Shocks to fish production: Identification, trends, and consequences

Jessica A. Gepharta,*, Lisa Deutschb, Michael L. Pacea, Max Troellb,c, David A. Seekelld,e,f

aDepartment of Environmental Sciences, University of Virginia, Charlottesville, VA, United Statesb Stockholm Resilience Centre, Stockholm University, Stockholm, SwedencBeijer Institute for Ecological Economics, Royal Swedish Academy of Sciences, Stockholm, SwedendDepartment of Ecology and Environmental Science, Umeå University, Umeå, SwedeneClimate Impacts Research Centre, Umeå University, Abisko, SwedenfArctic Research Centre, Umeå University, Umeå, Sweden

A R T I C L E I N F O

Article history:Received 22 July 2016Received in revised form 28 October 2016Accepted 14 November 2016Available online 25 November 2016

Keywords:FisheriesFood securityFood systemResilienceShocksTrade

A B S T R A C T

Sudden disruptions, or shocks, to food production can adversely impact access to and trade of foodcommodities. Seafood is the most traded food commodity and is globally important to human nutrition.The seafood production and trade system is exposed to a variety of disruptions including fisherycollapses, natural disasters, oil spills, policy changes, and aquaculture disease outbreaks, aquafeedresource access and price spikes. The patterns and trends of these shocks to fisheries and aquaculture arepoorly characterized and this limits the ability to generalize or predict responses to political, economic,and environmental changes. We applied a statistical shock detection approach to historic fisheries andaquaculture data to identify shocks over the period 1976–2011. A complementary case study approachwas used to identify possible key social and political dynamics related to these shocks. The lack of a trendin the frequency or magnitude of the identified shocks and the range of identified causes suggest shocksare a common feature of these systems which occur due to a variety, and often multiple andsimultaneous, causes. Shocks occurred most frequently in the Caribbean and Central America, the MiddleEast and North Africa, and South America, while the largest magnitude shocks occurred in Asia, Europe,and Africa. Shocks also occurred more frequently in aquaculture systems than in capture systems,particularly in recent years. In response to shocks, countries tend to increase imports and experiencedecreases in supply. The specific combination of changes in trade and supply are context specific, which ishighlighted through four case studies. Historical examples of shocks considered in this study can informpolicy for responding to shocks and identify potential risks and opportunities to build resilience in theglobal food system.ã 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND

license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction

Sudden and unexpected changes, or shocks, in food productionand distribution systems can limit access to food and adverselyimpact local nutrition and food security. Such events can initiate acascade of effects through the interlinked social-ecological foodsystem. The ability to respond and adapt to such disruptions whileundergoing change so as to still retain essentially the samefunction, structure, identity, and feedbacks describes the system’sresilience (Walker et al., 2004). Food systems with low resiliencehave limited responses and capacity for adaptation to disruptionsthrough mechanisms like trade, alternative food sources, backup

distribution, or emergency supplies, causing food shortages ofvarying degrees of intensity and duration (Schipanski et al., 2016).Even when food production shortages are temporary, periodswhere essential nutrients are lacking can adversely impact thehealth of vulnerable populations such as pregnant women,children, and the ill (Block et al., 2004). For example, the droughtin the Horn of Africa in 2011 contributed to the food insecurity andmalnutrition of over 11 million people, with one in three childrensuffering from food shortages, widespread decreases in farmer andagribusiness worker incomes, and increased unemployment(UNEP, 2011). Income and asset loss and unemployment through-out the food production chain have lasting impacts for poorfamilies and perpetuate poverty traps (Cuny and Hill, 1999).Therefore, characterizing the nature and frequency of disruptions,or shocks, to food systems is important to understanding thefactors contributing to global food security. Ideally, this insight can

* Correspondence to: National Socio-Environmental Synthesis Center, 1 Park PlSuite 300, Annapolis, MD 21401, United States.

E-mail address: jgephart@sesync.org (J.A. Gephart).

http://dx.doi.org/10.1016/j.gloenvcha.2016.11.0030959-3780/ã 2016 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Global Environmental Change 42 (2017) 24–32

Contents lists available at ScienceDirect

Global Environmental Change

journal homepa ge: www.elsev ier .com/locate /g loe nvcha

be leveraged to prevent or mitigate the effects of future shocks andbuild food system resilience.

Shocks to food production can limit local access to food, but canalso propagate through the international trade network, impactingprices and availability globally. The dynamics of this type of shockpropagation have recently been explored through network models(Gephart et al., 2016; Tamea et al., 2016; Marchand et al., 2016). The2008 grain crisis provides an example of a shock spreading throughthe trade network (Puma et al., 2015; Bren d’Amour et al., 2016).During this event, grain prices spiked due to increased demand forbiofuels, higher oil prices, decreasing grain stocks, and theweakened US dollar (Headey, 2011). Rising wheat prices led India,the second largest rice producer, to ban exports of non-Basmatirice in 2007, which subsequently led other rice exportingcountries, including China, Vietnam, and Egypt, to introduceexport bans (Christiaensen, 2009). Some major importers, includ-ing the Philippines, responded by purchasing additional rice atincreasing prices. Hoarding then further drove up the global priceof rice (Christiaensen, 2009). By the end of the crisis, the WorldBank reported over 130 million people were driven into povertyand the FAO estimated that an additional 75 million people becamemalnourished (Headey, 2011). This case illustrates the potential formultiple stressors (e.g. increasing biofuel demand and oil prices,changes in grain stock policies, and financial crises) to cause shockswhich propagate on large spatial scales, and also illustrates howdifferent sectors are increasingly interconnected (Homer-Dixonet al., 2015). A greater proportion of food is being tradedinternationally between more countries than ever before, andthis increases the potential for shocks to local food systems topropagate into global crises (D’Odorico et al., 2014; Bren d’Amouret al., 2016).

While droughts and the 2008 grain crisis illustrate theconsequences of shocks to agricultural production systems, shocksin fisheries systems are poorly characterized because temporalanalyses have tended to focus on long-term trends rather thansudden drops and their resulting impacts. However, the effect ofshocks is relevant to seafood production because seafood is amongthe most highly traded food commodities and is impacted bymultiple potential shocks including fishery collapses, naturaldisasters, oil spills, policy changes, and aquaculture diseaseoutbreaks (Gephart and Pace, 2015). Further, seafood is the sourceof almost 20% of animal protein consumed globally and anessential source of micronutrients in many coastal developingnations (FAO, 2014; Beveridge et al., 2013). As a result, it isimportant to identify historical cases of shocks to seafood systemsto assess their causes and impacts on trade and domestic seafoodsupply.

There are a variety factors that could contribute to either moreor fewer shocks over time or in particular regions or systems(Table 1). Increasing exploitation, intensification and connectivityof aquaculture, and natural or environmental disasters couldcontribute to more shocks while improved capture fisherymanagement or infrastructure, proactive avoidance measures, orstocks collapsing prior to the study period could contribute tofewer shocks (Table 1). Other factors could contribute to either

more or fewer shocks depending on the particular case, such as theincreasing connectivity of the global market (which could increasepressure on fisheries or provide a buffer) or increased stock dataavailability (which could allow for increased intensification orimproved management). Climate change also serves as a backdropto these factors, by potentially making fishery systems moresusceptible to shocks, by driving a redistribution of marine catches,and by causing more frequent extreme weather disruptions(Cheung et al., 2013; IPCC, 2014; Gattuso et al., 2015). A patternin historical shocks would identify potential vulnerabilities in theseafood production system. This creates opportunities to managemeasurable risks and supports the need to create buffers to hedgeagainst shocks arising from true uncertainty in these complexsystems—i.e. from unknown events impossible to predict (Sumaila,1998; Lauck et al., 1998). Further, patterns in the impact of shockson trade and supply inform whether and when a regional shockwill have distant impacts through international trade or mayimpact local human nutrition.

While shocks have been defined and identified in specificsystems with known causes or based on long time series, thesemethods cannot be applied in general when the shock cause isunknown and long time series data are unavailable. This isparticularly problematic for food production systems, includingfisheries, which are exposed to multiple environmental, policy, andeconomic shocks. One approach is to use expert or local knowledgeto identify events considered shocks to particular systems. Whilethis approach is valuable for studying individual systems, it isdifficult to standardize the definition of a shock across systems andmay be biased against shocks that are not widely reported on orthose which occurred in distant memory. As a result, a data-drivenapproach can complement system knowledge to identify shocksacross systems and over time.

Here we apply a statistical shock identification approach tonational fisheries production time series to answer the followingquestions: 1) have the frequency or intensity of shocks increased;2) do regions or production systems (capture versus aquaculture)have more, larger, or longer shocks; and 3) how are shocks dividedamong decreased exports, increased imports, and changes indomestic supply? We discuss four case studies in detail to illustratethe specific trade and seafood supply impacts of shocks which arisefrom different causes and occur within different contexts.

