Potential socioeconomic impacts from ocean acidification and climate change effects on Atlantic Canadian fisheries

RESEARCH ARTICLE Potential socioeconomic impacts from ocean acidification and climate change effects on Atlantic Canadian fisheries Tyler J. B. WilsonID1*, Sarah R. Cooley2, Travis C. Tai3, William W. L. Cheung3, Peter H. Tyedmers1 1 School for Resource and Environmental Studies, Dalhousie University, Halifax, NS, Canada, 2 Ocean Conservancy, Washington, DC, United States of America, 3 Changing Ocean Research Unit, Institute for the Oceans and Fisheries, The University of British Columbia, Vancouver, B.C., Canada * tjbwilson@dal.ca Abstract Ocean acidification is an emerging consequence of anthropogenic carbon dioxide emis- sions. The full extent of the biological impacts are currently not entirely defined. However, it is expected that invertebrate species that rely on the mineral calcium carbonate will be directly affected. Despite the limited understanding of the full extent of potential impacts and responses there is a need to identify potential pathways for human societies to be affected by ocean acidification. Research on these social implications is a small but developing field. This research contributes to this field by using an impact assessment framework, informed by a biophysical model of future species distributions, to investigate potential impacts facing Atlantic Canadian society from potential changes in shellfish fisheries driven by ocean acidi- fication and climate change. New Brunswick and Nova Scotia are expected to see declines in resource accessibility but are relatively socially insulated from these changes. Con-

versely, Prince Edward Island, along with Newfoundland and Labrador are more socially vulnerable to potential losses in fisheries, but are expected to experience relatively minor net changes in access. Introduction Ocean acidification (OA) is a facet of climate change driven by increasing carbon dioxide (CO2) concentrations in the atmosphere and ocean [1,2]. Recognition of the potential for OA to act as a biological stressor in the marine environment has only been widely recognised in the past 15 to 20 years. Since then, research addressing its biological implications has expanded rapidly [2–4], suggesting the potential for wide-ranging effects on marine organisms and eco- systems. To date, findings have been highly variable with some species showing positive physi- ological responses while many others respond negatively; responses have also been seen to vary between populations of individual species [4,5]. Despite ranges of expected impacts, cur- rent understandings suggest that negative responses from marine species are likely to be more widespread than neutral or positive responses [3,4]. Furthermore, it is widely agreed that PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 1 / 29 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 OPEN ACCESS Citation: Wilson TJB, Cooley SR, Tai TC, Cheung WWL, Tyedmers PH (2020) Potential socioeconomic impacts from ocean acidification and climate change effects on Atlantic Canadian fisheries. PLoS ONE 15(1): e0226544. https://doi. org/10.1371/journal.pone.0226544

Editor: Juan A. Añel, Universidade de Vigo, SPAIN Received: September 17, 2018 Accepted: November 28, 2019 Published: January 10, 2020 Copyright: © 2020 Wilson et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All relevant data are within the paper and its Supporting Information files. Funding: TJBW, TCT, WWLC and PHT acknowledge the support from the Marine Environmental Observation Prediction and Response Network (MEOPAR) http://www.meopar. ca/ - under the Canadian Ocean Acidification Research Program (COARp) https://meopar.ca/ projects/cycle-ii-research/. TJBW also acknowledges scholarship support from the Social Sciences and Humanities Research Council

invertebrate species that rely on calcium carbonate shells and exoskeletons (e.g., corals, oysters, crabs) will likely express negative responses [3,4]. Low pH events, driven by upwelling and exacerbated by OA, have already impacted shellfish production in the Pacific Northwest of the United States and Canada [1–3,6–8]. Globally, seafood is a critical dietary component for over 3 billion people and is also a highly traded commodity [9,10];changes in production could therefore have significant impacts for many and diverse communities. There is a body of emerging literature seeking to address how OA may affect human systems by linking projected biological responses to OA with ecosystem services such as fisheries production. To date, socioeconomic OA impact research has been undertaken at a range of scales and resolutions from international [11–13] to national [14–16] and sub-national [17]. These studies have generally focused on OA in isolation from other cli- mate change effects, and predicted declines in revenues and higher social risk in regions where shellfish fisheries are proportionally more important. Several previous assessments [e.g., 11,14,17], follow risk assessment frameworks to identify communities or regions that are most susceptible to impacts from OA. In these assessments risk is defined in conventions similar to that of the International Panel on Climate Change (IPCC) [18], and based on the socioeco- nomic structure of the social unit in question and the likelihood of an impact from OA.

