FISHING for PROTEINS How Marine Fisheries Impact on Global Food Security up to 2050

DAS PROJEKT WIRD THIS PROJECT IS PARTKOFINANZIERT-FINANCED DURCHBY THE DIE EUROPEAN UNION EUROPÄISCHE UNION REPORT

22001616

FISHING FOR PROTEINS How marine fisheries impact on global food security up to 2050. A global prognosis This publication has been produced with the financial contribution of the European Union. The contents of this publication are the sole responsibility of WWF and can in no way be taken to represent the views of the European Union.

Publisher: WWF Germany; International WWF Centre for Marine Conservation, Hamburg Date: October 2016 Authors: Prof. Dr. Martin Quaas, Dr. Julia Hoffmann, Katrin Kamin (all from Kiel University, Resource Economics Working Group), Dr. Linda Kleemann (Kiel Institute for the World Economy, GFA Consulting Group Hamburg); Karoline Schacht (WWF Germany) Translation: Katrin Kamin, Ann Marie Bohan Editing: Karoline Schacht (WWF) Coordination: Karoline Schacht (WWF), Thomas Köberich (WWF) Contact: [email protected] Design: Wolfram Egert/Atelier für Graphic Design Production: Maro Ballach (WWF) Print: Paper: Credits: F. Larrey/WWF Content

Summary of the Report 4

Background 4 Results 10 Key Findings of the study and WWF Comments 12 WWF Conclusion 14

Fishing for proteins. How marine fisheries impact on global food security up to 2050. 15

1 Structure of the Report 15

Climate Change 15 2 Fish Consumption 16

2.1 Fish Consumption in Selected Case Study Countries 18 2.2 Global Importance of Fish as a Protein Source 22 2.3 Food Security and Fish 25 2.4 Dependence on Fish 27 2.5 Fish Dependence Index 28 2.6 Current Fish Consumption and Fish Supply 32 3 The Bio-Economic Model 34

3.1 The Model Approach 34 3.2 Data and Estimation of Model Parameters 35 3.3 Global and Regional Demand Systems 37 3.4 Scenarios: Socio-Economic Pathways and Fishery Management 40 4 Results and Discussion 43

Appendix 50

National Recommended Intakes for Fish 50 Fish Supply Model 51 List of Large Marine Ecosystems 54 List of Protein-Rich Non-Fish Food Substitute Goods 54 List of Figures and Tables 55

Footnotes 56

References 57

Acknowledgement 59

Fishing for Proteins | 3 Summary of the Study Background The world’s population is growing and caring for it is now placing the earth’s natural resources under severe pressure. One of the most pressing questions concerning the future centres on the food security of what will soon be nine billion people: how will we all have enough to eat? Can we change fishing and agriculture in such a way that they will feed us but their negative effects on the environment will remain limited to an absolute minimum? Will we be in a position to resolve distribution issues fairly and peacefully?

According to estimates, global food requirements are set to double in the next 35 years. From a technological perspective, it seems possible that enough food can be produced for up to 10 billion people (Evans 1998). In terms of calories, farmers around the world harvest around one-third more food than is needed to feed the world’s population (BMEL 2015). Nevertheless, around a billion people go hungry every day. Their hunger is the result of a distribution problem and is a consequence of poverty and not of the lack of food availability.

Something that is lacking in some regions is needlessly wasted in others: globally, around 30 to 40% of all food along the production and supply chain ends up in the bin (WWF 2015). The possibility of expanding cultivated land for the agricultural production of staple foods seems very unlikely; on the contrary, this option has reached its limits, or has already exceeded them in many areas. Many farming systems generate huge harvests of products like corn, rice, cereals and meat while simultaneously degrading resources such as soil and water.

And what about fish? Fish plays a hugely important role in global food security. It provides more than 3.1 billion people with at least 20% of their animal protein but above all it is an important source of fatty acids and micronutrients (Thilstedt et al. 2016; FAO 2016; Béné et al. 2015). Fish currently supplies 17% of all the protein consumed in the world. This share will continue to grow because the rising income of consumers is accompanied by an increase in demand for high-quality fish (World Bank 2013). In addition to its importance as a source of food, fish is also of great socioeconomic importance: approximately 500 million individuals throughout the world make their living in some shape or form in the fishing industry (FAO 2014).

