Food for Thought Five centuries of cod catches in Eastern Canada

Five centuries of cod catches The ﬁshery for Northern Atlantic cod ( Gadus morhua ) oﬀ Newfoundland and Labrador, Eastern Canada, presents the most spectacular case of an exploited stock crashed in a few decades by an industrial bottom trawl ﬁshery under a seemingly sophisticated management regime after half a millennium of sustainable ﬁshing. The ﬁshery, which had generated annual catches of  to  tonnes from the beginning of the th century to the s, peaked in  at  tonnes, followed by a devastating collapse and closure  years later. Since then, stock recovery may have been hindered by premature openings, with vessels targeting the remains of the cod population. Previous research paid little attention towards using multicentury time series to inform sustainable catches and recovery plans. Here, we show that a simple stock assessment model can be used to model the cod population trajectory for the entire period from  to  for which catch estimates are available. The model suggests that if ﬁshing eﬀort and mortality had been stabilized in the s, precautionary annual yields of about  tonnes could have been sustained. Our analysis demonstrates the value of incorporating prior knowledge to counteract shifting baseline eﬀects on reference points and contemporary perceptions of historical stock status.


Introduction
For millennia, the biodiversity of the ocean has supported preindustrial fisheries, although they were already capable of extirpating easily accessible animals (Jackson et al., 2001). The ascent of industrial fishing, i.e. the deployment of large vessels fuelled by fossil energy (first coal, then diesel), however, radically changed fisheries, and made it increasingly possible to target accumulations of any desirable fish species at any distance from coastlines, depth of occurrence or season, all factors which created areas and times where fishing was not before possible (Swartz et al., 2010).
Thus, as industrial fishing spread across the world from the UK in the 1880s, global catch increased throughout most of the 20th century, and particularly after WWII, when bottom trawling became widespread. Although many coastal fish populations rapidly collapsed under the onslaught, this was long masked by the opening of new fishing grounds in hitherto unfished areas (Pauly et al., 2002;Cardinale et al., 2015). In the mid-1990s, however, the opening of new fisheries became unable to compensate for the overexploitation of the "old" fishing grounds (Froese et al., 2009), and the world catch peaked and began a decline which continues to this day (Pauly and Zeller, 2016;FAO, 2018), despite the increasing fishing effort and seafood demand. Recent intensive management efforts across the globe have started to show improvements in stock status for marine fish assessments that are based on science (Hilborn et al., 2020).
We now have reached a point where the only way to increase-or even sustain present fisheries catches-must involve the rebuilding C International Council for the Exploration of the Sea 2021. 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. of fish populations earlier depleted by overfishing. This is best illustrated by the Northern cod of Eastern Canada, i.e. Atlantic cod (Gadus morhua). The stock defined by this assessment includes all cod caught within NAFO-delineated Divisions 2J3KL.
Previously considered one of the world's largest and most important fish stocks (Hutchings and Rangeley, 2011), Northern cod have played a fundamental role in shaping the history, economy and culture of Atlantic Canada since the late 15 th century (Innis, 1940;Cell, 1982). In the 1960s, within a decade after European factory freezer trawlers began operation in Eastern Canada, catches peaked, and then plummeted. The declaration of a fishery exclusion zone in 1977, which largely eliminated foreign fishing, did not provide much of a respite, however, as Canada subsidised the building of a national fleet which continued overfishing. In 1992, the Canadian government declared a moratorium on Northern cod, as the stock had collapsed, followed in the next two years by cod moratoria in all eastern provinces of Canada, closing an entire economic sector.
The moratorium on directed commercial fishing was initially predicted by DFO to last two years to allow for sufficient stock recovery (Hutchings et al., 1997). Other factors such as temperature and prey availability may have contributed additional pressure on the stock's ability to recover (Rose and O'Driscoll, 2002;Buren et al., 2014). Almost 30 years later, all Canadian cod stocks remain in a critical state, their most recent estimates of population size being below their respective biomass limit reference points (all stock assessments for Canadian Atlantic cod are available through the Canadian Science Advisory Secretariat at https://www.dfo-mpo.gc .ca/csas-sccs/).

Applying a historical lens towards rebuilding
In order to estimate the full potential of an exploited resource, we must set our baseline near the start of its exploitation, and account for all withdrawals over time. Ignoring the past can lead to shifting baseline syndrome (Pauly, 1995), where we accept as baseline, a situation that does not account for the previous exploitation and its impact on stock size and dynamics. This can lead to underestimating fishing impacts and setting quotas too high, thus preventing a stock from rebuilding (Hutchings and Rangeley, 2011). The new discipline of historical marine ecology has emerged from attempts to counteract the shifting baseline syndrome, by demonstrating the value of recovering earlier abundance estimates and thus strengthening the management of marine populations (Jackson et al., 2001).
The case study presented here has two goals: (i) to demonstrate the usefulness of a 500+ year record of Northern cod catches for the current setting of stock rebuilding targets and (ii) to demonstrate, using the newly developed CMSY stock assessment method (Froese et al., 2017), that considering long time series does not necessarily require complex models with a multitude of free parameters.