In Canada, seafood production is concentrated on the Atlantic Coast, with over 80% of total landings and over 85% of the commercial fishing fleet based in the region [19]. Within Atlantic Canada, shellfish have become an increasingly important component of capture fish- eries, accounting for nearly 50% of regional landings by weight and over 75% of total landed value [20,21] (Fig 1). At the same time, Atlantic Canadian provinces and communities repre- sent some of the least wealthy segments of the Canadian economy, with most of the provinces being net recipients of federal equalization payments as well as receiving overall above average per capita federal support [22]. Furthermore, most of the provinces in the region have compar- atively rural populations, with many relatively small communities which are more highly dependent on employment from natural resource-linked sectors such as fisheries [23,24]. Given the high contribution of shellfish to total seafood production, together with the socio- economic background of the provinces in the region, Atlantic Canada presents a highly rele- vant setting for investigating how potential OA-driven changes in fisheries might affect human communities. In this paper, a biophysical model was linked with an impact assessment framework to investigate how OA and climate change might drive shifts in future availability of fisheries resources for Atlantic Canada, and which provinces in the region are most likely to be affected by these shifts. The assessment aims to highlight regions in Atlantic Canada where potential

impacts from OA and climate change is highest, under contemporary socioeconomic settings, given predicted changes in fisheries’ access. In the assessment framework OA and climate change are the sole drivers for changes in fisheries. Given the limits of scientific understanding and modelling capabilities more subtle, albeit important, factors such as potential for individ- ual species’ evolutionary responses to changing environmental conditions are not included. Methods Study area ‘Atlantic Canada’ typically refers to four Canadian provinces: New Brunswick (NB), New- foundland and Labrador (NL), Nova Scotia (NS) and Prince Edward Island (PEI). The prov- ince of Quebec (Que) also borders the Atlantic Ocean but it is often not included as part of ‘Atlantic Canada’ due to significant demographic and cultural differences. However, for the OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 2 / 29 (SSHRC) in the form of the Canadian Graduate Scholarship-Masters Program (CGS-M)- http:// www.sshrc-crsh.gc.ca/. TCT and WWLC also acknowledge support from the Social Sciences and Humanities Research Council of Canada (SSHRC) -http://www.sshrc-crsh.gc.ca/; and through the OceanCanada Partnership - https://oceancanada. org/. WWLC also acknowledges funding support from the Nippon Foundation-University of British Columbia Nereus Program - http://nereusprogram. org/. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests: The authors have declared that no competing interests exist.

purposes of this research, ‘Atlantic Canada’ will also include Quebec, as the province operates fisheries in Atlantic waters and is therefore exposed to potential changes in fisheries (Fig 2). All Canadian marine waters fall under federal jurisdiction and are managed by Fisheries and Oceans Canada (DFO). In Atlantic Canada, marine waters are divided into four manage- ment areas: Newfoundland and Labrador, Maritime, Gulf, and Quebec (Fig 2). In this analysis the Gulf and the Quebec management areas are treated as a single unit (henceforth, collectively referred to as the Gulf management area) because both management areas are comparatively small and the biophysical model used to project distribution changes does not have sufficient resolution to reliably differentiate between these management areas. The provinces of NS and NB border two management areas (Maritime and Gulf; Fig 2), and landings for each area are reported separately. Fig 1. Average annual landings by species groups by dollar value and tonnage. Arranged by species group for Atlantic Canadian fisheries for 1991–2010. Dollar values are normalised to year 2000 dollars and are in millions of dollars, weights are presented in thousands of tonnes (both data types use the same axis). Includes production from aquaculture. Note, in these data ‘shellfish’ encompasses all harvested marine invertebrate species. https://doi.org/10.1371/journal.pone.0226544.g001 Fig 2. Eastern Canada as it relates to the Northwest Atlantic Fisheries Organization (NAFO) fishing areas with

DFO management areas overlaid. For clarity approximate provincial borders have been emphasised in dark blue, U. S. and Canada border has been emphasised in black. In this assessment Gulf and Quebec management areas were treated as a single unit (figure modified from [25]). https://doi.org/10.1371/journal.pone.0226544.g002 OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 3 / 29

Species selection Fisheries in Atlantic Canada target a diverse array of marine species. Due to the variable nature of anticipated responses to OA a subset of commercially harvested species was selected to assess the potential impacts on fisheries and, ultimately, Atlantic Canadian communities. The selection of species was based on a combination of: a) their current economic importance (including capture fisheries and aquaculture production) in Atlantic Canada; and b) the cur- rent understanding that shellfish species are more susceptible to OA than finfish species [3,4]. To guide the selection of the specific fisheries that would be the focus of this research, total annual landings for all fisheries in each province and management area were compiled. Fisher- ies landing weights and values data were collected from DFO statistics [20]. Annual landing weights from 1991–2010 were averaged to obtain baseline annual weight for the year 2000. Landing values for each year were normalised to year 2000 dollars using the consumer price index from the Bank of Canada [26], and similarly averaged across 1991–2010. The compiled landing values included aquaculture production. While the biophysical model projections are not directly applicable to aquaculture, the production values were included here for two reasons. First, the majority of shellfish aquaculture in the region relies on wild populations for recruitment [27]. Consequently, changes that affect wild population distributions will also potentially affect levels of aquaculture production. And second, DFO