2. Methods

Shocks can be identified through qualitative approaches basedon literature, news reports, and expert knowledge, or throughquantitative approaches based on outliers or system-specificdefinitions. For example, both heat waves and floods are definedas extremes relative to the historical distribution of events, whiledroughts are identified by indices comparing supply and demandfor soil moisture (e.g. the Palmer drought index). However, thesemethods typically require long time series to generate a distribu-tion or are only relevant for specific types of shocks in a givensystem. While qualitative approaches are useful for studyingindividual systems, potential reporting biases, such as less

Table 1Possible reasons to expect an increase or decrease in the frequency or intensity of shocks in fisheries and aquaculture time series.

Reasons for more shocks Reasons for fewer shocks

Increasing exploitation Stocks already collapsedIncreasing intensification and connectivity of aquaculture Proactive avoidance measuresIncreasing natural or environmental disasters Improved infrastructureRestrictions to improve capture fishery management Improved capture fishery management in the pastIncreasing connectivity of the global market Increasing connectivity of the global marketIncreased stock data connection and availability Increased stock data connection and availability

J.A. Gephart et al. / Global Environmental Change 42 (2017) 24–32 25

reporting in some regions or over time, can limit the use for makingspatial or temporal comparisons. In order to reduce such a bias, weuse a quantitative approach to identify shocks and compliment thisidentification with a search of news, literature, and reports tomatch shocks with potential causes. Combining quantitative andqualitative approaches balances these advantages and disadvan-tages by integrating different strengths and limitations. Such acomplementary approach has previously been used to detectshocks in macroeconomic time series (Balke and Fomby, 1994).

We analyzed shocks in production time series from FAO FishStatfor each country (FAO FishStat, 2014). The “Global CommoditiesProduction and Trade” quantity data was used for the totalproduction shock analysis and the “Global Production by Produc-tion Source” quantity data was used for the production systemanalysis. While this data is known to be incomplete andunderestimate catch from small-scale fisheries (Pauly and Zeller,2016), this analysis is based on the time series patterns rather thanthe exact production estimates. Nevertheless, some limitations ofthe FAO statistics inhibit the detection of certain shocks.Specifically, inaccurate national reporting which masks drops inproduction would prevent the detection of shocks in these timeseries. As a result, shocks occurring in countries known toinaccurately report fishery data to the FAO, such as China, arenot expected to be detected in our analysis (Watson and Pauly,2001). Changes in national reporting practices could appear as ashock, but concerns of such false positives are minimized bypairing the statistical shock detection with a literature, report, andnews search for potential causes. An identified shock due only to areporting change would then likely have an ‘unknown’ cause.Additionally, the underestimation of small-scale fisheries produc-tion in the FAO data means that shocks primarily affecting small-scale fisheries may not be detected, while shocks primarilyaffecting industrial fisheries do not necessarily translate to impactson small-scale fisheries. Shock events which impact both fisheries,such as a natural disaster, would be detected, although the sectorsmay be disproportionately impacted by the event. Despite theselimitations, the FAO data provides global coverage of nationalproduction time series which enables a systematic detection ofshocks.

An existing method commonly used to identify outliers inexploratory spatial statistics was modified to detect shocks basedon deviations in the autocorrelation (Anselin, 1995, 1996). Theapproach is adaptable from spatial to temporal analysis because ofthe equivalence of the theoretical form of some autocorrelationcoefficients between these types of data. The autocorrelationcoefficient is an empirical representation of the relationshipbetween temporal measures of production. Deviations identifylocalized instabilities in the autocorrelation which, conceptually,represent sudden disruptions of the seafood production. Specifi-cally, shocks were identified as outlier deviations, or points withhigh Cook’s D values (>0.35), in a regression of the residuals andlag-1 residuals from a lowess fit of the time series with a smootherspan of 2/3 (see Fig. 1). The threshold of 0.35 was selected bycomparing the total number of shocks identified to the thresholdand selecting the point where the curve became relatively flat(Supplementary Fig. 1).

Shocks can be characterized by the frequency at which theyoccur in a system, the magnitude or intensity of the shock, and theduration or time to recovery. The frequency was defined as theinterval at which a shock occurs and the magnitude was defined asthe difference between a point and the previous 5-year average.For this analysis, only shocks representing a decrease were selectedbecause these are the most likely to adversely impact food securityand economic livelihoods. This means the magnitude is negativefor all shocks, but we display the absolute value of the magnitudein all figures. Recovery of production from a shock was defined as

the point where production returned to within at least 5% of thepre-shock production level. This measure of recovery does nothowever imply sustainable harvest. For example, a system may beoperating beyond a sustainable level and reduce catch down to asustainable level (at which point a shock would be detected) andnever return to the elevated, unsustainable level (i.e. neverrecover). Both temporary and lasting drops in production areconsidered shocks in this analysis because when a shock occurs it isnot generally known if or when catch will return to pre-shocklevels. Consequently, some shocks appear as a point change, whileothers appear as a step change.

Since shocks represent sudden drops in production, thedetection method does not identify long-term, more gradualreductions in fisheries, which are often of concern for thesustainability of a particular stock. Further, this method doesnot identify shocks in systems with high variability (the detectionlimit under different levels of variability is described in theSupplementary information). In systems with high variability,large deviations are frequent and are therefore not consideredshocks for this analysis. For example, although the drop in Peruviananchoveta catch during El Niño is well known, the reported catchdata has high variability and a shock is not detected using ourmethod for the strong El Niño event in 1997–1998, despite the dropin catch that year (Supplementary Fig. 2). Since such drops arecommon and likely more expected in these systems, they are not ashock in the same sense. In fact, regular fluctuations in anchovetacatch are well documented in industry reports and as a result Peruhas implemented coping strategies, including simultaneousownership of fishing fleet and processing factories, low costintensive monitoring, and rapid flexible management (Schreiberet al., 2001; FAO, 2016). Nevertheless, the high variability inproduction will have consequences for trade and seafood supply.For example, low catch periods in this fisheries will have ripplingeffects within the aquaculture sector since this is a unique andcritical component for aquafeed production. However, such

Fig. 1. Steps identifying shocks in time series. (a) A lowess regression was fit to thetime series data; (b) residuals were plotted against the time-lagged residuals; (c)Cook’s D was used to identify extreme points in the regression of residuals versustime-lagged residuals. Points with Cook’s D greater than 0.35 were identified asshocks.

26 J.A. Gephart et al. / Global Environmental Change 42 (2017) 24–32

impacts within systems with high variability are beyond the scopeof this analysis.

We compliment the analysis by searching the literature,reports, and news sources to identify the potential or likelycause(s) of each shock which occurred in the total seafoodproduction time series (Fig. 2, Supplementary Table 2). Positiveidentification of historical disruptions to fishery production thatco-occur with the set of identified shocks strengthens the data-driven shock detection approach. However, the identified causesare not intended to be an exhaustive list of factors contributing tothe observed shock. Instead, they only represent the eventsidentified as potential factors in the sources we were able to locateor major disruptions occurring in the country (identified withasterisk in Supplementary Table 2). Shock causes are classified asone or more of the following: political (i.e. country breaking up,war, financial crisis, etc.), overfishing, policy change (related tofisheries), aquaculture disease, natural disaster, or unknown. Thetrade and supply response was quantified based on the value ofimports, exports, and supply at the shock point relative to theprevious five-year reference period. FAOSTAT (2014) supply data iscalculated as the production and imports minus exports, domesticuse as animal feed and waste, plus any change in stocks, divided bythe population. Supply therefore represents a proxy of per capitaseafood available for consumption, but is not a direct measurementof actual consumption. We focus on the response to a shock withinthe seafood system and therefore buffering mechanisms, such asuse of grain stocks or imports of non-seafood commodities, arebeyond the scope of this analysis.

3. Results and discussion

3.1. Patterns and trends in shocks to national seafood production

We detected 48 shocks between 1976 and 2011 within 127national time series of total seafood production. While regionsgenerally experienced a similar number of total shocks, the shockrate (number of shocks divided by the number of time series) wasmuch higher in some regions (Fig. 3). For example, the shock rate in

Fig. 2. Shock magnitude for each year in total fisheries production time series for each country. Shocks were identified in FAO FishStat total national production time seriesusing the shock detection approach described in the methods. There is no significant trend in shock magnitude (p = 0.63, r2< 0.001) or number of shocks (p = 0.31, r2 = 0.005).Points are colored according to the identified shock cause.

Fig. 3. Shock rate (number of shocks divided by the number of time series in theregion), magnitude, number of recovered and not recovered cases, and recoverytime by region. Shocks were identified in FAO FishStat total national productiontime series using the shock detection approach described in the methods. Recoverywas defined as returning to within 5% of the previous 5-year average. Note thatAfrica refers to Sub-Saharan Africa, CaCA refers to the Caribbean and CentralAmerica, MENA refers to the Middle East and Northern Africa, N. Am refers to NorthAmerica, and S. Am refers to South America.

