (1995). 30. S. Jennings, J. L. Blanchard, Fish abundance with no fishing: Predictions based 59. C. Costello, D. Ovando, T. Clavelle, C. Kent Strauss, R. Hilborn, M. C. Melnychuk, on macroecological theory. J. Anim. Ecol. 73, 632–642 (2004). T. A. Branch, S. D. Gaines, C. S. Szuwalski, R. B. Cabral, D. N. Rader, A. Leland, Global fishery 31. V. Christensen, M. Coll, J. Buszowski, W. W. L. Cheung, T. Frölicher, J. Steenbeek, prospects under contrasting management regimes. Proc. Natl. Acad. Sci. U.S.A. 113, C. A. Stock, R. A. Watson, C. J. Walters, The global ocean is an ecosystem: Simulating 5125–5129 (2016). marine life and fisheries. Glob. Ecol. Biogeogr. 24, 507–517 (2015). 60. V. Christensen, C. J. Walters, R. Ahrens, J. Alder, J. Buszowski, L. B. Christensen, 32. D. Pauly, D. Zeller, Catch reconstructions reveal that global marine fisheries catches are W. W. L. Cheung, J. Dunne, R. Froese, V. Karpouzi, Database-driven models of the world’s higher than reported and declining. Nat. Commun. 7, 10244 (2016). Large Marine Ecosystems. Ecol. Model. 220, 1984–1996 (2009). 33. D. A. Carozza, D. Bianchi, E. D. Galbraith, The ecological module of BOATS-1.0: 61. L. Tremblay-Boyer, D. Gascuel, R. Watson, V. Christensen, D. Pauly, Modelling the effects A bioenergetically constrained model of marine upper trophic levels suitable for studies of fishing on the biomass of the world’s oceans from 1950 to 2006. Mar. Ecol. Prog. Ser. of fisheries and ocean biogeochemistry. Geosci. Model Dev. 9, 1545–1565 (2016). 442, 169–185 (2011). 34. D. A. Carozza, D. Bianchi, E. D. Galbraith, Formulation, general features and global 62. J. R. Watson, C. A. Stock, J. L. Sarmiento, Exploring the role of movement in determining calibration of a bioenergetically-constrained fishery model. PLOS ONE 12, e0169763 (2017). the global distribution of marine biomass using a coupled hydrodynamic—Size-based 35. T. A. Branch, O. P. Jensen, D. Ricard, Y. Ye, R. A. Y. Hilborn, Contrasting global trends ecosystem model. Prog. Oceanogr. 138, 521–532 (2014). inmarine fishery status obtained from catches and from stock assessments. Conserv. Biol. 63. R. Proud, N. O. Handegard, R. J. Kloser, M. J. Cox, A. S. Brierley, D. Demer, From 25, 777–786 (2011). siphonophores to deep scattering layers: Uncertainty ranges for the estimation of global 36. D. Pauly, R. Hilborn, T. A. Branch, Fisheries: Does catch reflect abundance? Nature 494, mesopelagic fish biomass. ICES J. Mar. Sci. 76, 718–733 (2019). 303–306 (2013). 64. J. H. Brown, J. F. Gillooly, A. P. Allen, V. M. Savage, G. B. West, Toward a metabolic theory 37. C. A. Stock, J. G. John, R. R. Rykaczewski, R. G. Asch, W. W. L. Cheung, J. P. Dunne, of ecology. Ecology 85, 1771–1789 (2004). K. D. Friedland, V. W. Y. Lam, J. L. Sarmiento, R. A. Watson, Reconciling fisheries catch 65. M. Edwards, J. P. W. Robinson, M. J. Plank, J. K. Baum, J. L. Blanchard, Testing and ocean productivity. Proc. Natl. Acad. Sci. U.S.A. 114, E1441–E1449 (2017). and recommending methods for fitting size spectra to data. Methods Ecol. Evol. 8, 57–67 38. J. Guiet, E. D. Galbraith, D. Bianchi, W. W. L. Cheung, Bioenergetic influence (2017). on the historical development and decline of industrial fisheries. ICES J. Mar. Sci. 77, 66. P. R. Boudreau, L. M. Dickie, Biomass spectra of aquatic ecosystems in relation to fisheries yield. 1854–1863 (2020). Can. J. Fish. Aquat. Sci. 49, 1528–1538 (1992). 39. V. H. Smith, S. B. Joye, R. W. Howarth, Eutrophication of freshwater and marine 67. R. W. Sheldon, A. Prakash, W. H. Sutcliffe, The size distribution of particles in the ocean. ecosystems. Limnol. Oceanogr. 51, 351–355 (2006). Limnol. Oceanogr. 17, 327–340 (1972). Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 9 of 10 SCIENCE ADVANCES | RESEARCH ARTICLE 68. J. F. Gillooly, J. H. Brown, G. B. West, V. M. Savage, E. L. Charnov, Effects of size 90. T. DeVries, T. Weber, The export and fate of organic matter in the ocean: New constraints and temperature on metabolic rate. Science 293, 2248–2251 (2001). from combining satellite and oceanographic tracer observations. Global Biogeochem. Cycles 69. S. K. Ernest, B. J. Enquist, J. H. Brown, E. L. Charnov, J. F. Gillooly, V. M. Savage, 31, 535–555 (2017). E. P. White, F. A. Smith, E. A. Hadly, J. P. Haskell, Thermodynamic and metabolic 91. T. Weber, J. A. Cram, S. W. Leung, T. DeVries, C. Deutsch, Deep ocean nutrients imply effects on the scaling of production and population energy use. Ecol. Lett. 6, 990–995 large latitudinal variation in particle transfer efficiency. Proc. Natl. Acad. Sci. U.S.A. 113, (2003). 8606–8611 (2016). 70. M. Hartvig, K. H. Andersen, J. E. Beyer, Food web framework for size-structured 92. V. M. Savage, J. F. Gillooly, J. H. Brown, G. B. West, E. L. Charnov, Effects of body size populations. J. Theor. Biol. 272, 113–122 (2011). and temperature on population growth. Am. Nat. 163, 429–441 (2004). 71. O. Maury, J.-C. Poggiale, From individuals to populations to communities: A dynamic 93. H. Andersen, J. E. Beyer, Size structure, not metabolic scaling rules, determines fisheries energy budget model of marine ecosystem size-spectrum including life history diversity. reference points. Fish Fish. 16, 1–22 (2013). J. Theor. Biol. 324, 52–71 (2013). 94. C. Barnes, D. Maxwell, D. C. Reuman, S. Jennings, Global patterns in predator-prey size 72. D. Pauly, The Sea Around Us Project: Documenting and communicating global fisheries relationships reveal size dependency of trophic transfer efficiency. Ecology 91, 222–232 impacts on marine ecosystems. AMBIO J. Hum. Environ. 36, 290–295 (2007). (2010). 73. L. von Bertalanffy, Problems of organic growth. Nature 163, 156–158 (1949). 95. E. Bissinger, D. J. S. Montagnes, D. Atkinson, Predicting marine phytoplankton maximum 74. H. Gislason, N. Daan, J. C. Rice, J. G. Pope, Size, growth, temperature and the natural growth rates from temperature: Improving on the Eppley curve using quantile mortality of marine fish. Fish Fish. 11, 149–158 (2010). regression. Limnol. Oceanogr. 53, 487–493 (2008). 75. E. L. Charnov, H. Gislason, J. G. Pope, Evolutionary assembly rules for fish life histories. 96. D. Dahlberg, A review of survival rates of fish eggs and larvae in relation to impact Fish Fish. 14, 213–224 (2013). assessments. Mar. Fish. Rev. 41, 1–12 (1979). 76. R. J. H. Beverton, S. J. Holt, On the Dynamics of Exploited Fish Populations (Springer Science 97. K. H. Andersen, M. Pedersen, Damped trophic cascades driven by fishing in model marine & Business Media, 1957), vol. 11. ecosystems. Proc. R. Soc. B 277, 795–802 (2010). 77. H. S. Gordon, The economic theory of a common-property resource: The fishery. 98. H. Pulkkinen, S. Mäntyniemi, Y. Chen, Maximum survival of eggs as the key parameter J. Polit. Econ. 62, 124–142 (1954). of stock–recruit meta-analysis: Accounting for parameter and structural uncertainty. 78. M. B. Schaefer, Some aspects of the dynamics of populations important to the Can. J. Fish. Aquat. Sci. 70, 527–533 (2013). Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 management of the commercial marine fisheries. Inter-American Trop. Tuna Comm. 99. R. A. Watson, A database of global marine commercial, small-scale, illegal and unreported Bull. 1, 23–56 (1954). fisheries catch 1950–2014. Sci. Data 4, 1–9 (2017). 79. D. Pauly, M. L. D. Palomares, “An empirical equation to predict annual increases in 100. K. T. Frank, B. Petrie, J. S. Choi, W. C. Leggett, Trophic cascades in a formerly cod- fishing efficiency,” Working Paper 2010-07, UBC Fisheries Centre, Vancouver, BC, dominated ecosystem. Science 308, 1621–1623 (2005). Canada, 2010. 80. D. Squires, N. Vestergaard, Technical change and the commons. Rev. Econ. Stat. 95, Acknowledgments 1769–1787 (2013). Funding: D.B. and J.G. acknowledge support from NASA awards 80NSSC17K0290, and 81. O. Maury, B. Faugeras, Y.-J. Shin, J.