data for American oyster merges aquaculture and wild harvest production statistics. Moreover, PEI blue mussel data reporting changed from being reported as capture to being reported as aquaculture midway through the baseline time period. Aquaculture production data is available at a provincial level but, in contrast to wild har- vest, it is not differentiated between management areas. Therefore, the data for NS shellfish aquaculture production values were evenly divided between the Gulf and Maritime manage- ment areas because aquaculture production in NS was distributed across the whole province. For NB, shellfish aquaculture was all counted under the Gulf management area because NB shellfish aquaculture is concentrated in that management area–in contrast to NB finfish aqua- culture, which occurs in the Bay of Fundy (i.e., the Maritime management area). Biophysical impact modeling Model selection. To highlight potential impacts of OA on the fishery sector, shellfish fish- eries were prioritised for assessment in this research. However, other aspects of climate change will also drive shifts in species abundances and distributions and will add to and potentially interact with OA effects [28–30]. Therefore, it was decided that inclusion of climate change related impacts along with OA would present a more holistic estimation of the potential future of Atlantic Canadian shellfish fisheries. A Dynamic Bioclimate Envelope Model (DBEM) [31,32], which integrates species’ ecophysiology and biogeographical responses to changes in

ocean conditions (including OA, warming and deoxygenation) to predict future species distri- butions, was selected to inform the underlying biophysical conditions driving potential changes in landings in Atlantic Canadian fisheries. Model data. The DBEM used in this study was initially developed in 2008 [32] and subse- quently modified and updated [31,33,34]. The DBEM uses outputs from earth system models as environmental drivers to project spatial and temporal changes in species distribution, abun- dance, and catch potential. It integrates the oxygen- and capacity-limited thermal tolerance [35] and the gill-oxygen supply [36] hypotheses to estimate the impacts of environmental stressors—including OA—on somatic growth and mortality rates. The model outputs used here incorporated OA as an impact on growth and survival for mollusc and crustacean species. Specifically, mollusc and crustacean species groups had different impact levels per unit change OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 4 / 29

in pH based on the mean impact findings of meta-analyses conducted in 2010 and 2013 [3,4] [31,34,37]. Potential for evolutionary responses to changing conditions are not accounted for in the current iteration of the DBEM. For additional details on the DBEM see S1 Note. Outputs from the model were structured as annual species-specific maximum catch poten- tials (a proxy for maximum sustainable yield) for cells on a global half-degree latitude by half- degree longitude grid for the period of 1950 to 2100. For each species and year combination, there were 12 datasets generated by the model representing: two Radiative Concentration Pathways (RCP) scenarios for future CO2 emissions [38,39] (RCP 2.6 –‘highly mitigated CO2 emissions’ and RCP 8.5 –‘business as usual CO2 emissions’); two OA treatments (‘with OA’; and ‘without OA’); and using three separate climate models (NOAA’s Geophysical Fluid Dynamics Laboratory (GFDL-ESM), Institute Pierre Simon Laplace Climate Modelling Centre (IPSL-ESM) and Max Planck Institute for Meteorology (MPI-ESM)) (Table 1). Data extraction for the species of interest along with the following data manipulation was performed using Mathworks MATLAB, version R2015b, and Microsoft Excel 2013. Initially, the 12 global datasets for each of the selected species were geographically constrained to include only data from grid cells corresponding to Canada’s Atlantic EEZ (NAFO areas 2, 3 and 4; Fig 2). Modifying from earlier methods, 20 year running means were calculated for

each decade (e.g., for the year 2000, values were averaged from 1991–2010) for each species’ 12 geographically truncated datasets in order to smooth interannual variability from the climate model projections [40]. This was performed for each grid cell in the truncated datasets. Median values from outputs using each of the three earth system models were used to limit the uncertainty surrounding model variability (the following steps were also performed for the datasets generated from the individual climate models, these are available in the Supporting Information S1 Table). This data manipulation resulted in a total of four datasets (two RCP scenarios and two OA treatments) for each species. The catch potentials were then summed across the DFO management areas (as per Fig 3) for each decade. Minor exceptions to this process were necessary where the median dataset projected future abundance of a given spe- cies in a management area dropping to zero. In these instances, the mean of the management area aggregated data from the three climate model values was substituted. To project future landings, DFO reported landings for the baseline period were multiplied by the relative change indicated by the outputs of the DBEM. The year 2000 was used as the reference year against which future changes in potential landings were assessed. Relative Table 1. Schematic representation of data configuration for each species, s. Each of the 12 configurations yielded a separate global distribution of catch potential in half-degree latitude by half-degree longitude cells for each year, Y that

was modelled (1950–2100). Species s, Year Y RCP 2.6 with OA GFDL-ESM IPSL-ESM MPI-ESM without OA GFDL-ESM IPSL-ESM MPI-ESM RCP 8.5 with OA GFDL-ESM IPSL-ESM MPI-ESM without OA GFDL-ESM IPSL-ESM MPI-ESM https://doi.org/10.1371/journal.pone.0226544.t001 OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 5 / 29