J.A. Gephart et al. / Global Environmental Change 42 (2017) 24–32 27

the Caribbean and Central America was two and a half times theshock rate in Africa and about twice the rate in Asia and Europe(Fig. 3). Although the Caribbean and Central America, and SouthAmerica had among the highest shock rates, they also had a higherpercent of cases where the production recovered to pre-shocklevels (80%) and the recovery times for these regions were amongthe lowest. This is compared to only 30% of cases returning to pre-shock levels in Europe, sub-Saharan Africa, and the Middle East andNorth Africa. Shock magnitudes tend to be fairly similar acrossregions, but the highest mean magnitudes occurred in Europe,Asia, and sub-Saharan Africa (Fig. 3). In general, the distributions ofmagnitudes and recovery times are asymmetric, such that mostshocks are small and most recoveries are quick, but when they arenot the largest shocks are much larger or longer than the medians.

Shocks did not become more frequent or larger in the nationalseafood production series over time (Fig. 2). This suggests thatshocks are a common feature of these production systems.Similarly, Sartori and Schiavo (2015) found no increase in thenumber of shocks in agricultural systems in the past 25 years. Theshocks to seafood production occurred due to a variety of identifiedcauses, but were dominated by political factors, fishery policychanges (e.g. new catch limits), and overfishing, often coupled witha policy change to limit fishing pressure (Fig. 2). No trend analysiswithin shock-cause categories can be reliably conducted due to thepotential bias of the cases with unknown causes. The lack of trendsand variety of factors supports a mixture of the hypothesizedreasons to expect an increase or decrease in shocks over time(Table 1).

Political factors, such as the breakup of a country, war, orfinancial crises, were frequently identified as a potential cause of aseafood production shock. In fact, the largest shock identified, thebreakup of the USSR (described in more detail below) was apolitical shock. However, such political disruptions occur atirregular intervals and therefore would not be expected to drivea trend over this period. We expected that overexploitation couldbe leading to more shocks or that improved management could bereducing shocks. While overfishing and policy measures to avoidoverfishing are frequently identified as causes of drops inproduction, there is not a clear trend over the period considered.Since the policy changes are typically aimed at reducingoverfishing, one would expect that these systems will experiencefewer shocks due to overfishing in the long run. These cases ofshocks may be examples of the short term cost of improvingmanagement (i.e. reduced harvests) for long term sustainability.

Countries are likely able to anticipate or prevent shocks tovarying degrees in different cases. Slow developing situationsleading to shocks, such as overfishing, can be monitored andexpected if action is not taken. Such slow drivers offer thepossibility of statistical early warning indicators through monitor-ing that would allow a response or preparations prior to a dramaticchange (Litzow et al., 2008; Carpenter et al., 2011; Seekell et al.,2012; Cline et al., 2014). Policy changes which are slowly phased inor have delayed implementation dates also allow the productiondrops to be expected. Anticipated decreases in seafood productionmay allow stakeholders to prepare in advance of the shock, whichwould mitigate the shock’s impacts. Other shock causes, such as adisease outbreak or natural disaster, are generally less predictable.This means there is less time for any management intervention orpreparation. As a result, the resilience of the food system could bediversified with additional food sources (Troell et al., 2014),maintaining backup distribution mechanisms, or emergencysupplies, and building capital to cope with crises in order toreduce the societal impacts of a shock.

When disaggregated into aquaculture and capture fishery timeseries, the shock magnitude and recovery tended to be similar forcapture and aquaculture production systems (Supplementary

Fig. 3). However, the shock rate tended to be higher for aquaculturethan capture fisheries from the late 1980s to the present, a periodof rapid aquaculture development (Supplementary Fig. 4). Whilethe cause of this difference is unknown, aquaculture is vulnerableto shocks from disease outbreaks and possibly also from rapidgrowth over-shooting environmental carrying capacity. Thepropensity for shocks to aquaculture adds caution to suggestionsthat aquaculture alone will be able to reliably meet future seafooddemands (Liu and Sumaila, 2008; Troell et al., 2014). However, theaquaculture sector is highly diverse with a multitude of species andsystems under different governance regimes and they differ greatlyfrom a resilience perspective. Further, by diversifying food sourcesand enabling greater control throughout the production system,aquaculture has the potential to add resilience to the overall foodsystem, particularly when the potential sources of shocks are takeninto consideration (Troell et al., 2014).

3.2. Shock impacts on trade and seafood supply

Countries are expected to respond to a shock to seafoodproduction in the short-run through a combination of increasedimports, decreased exports, and decreased supply. In the detectedshocks, imports increased in over one and a half times as manycases as it decreased. Exports increased and decreased equallycommonly, while supply decreased nearly twice as often as itincreased (Table 2). The most common combinations were: 1)imports and supply decreases with export increases; 2) increasesin all three, and; 3) import increases with export and supplydecreases. Counterintuitive increases in exports suggest that somecomponent of the seafood industry is unaffected by the shock. Thespecific impact on trade and supply is context dependent, withtrends in the trade balance and the fishery or fisheries beingaffected playing a large role in how the shock impacts trade andsupply. To illustrate this point, four shock cases and the impact ontrade and supply are described below.

3.2.1. Former USSR: a case of political and policy changesThe shock with the largest magnitude occurred in 1992 in the

former USSR countries and can be attributed to the breakup of theSoviet Union in 1991. From the 1970s up to 1991 the Soviet Unionsupported a large coastal and distant fishing industry with anestimated $30 billion in subsidies (Milazzo, 1998; Österblom andFolke, 2015). This led to an overcapitalization of the Russian fleetand supported high levels of total catch (Fig. 4). The subsidies also

Table 2Frequency of combinations of increases (+), decreases (�), and no change (o) inimports, exports, and supply at a shock point and total increases, decreases, and nochange observed for imports, exports, and supply.

Frequency Imports Exports Supply

10 � + �9 + + +7 + � �5 + � +5 � � �4 + + �4 + � o2 � � o1 � + +1 � � +1 + + o1 o � �

Total Increases 30 25 16Total No Changes 1 0 7Total Decreases 19 25 27

28 J.A. Gephart et al. / Global Environmental Change 42 (2017) 24–32

resulted in an inefficient fleet, with Soviet ships landing 1/5 of thecatch per ton of fishing fleet compared to the EU or Japan at the endof the Soviet Era (Kravanja and Shapiro,1993). Financial support forthe fisheries rapidly disappeared after the dissolution of the SovietUnion and the aging ships were divided among the newlyindependent states (Milazzo, 1998). Fish catch dropped precipi-tously during this transition (Fig. 4). Without the subsidies mostfishing operations were no longer profitable and 80–85% of theassessed fishing enterprises were filing or near filing bankruptcy(Milazzo, 1998).

Exports increased gradually during the 1980s after the USSRopened trade to socialist countries, but decreased during the early1990s during the dissolution of the USSR (Fig. 4). The exportsrebounded by the mid-1990s and continued to increase thereafter,coinciding with the former Soviet countries opening to trade withthe West. In the period immediately following the dissolution ofthe Soviet Union the exports as a percent of catch increaseddramatically. This situation is exemplified by Estonia, whereforeign trade opening within the country caused the price of fish togrow to nearly the level of Western Europe and for fish exports toincrease rapidly (Vetemaa et al., 2006). Further, Estonia’sindependence allowed access to fishing grounds that werepreviously tightly regulated by Soviet border controls (Vetemaaet al., 2006). The Estonian government passed policies aimed atincreasing access to fisheries for household consumption, but highprices for fish resulted in catches being sold to traders for export(Vetemaa et al., 2006). The decreased catch in conjunction with theincreased exports can explain the initial drop in per capita seafoodsupply in 1992 and the further decline through 1994 (Fig. 4). Therebound in supply corresponds to the gradual increase in importsand catch. Nevertheless, the supply and catch do not return to pre-shock levels by the end of the time series (Fig. 4). This caseillustrates a large political shock with lasting impacts throughoutthe fishing industry. Clearly, the breakup of the USSR had impactsfar beyond fishery catch, including the dramatic changes in tradepolicies which can help explain the increases in both imports andexports which may not otherwise be expected at a shock point.