-C. Poggiale, T. Ben Ari, F. Marsac, Modeling 80NSSC21K0420. California Department of Resources-Ocean Protection Council grant C0100400, environmental effects on the size-structured energy flow through marine ecosystems. and the Alfred P. Sloan Foundation. E.D.G. acknowledges support from the Spanish Ministry of Part 1: The model. Prog. Oceanogr. 74, 479–499 (2007). Economy and Competitiveness, through the María de Maeztu Programme for Centres/Units of 82. S. Jennings, J. K. Pinnegar, N. V. C. Polunin, T. W. Boon, Weak cross-species relationships Excellence in R&D (MDM-2015-0552). This project has received funding from the European between body size and trophic level belie powerful size-based trophic structuring in fish Research Council (ERC) under the European Union’s Horizon 2020 research and innovation communities. J. Anim. Ecol. 70, 934–944 (2001). programme (grant agreement no. 682602). Computational resources were provided by the Extreme 83. R. A. Locarnini, A. V Mishonov, J. I. Antonov, T. P. Boyer, H. E. Garcia, World Ocean Atlas Science and Engineering Discovery Environment (XSEDE) through allocation TG-OCE170017. Author 2005, Vol. 1: Temperature, S. Levitus, Ed. (U.S. Government Printing Office, 2006). contributions: D.B. designed the study, with inputs from all coauthors. D.A.C. and D.B. wrote the 84. K. H. Andersen, N. S. Jacobsen, K. D. Farnsworth, The theoretical foundations for size ecosystem model code. T.D. wrote the biogeochemical model code. J.G. and D.B. conducted the spectrum models of fish communities. Can. J. Fish. Aquat. Sci. 73, 575–588 (2015). analysis of stock assessment data. D.B. performed and analyzed the model simulations. All 85. D. Pauly, V. Christensen, Primary production required to sustain global fisheries. Nature coauthors contributed to the analysis of the results. D.B. wrote the article, with contributions from 374, 255–257 (1995). all coauthors. Competing interests: The authors declare that they have no other competing 86. J. H. Martin, G. A. Knauer, D. M. Karl, W. W. Broenkow, VERTEX: Carbon cycling interests. Data and materials availability: All data needed to evaluate the conclusions in the in the northeast Pacific. Deep Sea Res. Part A Oceanogr. Res. Pap. 34, 267–285 (1987). paper are present in the paper and/or the Supplementary Materials. 87. I. Kriest, A. Oschlies, On the treatment of particulate organic matter sinking in large-scale models of marine biogeochemical cycles. Biogeosciences 5, 55–72 (2008). Submitted 10 July 2020 88. T. DeVries, F. Primeau, Dynamically and observationally constrained estimates Accepted 18 August 2021 of water-mass distributions and ages in the global ocean. J. Phys. Oceanogr. 41, Published 8 October 2021 2381–2401 (2011). 10.1126/sciadv.abd7554 89. O. Duteil, W. Koeve, A. Oschlies, D. Bianchi, E. Galbraith, I. Kriest, R. Matear, A novel estimate of ocean oxygen utilisation points to a reduced rate of respiration in the ocean Citation: D. Bianchi, D. A. Carozza, E. D. Galbraith, J. Guiet, T. DeVries, Estimating global biomass interior. Biogeosciences 10, 7723–7738 (2013). and biogeochemical cycling of marine fish with and without fishing. Sci. Adv. 7, eabd7554 (2021). Bianchi et al., Sci. Adv. 2021; 7 : eabd7554 8 October 2021 10 of 10 Estimating global biomass and biogeochemical cycling of marine fish with and without fishing Daniele BianchiDavid A. CarozzaEric D. GalbraithJérôme GuietTimothy DeVries Sci. Adv., 7 (41), eabd7554. • DOI: 10.1126/sciadv.abd7554 View the article online https://www.science.org/doi/10.1126/sciadv.abd7554 Downloaded from https://www.science.org at Universitat Autnoma de Barcelona on May 18, 2022 Permissions https://www.science.org/help/reprints-and-permissions Use of this article is subject to the Terms of service Science Advances (ISSN ) is published by the American Association for the Advancement of Science. 1200 New York Avenue NW, Washington, DC 20005. The title Science Advances is a registered trademark of AAAS. Copyright © 2021 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC).