ICES Journal of Marine Science (2020), 77(5), 1854–1863. doi:10.1093/icesjms/fsaa044 Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 Original Article Bioenergetic influence on the historical development and decline of industrial fisheries 1,2 J. Guiet *, E. D Galbraith1,3,4, D. Bianchi1, and W. W. L. Cheung5 1 Institut de Ciència i Tecnologia Ambientals (ICTA-UAB), Universitat Autònoma de Barcelona, 08193 Cerdanyola del Vallès, Barcelona, Spain 2 Department of Atmospheric and Oceanic Sciences, University of California, Los Angeles, CA 90095, USA 3 Institució Catalana de Recerca i Estudis Avançats (ICREA), Barcelona 08010, Spain 4 Department of Earth and Planetary Science, McGill University, Montreal, QC, Canada 5 Nippon Foundation-Nereus Program and Changing Ocean Research Unit, Institute for the Oceans and Fisheries, University of British Columbia, Vancouver, BC V6T1Z4, Canada *Corresponding author: tel: þ1 310 600 0662; e-mail: jerome.c.guiet@gmail.com. Guiet, J., Galbraith, E. D, Bianchi, D., and Cheung, W. W. L. Bioenergetic influence on the historical development and decline of indus- trial fisheries. – ICES Journal of Marine Science, 77: 1854–1863. Received 11 June 2019; revised 11 February 2020; accepted 13 February 2020; advance access publication 10 May 2020. The global wild capture fishery expanded rapidly over the 20th century as fishing technology improved, peaking in the 1990s as most fisheries transitioned to fully- or over-exploited status. Historical records for individual large marine ecosystems (LMEs) tend to echo this same pro- gression, but with local variations in the timing and abruptness of catch peaks. Here, we provide objective descriptions of these catch peaks, which generally progressed from high- to low-latitude LMEs, and attribute the temporal progression to a combination of economic and eco- logical factors. We show that the ecological factors can be remarkably strong by using a spatially resolved, observationally-constrained, cou- pled macroecological-economic model to which we impose an idealized, globally homogeneous increase in catchability. The globally-uniform technology creep produces a spatial progression of fishing from high-to-low latitudes that is similar to observations, primarily due to the im- pact of temperature on ecosystem metabolism. In colder LMEs, low respiration rates allow the build-up of larger pristine standing stocks, so that high-latitude fisheries are profitable earlier, at lower levels of fishing technology. We suggest that these bioenergetic characteristics con- tributed significantly to the historical progression of this human-ecological system. Keywords: environmental driver, fisheries catch, large marine ecosystems, marine ecosystem model Introduction In general, the historical trajectories of most fisheries have Historical reconstructions of wild fish catches show a steady in- been characterized by an early developing phase during which crease in the global total until the 1990s, followed by a peak and catches increased, followed by a peak when the fishery was fully plateau, or perhaps a slight decline (Pauly and Zeller, 2016; exploited or over-exploited, followed by a decline in some dura- Watson, 2017). However, the smooth progression of globally tion (Grainger and Garcia, 1996). The decline could have been summed catches hides dramatically different histories of fish mitigated or eliminated where management efforts were able to stock exploitation at the local scale. For example, while catches in prevent overfishing (Caddy and Cochrane, 2001; Hilborn and the Western Central Pacific and Indian oceans are still growing, Ovando, 2014). As shown by Grainger and Garcia (1996), it catches in the Northern Atlantic, Mediterranean Sea and many appears that the sequential increase and decline in catches has other regions have been shrinking (Food and Agriculture shifted geographically over time, including a migration from Organization, 2014). higher- to lower-latitude regions. C International Council for the Exploration of the Sea 2020. V This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/ licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited. Bioenergetic influence on the development of fisheries 1855 Such regional differences in the historical trajectories of wild emerge exclusively from environmental spatial variability. We fish catch could have emerged from two types of factors: (i) spa- then compare both the observed and modelled progressions with tial variation in human factors, including fish capture and proc- spatial patterns of temperature and PP and discuss mechanisms essing technology, market forces, access to capital, or the by which environmental drivers could have complemented socio- imposition of regulations (Caddy and Cochrane, 2001; Watson economic drivers in shaping the historical development and de- et al., 2013) and/or (ii) spatial variation in the environment- cline in industrial fisheries. dependent characteristics of fish communities (Stock et al., 2017; Van Denderen et al., 2018). Unravelling the contributions of these Material and methods Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 different factors to the historical increase and decline in catches Observed catch in LMEs could provide an improved understanding of the underlying LMEs account for 95% of the total wild fish catches while cover- human-ecosystem dynamics. ing 22% of the global ocean surface (Stock et al., 2017). These Among other human factors, geographical differences in the regions are differentiated based on environmental characteristics accumulation of productive capital and the development of fish- such as bathymetry, hydrography, PP and trophic relationships ing technology clearly played a major role (Gelchu and Pauly, (Sherman and Duda, 1999), making them appropriate regional 2007). The level of capital and technologies for industrial fisheries units to investigate the impact of the environment on the regional to develop was first reached near North America after World War variability of historical time-series of fish catches. For each LME, II, and subsequently near Europe and Japan. From there, they ex- the time-series of total catches over the period 1950–2013 has panded to new ecosystems in other regions as fish demand in- been reconstructed by SAUP (Pauly and Zeller, 2015) including creased and as stocks were depleted (Anticamara et al., 2011; Bell industrial, artisanal, subsistence and recreational fishing as well as et al., 2017). This diffusion of fisheries from high-to-low latitudes discarded catches. The high-seas regions outside LMEs are not in- (Swartz et al., 2010) is likely to have contributed to the sequential cluded in the present study because of their low contribution to progression of catches from high-to-low latitude. As high-latitude total wild catch. fisheries became depleted and as management measures were im- There are 66 LMEs, all but one of which (the Central Arctic posed to reduce effort, there was also a shift of fishing vessels to Ocean) are associated