3.2.2. Ghana: a case of overfishingGhana has historically been a major fishing nation in West

Africa, with a high reliance on seafood for nutrition, employment,

and the national economy (Atta-Mills et al., 2004). Ghana’sproductive coastal waters in the Gulf of Guinea result from theCentral West African upwelling system. There is seasonalvariability in the fishery’s productivity due to annual upwellingcycles, while interannual variability is driven by large-scaleatmospheric pressure systems in the South Atlantic and El Niñoevents in the tropical Pacific (Perry et al., 2011). Despite this naturalvariability, the year 2000 represents a shock that falls outside thenormal variability of the system and over-exploitation is the mostlikely explanation for this drop in catch (Fig. 4; Atta-Mills et al.,2004). By the mid-1990s landings of pelagic fish had leveled offand inshore marine resources were fully- or over-exploited (Perryet al., 2011; Atta-Mills et al., 2004). Further, catch per unit effort fordemersal species declined through the 1980s and 1990s (Koran-teng, 2002). Total catches were maintained through the 1990s byfishing farther off shore or switching gear to target different or newspecies. Historically, Ghanaian fishermen had adapted to periods oflow catch by migrating to new fishing areas, but enforcement of theEconomic Exclusive Zones and other policy actions by neighboringcountries limited migration opportunities (Atta-Mills et al., 2004).

Despite the drop in total production in 2000, seafood exportsremained relatively constant, thereby representing an increase inthe percent of seafood exported (Fig. 4). During the mid-1990swhen catch per unit effort was declining, imports increased andGhana became a net importer of seafood. Now, Ghana primarilyexports high value species (e.g. shrimp, tuna, cuttlefish) whileimporting lower value, frozen seafood (Atta-Mills et al., 2004).Around this time per capita supply of seafood began to track thepattern in imports (Fig. 4). Although the catch and supply numbersfor Ghana likely underestimate the role of subsistence fisheries(Nunoo et al., 2006; Pauly and Zeller, 2016), the data capture thedominant patterns in production, trade, and supply. This caseillustrates historical overfishing, possibly in combination withlimitations on fishing in neighboring waters, leading to a drop inproduction and a long-term trend of increasing imports compen-sating for stagnating catches.

3.2.3. Saint Pierre and Miquelon: a case of political dispute, policychange, and overfishing

Saint Pierre and Miquelon are French territorial islands off thecoast of Newfoundland. The islands’ economy has traditionally

Fig. 4. Time series of production (black), imports (green), exports (red), and per capita supply (blue; right axis scale) for the former USSR, Ghana, Saint Pierre and Miquelon,and Sri Lanka. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

J.A. Gephart et al. / Global Environmental Change 42 (2017) 24–32 29

been based on fishing and servicing fishing vessels (The WorldFactbook, 2016). Cod is the most important fishery for the islandsand this case illustrates a situation where seafood catch is largelydestined for export and the exports trace the annual catch veryclosely (Fig. 4). France’s fishing rights in the waters offNewfoundland date back to the Treaty of Utrecht of 1713, butbecame a source of dispute in the 1977 when both France andCanada extended their fishing zones to 200 NM from their coasts(McDorman, 1990). This resulted in overlapping claims to waterswith productive fisheries and potential hydrocarbon resources(McDorman, 1990). A 1972 agreement allowing France access to17,500 tons of catch kept the dispute at bay. But, the territorialstruggle escalated and peaked in 1987–1988 when Canada claimedFrance was exceeding its fishing quota, denied the agreementrenewal, and blocked French vessels from ports and the fishinggrounds (Burns, 1988). Canada then arrested the crew of a vesselregistered in Saint Pierre and Miquelon and France retaliated byexpelling the Canadian ambassador to France and denyingCanadian citizens entry at Parisian airports (McDorman, 1990).The 1988 dispute is identified as the first shock point in Fig. 4.Canada and France reached an agreement through mediation in1989 and fish catch in Saint Pierre and Miquelon rebounded(McDorman, 1990; Fig. 4). However, in 1993, a much larger shockoccurred when the cod stock was near commercial extinction andthe entire fishery was closed to rebuild stocks (Hutchings andMyers, 1995). Immediately following the reductions in catch, SaintPierre and Miquelon’s seafood imports increased, before catchincreased to a moderate level compared to the pre-shockconditions (Fig. 4). FAO seafood supply information is unavailablefor these islands, but the collapse and closure of the cod fisheryseverely impacted the livelihoods of the people in the region(Milich, 1999). This case illustrates a political dispute and a policychange, both with a back drop of overfishing.

3.2.4. Sri Lanka: a case of a natural disasterPrior to the December 2004 tsunami, Sri Lankan fisheries

employed around 163,000 people, with subsistence fishingproviding a livelihood for many unemployed people (Ministry ofFisheries and Aquatic Resources, 2003). Sri Lanka’s fisheries areknown to have been stressed prior to 2005, but the tsunami andresulting devastation was directly associated with the 2005 shockto production. Ten of the twelve main fishing harbors wereseverely damaged, along with 65% of the fishing fleet (De Silva andYamoa, 2007). Damage to fishing craft and gear was particularlysevere because the event occurred on a holiday when the boatswere inshore and received the full impact of the tsunami (Stirrat,2006). There was also significant damage to the post-harvestsector, including markets and retail stalls, as well as concerns aboutdamaged waste water systems leaking into fishing grounds (DeSilva and Yamao, 2007). Immediately following the disaster, a vastrange of relief organizations became active in Sri Lanka. Afternatural disasters, there is an incentive for NGOs to spend money onvisible aid, including in this case distributing new fishing gear(Stirrat, 2006). Such actions resulted in the number of boats insome areas of Sri Lanka to exceed the number of boats prior to thetsunami (FAO, 2007). This, along with the new boats having highercatching power than the old boats, can likely explain the sharpjump in seafood production the year following the tsunami (FAO,2007; Fig. 4).

The majority of seafood produced in Sri Lanka is through small-scale fisheries and is destined for domestic consumption. This isreflected in Fig. 4 where the patterns of per capita supply mirrorthe patterns of seafood production. Sri Lanka is highly dependenton seafood, with 52% of animal protein derived from seafood andmuch higher levels of dependency in coastal fishing communities(De Silva and Yamoa, 2007). Imported dry and canned fish flooded

the retail markets immediately after the tsunami, but the averageprices were substantially higher than the average prices for localfish (Subasinghe, 2005). Overall, there was not an increase in theimports for 2005, but there was a substantial drop in per capitaseafood supply at the shock point (Fig. 4). This case illustrates ashock from a natural disaster and the impacts this type of shock canhave throughout the seafood system. It also illustrates a case wherechanges in trade fail to compensate for the drop in production,resulting in a temporary decrease in local seafood supply withpossible impacts on the local food security.

3.3. Impacts beyond national seafood supply and trade balance

Shocks to seafood production extend beyond the per capitaseafood supply and national trade balance. Capture fisheriesemployed 58.3 million people in 2012, with 37% of those peopleemployed full time (FAO, 2014). Employment in the fishery sectorhas grown at a faster rate than the world population and traditionalagriculture sector (FAO, 2014). A majority of people employed infisheries live in Asia and Africa and the FAO estimates that thefisheries sector assures the livelihoods of 10–12 percent of theworld’s population (FAO, 2014). These figures may underestimatethe number of people in the developing world employed throughsubsistence fishing (Teh and Sumaila, 2013) as well as employmentbeing generated throughout the value chains and associatedbusinesses (Béné et al., 2016). As a result, shocks can impact GDPand unemployment levels at the national scale and can havelasting impacts on those relying on fisheries income.

Declines in fishery catch can also cause shifts in employment,crime, and sources of food. For example, negative economic shocksto fisheries are correlated with an increase in piracy and decliningfish harvests have been linked to increases in human traffickingwhen fishers attempt to minimize production costs (Flückiger andLudwig, 2014; Brashares et al., 2014). Declines in seafood catchhave also been linked to increased hunting in nature preserves andthe sale of bushmeat in local markets in West Africa (Brashareset al., 2004). Thus, fishery shocks can reach beyond trade andnutrition, spill over negatively into other resource systems, andimpact human trafficking, organized crime, and biodiversityconservation.

While this study focused on the trade balance impacts at thenational scale, changes in imports and exports imply shifts in thetrade partners’ trade balances. A series of recent studies haveexplored the distant impacts of shocks to primary commodities inthe global trade network during the 2008 grain crisis and throughnetwork models (Puma et al., 2015; Gephart et al. 2016; Tameaet al., 2016; Bren d’Amour et al., 2016; Marchand et al., 2016). Thesestudies have found import-dependent countries, countries withlow production diversity, and regions with low willingness to payas being more vulnerable to external shocks in the network.Marchand et al. (2016) found that national reserves dampen thepropagation of a shock. This suggests that since seafood is not heldin reserves in the way grains are, the propagation of a shockoriginating from the fishery system could be more far-reaching.There are likely other knock-on effects of shocks in substitute foodcommodity systems which were not studied here, but areimportant to the overall food security impacts of shocks.Evaluating these distant and knock-on impacts of shocks inhistorical examples, such as those identified here is an importantnext step in understanding how shocks may alter the global tradenetwork and impact food security and human well-being.