Stock assessment
We used a new open source stock assessment tool (CMSY) (Froese et al., 2017), which is based on surplus-production modelling (Schaefer, 1954(Schaefer, , 1957. The CMSY tool (Froese et al., 2017) includes two methods: the first (named CMSY, same as the overall tool) derives fisheries reference points from catch data and priors with a Monte Carlo approach, while the second (named BSM) is a Bayesian state-space implementation of a traditional surplus production model which derives its estimates from catch plus abundance or effort data, if available. Since the BSM assessment is based on more information, it usually produces narrower estimates of changes in population biomass trends over time. Overall, both methods show good agreement with more data-demanding assessments (Martell and Froese, 2013). The most recent CMSY R-code is available from http://oceanrep.geomar.de/33076/. The R-code used for this analysis is available in the supplementary material.
Equation (1) below describes how parameters for the intrinsic rate of population increase (r), carrying capacity (k), and biomass in a given year (B t ) and catch in the same year (C t ) can be used to determine biomass (B) in the subsequent year (t + 1). Bias-correcting lognormal errors (e s 1 and e s 2 ) are assigned to surplus production and catch, respectively.
The above equation is modified (Equation 2) when a stock size is severely depleted (biomass below 0.25k or 0.5B/B MSY ) to account for depensation-the reduction of recruitment at a small stock size Maroto and Moran, 2014;Perälä and Kuparinen, 2017;Neuenhoff et al., 2019). This differs from the latest assessment model (Cadigan, 2015) used by DFO (Brattey et al., 2018), which does not consider depensatory population dynamics, but does report periods of very low productivity for the Northern cod stock after the collapse (Morgan, 2019).
Based on this theoretical framework, the CMSY method estimates likely biomass trajectories that correspond to the biomass reductions caused by fishing, the range for carrying capacity (k) and intrinsic rate of population increase (r). Uniform ranges for r and k were translated into prior densities with central values (Froese et al., 2017). The most probable "viable" r-k pair is selected from the tip of a triangle-shaped bivariate plot of r vs. k (Froese et al., 2017). When relative biomass data are known, an additional parameter (i.e. catchability or q) is estimated to convert catch-per-unit-ofeffort into biomass. Each tentative biomass trajectory is compared with the available relative biomass trend, which usually results in narrower confidence intervals.

Selection of priors
In the present study, a reconstructed catch time series  Supplementary Information) starts in 1508 and was updated to 2017, using Northwest Atlantic Fisheries Organization (NAFO) annual reports (NAFO, 2021), and further updated to 2019 from the most recent DFO stock status update (DFO, 2021b) ( Table 1).
Resilience corresponds to the intrinsic rate of population increase (r). We used a lower (Hutchings and Rangeley, 2011) bound of r set at 0.095 year −1 and an upper (Hutchings, 1999) bound set at 0.3 year −1 . Other studies also present the intrinsic rate of population increase within the chosen range (Myers et al., 1997;Rose, 2004).
Independent prior knowledge about the reduction of biomass by fishing from the start of the fishery to the end of the time series was translated into broad ranges of biomass relative to unexploited biomass (Table 2). At the start of the time series with very little fishing in 1508, the biomass range relative to unexploited biomass was set at 0.9-1.0 (very low depletion (Rose, 2004)). The end of the time series in 2019 corresponds to a biomass range of 0.01-0.20, as justified by expert knowledge that the stock's biomass is below critical levels (very strong depletion ( Table 3). The state-space model implementation of the BSM (Millar and Meyer, 2000) accounts for process error in population dynamics and observation error in measurement and sampling (Thorson et al., 2012). The standard deviation of the process error is specified in the code as sigmaR with a default value of 0.1, which we varied, to evaluate sensitivity, between 0.1 and 0.4. The alternative values of the process error had minimal effect on the model output, thus the default value of 0.1 was used in the final analyses. Process error is sampled anew for every year of the time series, accounting for uncertainty in the modelled productivity. The code also models error in catch, with a lognormal distribution.