with historical total catch time-series. developing nations at lower latitudes (Alder and Sumaila, 2004). Following Friedland et al. (2012), we exclude from our study nine Thus, while management stimulated a stabilization or decline in LMEs where data are incomplete or unreliable. These regions are catches at higher latitudes, it may indirectly have led to an in- the Antarctic, the Beaufort Sea, Chukchi Sea, East China Sea, East crease in catches at lower latitudes (Watson et al., 2014). Siberian Sea, Hudson Bay, Kara Sea, Laptev Sea and Yellow Sea. At the same time, environmental features undoubtedly con- The remaining 56 LMEs present various shapes of catch time- tributed to variations between ecosystems by influencing the series (see Figure 1 and Supplementary Information S1). Most available fish biomass. For instance, it has been shown that fish time-series show a widespread increase in catches explained by in- production can be limited by the amount of energy available to creasing effort and an improvement in technologies from 1950 coastal ecosystems from primary production (PP) (Jennings onwards, but the detailed trajectories display diverse characteris- et al., 2008; Chassot et al., 2010), as well as water temperature (T), which influences the metabolic rates of ectotherms (Brown tics, with some regions passing through multiple large peaks of et al., 2004; Clarke and Fraser, 2004). In multiple regions of the catches, like in the California Current (# 3), the Humboldt global ocean, temperature and PP differences have been shown to Current (# 13), or the Faroe Plateau (# 60). In these ecosystems, influence yields, along with factors related to the export of energy small pelagic fish are an abundant component of landings and from pelagic food webs (Chassot et al., 2010; Friedland et al., sharp variations in these target species control fisheries yields 2012). In addition, there is growing appreciation for how the in- (Pauly and Zeller, 2015). Other regions present an increase in fluence of climate change on T and PP will alter fish catches, with catches up to a plateau, like in the East Bering Sea (# 1). These both single-stock and ecosystem-based models predicting signifi- examples do not represent an exhaustive list of historic catch cant declines under continued ocean warming (Carozza et al., time-series trajectories [see Conti et al. (2012) for more details 2019; Free et al., 2019; Lotze et al., 2019). However, to our knowl- about the trends]. The most widespread feature of the time-series edge, the impact that spatial variations in these environmental in LMEs is a monotonic increase in catches up to a maximum, conditions may have had on the historical progression of fisheries like in the Bay of Bengal (# 34) or the Sulu-Celebes Sea (# 37), of- development has not previously been considered. ten followed by a monotonic decrease, like in the Sea of Japan (# In this analysis, we provide objective descriptions of the catch 50) or the Central Pacific American Coast (# 11). A rebound in reconstructions provided by the Sea Around Us Project (SAUP; the catches may happen after this monotonic decrease, like in the Pauly and Zeller, 2015) at the scale of Large Marine Ecosystems South Brazil Shelf (# 15). (LMEs), 66 eco-regions of the World Ocean with distinct envi- In this study, we focus on the catch time-series that present a ronmental characteristics. We first compare the resulting spatial clear peak in the total catch rate. We use objective criteria to ex- progression of catch with an LME-level Human Development tract these peaks from the 64 years of available data, following Index (HDI; United Nations Development Programme, 2016), smoothing with a 6-year running average to filter out inter- with the expectation that a higher HDI would serve as a proxy for annual variations. An LME is determined to include a clear peak a greater degree of industrialization of fisheries. We also compare when the 21 years of highest catches in the time-series occur al- the observations with time-resolved numerical simulations using most sequentially within a 23-year period. With this criterion, we the global bioeconomic model BOATS (Carozza et al., 2016, discard LMEs where there is not a clear dominant peak, as in the 2017). In the model, globally uniform technological progress in Faroe Plateau (# 60), the North Brazil Shelf (# 17), and the the fishery serves as the primary driver of long-term changes in Aleutian Islands (# 65), and where the peak is embedded in a fish catches (Galbraith et al., 2017) so that spatial catch variations trend of rising catches, as in the California Current (# 3). We 1856 J. Guiet et al. retain 37 ecosystems out of the initial selection of 56 (Figure 1). (in twB=m2 =second, where q is the biomass catchability per unit Note that, with this method, a few of the identified peaks may be of effort in m2 =W=second) (Gordon, 1954; Schaefer, 1954): described as plateaux, without clear catch declines after the max- ima. The conclusions of this study have little sensitivity to varia- dE r c ¼ je ; (1) tions in the set of peaks selected (see Supplementary Information dt E S2), or the details of the method for extracting the features of the peaks (Metrics for catch time-series and Supplementary where je (W2 =m2 =$) is a relaxation parameter representing fleet Information S3). dynamics. The variation in effort per unit of profit (r  c) Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 decreases with increasing fishing effort (as 1=E). The inverse rela- Simulated catch with BOATS tionship between the rate of change and fishing effort reflects the The BOATS model simulates global wild capture fisheries on a 1 fact that the portion of the total profit to be gained by individual grid of the global ocean, as described in detail by Carozza et al. fishers shrinks with total effort, weakening the motivation to (2016, 2017). It consists of two modules, an ecological and an move to the fishing grounds and thus slowing the adjustment of economic module. effort (Gordon, 1954). The model represents the propagation of biomass B(m) from We assume that, in most LMEs, there was minimal regulation lower- to upper-trophic levels assuming a direct link between until some time after peak LME catches, since the motivation to individuals’ size m and trophic level. This propagation corre- establish effective management was commonly provided by de- sponds to the growth of individuals and is constrained by PP and clining catches following overfishing (Worm et al., 2009; Hilborn water temperature [see Environmental drivers (PP and T)]. Both and Ovando, 2014). Thus, the historical development, peak and variables directly influence the bioenergetics of individuals and initial decline in catches at the LME level can be approximated by thus indirectly control the biomass distribution. The ecological this open-access dynamics (Galbraith et al., 2017). The model module provides the monthly distribution of biomass B(m) simulates effort independently in each grid cell and, therefore, (in