Yet the state of global fish stocks is cause for concern. Among the scientifically assessed fish stocks, 31% are considered to be overfished and another 58% to be yet fully fished (FAO 2016; Costello et al. 2016). A further increase in fishing pressure could gravely jeopardise the health of the fully fished stocks (FAO 2016).

In WWF’s view, the discussion about supplying the world’s population with high-quality protein neglects the fact that both food productions systems – the sea and the land – are closely interconnected and in terms of their capacity and natural limits must be viewed as one. Protein-rich soya is used in fish food whereas fish meal and fish oil are in turn part of the animal food of pigs and poultry. Marine catch rates can obviously not be increased, they have in fact been stagnating for almost 30 years. The demand for fish is currently much greater than can be covered by marine fish alone and already today half of all fish in the world is farmed or comes from aquaculture. This branch of the food industry, which has grown hugely over the last 40 years, requires both sea and land (see the box: Aquaculture).

4 The purpose of fishery management is to safeguard fish resources and ensure their sustainable and environmentally friendly use in the long term. It is the responsibility of policy makers to ensure that this happens. A number of researchers are convinced that this management must be improved significantly to strengthen global food security and prevent the imminent collapse of fish stocks (Pauly et al. 2005; Worm et al. 2006, 2009; Branch 2008; Branch et al. 2010; Allison et al. 2012; Quaas et al. 2016). Such reforms in management could prove very costly in the short term. However, the measures would be ultimately worthwhile if stocks were to reach a healthy size again (Quaas et al. 2012; Sumaila et al. 2012). Con- sistent, effective fishery management that pursues an ecosystem-based approach, ensures enforcement of the rules, severely restricts illegal fishing and embeds the concept of sustainable management in all fisheries will improve the global fish supply. This is vital in meeting the continuing growth in demand for fish and maintaining marine biodiversity and ecosystem functions (Worm et al. 2009; Fro- ese and Proelss 2010). After all, healthy fish stocks can only live in healthy seas.

In this study, WWF is seeking to find and consolidate answers to three questions: »»What is the maximum quantity of fish that can be obtained from the seas in 2050 under sustainable conditions? »»How will fish demand develop gloablly and regionally up to 2050? »»How will these projections affect the consumption of fish? For example, do we face the threat of a fish protein gap?

Aquaculture Growing numbers of people are eating increasing volumes of fish. In order to meet the growing worldwide demand, fish is also farmed. In fact, were it not for the strong expansion in aquaculture seen in recent decades, the demand for fish could not have been met as the yields from global marine fishery have been stagnating for around 30 years. With an average annual growth of 9% since 1970, aquaculture is the fastest growing branch of the global food industry. The Food and Agricultural Organization of the UN (FAO) calculated total aquaculture production of over 90 million tons in 2014. Today, more than half the edible fish consumed in the world is farmed. However, the enormous growth in the farmed sector is problematic for several reasons. For one thing, aquaculture is overwhelmingly practised in countries that have little or no statutory frameworks for regulating aquaculture or protecting the environment. For another, it causes major marine pollution, if chemicals, food remains, faeces and medications from the open cages reach the rivers and seas. Feeding predatory fish in breeding facilities requires primarily wild fish; herbivorous fish rely more on agricultural protein. In the past, as a result of the construction of facilities for shrimp farming in the coastal regions of tropical and subtropical countries, valuable habitats like mangrove forests were lost. Their destruction had huge consequences for the operation of coastal ecosystems, coastal protection and fishing. In this study, we are focusing on the future of fish from the sea. The future of aquaculture is dealt with in a separate report.

Fish in the Diet

The unique combination of high-quality protein and important nutrients makes fish an exceptionally valuable food. For one thing, it is a good source of animal protein – 150 g of fish provides approximately 50 to 60% of an adult’s daily

Summary | 5 requirements. It also provides fatty acids, vitamins and other vital nutrients like iodine and selenium, which do not exist in this quantity or variety in any other cereal or meat (Beveridge et al. 2013; Kawarazuka and Béné 2011; WOR2 2013). Food diversity and quality are important elements in the fight against hunger and malnutrition. Poverty is correlated with an excessive intake of staple foods like rice, corn and cereals and an insufficient share of proteins, fats and nutrients.