In addition to adapting to changes in domestic seafoodproduction through trade, countries can replace seafood consump-tion by increasing the production of agriculture and livestocksubstitute foods. Such changes operate on a longer timescale than asingle year shock and are likely important for step-changes in the

30 J.A. Gephart et al. / Global Environmental Change 42 (2017) 24–32

level of seafood production. However, countries’ abilities toproduce substitute foods is limited by their available land andwater resources. For example, Gephart et al. (2014) evaluates thewater cost and ability of countries to replace marine protein withterrestrial foods based on current consumption patterns and waterresources. Developing coastal African and island nations were themost limited in their ability to replace marine protein throughdomestic agriculture and livestock production. Projections ofchanges in consumption of seafood and its substitutes underseafood production declines requires supply and demand model-ing. For example, both the International Food Policy ResearchInstitute’s International Model for Policy Analysis of AgriculturalCommodities and Trade and WorldFish’s AsiaFish Model providetools for projecting changes in consumption patterns underscenarios of seafood production declines (Delgado et al., 2003;Briones et al., 2004). However, such projections focus on gradualchange and do not fully capture the short-run changes thatimmediately follow a shock.

Sudden decreases in seafood production propagate shocksthrough multiple components of the food production system.Immediate responses are that countries may import more orexport less seafood, or domestic seafood supply may drop. Thoseemployed in fishing or fish processing may become unemployed orseek work elsewhere, resulting in potentially unexpected con-sequences. Decreases in seafood supply likely result in increasedconsumption of substitute foods, but may also result in restrictednutritional access for those with limited access to substitute foods.In the longer-run, countries which do not recover from the shockmay increase the production of substitute foods. However, theability to do so is limited by available natural resources. Thecapacity of countries and communities to respond and adapt toshocks to seafood production speaks to their resilience. Learningfrom historical examples of shock causes, impacts, and responsesprovide opportunities to build resilience.

4. Conclusion

Shocks are a common feature of seafood production systemswhich occur due to a variety, and often multiple and simultaneous,causes. Shocks occurred the most frequently in Central Americaand the Caribbean, the Middle East and North Africa, and SouthAmerica, but the largest magnitude shocks occurred in Asia,Europe, and Africa. Shocks also occurred more frequently inaquaculture than in capture systems. The complementary quanti-tative and qualitative methods employed here provide a system-atic approach to look back in time to identify shocks and evaluatetheir impacts in an increasingly globalized system. While the tradebalance and food supply response to shocks is context specific,there is a tendency for the imports to increase and the supply todecrease. Historical examples of shocks can inform policyconsiderations for responding to shocks and learning from theseexamples helps identify potential risks and opportunities tobuilding resilience in the global food system.

Acknowledgements

Wewouldliketothank thevolunteersat theStockholmResilienceCentre who gave valuable feedback during the development of thisproject. This research was supported in part by the Department ofEnvironmental Sciences at the University of Virginia, the NationalScience Foundation Graduate Research Fellowship, the NationalScience Foundation Graduate Research Opportunities Worldwideprogram, the Swedish Research Council, and the National Socio-Environmental Synthesis Center under funding received from theNational Science Foundation DBI-1052875.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.gloenvcha.2016.11.003.

References

Österblom, H., Folke, C., 2015. Globalization, marine regime shifts and the SovietUnion. Philos. Trans. R. Soc. B Biol. Sci. 370, 20130278.

Anselin, L., 1995. Local indicators of spatial association––LISA. Geog. Anal. 27 (2),93–115.

Anselin, L., 1996. The Moran scatterplot as an ESDA tool to assess local instability inspatial association. Spatial Analytical Perspectives on GIS, , pp. 111–125.

Atta-Mills, J., Alder, J., Sumaila, U.R., 2004. The decline of a regional fishing nation:the case of Ghana and West Africa. Nat. Resour. Forum 28, 13–21.

Béné, C., Arthur, R., Norbury, H., Allison, E.H., Beveridge, M.C.M., Bush, S., Campling,L., Leschen, W., Little, D., Squires, D., Thilsted, S., Troell, M., Williams, M., 2016.Contribution of fisheries and aquaculture to food security and povertyreduction: assessing the current evidence. World Dev. 79, 177–196.

Balke, N.S., Fomby, T.B., 1994. Large shocks, small shocks, and economicfluctuations: outliers in macroeconomic time series. J. Appl. Econometrics 9 (2),181–200.

Beveridge, M.C.M., Thilsted, S.H., Phillips, M., Metian, M., Troell, M., Hall, S.J., 2013.Meeting the food and nutrition needs of the poor: the role of fish and theopportunities and challenges emerging from the rise of aquaculture. J. Fish Biol.83 (4), 1067–1084.

Block, S.A., Kiess, L., Webb, P., Kosen, S., Moench-Pfanner, R., Bloem, M.W., Timmer,C.P., 2004. Macro shocks and micro outcomes: child nutrition during Indonesia’scrisis. Econ. Hum. Biol. 2, 21–44.

Brashares, J.S., Arcese, P., Sam, M.K., Coppolillo, P.B., Sinclair, A.R.E., Balmford, A.,2004. Bushmeat hunting, wildlife declines, and fish supply in West Africa.Science 306 (5699), 1180–1183.

Brashares, J.S., Abrahms, B., Fiorella, K.J., Golden, C.D., Hojnowski, C.E., Marsh, R.A.,McCauley, D.J., Nuñez, T.A., Seto, K., Withey, L., 2014. Wildlife decline and socialconflict. Science 345 (6195), 376–378.

Bren d’Amour, C., Wenz, L., Kalkuhl, M., Steckel1, J.C., Creutzig, F., 2016.Teleconnected food supply shocks. Environ. Res. Lett. 11, 035007.

Briones, M., Dey, M.M., Ahmed, M., 2004. The future for fish in the food andlivelihoods of the poor in Asia, NAGA. WorldFish Center Q. 27 (3–4), 48–50.

Burns, J.F., 1988. Canada-France Fishing Dispute Runs Deep. The New York Times (8May 1988).

Carpenter, S.R., Cole, J.J., Pace, M.L., Batt, R., Brock, W.A., Cline, T., Coloso, J., Hodgson,J.R., Kitchell, J.F., Seekell, D.A., Smith, L., Weidel, B., 2011. Early warnings ofregime shifts: a whole-ecosystem experiment. Science 332, 1079–1082.

Cheung, W.L.W., Watson, R., Pauly, D., 2013. Signature of ocean warming in globalfisheries catch. Nature 497, 365–386.

Christiaensen, L., 2009. Revisiting the Global Food Architecture: Lessons from the2008 Food Crisis. World Institute for Development Economics Research(Discussion Paper No. 2009/04).

Cline, T.J., Seekell, D.A., Carpenter, S.R., Pace, M.L., Hodgson, J.R., Kitchell, J.F., Weidel,B.C., 2014. Early warnings of regime shifts: evaluation of spatial indicators froma whole-ecosystem experiment. Ecosphere 5 (8) (article 102 13 pages).

Cuny, F.C., Hill, R.B., 1999. Famine, Conflict, and Response: A Basic Guide. KumarianPress, West Hartford CT.

D’Odorico, P., Carr, J., Francesco, L., Ridolfi, L., Vandoni, S., 2014. Feeding humanitythrough global food trade. Earth’s Fut. 2 (9), 458–469.

De Silva, D.A.M., Yamao, M., 2007. Effects of the tsunami on fisheries and coastallivelihood: Aa case study of tsunami-ravaged southern Sri Lanka. Disasters 31(4), 386–404.

Delgado, C.L., Wada, N., Rosegrant, M.W., Meijer, S., Ahmed, M., 2003. Fish to 2020:Supply and demand in changing global markets. World Fish Center Tech. Rep.62.

Food and Agriculture Organization, FishStat database. (2014). Available at: http://www.fao.org/fishery/topic/166235/en.

Food and Agriculture Organization of the United Nations (FAO) Regional Office forAsia and the Pacific. (2007). An overview of the impact of the tsunami onselected coastal fisheries resources in Sri Lanka and Indonesia, Bangkok.

Food and Agriculture Organization of the United Nations (FAO). (2014). The State ofWorld Fisheries and Aquaculture 2014. Rome, 223 pp.

Food and Agriculture Organization. (2016). GLOBEFISH—Analysis and informationon world fish trade: Fish oil and fish meal market reports. Available at: http://www.fao.org/in-action/globefish/market-reports/fish-oil-and-fishmeal/en/.

Food and Agriculture Organization. (2014). FAOSTAT database, Available at: http://faostat3.fao.org/faostat-gateway/go/to/home/E.

Flückiger, M., Ludwig, M., 2014. Economic shocks in the fisheries sector andmaritime piracy. J. Dev. Econ. 114, 107–125.

Gattuso, J.-P., Magnan, A., Billé, R., Cheung, W.W.L., Howes, E.L., Joos, F., Allemand, D.,Bopp, L., Cooley, S.R., Eakin, C.M., Hoegh-Guldberg, O., Kelly, R.P., Pörtner, H.-O.,Rogers, A.D., Baxter, J.M., Laffoley, D., Osborn, D., Rankovic, A., Rouchtte, J.,Sumaila, U.R., Treyer, S., Turley, C., 2015. Contrasting futures for ocean andsociety from different anthropogenic CO2 emissions scenarios. Science 349(6243) doi:http://dx.doi.org/10.1126/science.aac4722.