tonnes of wet biomass, twB=m2 ) of potential wild catch does not explicitly account for the spatial redistribution of effort available to fishers between a minimum and maximum size from depleted regions to developing ones. Instead, spatio- m 2 ½mmin ; mmax . temporal patterns emerge from differences in local dynamics The economic module simulates the dynamics of the effort E (Galbraith et al., 2017). (in W=m2 ) as a function of the profit, increasing in a grid cell Each BOATS simulation is a combination of a spin-up simula- when profit is positive and decreasing when profit is negative. tion that initializes the pristine (unfished) ocean, and a coupled Profit depends Ðon the harvest that can be achieved from the biological–economic simulation reproducing the sequential de- mmax biomass Btot / mmin BðmÞdm of wild catch available and is calcu- velopment and decline in fishing in different regions (Carozza lated as the balance between the costs c ¼ CE (C, cost per unit et al., 2017). An ensemble of five optimized sets of biological pa- effort in $=W=second) and revenues r ¼ pH (p, price per bio- rameter values is used to span parameter uncertainty for the pre- mass unit in $=twB) generated from the harvest H / qBtot E sent study, as in Galbraith et al. (2017). In the economic module, Figure 1. Total catch time-series per LME. Catch time-series from 1950 to 2013: blue lines, SAUP data; red lines, 6 years running average for the identification of the timing of catch peaks; black dots, 21 years of highest catches. Numbers refer to the list of LMEs detailed in Supplementary Table S1, and numbers in red correspond to the 37 time-series where a peak is recognized. Note that five North Atlantic LMEs are plotted separately in the top right due to the lack of space on the map. Bioenergetic influence on the development of fisheries 1857 the catchability parameter q increases homogeneously at a rate rq technology level of fleets fishing in the LME. We reconstruct the while cost and price (c and p) are held constant, since their histor- HDI time-series for each LME, HDILME ðyÞ, from the HDI of each ical role was likely secondary (Galbraith et al., 2017). country HDIcountry, and the historical catches per LME of each We analysed the catch time-series by taking the mean of the country H LME ðcountry; yÞ between the year y ¼ 1950 and 2013, five ensemble members and summing at the LME level. In addi- as provided by SAUP (Pauly and Zeller, 2015): tion, to explore the sensitivity of our results to different aspects of X the model, we repeated the experiments using two additional LME LME country HDIcountry H ðcountry; yÞ rates of catchability increase rq, and excluding the effects of tem- HDI ðyÞ ¼ X : (2) Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 country H LME ðcountry; yÞ perature and PP from specific components of the model (see Supplementary Information S4). The resulting time-series indicates the average level of develop- ment for each LME over time, due to changes in the relative con- Metrics for catch time-series tribution of different national fleets to the total catches. We Year of peak catches compare the mean HDI per LME between 1950 and 2013, LME The year of peak catches in each LME, ypeak , is estimated for both  LME , as well as variation DHDILME ¼ HDILME ð2013Þ  HDI observations and simulations. Observed catch time-series were HDILME ð1950Þ, with the ypeak LME . smoothed with a 6-year running average. For each time-series, the year of peak catch is defined as the first year when 95% of the Results maximum catch is reached. This criterion is used to reduce the Observed history of peak catches uncertainty in identifying the onset of long catch plateaux. We Among the 56 historical catch time-series analysed (Figure 1 and LME tested multiple methods of the selection of ypeak (see Supplementary Information S1), our objective criteria identify 37 Supplementary Table S3a) and found relatively small sensitivity catch peaks (see Supplementary Table S2 for results on different that would not influence our conclusions. LME sets of LMEs). The year of peak catches for each, ypeak , shows a discernible spatial progression between different parts of the Time-series slopes global ocean (see Figure 2a). The peaks occurred first in the 1960s LME The slopes of pre-peak increase, sincrease , and post-peak decrease, in the Northwest Atlantic. Tropical LMEs are generally the last to LME sdecrease , are also determined on observed time-series smoothed reach their peaks with some still developing in 2013, such as the with a 6-year running average and directly calculated from the Sulu-Celebes Sea (# 37). model without applying any smoothing. The time-series are nor- malized by the maximum yield reached, allowing comparison of Comparison with HDI the relative speed of development and decline before and after the Geographical variations in social and economic characteristics of peaks. The increasing and decreasing time segments are identified fishing nations have been widely documented (Gelchu and Pauly, by comparison to the maximal rates of change pre- and post-peak 2007; Anticamara et al., 2011; Bell et al., 2017). We therefore car- using the thresholds [dH=dt > maxðdH=dtÞ=2:5] and ried out a comparison of our catch peak years and the corre- [dH=dt < minðdH=dtÞ=2:5]. The slopes over the identified seg- sponding catch-weighted HDI in each LME by calculating their ments are then calculated by linear regression. This method was pairwise coefficients of determination. empirically devised to provide robust diagnostics (see When we compare the year of peak catch ypeak LME to the historical Supplementary Table S3b). level of societal development of LMEs encapsulated in HDI  LME , 2 we found a significant negative correlation (R ¼ 0:13 and p < Investigated drivers 0.05, see Figure 3 and summary Table 1). This is consistent with Environmental drivers (PP and T) the expectation that LMEs exploited at higher HDI would have PP and temperature are the environmental drivers of fish produc- reached peak catches earlier. tion in BOATS. Following Carozza et al. (2016), we use monthly cli- In addition, changes in HDILME between 1950 and 2013 can be matologies of temperature from the World Ocean Atlas 2005 constructed from changes in the composition of nations fishing (Locarnini et al., 2006). In each 1o  1o grid cell, we use average within the LME. We use the DHDILME ¼ HDILME ð2013Þ  temperature over the upper 75 m. For PP, we take the average of HDILME ð1950Þ to test for the invasion of long-distance fleets in three satellite-based net PP estimates (Behrenfeld and Falkowski, low-latitude waters. Relatively few LMEs appear to have had large 1997; Carr et al., 2006; Marra et al., 2007) to capture some of the invasions of long-distance fleets by this measure (see size of variability that exists in different models. Similar to temperature, we markers Figure 3 and Supplementary Information S5), among use monthly climatologies on a 1o  1o grid. For comparison to ob- them the Guinea Current (# 