Fish is frequently the only available and affordable source of animal protein in the coastal regions of developing countries. In a worldwide comparison, rather less fish is consumed in poorer countries (approximately 10 kg per capita per year), whereas the per capita consumption of around 22 kg per year in Asia, North America and Europe is higher than the global average of 20 kg. This reflects the various factors that affect fish consumption: how available fish is, how expensive it is, whether there are dietary traditions in relation to fish and how developed the country is. Generally, the lower the income, the lower the consumption of fish.

The World Health Organization (WHO) recommends the regular consumption of fish – one to two portions a week (WHO 2002).1 With an average portion size of 150 g, this results in a worldwide recommended annual consumption of 11.7 kg fish per capita. Several national nutrition guidelines were also analysed for this study. They operate within a similar range, averaging 10.6 kg of fish per capita per year (see Table 6 in the appendix).

However, this rough guideline only applies in Africa and Latin America; all other regions in the world consume significantly more fish (Figure Z1). Globally, an average of more than 20 kg of fish is currently consumed per capita per year (FAO 2016). The average German also consumes roughly 14 kg of fish per year, more than the recommended intake.2 In general, Germans eat too much protein. Depending on the individual age group, they consume between 130 and 160% of the recommended amount (MRI 2008). We are thus eating more protein and more fish than we really need. As the world’s population grows and the population density constantly increases in coastal areas, the question arises of whether we are satisfying our need for fish at the expense of those who actually need it. Viewed at a global level, fish is already distributed unequally and too much fish per capita is eaten in the Northern hemisphere.

Staple foods like corn, rice and other cereals account for a large share of the dietary pattern of poor people. The consumption of fish is important in correcting the imbalance between calories and protein. Fish is generally not only cheaper than other animal protein but is also often a basis of local and/or traditional recipes. In countries like Senegal or Indonesia, fish accounts for up to 40% of the total intake of animal protein.

In absolute figures, the consumption of animal protein in developing countries is lower than in developed countries. However, the share of animal protein in total protein is growing very rapidly. This is due primarily to economic development and the way in which developing countries in Africa and Asia are ‘catching up’. If we make a distinction between fish and meat in the consumption of animal protein, it becomes clear that the contribution made by fish to the supply of animal protein has fallen slightly since 1990 – primarily in favour of meat.

In poor countries, where fish is traditionally eaten, rising income leads to an increase in the consumption of meat and higher-quality fish species. Conse-

6 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry BerryDark BerryMedium BerryLight

11. Grey GreyDark GreyMedium GreyLight

Base Colours BaseColoursBackground BaseColoursTintedBox Black

Fig. Z1 120 Total protein intake in the eight case study countries, subdivided into total protein 100 (excluding fish, light blue) and fish protein (dark blue). Source: FAOSTAT 80 Fish protein 2009-11 in gr/caput/day Fish protein (2009-11). 60 Non-Fish protein 2011 (in g/caput/day) in gr/caput/day Non-Fish protein 40 World average (2011) protein supply (in g/caput/day) .... Nutrition 20 Nutrition recommendation recommendation for total protein intake for total protein intake (in g/caput/day) 0 South Senegal Peru USA Indo- Germany France China World average Africa nesia protein supply (g/caput/day) quently, small pelagic fish (those that live in the open sea between the surface of the water and the bottom of the sea) are replaced by larger demersal species.

Between 1990 and 2012, the consumption of wild-caught fish remained almost constant while the consumption of fish cultivated through aquaculture increased five-fold. In 2015, half the volume of fish produced for human consumption came from aquaculture, compared with just 5% in 1962 and 37% in 2002 (FAO 2015).

From a global perspective, there is enough food to feed everyone in the world. If we also take the current protein supply as a basis, there is no protein gap. Food distribution problems are actually at the heart of hunger issues.

The global average supply of protein was 79 g per capita per day in 2011, while the average protein requirement was 49.6 g per capita per day. The latter figure was calculated on the basis of the recommended 0.8 g per kilogram of body weight and the average weight of a person in 2011 (62 kg). Measured against the WHO’s recommended intake, the 79 g corresponds to an oversupply of protein of around 30%.

Figure Z1 shows the protein supply in the countries that were selected as examples for this study: South Africa and Senegal, Peru and the USA, China and Indonesia, Germany and France. The height of each bar represents the total supply of protein, subdivided into dark blue areas for fish and light blue areas for other proteins.

The New Fish Dependence Index

Our fish dependence index measures the level of dependence on fish as a source of income and nutrition (especially protein). It is based on the composition of a number of factors: a) food security (incidence of malnutrition in % of the population); b) fish consumption (share of fish in the total consumption of animal protein in %); c) national catch quantity per capita; and d) gross domestic product (GDP) (in USD; capacity to replace fish by other protein-rich food). See section 2.5 for further details on the index.

Summary | 7 Fig. Z2 Overview of global fish dependence.

high medium high medium low no data

In Figure Z2, we link the country-specific situation of food (in)security and the general situation regarding health and hunger with the value of fish and fisheries to the country’s socioeconomic status and the livelihoods of its citizens in order to describe the fish dependence of individual countries. The index shows that countries with a high share of fish in their diet are particularly dependent on fish. More importantly, however, these countries (in dark blue) are precisely the countries that tend to have a large fisheries sector and are neither wealthy nor particularly food secure.

According to this index, Senegal, for example, appears to be particularly dependent on fish. At the same time, Senegal is also an example of the complexity reflected in this statement. According to estimates based on FAO figures, approximately one million people are directly or indirectly dependent on fishing in the country. Fish accounts for 44% of animal protein intake but just 12% of total protein. If the global recommendation of 11.7 kg of fish per person per year is taken as a reference, the annual average per capita consumption of 24 kg of fish in Senegal is ‘too much’. At 60 g per capita per day, protein supply is also above the required value of 49 g. Thus, on the one hand, a moderate decline in fish intake would not lead to a protein gap in Senegal. Nevertheless, 10% of the population is undernourished and fishing is the main source of income in rural coastal regions (Thiao et al. 2012). So even though the protein supply would be sufficient, a shrinking fisheries sector would probably see an increase in poverty and hunger in the coastal regions (Lam et al. 2012) with the potential consequence of political instability.

Fish Demand and Fish Supply

We wanted to know which regions in the world can meet their requirements through their own production now and in the future and where there is a growing dependency on imports to meet demand. To do this, we subdivided the world’s seas into 64 large marine ecosystems (LMEs). These 64 ecosystems supply up to 95% of the annual global fish catch (Sherman et al. 2009) and present quite specific challenges for regional, and in some cases, multinational management. Then we calculated whether the fish catches in these regions in 2010 were able to meet the local demands of people in the neighbouring countries for fish. To this end, we drew on the data from the Sea Around Us project conducted by the University of Vancouver (Sea Around Us database).

8 Fig. Z3 Amount of per capita fish consumption, fish catches and population size on an LME basis in 2010. Data: Sea Around Us database/own maps

Pop. 2010 Scenario (mn) > 50 Figure Z3 shows the LMEs. The productivity of the regions differs significantly: 50–150 red or yellow means ‘does not supply enough fish to meet local demand’; light 150–500 green and green mean ‘supplies enough/more than enough fish to meet local 500–1,000 demand’.

Fraction LMEs with several neighbouring countries (such as the Mediterranean, Carib- 0–80% bean Sea and Baltic Sea) appear to be less able to cover local demand, whereas 80–100% LMEs with only one or a few neighbouring states perform better. Moreover, 100–500% > 500% the highly productive LMEs in the North Atlantic and East Pacific are generally better able to meet local demand. This also applies to Europe, the East and Catches (mn tons) West Coast of the USA and the western coast of Latin America. By contrast, fish no data production in the LMEs around Africa (with the exception of Northwest Africa) 0.01 – 0.60 and along the Asiatic and Australian coasts is inadequate when compared to 0.61 – 1.50 current demand. 1.51 – 4.00 4.01 – 8.00 8.01 –12.86 The Bio-economic Model

Looking ahead to 2050, we are projecting future global fish catches and possible effects on fish consumption. As fish catches are generally affected by fishery activity and the productivity of stocks, we need to apply a bio-economic model to determine future catches. This model combines an ecological aspect, which describes the productivity of fish stocks, and an economic aspect, which describes the economic incentives for carrying out fishery activity and the distribution of fish catches across the markets.