change in the modelled catch potential for each species in subsequent decades was calculated for each management area, as well as for the entire Atlantic Canadian region (Eq 1). To assess patterns across the region, the relative change for each grid cell was also calculated. DA;Y;r;s ¼ Ydata 2000data 2000data Eq 1 Where ΔA,Y,s is the relative change compared to the year 2000 for management area A, in decade Y, under RCP scenario r, and species s. Analysis of potential future landings. Two future time-steps, representing the middle and end of the 21st century (2050 and 2090, respectively) were selected as endpoints for the assessment of changes in landings. Values for both time-steps were calculated relative to the reference year (2000). The coupled impact of climate change including OA was determined to be a more relevant focus for the investigation of potential impacts across Atlantic Canada and was therefore used for the main analysis. The 2050 and 2090 potential landings for each species within each man- agement area were compared with baseline landings. Differences in future landings under the two climate scenarios (RCP 2.6 and RCP 8.5) were also compared to identify how different emission scenarios might affect future fisheries. Individual species changes, as well as cumula- tive changes across species, were investigated within and between the management areas. Constructing the socioeconomic impact index An impact assessment framework was developed to evaluate the potential socioeconomic

impacts Atlantic Canada may experience through changing fisheries landings driven by ocean acidification and climate change. A range of approaches to modelling potential socioeconomic impacts posed by hazards such as OA and climate change have emerged in the literature in recent years [14,17,18]. The framework constructed here broadly follows the methodology of Fig 3. Geographic coverage of datasets. Grid lines indicate individual half-degree latitude/longitude data cells. White boundary line outlines study area (i.e., Canadian Atlantic EEZ). Data outside of this boundary was truncated from the datasets before aggregating management area data. Cell colours define data aggregation layer for DBEM outputs: Newfoundland and Labrador management area is represented by dark green cells; Maritime management area is represented by light green cells, and Gulf (including Quebec) management area is represented by the yellow-green cells. Inclusion of land area within data aggregation is irrelevant, as these cells do not contain any data. Coastline data from: [41]. https://doi.org/10.1371/journal.pone.0226544.g003 OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 6 / 29

risk assessments addressing ocean acidification’s impacts on fisheries [17], wherein risk is defined as the intersection of a hazard, exposure to the hazard, and vulnerability to the hazard (following the IPCC Working Group II report: Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation—Chapter 2: Determinants of Risk: Expo- sure and Vulnerability. [18]). Vulnerability in turn encompassed two sub-components: sensi- tivity and adaptive capacity. This assessment follows similar definitions with the potential impact being defined as the combination of social vulnerability and biophysical exposure to the hazard of climate change and ocean acidification. The potential impact is not an absolute measurement, rather it is a metric for a relative comparison of how susceptible the provinces are to changing conditions. In the framework developed for Atlantic Canada, the exposure component represented the modelled OA and climate change-driven effects on the fisheries as determined through analy- sis of DBEM outputs. The sensitivity and adaptive capacity terms were used to describe the social factors that may obstruct or fortify communities’ responses to changes in fisheries. In total, the impact index was informed by six separate indicators: one for exposure, two for sen- sitivity, and three for adaptive capacity (Fig 4). Within each individual component of the framework, the sub-components or indicators were equally weighted. This weighting inher- ently implies that the indicators contribute equally to potential impacts in the region. In reality,

this is unlikely to be the case, however without strong support for preferential weights, equal weighting was selected as the most straightforward and neutral system [42,43] (Fig 4). Although a cumulative impact index was constructed, the biophysical and social factors should be considered as representing two distinct aspects of potential impact [14,18]. The combined index represents where these two dimensions overlap to have the highest effect. The biophysical indicator (i.e., exposure) represents where the most prominent changes in the resource will occur; and the social indicators (i.e., vulnerability) highlight where the expected changes are likely to have the highest impact given the current socioeconomic setting. Two of the provinces, NB and NS, border two separate DFO management areas (the Gulf and the Maritime management areas; Fig 2). For the purposes of this analysis, these two prov- inces were sub-divided into gulf (NB-gulf; NS-gulf) and maritime (NB-mar; NS-mar) provin- cial sub-sections. However, both sub-sections of the respective provinces rely on the same social data because these data were not available at a resolution that would have allowed sepa- rate treatments within the provinces. The sub-sections were exposed separately to potential changes in both relevant DFO management areas. Exposure indicator construction. The most significant divergence from previous risk/ vulnerability assessments of potential impacts from OA was the inclusion of a biophysical model to inform the exposure and hazard components of the framework simultaneously. Past