28), the Somalia Current (# 31), and served peaks, we take the annual mean of these drivers over LMES. the Red Sea (# 33). These invasions should have had a slight in- LME fluence on the correlation between ypeak  LME (Table 1). and HDI Human Development Index The HDI attempts to capture the average achievement in terms of Simulated history of peak catches life expectancy, level of knowledge, and standard of living and is The BOATS model simulates the temporal evolution of fisheries available at the country level (United Nations Development as driven by a globally homogeneous increase in catchability, Programme, 2016). Here, it is used as a proxy for the level of which represents the net outcome of all aspects of technological technological development of the fisheries of distinct nationalities progress on the ability to catch the available fish with a given involved in a given LME. Weighted according to the catches of effort. Given the global uniformity of catchability change, the each nation operating in the LME, we take it as a proxy of the differences in simulated peak years among LMEs depend only on 1858 J. Guiet et al. Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 Figure 3. Socio-economic influence on the observed and simulated catch peaks. Year of maximum catches yLME peak as a function of mean  LME in the 37 selected LMEs (y ¼ 40:8x þ 2020; R2 ¼ 0:13; HDI p < 0:05). The diameter of the markers is proportional to DHDILME ¼ HDILME ð2013Þ  HDILME ð1950Þ, which expresses changes in the composition of nations fishing within the LME, red positive, blue negative. Table 1. Summary of linear regressions: slope, R2, and p-value. Variable 1 Variable 2 Slope R2 p-Value Human Development Obs. yLME peak 40.8 0.13 <0.05  LME ) Index (HDI Figure 2. Observed and simulated years of maximum catch. Spatial Obs. yLME Sim. yLME 9:32  101 0.30 < 103 peak peak progression of the year of peak catches yLMEpeak in the 37 LMEs selected: (a) observations, (b) simulations, and (c) simulated vs. observed year Temperature (T) Sim. yLME peak 1.53 0.44 < 105 of peak catches (y ¼ 9:32  101 x þ 1700; R2 ¼ 0:30; Obs. yLME peak 7:73  101 0.33 < 103 p < 103 ). The solid line is a linear fit through the observed and LME Sim. sincrease 1:62  104 0.12 <0.05 simulated timings of peaks, and the dotted line is the 1:1 slope. The Obs. sLME increase 6:89  104 0.09 >0.05 empty circles indicate LMEs in the Southern Hemisphere, and the Sim. sLME increase 1:68  104 0.29 < 103 filled circles indicate LMEs in the Northern Hemisphere. Obs. sLME increase 1:87  103 0.33 <0.01 Primary production (PP) Sim. yLME peak 9:56  102 0.09 >0.05 Obs. yLME peak 1:46  102 0.01 >0.05 environmental differences between grid cells. Perhaps surpris- ingly, despite globally homogeneous technological and economic conditions, the simulations show a global spatial progression of fishing that is broadly similar to the observations, with peaks oc- effect that is absent in the model. In an attempt to detect this, we curring first at high latitudes, and subsequently at low latitudes removed particularly well-managed LMEs such as the Australian (see Figure 2b). The simulated LME peak years are significantly LMEs (# 39 and 41–45; Flood et al., 2014). However, this did not correlated with observations where a peak of catches is identified increase the coefficient of determination between simulations and (R 2 ¼ 0:30; p < 103 , Figure 2c and Table 1), despite the large observations (R 2 ¼ 0:29; p < 0:01, see Supplementary Figure number of confounding processes. This result holds for different S6b). sets of LMEs (Supplementary Table S2), as well as when we An additional consideration is that some LMEs contain small include all 56 LMEs, although it is weaker in this case regions with very high or low PP as inferred from satellite, a fact (R 2 ¼ 0:15; p < 0:01, see Supplementary Figure S6a). The that may challenge the model’s ability to correctly resolve biomass model suggests a slower sequential development of fishing than production. For example, in some LMEs, the PP is concentrated observed, as revealed by the larger spread of the years of peak in the vicinity of the mouth of major rivers, such as in the catch in the simulations (Figure 2c). This feature is related to the Patagonian Shelf (# 14), the Guinea current (# 28), or the rate of catchability increase, here rq ¼ 5%. A higher rate would Gulf of Thailand (# 35). Because the model does not resolve compress the interval between the first and the last occurrences of the movement of fish or their prey, it may overestimate fish LMEs maximum harvests, while a lower rate would have the op- density in these localized highly productive regions, leading to posite effect (see Supplementary Information S4). an earlier peak in catches compared with observations. Indeed, It might be expected that fisheries management has had an in- when we remove the most heterogeneous LMEs from the 37 fluence on the year of maximum catches in some ecosystems, an selected, the coefficient of determination increases substantially Bioenergetic influence on the development of fisheries 1859 (R 2 ¼ 0:43; p < 104 , see Supplementary Figures S6c and S7 (a) summarizing the heterogeneity of T and PP across each LME). We note that the inclusion of unreported catch estimates in the SAUP reconstructions influences the results to some degree. When only reported catches are considered to determine the year of peak catch, the coefficient of determination between simula- tions and observations is weaker (R 2 ¼ 0:15; p < 0:05, see Supplementary Figure S6d). However, the removal of unreported Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 LME catches does not strongly alter the sequence of ypeak for most LMEs and the low coefficient of determination is largely attribut- able to the Insular Pacific-Hawaiian LME (# 10). Thus, our (b) results generally apply to reported catch time-series as well. Comparison with environmental drivers Given that the model broadly reproduces the observed trends of LME ypeak , we next consider how the bioenergetic mechanisms at play in the model might have contributed to the historical timing of catch peaks, while bearing in mind that socio-economic factors are certain to have also contributed in reality. As with the HDI above, relationships between the catch peaks and the environmental drivers can be roughly assessed by calcu- (c) lating their pairwise coefficients of determination. In the model, LME ypeak is strongly correlated with T in the 37 LMEs with unambig- uous catch peaks (R 2 ¼ 0:44; p < 105 , see Figure 4a, in black) but is not significantly correlated with PP (see Table 1 and Supplementary Figure S8). A very similar pattern emerges in the LME observations, where no correlation is found between ypeak and PP, but a stronger coefficient of determination is found with T (R 2 ¼ 0:33; p < 103 see Figure 4a, in blue, or Supplementary Information S9 for the same figure with LME numbers, and LME Table 1). The correlation between ypeak and T is largely indepen- dent of the set