J.A. Gephart et al. / Global Environmental Change 42 (2017) 24–32 31

Gephart, J.A., Pace, M.L., 2015. Structure and evolution of the global seafood tradenetwork. Environ. Res. Lett. 10 (12), 125014.

Gephart, J.A., Pace, M.L., D’Odorico, P., 2014. Freshwater savings from marine proteinconsumption. Environ. Res. Lett. 9. doi:http://dx.doi.org/10.1088/1748-9326/9/1/014005.

Gephart, J.A., Å. Brännström, Rovenskaya, E., Diekmann, U., Pace, M.L., 2016.Vulnerability to shocks in the global seafood trade network. Environ. Res. Lett.11 (3), 035008.

Headey, D., 2011. Rethinking the global food crisis: the role of trade shocks. FoodPolicy 36, 136–146.

Homer-Dixon, T., Walker, B., Biggs, R., Crépin, A.S., Folke, C., Lambin, E.F., Peterson, G.D., Rockström, J., Scheffer, M., Steffen, W., Troell, M., 2015. Synchronous failure:the emerging causal architecture of global crisis. Ecol. Soc. 20 (3), 6.

Hutchings, J.A., Myers, R.A., 1995. The biological collapse of Atlantic cod offNewfoundland and Labrador: an exploration of historical changes inexploitation, harvesting technology, and management. In: Arnason, R., Felt, L.(Eds.), The North Atlantic Fisheries: Successes, Failures, and Challenges.Institute of Island Studies, University of Prince Edward Island, Charlottetown, P.E.I, pp. 37–93.

IPCC, 2014. Climate change 2014: synthesis report. In: Core Writing Team, Pachauri,R.K., Meyer, L.A. (Eds.), Contribution of Working Groups I, II and III to the FifthAssessment Report of the Intergovernmental Panel on Climate Change. IPCC,Geneva, Switzerland 151 pp..

Koranteng, K.A., 2002. Status of demersal fishery resources on the inner continentalshelf off Ghana. In: McGlade, J.M., Cury, P., Koranteng, K.A., Hardman-Mountford, N.J. (Eds.), The Gulf of Guinea Large Marine Ecosystem. ElsevierScience, Amsterdam, pp. 261–274.

Kravanja M., E. Shapiro. (1993). The Baltic States, The Commonwealth ofIndependent States, and Eastern Europe (fishing fleets). Published in: WorldFishing Fleets: An Analysis of Distant-water Fleet Operations. Past-Present-Future. Volume V. Prepared by the Office of International Affairs, NationalMarine Fisheries Service, NOAA, U.S. Department of Commerce Silver SpringMaryland, November 1993.

Lauck, T., Clark, C.W., Mangel, M., Munro, G.R., 1998. Implementing theprecautionary principle in fisheries management through marine reserves. Ecol.Appl. 8 (1), S72–S78.

Litzow, M.A., Urban, J.D., Laurel, B.J., 2008. Increased spatial variance accompaniesreorganization of two continental shelf ecosystems. Ecol. Appl. 18, 1331–1337.

Marchand, P., Carr, J., Dell’Angelo, J., Fader, M., Gephart, J.A., Kummu, M., Magliocca,N., Porkka, M., Puma, M., Ratajczak, Z., Rulli, M.C., Seekell, D.A., Suweis, S.,Tavoni, A., D’Odorico, P., 2016. Reserves and trade jointly determine exposure tofood supply shocks. Environ. Res. Lett..

McDorman, T.L., 1990. The Canada-France maritime boundary case: drawing a linearound St. Pierre and Miquelon. Am. J. Int. Law 84 (1), 157–189.

Milazzo, M., 1998. Subsidies in world fisheries: a reexamination. World BankTechnical Paper; No. WTP 406. Fisheries Series. The World Bank, Washington, D.C.

Milich, L., 1999. Resource mismanagement versus sustainable livelihoods: thecollapse of the Newfoundland cod fishery. Soc. Nat. Resour. 12, 625–642.

Ministry of Fisheries and Aquatic Resources, 2003. Annual Report. Ministry ofFisheries and Aquatic Resources, Colombo.

Nunoo, F.K.E., Asiedu B., Amador K., Belhabib D., Pauly D. (2006). Reconstruction ofmarine fisheries catches for Ghana, 1950–2010. The University of BritishColumbia Fisheries Centre Working Paper Series, Working Paper #2014–13.

Pauly, D., Zeller, D., 2016. Catch reconstructions reveal that global marine fisheriescatches are higher than reported and declining. Nat. Commun. 7, 10244.

Perry, R.I., Ommer, R.E., Barange, M., Jentoft, S., Neis, B., Sumaila, U.R., 2011. Marinesocial-ecological responses to environmental change and the impacts ofglobalization. Fish Fish. 12, 427–450.

Puma, M.J., Bose, S., Young Chon, S., Cook, B.I., 2015. Assessing the evolving fragilityof the global food system. Environ. Res. Lett. 10, 24007. doi:http://dx.doi.org/10.1088/1748-9326/10/2/024007.

Sartori, M., Schiavo, S., 2015. Connected we stand: a network perspective on tradeand global food security. Food Policy 57, 114–127.

Schipanski, M.E., MacDonald, G.M., Rosenzweig, S., Chappell, M.J., Bennett, E.M.,Kerr, R.B., Blesh, J., Crews, T., Drinkwater, L., Lundgren, J.G., Schnarr, C., 2016.Realizing resilient food systems. BioScience doi:http://dx.doi.org/10.1093/biosci/biw052.

Schreiber, M.A., Ñiquen, M., Bouchon, M., 2001. Coping strategies to deal withenvironmental variability and extreme climatic events in the Peruvian anchovyfishery. Sustainability 3, 823–846.

Seekell, D.A., Carpenter, S.R., Cline, T.J., Pace, M.L., 2012. Conditionalheteroskedasticity forecasts regime shift in a whole-ecosystem experiment.Ecosystems 15, 741–747.

Stirrat, J., 2006. Competitive humanitarianism. Anthropol. Today 22 (5), 11–16.Subasinghe, S., 2005. Sri Lanka: Assessment of Rehabilitation and Re-construction

Needs in the Tsunami Affected Post-harvest Fisheries Sector. FAO, Rome.Sumaila, U.R., 1998. Protected marine reserves as fisheries management tools: a

bioeconomic analysis. Fish. Res. 37, 287–296.Tamea, S., Laio, F., Ridolfi, L., 2016. Global effects of local food-production crises: a

virtual water perspective. Sci. Rep. 6 doi:http://dx.doi.org/10.1038/srep18803.Teh, L.C.L., Sumaila, U.R., 2013. Contribution of marine fisheries to worldwide

employment. Fish Fish. 14, 77–88.The World Factbook: Saint Pierre and Miquelon, Washington, DC. Central

Intelligence Agency, Updated February (2016). https://www.cia.gov/library/publications/the-world-factbook/index.htm (Accessed 17 February 2016).

Troell, M., Naylor, R.L., Metian, M., Beveridge, M., Tyedmers, P.H., Folke, C., Arrow, K.J., Barrett, S., Crépin, A.-S., Ehrlich, P.R., Å. Gren, Kautsky, N., Levin, S.A., Nyborg,K., Österblom, H., Polasky, S., Scheffer, M., Walker, B.H., Xepapadeas, T., deZeeuw, A., 2014. Does aquaculture add resilience to the global food system?Proc. Natl. Acad. Sci. 111 (37), 13257–13263. doi:http://dx.doi.org/10.1073/pnas.1404067111.

UNEP. (2011). Drought in the Horn of Africa: challenges, opportunities andresponses. African Ministerial Conference on the Environment, Bamako, 13–14September 2011.

Vetemaa, M., Eschbaum, R., Saat, T., 2006. The transition from the Soviet system to amarket economy as a cause of instability in the Estonian coastal fisheries sector.Marine Policy 30, 635–640.

Walker, B.H., Holling, C.S., Carpenter, S.R., Kinzig, A., 2004. Resilience, adaptabilityand transformability in social-ecological systems. Ecol. Soc. 9 (2), 5.

Watson, R., Pauly, D., 2001. Systematic distortions in world fisheries catch trends.Nature 414, 534–536.