assessments have used an expected change in ocean chemistry to represent OA (the hazard) and used this in concert with the importance of OA susceptible fisheries (the exposure) to link the OA phenomenon with fisheries [11,14,15,17]. In this assessment, the biophysical model directly applied the expected climate change impact (including OA) onto the relevant fisheries. This allowed the change in fisheries landings to be explicitly incorporated into the framework. Thus the hazard and exposure terms were essentially incorporated into a single term (collec- tively referred to as ‘exposure’ throughout this assessment). A single indicator to account for all of the assessed species was constructed. To reflect the very different current economic contributions of individual species to total landed value, the proportion of the total shellfish value derived from each species was calculated and multiplied by its DBEM-predicted relative change for each management area. These value-scaled changes were summed across all assessed species (Eq 2). This was repeated separately for 2050 and 2090 and both RCP climate scenarios to obtain four exposure indicator scores per OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 7 / 29

management area. From here forward, the term ‘quartet’ will be used to refer to the four expo- sure scenarios, as well as the subsequent potential impact scores. EA;Y;r ¼ X DA;Y;r;s � VA;s VA;tot ! " # S Eq 2 Where EA,Y,r is the exposure indicator score for area A, in time-step Y, and RCP scenario r. ΔA, Y,s is the DBEM predicted relative change in landings for time-step Y, RCP scenarios r, and species s (from Eq 1). VA,s is the year 2000 average annual value for species s, and VA,tot is the total annual landings of the seven species value for the management area (Table 2). The impact index was constructed so that high scores implied higher potential for impact. To align the exposure scores with this orientation, the values were multiplied by (-1) so that losses in potential future landings became positive. Additionally, in order to combine the sepa- rate index indicators, it was necessary to first normalise each separate indicator to score between 0 and 1 [42,43]. Therefore, the 12 exposure values (three management areas, each with a quartet of scores) were normalised following a min-max calculation to score between 0 and 1 (Eq 3) [43]. Min-max normalization is a widely used method to allow distinct indicators with different units to have the same range of values (i.e., 0–1) for more straightforward inte- gration into an index [43]. In the index, the exposure indicator scores for management areas Fig 4. Framework for the impact index demonstrating relationship between components. Elements at each branch

are equally weighted to other items at the same level as indicated by bracketed numbers. https://doi.org/10.1371/journal.pone.0226544.g004 Table 2. Cumulative exposure indicator of future changes in fisheries (before normalisation of values). Values in fourth and fifth columns are the indicator scores (i.e., the relative change in landings scaled by proportional value of the individual species). For integration into the impact index, the exposure scores were reversed and then normalised to score between 0 and 1, so that declines in landings earned a high score and implied a high exposure to OA and climate change. Management region Exposed provinces Climate scenario Exposure indicator scores 2050 2090 Maritime NS-mar; NB-mar RCP 2.6 -0.114 -0.166 RCP 8.5 -0.202 -0.498 Gulf NS-gulf; NB-gulf; PEI; Que RCP 2.6 0.042 0.017 RCP 8.5 0.070 0.033 Newfoundland and Labrador NL RCP 2.6 -0.012 -0.026 RCP 8.5 0.015 0.060 Regional total n/a RCP 2.6 -0.014 -0.034 RCP 8.5 -0.018 -0.069 https://doi.org/10.1371/journal.pone.0226544.t002 OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 8 / 29

were combined with the social vulnerability indicator scores for adjacent provinces–the Gulf and the Maritime management areas’ scores were applied to more than one province (NB- gulf, NS-gulf, PEI, and Que; and NB-mar, NS-mar; respectively) while the Newfoundland and Labrador management area scores were only applied to NL (Fig 3). IE;A;Y;r ¼ EA;Y;r MinðEÞ � MaxðEÞ MinðEÞ ð Þ Eq 3 Min-Max approach for normalising exposure indicator scores [43]. Where IE,A,Y,r is the nor- malised index score for the exposure score EA,Y,r as calculated in Eq 2. Min(E) and Max(E) indicate the respective minimum and maximum exposure scores obtained. The normalisation sets the maximum score to be 1 and the minimum score to be 0. Social vulnerability. Vulnerability represents a province’s reliance on fisheries (sensitiv- ity) as well as the community’s broader ability to respond to, and absorb, changes (adaptive capacity). Vulnerability was assessed at the provincial level as this was the finest level of politi- cal organization for which most relevant data were consistently available across the region. To align the social data with the exposure data, the year 2000 was used as the baseline where possi- ble. However, some of the data (e.g., crew size) was drawn from reports which are not pub- lished annually, and consequently these data are from the nearest year available (in all instances this was within 10 years of the year 2000). The social data was centered on the year 2000 baseline to align with the biophysical data.