of LMEs selected, with the correlations remaining significant when we consider alternate sets of LMEs, or even all Figure 4. Influence of temperature on the features of observed and 56 LMEs with reliable catch time-series (see Supplementary simulated catch peaks. Features in selected LMEs as a function of the mean annual water temperature T: (a) year of peak yLME peak , simulations Information S2 and S10). in black (y ¼ 1:53x þ 124; R2 ¼ 0:44; p < 105 ) and observations Surprisingly, the correlation between HDI  LME and y LME is not as peak in blue (y ¼ 7:73  101 x þ 1974; R2 ¼ 0:33; p < 103 ); (b) slopes LME strong as the correlation between ypeak and LME temperature (see sLME 4 increase , simulations in black (y ¼ 1:62  10 x þ 2:54  10 ; 2 Table 1). In addition, while HDI LME correlates with ecosystem tem- R ¼ 0:12; p < 0:05) and observations in blue (y ¼ 6:89 104 x 2 perature, when we carried out a multiple linear regression of ypeak LME þ4:42  102 ; R2 ¼ 0:09; p > 0:05); (c) slopes sLME decrease , simulations in with both temperature and HDI  LME , there was no improvement black (y ¼ 1:68  104 x  1:71  102 ; R2 ¼ 0:29; p < 103 ) and over the regression using only temperature (see adjusted R2 in observations in blue (y ¼ 1:87  103 x  7:08 102 ; R2 ¼ 0:33; p < 0:01). In (a), the size of the black markers reflects the standing Supplementary Table S5a) and only temperature was a significant biomass at the beginning of exploitation. The solid lines are linear predictor (Supplementary Table S5b). Thus, HDI does not appear fits through the observed and simulated features of the peaks. LME to be a better predictor of ypeak than ecosystem temperature. We also tested whether or not the slopes associated with Although these correlations do not provide direct evidence of catch peaks are correlated with the environmental drivers. The LME a causal relationship, they are consistent with metabolism-based modelled sincrease is only weakly correlated with T, and the ob- expectations that water temperature should exert a direct bioen- served slope is not significantly correlated (see Table 1, Figure 4b, ergetic control on fish abundance and production rates, thereby and Supplementary Information S2 for sensitivity tests), suggest- interacting with fishing. A strong role for temperature in the LME ing negligible influence on sincrease . In contrast, the rate of rela- model is further confirmed by an alternate simulation where LME tive decrease sdecrease is significantly correlated with T, both in all effects of temperature are removed (see Supplementary the model (R2 ¼ 0:29; p < 103 , see Figure 4c, in black) and Information S4). In this case, the correlation between modelled in observations (R 2 ¼ 0:33; p < 0:01, see Figure 4c, in blue, LME and observed ypeak disappears. The lack of a significant role for Supplementary Information S9 for the figure with LME num- PP does not reflect a lack of importance, since it plays a major bers, Table 1, and Supplementary Information S2 for sensitivity role in limiting fish production in the model Carozza et al. tests). Note that, because of the irregularity of observed catch (2016); however, PP does not emerge as a useful predictor of the time-series, the determination of the slopes remains more uncer- timing of harvest peaks. tain than the determination of the peak years (see Supplementary Water temperature can alter many vital aspects of marine ecto- Information S3). thermic animals, including feeding rates (Rall et al., 2012), 1860 J. Guiet et al. growth, mortality and reproduction (Brown et al., 2004; Clarke (a) and Fraser, 2004), and respiration (Clarke and Johnston, 1999) with consequent impacts on the dissipation rate of biomass energy (Guiet et al., 2016). These direct temperature effects are represented in the model by Arrhenius dependences (Carozza et al., 2016), which predict averaged rate increases of 60–110% for every 10 C temperature increase, for different parameter ensembles, and 87% for the ensemble mean. They affect growth, Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 reproduction, and mortality. BOATS does not explicitly includes the temperature-dependent respiration associated with activity and reproduction, but these processes are included implicitly in the losses due to temperature-dependent mortality (Carozza (b) et al., 2019). Water temperatures also influence the modelled tro- phic transfer of energy because cold waters tend to be character- ized by larger phytoplankton cells, which are in turn consumed by larger predators, shortening food webs (Stock and Dunne, 2010). In the model, this size of phytoplankton is computed from PP and temperature with an empirical formulation (Dunne et al., 2005). These temperature dependencies alter both the initial standing biomass and the replacement rate of fished biomass by recruitment and growth. Each of these two factors impacts the in- teraction with fishing effort in a different way, by modifying the year of peak catches and the rate of decline, respectively, as we (c) discuss next. Temperature influence on initial biomass In the pristine state (i.e. prior to fishing), simulated cold LMEs support a higher biomass density than warm LMEs (see cold warm Figure 5a, Btot > Btot and Figure 4a, size of black markers). In the model, T influences the pristine biomass through its effects on mortality, representative size of phytoplankton and growth rates. Among the effects, the temperature impact on mortality is the most important. Phytoplankton size and growth rates have a secondary impact (see sensitivity experiments in the Supplementary Information S4). The general model prediction Figure 5. Schematic representation of the history of simulated agrees with theoretical (Ursin, 1984) and other modelling studies catch time-series in cold and warm ecosystems. Comparison of (Jennings et al., 2008). simulated biomass decrease (a), total catch peak (b) and catch peaks normalized by the maximum catches and centred around the mid- All else being equal, high biomass densities would be expected catch peak (c): in a colder (blue) ecosystem, the Benguela Current to yield a greater catch for a given amount of effort and are there- (# 29), T ¼ 18 C; in a warmer (red) ecosystem, the Caribbean Sea fore more likely to be profitable with relatively primitive fishing (# 12), T ¼ 27 C. In green, the logarithm of the catchability technologies. Thus, cold locations with larger pristine biomass increases over years. While catchability increases, ecosystems with a cold=warm densities would be expected to have been exploited first. This can lower standing biomass Btot begin to be exploited when the be understood by considering a critical level of biomass Bcrit, de- biomass is equal to the critical biomass level allowing the onset of fined as the minimum biomass at which exploitation is profitable. fishing Bcrit ¼ C=pq, black dashed line. The vertical dashed lines Under the simple assumption of economic open access, this indicate the beginning of exploitation. In (a), the black vertical arrow