Assessment results
The CMSY analysis produces proxies for MSY, F MSY , B MSY , and indicators like stock size (B/B MSY ) and exploitation (F/F MSY ) ( Table 4). The outputs of both CMSY and BSM analyses were similar for Northern cod (Table 4), thus building confidence in the results. Since BSM estimates are based on more data, they were used for the estimates presented in the text below ( Figure 1). Analysing 512 years of catch data (Figure 1a) and 37 years of relative biomass data (Table 3) produced an estimate of the intrinsic rate of population increase (with 95% confidence intervals) of Northern cod of r = 0.25 year −1 (0.14-0.46 year −1 ) and a carrying capacity of k = 6.0 million tonnes (4.0-8.9 million tonnes). Maximum Sustainable Yield (MSY) can then be computed from r·k/4, which yields 380000 t·year −1 , (290000-490000 tonnes), while biomasses can be expressed relative to the biomass that can produce MSY (B/B MSY ; see Figure 1b). Figure 1b shows that, for 200 years, the fishery for Northern cod impacted its biomass only lightly, and that it began to be noticeably reduced from 1700 on; however, it remained well over B MSY and thus capable of producing MSY as well as fulfilling its ecosystem role as a major predator in the waters off what is now Eastern Canada. With catches increasing from the 18th to the middle of the 20th century, the biomass decline accelerated, but it was only in the 1960s, with the onset of the industrial trawl fishery, that Northern cod biomass precipitously declined below B MSY , and specifically after 1968, when the peak reported catch of 810000 tonnes was extracted ( Figure 1b). Figure 1c and d show that the 1977 declaration of a fishery exclusion zone and the departure of foreign fleets led to a brief stabilization, at a suboptimal level, of the biomass of Northern cod in  R. Schijns et al.  . * Stock-recruitment model (Ricard et al., )  * . Shelton model (Shelton, ) . . * . * Assumed a : ratio for total biomass to SSB according to DFO () and Ricard et al. ().
the late 1970s and early 1980s. At this time, precautionary annual yields of around 200000 tonnes may have been sustained, but this opportunity to let the stock rebuild was not used. Rather, a newly built, heavily subsidized Canadian trawler fleet replaced the fishing mortalities previously exerted by foreign fleets, leading to a second collapse of catches ( Figure 1e) and biomass, which fell below 1% of its original biomass ( Figure 1f). Remarkably, the 1992 fisheries "moratorium" did not lead to a cessation of the fishery. Rather, post-moratorium catches, ranging between 400 and 13000 tonnes per year, continued to be taken (DFO, 2021b), consisting of subsistence and recreational catches, by-catch, occasional 'sentinel surveys' , and a stewardship fishery (i.e., small-scale commercial fisheries by any other name), and exerting just enough pressure to forestall a rebuilding of the population (Rose and Walters, 2019). The rebuilding plan, released in 2020, states that fishery removals are to be kept at the "lowest possible level" until stock biomass has grown above the critical zone (DFO, 2021a). This plan has been criticized for not restricting catches sufficiently and explicitly allowing increase in quotas before the stock has reached the limit reference point .
In addition to the time series of catch, the BSM method used a relative index of abundance available from fisheries-independent surveys conducted by the Canadian Department of Fisheries and Oceans since 1983 (DFO, 2021b) ( Table 3). The results of BSM are similar to published estimates of more data-intensive models ( Table 5). The model estimates biomass in 2019 is 310 (131-570) kt. Although our estimate is lower than the 480 kt reported by DFO (2021b), it is consistent with Rose and Walters' (2019) estimates of 300 kt in 2015 declining to about 250 kt in 2017. As well, the intrinsic rate of population increase (r) estimated here is similar to that estimated in another long-term assessment of Northern cod, but which explicitly accounted for climate effects (Rose, 2004).
To explore changes in carrying capacity over the 512 years, we repeated the assessment for two recent periods, 1930-2019 and 1970-2019. The estimate of carrying capacity for the 1970-2019 period of 3.0 (2.0-4.4) million tonnes is lower than the estimate for the entire period of 6.0 (4.0-8.9) million tonnes, though the difference is not significant, with both estimates being included in their respective 95% confidence limits. This decline may indicate a true change in carrying capacity (Palomares et al., 2018) or it may stem from non-consideration or under-reporting of previous catches and then present a case of a shifting baseline, where a rebuilding target such as B MSY is underestimated because only recent data were included in the analysis (Préfontaine, 2009). For a lesser known stock, the shifting baseline syndrome (Pauly, 1995) may be more prevalent, especially if consecutive assessments selected more recent years without incorporating knowledge of past exploitation. For the case of Northern cod, the use of well-informed priors prevents or limits shifting baselines, as reflected by the limited shifts in reference points for the selected time periods (Table 4). The estimates of F MSY in Table 4 (with and without depensation) are consistent with the hypothesis that Northern cod is not capable of sustaining levels of fishing mortality as high as those of other cod stocks (Myers et al., 1996;Rose, 2019). These findings suggest that management strategies should strive to include historical data in order to provide realistic reference points as targets for rebuilding.