nated by far more numerous but smaller vessels with fewer than five crew members) they were treated as representing these two distinct aspects of the provincial socioeconomic structure. The first indicator was assessed using year 2000 total annual landed value (averaged from 1991–2010) of the assessed species as a fraction of total provincial GDP (averaged between 1997 and 2003). To make the dollar values comparable across years all dollar values were nor- malised to year 2000 dollars [26,44]. Scaling the fisheries value relative to provincial GDP allowed the indicator scores to be relatable between the provinces, which have a broad range of economic scales. It is difficult to obtain realistic employment estimates for individual fisheries because of the seasonal nature of the sector. This type of data is further convoluted by different methods of reporting employment statistics. For example, employment data are often presented as full- OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 9 / 29

time equivalent jobs (a measure of the number of hours of work) in government reports. This is not a particularly relevant metric for seasonal employment–especially when it can generate enough income that additional employment is not necessary to support a desired quality of life (as can be the case for high-value fisheries). The estimates used here attempt to reflect the number of people employed with fisheries as their main source of income. An estimate of employment was derived from the number of licences per species multiplied by average crew sizes for the applicable fishing fleet. Licence numbers were obtained from DFO statistics [45]. Where possible crew size estimates were collected from industry reports and assessments [46–50] (see S2 Note for an aside on ‘inshore’ and ‘offshore’ fleets and special cases regarding data availability). As well as through discussion with Michael Gardner, of Gardner Pinfold Consulting (personal communication, September 11, 2017). Crew size esti- mates for most of the assessed fisheries (in their respective management areas) were obtained from studies conducted in the early 2000s but these were the closest available data to the base- line year (see S2 Table for data sources for crew size estimates). Where crew size estimates were not available for a given species within a specific management area, the average crew size from fisheries for that species from other areas was used. Total employment in the relevant fisheries was summed for each province and divided by provincial population.

Aquaculture employment was not considered in this indicator because available data does not readily differentiate between finfish and shellfish aquaculture employment. Furthermore, employment from aquaculture is relatively small compared to wild harvest employment. As a result employment dependence was set to zero for species whose production was derived entirely or almost entirely from aquaculture. For similar reasons, estimates of secondary employment such as processing and retail were not considered in this assessment. For many fisheries in the Atlantic region licences are not fully utilised. However, given the high value of the species assessed in this study, it was considered reasonable to assume all or a very high proportion of all licences in these fisheries were active. Crossover between fisheries, where a licence holder (as well as crew) may operate in more than one of the assessed fisheries was not accounted for (i.e., double counting was a possibility). Lacking a much more thorough social investigation this indicator provides an approximation of the employment derived from shellfish harvesting in the region. Assuming employment patterns in the industry are similar between provinces, using primary harvesting as a proxy for total employment should provide a reasonable first-order estimate for comparing relative importance of the fisheries between provinces. The values obtained for both sensitivity indicators were normalised (min-max) to score between 0 and 1, with the province where the fisheries are proportionally most important scor-

ing 1 for each indicator. For each province the two indicator scores were averaged (Fig 4) to obtain a final sensitivity score with a maximum possible score of 1 (Eq 4). A high score indi- cated a higher reliance on the fisheries, and a more sensitive social unit. SenP ¼ ðEmpNP þ ValNPÞ 2 Eq 4 Where Sen is the sensitivity score for province P, and EmpNP is the normalised employment indicator and ValNP is the normalised economic value indicator for province, P. Adaptive capacity: To estimate the ability of each province to potentially respond to changes in fisheries landings, three indicators of adaptive capacity were compiled. Adaptive capacity is typically representative of positive aspects of a society [18], so the indicators were collected with higher scores indicating greater adaptive capacity. The indicator values were then reversed so that a lower adaptive capacity contributed to a higher vulnerability. OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 10 / 29

In other socioeconomic assessments of potential OA effects on fisheries, an indicator repre- senting alternative employment options has been used as a key component of adaptive capacity [14,17]. This is similarly relevant for fisheries in Atlantic Canada, if the availability of fisheries resources decline people currently employed in the industry will need to shift to other employ- ment sectors. As previously mentioned, due to the seasonal nature of fisheries employment in Atlantic Canada it is difficult to distinguish patterns of employment in fisheries as separate from employment in other sectors. Thus, provincial unemployment rates were used to indicate the potential for alternative employment options to fishing [15,17] following an assumption that low provincial unemployment would mean a higher demand for potentially displaced fishers. Provincial unemployment is particularly relevant in Atlantic Canada where rates were 3% to 10% above national rates during the assessment baseline year [51]. Data were averaged across a 10 year period bracketing the year 2000 baseline for the assessment (i.e. 1996–2005 - [51]). The reciprocal of provincial unemployment values were used so that lower provincial unemployment increased adaptive capacity. Education within a society provides an indication of how well a community will be able to respond to changing conditions and has been used in other socioeconomic assessments of OA [15,17]. Higher education levels are generally seen as presenting more opportunities to adapt,