occurs when the revenue r ¼ pH ¼ pqBcrit E (where p is the ex- () indicates the degree to which the cold water biomass Bcold is out of equilibrium with the evolving q at the point where fishing vessel price, H is the wild fish harvest, q is the catchability, and E begins on Bwarm. is the effort) is equal to the associated costs of fishing, c ¼ CE (C ¼ cost per unit effort; see model description in Material and Methods). This implies that Bcrit ¼ C=pq, so that given fixed C and the colder Benguela Current (# 29). Fisheries develop first in and p, the Bcrit is determined by q alone (Carozza et al., 2017), cold regions with a higher standing biomass, when Bcrit ¼ Bcold and decreases with increasing q. In our increasing-technology (at q ¼ qcold ). Subsequently, when the catchability has increased experiments, fishing begins at any given location when q rises to sufficiently such that Bcrit ¼ Bwarm (at q ¼ qwarm ), fishing begins the point at which Bcrit is equal to the local standing biomass. in the warm region. Note that the simulated extraction of These mechanisms bear some resemblance to “basin models”, in biomass illustrated in Figure 5a proceeds slowly enough that the which a geographically resolved habitat suitability controls pro- biomass Bcold remains higher than Bcrit throughout. Because of duction, and fishing activity is proportional to the local biomass this non-equilibrium behaviour, the Bcold at the time that q (MacCall, 1990). reaches qwarm remains higher than the corresponding Bwarm. Figure 5a illustrates this mechanism, comparing two regions Moreover, for a continuous increase in catchability q, Bcrit inexo- with different temperatures, the warmer Caribbean Sea (# 12) rably decreases, as smaller and smaller biomasses become Bioenergetic influence on the development of fisheries 1861 profitable to fisheries, driving biomass and ultimately catches to- potential mismatches between spawning grounds, feeding wards zero. Thus, by influencing the pristine biomass of ecosys- grounds, and the regions where fish are caught. A complementary tems, the temperature is expected to influence the technology analysis is provided in Supplementary Information S11 by look- level under which fishing first becomes profitable. ing at the catch time-series according to exclusive economic zones (EEZs). This analysis shows weaker correlations with temperature Temperature influence on biomass replacement rate than at the LME scale, which may reflect the fact that some EEZs In the model, higher water temperatures lead to faster growth span multiple ecosystems, blurring environmental distinctions, while in other cases, a single ecosystem is partitioned between Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 and reproduction rates. As a result, despite large pristine biomasses, cold LMEs have slower biomass production per unit multiple EEZs, potentially separating fishing grounds from of biomass compared with warm LMEs with similar primary pro- regions of biomass production. The weaker correlations may also ductivity. Thus, warm ecosystems have faster biomass turnover, reflect an accentuation of country-specific social and economic replacing their standing biomass more quickly. In the meantime, factors that cloud the environmental influence. the model predicts smaller phytoplankton in warmer waters, so Our analysis also relies on reconstructions of catch time-series that the energy available for recruits and growth at a given fish from SAUP with their own inherent uncertainties. We assessed size is reduced in warmer ecosystems. The mortality rate also the degree to which the correlation between catch peak years and increases in warmer waters. When fishing begins, these effects LME temperature is robust to the set of LMEs selected by varying influence the shape of peaks in the catch time-series. the parameters of our method to identify large catch peaks in In Figure 5b, the simulated peak of fishing is more gradual at a time-series (see Supplementary Information S2) and by isolating warm site than at a cold site. When normalized by the maximum distinct sets according to the characteristics of the environment total catches (see Figure 5c), the increasing slopes at the two sites and the management (see Supplementary Information S6). are similar, but the peak is longer and the decline is more gradual The correlation remains significant when our method is applied at the warm site. This dynamic is consistent with the correlation to the catch time-series before reconstruction (Supplementary LME between water temperature and sdecrease (cf Figure 4c), despite the Information S6) or even when applied to the 56 LMEs lack of correlation with the rate of increase (cf Table 1 and (Supplementary Information S10). Thus, the covariation of the year Figure 4b). We suggest that increasing catchability dominates of maximum catch at LME with ecosystem temperature appears to the catch increase, therefore supporting similar trends of rising be robust. catch, while temperature-dependent biomass replacement rates influence the rate of catch decrease. Sensitivity tests with the Individual stocks vs. LME-level changes model confirm the role of temperature on the biomass turnover In the analysis above, we only consider aggregated time-series of in warm waters, arising from faster reproduction and growth of the total catch. The aggregation conceals the fact that different ectotherms, convolved with variations in the phytoplankton rep- species and stocks within a given ecosystem can vary in terms resentative size that limits recruitment (see Supplementary of their economic features, catchability, and vulnerability to Information S4). exploitation (Cheung et al., 2007). The vulnerability depends on stock-specific factors including the maximum length, the age at Discussion first maturity, the fecundity, and the growth rate. Conceivably, The observed temporal progression of catches across the global these different features of individual stocks could influence the ocean is largely consistent with the expectation from spatial varia- development and decline in fisheries preferentially targeting these tions in economic development, but also with the expected stocks, influencing the sequential succession of stock exploita- impacts of spatial variations in water temperature. Specifically, tions and thereby influencing the shape and timing of the LME the observed year of peak catch occurred earlier in ecosystems catch peak (see Supplementary Information S12). with higher HDI and in colder water, while the rate of post-peak However, an analysis of SAUP catch records carried out at the catch decline was more rapid in colder water. PP alone was not species level does not indicate systematic features that would bias a good predictor of the year of historical catch peaks. While the LME-level aggregation (Supplementary Information S12). differences in the accumulation of capital and technological de- The timing of peak catch years for individual stocks was not velopment undoubtedly explain a good part of these