32 J.A. Gephart et al. / Global Environmental Change 42 (2017) 24–32

Supplementary Information 1

Detection Sensitivity 2

The sensitivity of the shock detection method used in this study depends on both the 3

variability of a time series and the magnitude of the shock being identified. Time series were 4

simulated with varying shock magnitudes and different levels of random variation in order to test 5

the proportion of shocks detected for each combination (Supplementary Table 1). Time series 6

were modelled using an ARIMA model with an AR term of -0.1 and one degree of differencing. 7

The model order and AR term were determined by selecting the most common parameters after 8

fitting ARIMA models to each seafood production time series using the auto.arima function from 9

the forecast package and the arima function in R. Each combination of standard deviation 10

(ranging from 0 to 1) and shock magnitude (ranging 0 to 6) was simulated 1000 times. The shock 11

detection approach described in the methods was then applied to each time series to calculate the 12

proportion of times the imposed shock was detected. The detection method had a very low false 13

positive rate, with essentially no shocks being detected when no shock was imposed (magnitude 14

equal to zero). The ability to detect a shock decreases as standard deviation of the time series 15

increases and increases as the shock magnitude increases. Consequently, the ratio of the shock 16

magnitude to the standard deviation determines the ability to reliably detect a shock. The ratio of 17

shock magnitude to time series standard deviation for the shocks detected in the seafood 18

production data ranged from 0.03 to 6, with most values falling around 2. 19

Supplementary Tables 20

Supplementary Table 1: Proportion of imposed shocks that were detected for different degrees 21

of variability (standard deviation varies across columns) and shock magnitude (degree of shock 22

varies down rows). 23

Simulated Time Series Standard Deviation

Shoc

k m

agni

tude

0.0 0.1 0.2 0.4 0.5 0.6 0.7 0.9 1.0 0 NaN 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00

0.5 1.00 0.25 0.06 0.02 0.01 0.01 0.01 0.00 0.00 1 1.00 0.49 0.21 0.11 0.05 0.03 0.03 0.01 0.01

1.5 1.00 0.76 0.37 0.24 0.14 0.07 0.05 0.04 0.02 2 1.00 0.96 0.47 0.31 0.25 0.15 0.11 0.09 0.06

2.5 1.00 1.00 0.64 0.38 0.28 0.23 0.15 0.12 0.09 3 1.00 1.00 0.77 0.49 0.37 0.29 0.24 0.15 0.13

3.5 1.00 1.00 0.91 0.57 0.41 0.34 0.27 0.24 0.18 4 1.00 1.00 0.96 0.68 0.47 0.40 0.32 0.26 0.22

4.5 1.00 1.00 0.99 0.78 0.56 0.44 0.39 0.30 0.25 5 1.00 1.00 1.00 0.85 0.61 0.47 0.41 0.33 0.31

5.5 1.00 1.00 1.00 0.93 0.71 0.53 0.45 0.38 0.31 6 1.00 1.00 1.00 0.96 0.79 0.58 0.47 0.38 0.35

Supplementary Table 2: Shocks identified (country, year, magnitude, and recovery time and status) in total production time series 1with identified likely cause and cause category. Asterisks behind a cause indicate a major event that occurred that year in the country 2for which no reference directly related the event to a change in fisheries. 3

Country Year Magnitude (T)

Recovery Time (y)

Recovered Cause(s) Cause Category

Reference

Former USSR 1992 -2142821 20 N

Break up of USSR in 1991, loss of fisheries subsidies, dispersal of fishing fleet among independent states Political Milazzo 1998

Democratic People's Republic of Korea

1994 -464885.2 18 N

Combination of regional overfishing, low industry investment by the state, and oil shortages

Overfishing and policy change

Kim et al 2007; FAO 1996

Republic of Korea 1991 -315902.4 18 Y Combination of overfishing and

new regulations

Overfishing and policy change

Shon et al. 2014; Kim et al 2007

Democratic People's Republic of Korea

1991 -214383.4 21 N

Combination of regional overfishing, low industry investment by the state, and oil shortages

Overfishing and policy change

Kim et al 2007; FAO 1996

India 1999 -172412.6 1 Y Overfishing and series of changes in export standards

Overfishing and policy change

Devaraj and Vivekanandan 1999

Argentina 1998 -66867.6 8 Y Beginning of Argentinian great depression* Political n/a

Morocco 2011 -63091.4 1 N

Regional overfishing and expiration of fisheries partnership agreement with the EU (renewed in 2013)

Overfishing and policy change

Belhabib et al. 2013

Ghana 2000 -54592.8 12 N Overfishing cited, with general decline after 2000 Overfishing

Atta-Mills et al. 2004

Bulgaria 1990 -48310.8 22 N Ended distant water fishery and Policy Popescu 2011;

refocused on the Black Sea; Transition of aquaculture from public to private ownership, with reforms that caused a decrease in FW aquaculture production and exports

change NALO 2005

Portugal 1992 -46122.8 3 Y

Newfoundland cod moratorium in response to over-exploitation, changes in fishing regulations, and decline in fishing capacity.

Overfishing and policy change

Leitão et al. 2014

Panama 2011 -42573.8 1 N Overfishing cited Overfishing Bahri 2012 Venezuela (Bolivarian Republic of)

1994 -41017.4 18 N Venezuela banking crisis* Political n/a

Former Yugoslav States

1991 -39092 21 N Yugoslav wars* Political n/a

Angola 2006 -38694.6 1 Y

Overexploitation and change in hydrological conditions; No horse mackerel licenses issued for preservation of this fishery

Overfishing and policy change FAO 2007

Seychelles 2007 -33084.8 5 N Steep drop in tuna catches following abnormally high years Overfishing Martín 2011

Bermuda 1981 -21858 31 N Declines in marine fisheries resulted in a ban on new entrants to the commercial fishery

Overfishing and policy change

Luckhurst and Ward 1996

Côte d'Ivoire 2003 -21664 9 N Civil war began (ended by 2004)* Political n/a

Former Czechoslovakia 1991 -14490.6 21 N

Velvet revolution at end of 1989, formal breakup at beginning of 1993 Political n/a

Sri Lanka 2005 -13683.8 1 Y Tsunami caused extensive fishing gear and infrastructure damage

Natural disaster

De Silva and Yamoa 2007

Kenya 2005 -12685.2 7 N Long-term declines from Overfishing FAO 2007b;

overfishing in coastal and inland waters

McClanahan et al. 2008

Egypt 2004 -9437.2 8 N Declines in both marine and freshwater fisheries attributed to weak regulation and overfishing Overfishing

Mahmoud et al. 2015; Mohamed et al. 2010

Former Yugoslav States

1992 -7479.4 20 N Breakup of Yugoslavia* Political n/a

Mauritius 2011 -7022.2 1 N Unknown n/a Saint Pierre and Miquelon 1993 -6481.4 19 N Newfoundland cod moratorium Overfishing

Hutchings and Myers 1995

Belize 2007 -6359 5 N Hurricane Dean caused extensive fishing gear damage

Natural disaster

National Emergency Management Organization 2007

Algeria 1996 -5905 16 N Algerian Civil War* Political n/a Israel 1987 -4569.4 25 N Unknown n/a

Netherlands 1994 -4390.6 17 N Series of fishery management changes in 1993 to both capture fisheries and aquaculture

Policy change NALO 2005

Finland 1998 -4169.6 1 Y Unknown

Honduras 2007 -2432.2 2 Y Hurricane Felix; general fish declines with overfishing

Overfishing and natural disaster

Funes et al. 2015

Guinea 2004 -2400 5 Y Unknown n/a

Nicaragua 2001 -2092.2 2 Y Hurricane Mitch and White Spot Viral Syndrome outbreak in shrimp aquaculture.

Disease and natural disaster NALO 2005

Libya 1996 -1898 16 N Unknown n/a

Suriname 2008 -1653.8 2 Y Has periodically suffered problems with overfishing* Overfishing

Hornby et al. 2015

Saint Pierre and Miquelon 1988 -1490.2 1 Y France-Canada dispute and ban

due to overfishing accusations

Overfishing and policy change

McDorman 1990

Sao Tome and Principe 1986 -1322 26 N

Economic downturn and political restructuring; Fisheries agreement with EU began in 1984 (and concluded in 2007)

Political and policy change

European Commission 2015

Guyana 2005 -1223.8 1 Y

As early as July 1994 there were indications that fisheries resources were on the decline. In 2005, there was strict enforcement catch limits Overfishing FAO 2005

Mali 2011 -1184.4 1 N Northern Mali conflict in 2012* Political n/a

Belize 2006 -750.2 6 N Declining catch per unit effort in 2000s attributed to over exploitation Overfishing

Zeller et al. 2011

Suriname 2009 -744.6 1 Y Declining regional stocks and inadequate management* Overfishing FAO 2008

Libya 1983 -690.4 11 Y Unknown Switzerland 1988 -676 24 N Unknown Bahamas 2006 -422.6 1 Y Unknown

Uganda 2003 -312.2 0 Y

Overfishing in Lake Victoria addressed through ad hoc interventions prior to the National Fisheries Policy in 2004