The broad confidence limits in our estimates of k reflect a legitimate challenge in estimating carrying capacity based on historical data. Although not statistically significant, we cannot discount the possibility that the changes in k are biologically informative and indicative of changing production regimes. One putative correlate of Northern cod productivity is water temperature (Lilly et al., 2008). For example, citing Colonial Office export records, Innis (1940) reported low catches during the 1713-1720 and 1789-1792 periods, ostensibly because cold water had reduced the availability of cod to inshore fisheries. In contemporary times, water temperatures were colder in the 1970 to 2000 period when compared to the 1940 to 1970 period, and this might have contributed to lower productivity in the short term. But if one examines temperatures with a longer historical lens (as we have done with the catch data), the colder temperatures of the late 1980s and early 1990s were experienced by Northern cod from the 1850s to the 1930s (Hutchings and Myers, 1994;NCAR, 2021) with no discernably negative effects on catches.
Exploring the results of Table 4 further, the CMSY method produces lower estimates of r (closer to the prior) and consequently higher estimates of k (r and k are inversely related in the context of a Schaefer model). The observation that the CMSY output is closer to the r prior than the output from the BSM model stems from the fact that the CMSY model has no information on stock abundance. In other words, the higher r values estimated by BSM stem from the incorporation of highly informative CPUE data. We note, however, that the confidence limits of r from the CMSY and BSM outputs overlap, suggesting that the differences in r produced by the two methods are not substantial.
Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsab153/6359257 by guest on 29 August 2021 Five centuries of cod catches in Eastern Canada  Our assessment suggests that the biomass of Northern cod is currently (in 2019) 2% of carrying capacity and less than 0.05 B MSY . Independently of the accuracy of these quantitative estimates, the biomass of Northern cod is clearly far lower than the historical biomass that was capable of sustaining annual catches of 150000 to 200000 tonnes ( Figure 1). There is a scientific consensus that the stock is currently well below its biomass limit reference point (0.48 B LIM , according to DFO, 2021b) and that periodic inshore fisheries since the 1992 moratorium have had (Hutchings and Rangeley, 2011;Rose and Walters, 2019), and continue to have (DFO, 2021b), an inhibitory effect on stock rebuilding.
All else being equal, the smaller a population, the greater its susceptibility to stochastic environmental change, resulting in increased variability in mortality in fishes (Minto et al., 2008); the greater the magnitude of population reduction, the longer and more uncertain the rebuilding period (Neubauer et al., 2013). Such impairments to recovery can be caused by depensation or Allee effects (Perälä and Kuparinen, 2017;Neuenhoff et al., 2019). Manifest by a decline in realized per capita population growth rate with declining population size, depensation in marine fish populations can be the result of declining recruits per spawner, increased natural mortality, or both (Maroto and Moran, 2014;Hutchings, 2015). Depensation is built into the principal Equation 2 of CMSY and reflected by a linear decline of curFmsy when biomass falls below 0.25k, a threshold consistent with previous estimates of where the Alleeeffect threshold might exist for marine fishes, including cod (Hutchings, 2014(Hutchings, , 2015. Our incorporation of depensation draws explicit attention to the possibility that per capita population growth, and consequently F MSY (Table 4), declines with declining abundance at low population size, a caveat that is not reflected by current management strategies for Northern cod (Winter and Hutchings, 2020).

Conclusion
The CMSY tool may be useful to assess both data-limited stocks (those with only catch available) and data-rich stocks (e.g. Northern cod), as it can provide longer term estimates of stock status by incorporating past data-limited periods. Centuries-old catch data exist for several stocks, such as Bluefin tuna (Thunnus thynnus) in the Mediterranean (commercialized around the 8th century (Lleonart et al., 1998;Addis et al., 2009)), Atlantic herring (Clupea harengus) in the Baltic Sea (fishery started in the 13th century (MacKenzie et al., 2002)), and Atlantic salmon (Salmo salar) in the Celtic Sea (fishery started in the 14th century (Manx Heritage Foundation, 1991)). By integrating historical data into stock assessments, we may better understand the total impact of fisheries on marine ecosystems and effectively manage marine populations for a long-term future.

Supplementary Data
Supplementary material is available at the ICESJMS online version of the manuscript.

Data availability statement
All data used in this paper can be found in the Supplementary Data. The full CMSY package developed by Froese et al. (2017) is available from: https://oceanrep.geomar.de/33076/.