historical systematically related to the year of peak catch, and post-peak progressions, our results suggest that the impact of water temper- declines in individual stocks were similar to those of the LMEs in ature on ecosystem metabolism played a complementary role. which they are found. We therefore conclude that species-level Most importantly, greater pristine standing biomass density effects do not play a major role in determining the LME-level should have allowed an earlier development of profit-driven peak year and post-peak decline. industrial fisheries in cold waters, while fishing technology was less advanced. More rapid turnover of biomass and recruitment Mechanisms beyond bioenergetics in warm waters is likely to have played a secondary role, contrib- Any model presents, by necessity, a simplified view of a complex uting to slower post-peak declines in catches at low latitudes. global ecosystem. In the model used here, the temperature effect on growth, recruitment, and mortality rates places the dominant Uncertainty in observed LME-level correlations control on standing biomasses, and thereby the development and The analysis above was carried out at the coarse scale of LMEs decline of fishing, but temperature also covaries with other envi- and annually averaged observations, potentially masking finer- ronmental factors that can influence standing biomasses that are scale spatio-temporal variations that could impact the strength of not explicitly included in the model. For example, extensive con- our correlations. However, since many fish species are migratory, tinental shelves allow high fractions of benthic production that an analysis at finer spatial scales would raise other issues, such as can accumulate high biomass and are also easily exploited (Van 1862 J. Guiet et al. Denderen et al., 2018). In addition, the higher diversity of ex- Anticamara, J., Watson, R., Gelchu, A., and Pauly, D. 2011. Global ploitable fish species in warm, low-latitude ecosystems could fishing effort (1950–2010): trends, gaps, and implications. cause harvest peaks to be extended as fishers shift from one stock Fisheries Research, 107: 131–136. to the next, decreasing the slope of post-peak declines (Maureaud Behrenfeld, M. J., and Falkowski, P. G. 1997. Photosynthetic rates de- rived from satellite-based chlorophyll concentration. Limnology et al., 2019). Finally, the simple features of our economic model and Oceanography, 42: 1–20. place technological development as the dominant drivers and Bell, J. D., Watson, R. A., and Ye, Y. 2017. Global fishing capacity and ignore other such as changes in profitability that may have fishing effort from 1950 to 2012. Fish and Fisheries, 18: 489–505. Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 occurred when fisheries change targeted species as they move Brown, J. H., Gillooly, J. F., Allen, A. P., Savage, V. M., and West, G. equatorward, or changes in cost as steaming distance from port B. 2004. Toward a metabolic theory of ecology. Ecology, 85: increases. Accounting for these additional effects would add fur- 1771–1789. ther nuance to the historical progression of global fisheries. Caddy, J., and Cochrane, K. 2001. A review of fisheries management past and present and some future perspectives for the third mil- lennium. Ocean & Coastal Management, 44: 653–682. Concluding remarks Carozza, D. A., Bianchi, D., and Galbraith, E. D. 2016. The ecological While spatially variable socio-economic drivers certainly influ- module of BOATS-1.0: a bioenergetically constrained model of enced the historical progression of industrial fishing from high marine upper trophic levels suitable for studies of fisheries and latitudes to the tropics, our results indicate the likelihood that ocean biogeochemistry. Geoscientific Model Development, 9: ecosystem temperature also played a role through its influence on 1545–1565. the biological component of these coupled ecological–economic Carozza, D. A., Bianchi, D., and Galbraith, E. D. 2017. Formulation, systems. The observed correlations between the year of peak catch general features and global calibration of a and HDI are not stronger than those with temperature, and our bioenergetically-constrained fishery model. PLoS One, 12: 1–28. bioeconomic model produced a progression similar to the obser- Carozza, D. A., Bianchi, D., and Galbraith, E. D. 2019. Metabolic impacts of climate change on marine ecosystems: implications for vations when forced only with a globally homogeneous progres- fish communities and fisheries. Global Ecology and Biogeography, sion of technological development. Together, these findings 28: 158–169. emphasize the role that ecosystem temperature could have played Carr, M.-E., Friedrichs, M. A., Schmeltz, M., Aita, M. N., Antoine, in shaping the historical development of fisheries over the past D., Arrigo, K. R., Asanuma, I. et al. 2006. A comparison of global 70 years. estimates of marine primary production from ocean color. Deep The recognition that temperature, via its influence on Sea Research Part II: Topical Studies in Oceanography, 53: ecosystem-level metabolism, could have contributed to the 741–770. historical progression of fisheries provides a new perspective on Chassot, E., Bonhommeau, S., Dulvy, N. K., Mélin, F., Watson, R., Gascuel, D., and Le Pape, O. 2010. Global marine primary pro- these complex human-ecological systems. Importantly, it is con- duction constrains fisheries catches. Ecology Letters, 13: 495–505. sistent with the expected changes in fisheries productivity under Cheung, W. W. L., Lam, V. W. Y., Sarmiento, J. L., Kearney, K., climate change, with warmer ecosystems supporting less fish bio- Watson, R., Zeller, D., and Pauly, D. 2010. Large-scale redistribu- mass and lower maximum sustainable yields (Cheung et al., 2010; tion of maximum fisheries catch potential in the global ocean un- Carozza et al., 2019; Lotze et al., 2019; Free et al., 2019). In addi- der climate change. Global Change Biology, 16: 24–35. tion, the fact that water temperature has opposing influences on Cheung, W. W., Watson, R., Morato, T., Pitcher, T. J., and Pauly, D. biomass accumulation vs. biomass replacement suggests that the 2007. Intrinsic vulnerability in the global fish catch. Marine outcome of future warming on fish catches will vary with fishing Ecology Progress Series, 333: 1–12. pressure. These findings provide support for a greater emphasis Clarke, A., and Fraser, K. P. P. 2004. Why does metabolism scale with temperature? Functional Ecology, 18: 243–251. on temperature-dependent ecosystem bioenergetics, which con- Clarke, A., and Johnston, N. M. 1999. Scaling of metabolic rate with siders the long-term outlook for fisheries. body mass and temperature in teleost fish. Journal of Animal Ecology, 68: 893–905. Supplementary data Conti, L., Grenouillet, G., Lek, S., and Scardi, M. 2012. Long-term Supplementary material is available at the ICESJMS online ver- changes and recurrent patterns in fisheries landings from large sion of the manuscript. marine ecosystems (1950–2004). Fisheries Research, 119: 1–12. Dunne, J. P., Armstrong, R. A., Gnanadesikan, A., and Sarmiento, J. Funding L. 2005. Empirical and mechanistic models for the particle export This study has received funding from the European Research ratio. Global Biogeochemical Cycles, 19, GB4026. Council (ERC) under the European Union’s Horizon 2020 re- Flood, M., Ilona, S., Andrews, J., Ashby, C., Begg, G., Fletcher, W. J., search and innovation programme (682602) and support from Gardner, C. et al. 2014. Status of Key Australian Fish Stocks Reports 2014. Fisheries Research and Development Corporation. the Spanish Ministry of Economy and Competitiveness, through Food and Agriculture Organization. 