while lower education rates limit options and capabilities to respond to change. The percent of adult populations with at least some post-secondary education was used to estimate the educa- tion level of the five assessed provinces [52]. As with the unemployment data, provincial edu- cation were averaged from 1996 to 2005. The final element that was considered as part of each province’s adaptive capacity was the extent to which potentially OA-impacted species are currently cultured in the region. Aquacul- ture production is expected to be more resilient to OA and climate change than wild harvest, since many environmental conditions can be at least partially controlled or compensated for– especially when hatcheries are used to rear organisms through the most susceptible life-stages [53]. Thus, a strong aquaculture sector could potentially strengthen a community’s ability to respond to OA and climate change. In Atlantic Canada there is no aquaculture production of crustacean species so the aquacul- ture indicator only includes mollusc production. The indicator was constructed to reflect pat- terns in development within the industry as well as relative contribution to total provincial shellfish production. First, the trend in mollusc aquaculture production was estimated for each province by comparing the average annual production for 2001 to 2010 against average annual production for 1991–2000. Second, the fraction of total shellfish production (including both wild harvest and aquaculture) sourced from mollusc species (using the year 2000 average

annual production) was calculated. These two terms were then multiplied together to obtain the indicator value (Eq 5). Aqp ¼ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Aqua2001 2010 Aqua1991 2000 � � � Mol1991 2010 Tot1991 2010 � � s Eq 5 Where Aqp is the aquaculture indicator score for province p. Aqua denotes the average aqua- culture production for the subscripted time-period; Mol indicates mollusc production; and Tot indicates total shellfish production. As with the previous index components, each of the adaptive capacity indicators were trans- formed linearly to score between 0 and 1. To align with the directionality of the other compo- nents in the framework the normalised scores were subtracted from 1 so that weaker adaptive capacity indicators had higher scores (i.e., closer to 1) and therefore contributed to the poten- tial impact. As with the indicators for sensitivity, the individual adaptive capacity indicators OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 11 / 29

were treated as having equal weights. The three adaptive capacity indicators (for employment, education and aquaculture production) were averaged to obtain a cumulative adaptive capac- ity score with a maximum potential value of 1 Eq 6. ACP ¼ ðUENP þ EdNP þ AqNPÞ 2 Eq 6 Where AC is the adaptive capacity score for province P, and UENP is the normalised provincial unemployment indicator, EdNP is the normalised education level indicator and AqNP is the normalised aquaculture indicator. Vulnerability: Sensitivity and adaptive capacity were treated as having equal weights, and their values were averaged to arrive at a social vulnerability score for each province, with a potential maximum value of 1 (Fig 4). Weighting the sensitivity and adaptive capacity compo- nents equal to each other indirectly applied different weights to their composite indicators (Fig 4). However, the concepts of sensitivity and adaptive capacity represent different aspects of social vulnerability so their overall contributions were treated as equally relevant regardless of the number of indicators. Vulnp ¼ ðSenP þ ACPÞ 2 Eq 7 Where VulnP is the vulnerability score, SenP is sensitivity (Eq 4) and ACP is adaptive capacity (Eq 6) for province P. Potential socioeconomic impact. To calculate the final impact index scores the exposure and vulnerability scores were equally weighted and then combined. Each province (or sub-sec- tion thereof for NB and NS) received a quartet of potential impact scores following the quartet

of exposure indicator scores that were applied. The 28 final index scores were divided into five categories of potential impact, and described as ‘high,’ ‘moderate,’ ‘low,’ ‘minimal’ and ‘least’ potential for impact from OA and climate change. Each category represented six separate index scores, except the middle category (low), which only included four scores. Categories were not interpreted as absolute descriptors of the risk posed to the provinces, rather they describe relative levels of potential for impact within the Atlantic Canadian region. PP ¼ ðIP þ VulnPÞ 2 Eq 8 Where P is the potential impact index score, I is the normalised exposure score (Eq 3) and Vuln is the vulnerability score (Eq 7) for province P. Results Species Shellfish make up nearly half of the total annual landing weight in Atlantic Canada. Further- more, shellfish species typically receive a higher value per tonne than finfish, such that shellfish landings make up nearly three quarters of the total annual fisheries value (Fig 1 and S3 Table). Within shellfish landings, crustacean species are dominant, with northern shrimp (Pandalus borealis), snow crab (Chionoecetes opilio) and American lobster (Homarus americanus) accounting for over 60% of the total shellfish landed weight (30%, 21%, and 12% by weight respectively; and 17%, 26% and 41% by landings value, Fig 5). Nonetheless, some mollusc spe- cies also contribute substantially to landings, with sea scallop (Placopecten magellanicus), Stimpson’s surf clam (Mactromeris polynyma), eastern blue mussel (Mytilus edulis) and Ameri-

can oyster (Crassostrea virginica) representing 19%, 7%, 5%, and 3% of all shellfish landings by OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 12 / 29