Overfishing and policy change

Kjӕr et al. 2012

Albania 1991 -194.6 8 Y Unknown Bahrain 2000 -83.2 2 Y Unknown Mongolia 2005 -53 7 N Unknown Cayman Islands 1988 -49 1 Y Unknown 4

Supplementary Figures 1

2

Supplementary Figure 1: Number of shocks identified versus threshold set for the Cook’s D 3

value. 4

5

Supplementary Figure 2: Peruvian herring, anchovy and sardine catch time series. Despite the 6

visual drop in catch during the 1982–1983 and 1997–1998 El Niño events (indicated with 7

vertical dashed lines), the Cook’s D values (0.02 and 0.19) are less than the threshold of 0.35, 8

which can be attributed to the relatively high variability in the catch. 9

10

Supplementary Figure 3: Shock magnitude, number of recovered and not recovered cases, and 11

recovery time for by region for shocks identified in the 204 capture and 169 aquaculture time 12

series. Shocks were identified in FAO FishStat total national capture and aquaculture production 13

time series using the shock detection approach described in the methods. Recovery was defined 14

as returning to within 5% of the previous 5 year average. 15

16

17

Supplementary Figure 4: Shock rate in capture fishery (black) and aquaculture (blue) 18

production time series for each country. Shocks were identified in FAO FishStat total national 19

capture and aquaculture production time series using the shock detection approach described in 20

the methods. 21

22

References 23

Bahri, T (2012) Review of the State of Fisheries in the WECAFC region. Food and Agriculture 24Organization of the United Nations, Western Central Atlantic Fishery Commission, 25Fourteenth Session, Panama City, Panama. 26

Belhabib, D, S Harper, D Zeller and D Pauly (2013) Reconstruction of marine fisheries catches 27from Morocco (north, central and south), 1950-2010. pp 23-40. In: Belhabib, D, D Zeller, 28S Harper, and D Pauly (eds.), Marine fisheries catches in West Africa, 1950-2010, part I. 29Fisheries Centre Research Reports 20 (3). Fisheries Centre, University of British 30Columbia, Canada [ISSN 1198-6727]. 31

Devaraj, M and E Vivekanandan (1999) Marine capture fisheries of India: Challenges and 32opportunities. Current Science, 76(3): 314–332. 33

European Commission. São Tomé and Principe: Fisheries partnership agreement. Updated 18 34May 2015 [Cited 2 May 2016] 35http://ec.europa.eu/fisheries/cfp/international/agreements/sao_tome/index_en.htm. 36

Food and Agriculture Organization (1996) The State of World Fisheries and Aquaculture 1996. 37Available at http://www.fao.org/docrep/003/W3265E/w3265e00.htm 38

Food and Agriculture Organization (2005) Fishery country profile: The Kingdom of Sweden. 39Available at http://www.fao.org/fi/oldsite/FCP/en/GUY/profile.htm 40

Food and Agriculture Organization (2004) Fishery country profile: The Republic of Guyana. 41Available at http://www.fao.org/fi/oldsite/FCP/en/SWe/profile.htm 42

Food and Agriculture Organization (FAO 2007a) Fishery country profile: The Republic of 43Angola. Available at ftp://ftp.fao.org/Fi/DOCUMENT/fcp/en/FI_CP_AO.pdf 44

Food and Agriculture Organization (FAO 2007b) Fishery country profile: The Republic of 45Kenya. Available at ftp://ftp.fao.org/fi/DOCUMENT/fcp/en/FI_CP_KE.pdf 46

Food and Agriculture Organization (FAO 2008) Fishery country profile: The Republic of 47Suriname. Available at ftp://ftp.fao.org/FI/DOCUMENT/fcp/en/FI_CP_SR.pdf 48

Funes, M, K Zylich, E Divovich, D Zeller, A Lindop, D Pauly and S Box (2015) Honduras, a 49fish exporting country: Preliminary reconstructed marine catches in the Caribbean Sea 50and the Gulf of Fonseca, 1950 – 2010. Working Paper #2015 – 90, Fisheries Centre, The 51University of British Columbia. 52

Hornby, C, S Harper, J MacDonald and D Zeller (2015) Reconstruction of Suriname’s marine 53fisheries catches from 1950-2010. Working Paper #2015 – 49, Fisheries Centre, The 54University of British Columbia. 55

Kim, S, CI Zhang, JY Kim, JH Oh, S Kang and JB Lee (2007) Climate variability and its effects 56on major fisheries in Korea. Ocean Science Journal, 42(3): 179–192. 57

Kjӕr, AM, F Muhumuza, T Mwebaze and M Katusiimeh (2012) The political economy of the 58fisheries sector in Uganda: ruling elites, implementation costs and industry interests. DIIS 59Working Paper 2012:04, Danish Institute for International Studies, Copenhagen, 60Denmark. 61

Leitão, F, V Baptista, K Erzini, D Iritani and D Zeller (2014) Reconstruction of mainland 62Portugal fisheries catches 1950-2010. Working Paper #2014 – 08, Fisheries Centre, The 63University of British Columbia. 64

Luckhurst BE and JAD Ward (1996) Analysis of trends in Bermuda’s fishery statistical database 65from 1975 to 1990 with reference to fishery management measures implemented during 66this period. Proceedings of the Gulf and Caribbean Fisheries Institute 44:306–324. 67

Macikenaite, V (2014) The Implications of China’s Fisheries Industry Regulation and 68Development for the South China Sea Dispute. In Rozdział z: The Quandaries and 69Foreign Development, ed. D Mierzejewski, “Contemporary Asian Studies Series.” 70

Mahmoud, HH, L Teh, M Khalfallah and D Pauly (2015) Reconstruction of marine fisheries 71statistics in the Egyptian Mediterranean Sea, 1950-2010. Working Paper #2015 – 85, 72Fisheries Centre, The University of British Columbia. 73

Marín, A, S Gelcich, G Araya, G Olea, M Espíndola and JC Castilla (2010) The 2010 tsunami in 74Chile: Devastation and survival of coastal small-scale fishing communities. Marine 75Policy, 34: 1381–1384. 76

Martín, JI (2011) Fisheries in the Seychelles and Fisheries Agreements with the EU. Directorate-77General for Internal Policies of the Union, Policy Department B: Structural and Cohesion 78Policies, European Parliament, Brussels, Belgium. 79

McClanahan, CC Hicks and ES Darling (2008) Malthusian overfishing and efforts to overcome it 80on Kenyan coral reefs. Ecological Applications, 18(6): 1516–1529. 81

Milazzo, M (1998). Subsidies in World Fisheries: A Reexamination. World Bank technical 82paper; no. WTP 406. Fisheries series. Washington, D.C.: The World Bank. 83

Mohamed, SM, MK Ahmed and KIA Aly (2010) An analytical economic study of Egyptian 84fisheries. Journal of American Science, 6(9): 768–772. 85

National Aquaculture Sector Overview. Bulgaria. National Aquaculture Sector Overview Fact 86Sheets. Text by Hubenova, T. In: FAO Fisheries and Aquaculture Department [online]. 87Rome. Updated 3 September 2007. [Cited 3 May 2016]. 88http://www.fao.org/fishery/countrysector/naso_bulgaria/en 89

National Aquaculture Legislation Overview. Netherlands. National Aquaculture Legislation 90Overview (NALO) Fact Sheets. Text by Spreij, M. In: FAO Fisheries and Aquaculture 91Department [online]. Rome. Updated 27 January 2005. [Cited 3 May 2016]. 92

National Aquaculture Sector Overview. Nicaragua. National Aquaculture Sector Overview Fact 93Sheets. Text by Saborio Coze, A. In: FAO Fisheries and Aquaculture Department 94[online]. Rome. Updated 1 February 2005. [Cited 3 May 2016]. 95http://www.fao.org/fishery/countrysector/naso_nicaragua/en 96

National Emergency Management Organization (2007) Damage assessment and needs analysis: 97Hurricane Dean, Preliminary Report, Version 3, 3 September, 2007. 98

Popescu I (2011) Fisheries in Bulgaria. European Parliament, Directorate-General for 99International Policies, Policy Department B: Structural and Cohesion Policies. Fisheries, 100Brussels. 42 p. 101

Rosenblum, M and M Cabra (2012) In mackerel's plunder, hints of epic fish collapse. The New 102York Times, January 25, 2012. 103

Shon, S, S Harper and D Zeller (2014) Reconstruction of marine fisheries catches for the 104Republic of Korea (South Korea) from 1950–2010. Working Paper #2014 – 19, Fisheries 105Centre, The University of British Columbia. 106

Zeller, D, R Graham and S Harper (2011) Reconstruction of total marine fisheries catches for 107Belize, 1950-2008. In: Palomares, MLD, D Pauly (eds.), Too Precious to Drill: the 108Marine Biodiversity of Belize, pp. 142-151. Fisheries Centre Research Reports 19(6). 109Fisheries Centre, University of British Columbia [ISSN 1198-6727] 110

111