2014. The State of World the Maria de Maeztu Programme for Centres/Units of Excellence Fisheries and Aquaculture. Technical Report of the Food and in RD (MDM-2015-0552). JG and DB acknowledge support from Agriculture Organization of the United Nations. Rome, Italy. NASA grant TunaScape (award 80NSSC17K0290) and California Free, C. M., Thorson, J. T., Pinsky, M. L., Oken, K. L., Wiedenmann, Department of Resources-Ocean Protection Council (C0100400). J., and Jensen, O. P. 2019. Impacts of historical warming on ma- rine fisheries production. Science, 363: 979–983. References Friedland, K. D., Stock, C., Drinkwater, K. F., Link, J. S., Leaf, R. T., Alder, J., and Sumaila, U. R. 2004. Western Africa: a fish basket of Shank, B. V., Rose, J. M. et al. 2012. Pathways between primary Europe past and present. The Journal of Environment & production and fisheries yields of large marine ecosystems. PLoS Development, 13: 156–178. One, 7: 1932–6203. Bioenergetic influence on the development of fisheries 1863 Galbraith, E. D., Carozza, D. A., and Bianchi, D. 2017. A coupled Pauly, D., and Zeller, D. 2015. Sea around Us Concepts, Design and human-earth model perspective on long-term trends in the global Data. www.seaaroundus.org. marine fishery. Nature Communications, 8: 14884. Pauly, D., and Zeller, D. 2016. Catch reconstructions reveal that Gelchu, A., and Pauly, D. 2007. Growth and Distribution of global marine fisheries catches are higher than reported and de- Port-based Global Fishing Effort within Countries’ EEZs from clining. Nature Communications, 7: 10244. 1970 to 1995. Faculty Research and Publications. Fisheries Centre. Rall, B. C., Brose, U., Hartvig, M., Kalinkat, G., Schwarzmüller, F., University of British Columbia. https://open.library.ubc.ca/collec Vucic-Pestic, O., and Petchey, O. L. 2012. Universal temperature tions/facultyresearchandpublications/52383/items/1.0074748 (last and body-mass scaling of feeding rates. Philosophical accessed 20 March 2020). Downloaded from https://academic.oup.com/icesjms/article/77/5/1854/5835400 by Universitat Autònoma de Barcelona user on 14 April 2021 Transactions of the Royal Society of London B: Biological Gordon, H. S. 1954. The economic theory of a common-property re- Sciences, 367: 2923–2934. source: the fishery. Journal of Political Economy, 62: 124–142. Schaefer, M. B. 1954. Some aspects of the dynamics of populations Grainger, R., and Garcia, S. M. 1996. Chronicles of Marine Fishery important to the management of the commercial marine fisheries. Landings (1950-1994): Trend Analysis and Fisheries Potential. Inter-American Tropical Tuna Commission Bulletin, 1: 23–56. FAO, Rome. Sherman, K., and Duda, A. M. 1999. Large marine ecosystems: an Guiet, J., Aumont, O., Poggiale, J.-C., and Maury, O. 2016. Effects of emerging paradigm for fishery sustainability. Fisheries, 24: 15–26. lower trophic level biomass and water temperature on fish com- munities: a modelling study. Progress in Oceanography, 146: Stock, C., and Dunne, J. 2010. Controls on the ratio of mesozoo- 22–37. plankton production to primary production in marine ecosys- Hilborn, R., and Ovando, D. 2014. Reflections on the success of tradi- tems. Deep Sea Research Part I: Oceanographic Research Papers, tional fisheries management. ICES Journal of Marine Science, 71: 57: 95–112. 1040–1046. Stock, C. A., John, J. G., Rykaczewski, R. R., Asch, R. G., Cheung, W. Jennings, S., Mélin, F., Blanchard, J. L., Forster, R. M., Dulvy, N. K., W. L., Dunne, J. P., Friedland, K. D. et al. 2017. Reconciling fish- and Wilson, R. W. 2008. Global-scale predictions of community eries catch and ocean productivity. Proceedings of the National and ecosystem properties from simple ecological theory. Academy of Sciences, 114: E1441–E1449. Proceedings of the Royal Society of London B: Biological Swartz, W., Sala, E., Tracey, S., Watson, R., and Pauly, D. 2010. The Sciences, 275: 1375–1383. spatial expansion and ecological footprint of fisheries (1950 to Locarnini, R. A., Mishonov, A. V., Antonov, J. I., Boyer, T. P., and present). PLoS One, 5: e15143. Garcia, H. E. 2006. World Ocean Atlas 2005, Volume 1: United Nations Development Programme. 2016. Human Temperature. S. Levitus, Ed. NOAA Atlas NESDIS 61, U.S. Development Report 2016: Human Development for Everyone. Government Printing Office, Washington, DC. 182 pp. http://hdr.undp.org/en/2016-report (last accessed 20 March Lotze, H. K., Tittensor, D. P., Bryndum-Buchholz, A., Eddy, T. D., 2020). Cheung, W. W. L., Galbraith, E. D., Barange, M. et al. 2019. Ursin, E. 1984. The tropical, the temperate and the arctic seas as me- Global ensemble projections reveal trophic amplification of ocean dia for fish production. Dana, 3: 43–60. biomass declines with climate change. Proceedings of the Van Denderen, P. D., Lindegren, M., MacKenzie, B. R., Watson, R. National Academy of Sciences of the United States of America, A., and Andersen, K. H. 2018. Global patterns in marine preda- 116: 12907–12912. tory fish. Nature Ecology & Evolution, 2: 65. MacCall, A. D. 1990. Dynamic Geography of Marine Fish Watson, R., Zeller, D., and Pauly, D. 2014. Primary productivity Populations. Washington Sea Grant Program, Seattle, WA. demands of global fishing fleets. Fish and Fisheries, 15: 231–241. Marra, J., Trees, C. C., and O’Reilly, J. E. 2007. Phytoplankton pig- Watson, R. A. 2017. A database of global marine commercial, ment absorption: a strong predictor of primary productivity in the surface ocean. Deep Sea Research Part I: Oceanographic small-scale, illegal and unreported fisheries catch 1950–2014. Research Papers, 54: 155–163. Scientific Data, 4, 170039. Maureaud, A., Hodapp, D., van Denderen, P. D., Hillebrand, H., Watson, R. A., Cheung, W. W., Anticamara, J. A., Sumaila, R. U., Gislason, H., Spaanheden Dencker, T., Beukhof, E. et al. 2019. Zeller, D., and Pauly, D. 2013. Global marine yield halved as fish- Biodiversity–ecosystem functioning relationships in fish commu- ing intensity redoubles. Fish and Fisheries, 14: 493–503. nities: biomass is related to evenness and the environment, not to Worm, B., Hilborn, R., Baum, J. K., Branch, T. A., Collie, J. S., species richness. Proceedings of the Royal Society. Biological Costello, C., Fogarty, M. J. et al. 2009. Rebuilding global fisheries. Sciences, 286: 20191189. Science, 325: 578–585. Handling editor: Ken Andersen