tonnage, respectively (9%, 3%, 2%, and 1% of total shellfish value, Fig 5). Remaining shellfish species that are harvested commercially combine to make up about 5% of the total shellfish landed weight (and less than 2% by value). Due to their significant contribution to the total commercial fisheries production in Atlantic Canada, changes in landings for the above-identi- fied species would represent a significant change for the value of the whole region’s fisheries. Biophysical results Modelled ocean acidification impacts. When the model outputs ‘with OA’ were com- pared against the outputs ‘without OA’ the overall trends in landing changes were similar. However, the OA effects in all scenarios for all species moderated increases and exacerbated declines (Fig 6). The implementation of OA in the model lead to mollusc species being more strongly influenced (eastern blue mussel were an exception where the treatments were nearly indistinguishable, possibly due to low overall abundance or the eastern blue mussel physiology in the model being more affected by other environmental factors). For crustacean species the Fig 5. Breakdown of total shellfish average annual landings value. Data represent the average annual value from 1991–2010. Top seven most valuable species are indicated by data labels. The remaining shellfish species include other crab species (i.e., not snow crab), sea urchin, whelks and sea cucumber as well as ‘other shellfish,’ which together amount to 1.5% of the total value. Note that ‘clams’ refers to several species, however Stimpsons’ Surf Clam is by far the

largest fraction of this group. https://doi.org/10.1371/journal.pone.0226544.g005 OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 13 / 29

‘with OA’ treatment tracked the ‘without OA’ treatment more closely, but were nonetheless seen to be somewhat lower throughout the simulations. Under the RCP 8.5 scenario the OA impacts were more pronounced than under the RCP 2.6 in all species (except eastern blue mussel) (Fig 6). Atlantic Canadian regional changes. Across the whole Atlantic Canadian region, change in the cumulative net landings in 2090 were anticipated to be relatively neutral across the seven assessed species, regardless of the climate scenario (Fig 7). The largest relative changes were predicted for species with the lowest total landing weights in 2000: American oyster, east- ern blue mussel and Stimpson’s surf clam all have changes exceeding 15% (positive for oyster and mussel, negative for clams) under RCP 8.5, but all have region-wide annual landings under 30,000 tonnes (Fig 7). Thus, relatively large percent changes resulted in comparatively small absolute changes in landings for these species. In terms of absolute change at the regional Fig 6. Comparison of model outputs with OA (red) and without OA (blue) effects through the century, under both climate scenarios (RCP 2.6 as solid lines, RCP 8.5 as dotted lines) for whole Atlantic Canadian EEZ. Values are relative changes in catch potential. Note different axes between sub-plots. (S1 Fig. contains relative changes in distribution for both ‘with OA’ and ‘without OA’ treatments). Lines represent multi-model medians from outputs of three earth system models (GFDL, IPSL, MPI). https://doi.org/10.1371/journal.pone.0226544.g006

OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 14 / 29

scale, northern shrimp were projected to experience the largest absolute increases in produc- tion by the end of the century (+2,000T to +10,000T for RCP 2.6 and RCP 8.5, respectively, out to 2090). However, since northern shrimp represented the highest landing weight species in 2000, the relative gains are modest (2 and 8% for RCP 2.6 and RCP 8.5, respectively). In con- trast to both of these scenarios, snow crab had both high relative change and high absolute landings, with a predicted decline of 16–17% (RCP 2.6 and RCP 8.5, respectively) by 2090, on top of the second highest year 2000 landed weight (~80 000 T) (Fig 7). Results of the DBEM under the RCP 8.5 emissions scenario tend to show more extreme changes in potential future landings, whether positive or negative, than the predictions under the RCP 2.6 scenario (Fig 7). Similarly, by the end of the century (2090) changes were generally expected to be larger than the middle century (2050) (Fig 7). There were some instances where these trends were reversed (e.g., sea scallop across the entire region—(Fig 7)); however, this generally only occurred when projected impacts were very small. Relative changes that are very small, especially when the actual landing weights are also low, are best interpreted as no change or insignificant change. Small relative changes predicted by the model may be the result of interannual variability in the underlying climate models. Nevertheless, for some of the highly landed species (e.g., northern shrimp), even a small percent change implies a sizable change in tonnes landed.

When all seven species are considered in combination across the entire region out to 2090, total landings are projected to decline, if very slightly. Under the RCP 2.6 scenario a total loss of 6,400 tonnes is predicted relative to year 2000 landings (representing a 1.8% reduction in Fig 7. Changes in landing weight for projection scenarios for each species in the three DFO management areas and across the total Atlantic region. Each set of bars from left to right presents the year 2000 catch weight (pale blue), 2050 projected catch (yellow), and 2090 projected catch (green). Sections of each bar with diagonal hashes denote the RCP 8.5 climate scenario projections; the solid (semi-transparent) sections indicate the RCP 2.6 projections. The projected changes for the total regional landings are based on the relative change calculated across the whole study region and therefore do not present the sum of the projected landings from the individual management areas. DBEM relative change projections as percent change for each species in each time-step/climate scenario are available in S1 Table. https://doi.org/10.1371/journal.pone.0226544.g007 OA and climate change impacts on Atlantic Canadian fisheries PLOS ONE | https://doi.org/10.1371/journal.pone.0226544 January 10, 2020 15 / 29