The model is designed to explain how the total volume of fish catches changes in different economic and fishery management scenarios and how the total global catch is distributed in terms of regional catches and regional consumption quantities.

We based the modelling framework on various current fishery management systems. A new element of this approach is that we include interactions in the sea. The fish include predator and prey species and both are caught. Previous studies with comparable global research approaches ignored the biological interactions and either included all fish species in one model (World Bank 2009) or considered stocks that are biologically independent of one another (Quaas et al. 2016; Costello et al. 2016).

Summary | 9 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry BerryDark BerryMedium BerryLight

11. Grey GreyDark GreyMedium GreyLight

Results Base Colours BaseColoursBackground BaseColoursTintedBox Black For our projection we assume a Maximum Sustainable Yield (MSY) management scenario for all fisheries. The projection calculates the MSY, which provides an indication of the maximum contribution that global fish stocks could theoretically make in supplying the world’s population with protein in 2050. The estimated MSYs for the global fish stocks using three different model approaches are presented below.

Fig. Z4 250 Estimated MSY that the global fish stocks could supply using three different 200 model approaches. (in million tons) 150 Global catch quantities

100

0 Yield-oriented Global surplus Total regional predator-prey model model surplus models

The first bar shows the global catch quantity for a purely yield-oriented predator-prey model. The first model determines the productivity of global fish stocks based on the interactions between the predatory and prey fish: only when the stocks of large predatory fish are depleted can the catch of their prey fish be increased significantly, thus increasing the total volume of the overall catch. Con- sequently, the objective of fishery management in this model is to maximise catch quantities. However, an MSY of 160 million tons in 2050 can in principle only be achieved at the expense of marine biodiversity. The increased catch quantity is accompanied by a high level of uncertainty (+/- 90 million tons). This is a typical effect following the destabilisation of the predator-prey balance. If all other target values for healthy seas as a prerequisite for healthy fish stocks are disregarded – for example intact habitats or the minimisation of unwanted bycatch – a higher catch quantity would be possible but would be neither desirable nor sustainable from an ecological perspective.

The second and third bars, on the other hand, show a stable maximum catch for Schaefer surplus models. Such a model specifies that the utilisation rate may not be higher than the natural growth rate of renewable resources. We first calculated the surplus model for the entire sea and assumed one global stock (second bar); we then calculated it for the 64 individual LMEs (third bar) and assumed one stock for each of them. When added up, the result of the third model agrees with the result of the second model: both project around 112 million tons of fish for 2050. We use the surplus model to analyse the potential contribution that the LMEs could make to meeting the global and regional demand for fish protein.

We retrospectively calculated a total global catch of 101 million tons of fish for 2010. This means that current catch quantities cannot be increased by more than 10% in the future. Accordingly, marine resources already seem to be almost fully exploited.

10 1. Red RedDark RedMedium RedLight

BlueDark BlueMedium BlueLight 2. Orange OrangeDark OrangeMedium OrangeLight Senegal 8. Aqua South Africa AquaDark AquaMedium AquaLight 3. Yellow YellowDark YellowMedium YellowLight Peru 9. Pink USA PinkDark PinkMedium PinkLight 4. Green GreenDark GreenMedium GreenLight 10. Berry Indonesia BerryDark BerryMedium BerryLight 5. Earth China EarthDark EarthMedium EarthLight 11. Grey GreyDark GreyMedium GreyLight 6. Brown France BrownDark BrownMedium BrownLight Germany Base Colours BaseColoursBackground BaseColoursTintedBox Black 7. Blue

0,5 8

0,4 6 0,3 4 0,2 2 0,1 0 0 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

50 3 40 2 30 20 1 10

0 0 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

30 140

25 120 100 20 80 15 60 10 40

5 20

0 0 20 30 40 50 60 70 80 90 100 20 30 40 50 60 70 80 90 100 Management effectiveness (%) Management effectiveness (%)

Fig. Z5 We also studied how various levels of fishery management effectiveness affect Global fish catches in 2050 catch quantities. Our analysis concluded that if fishery management effectiveness according to global bio-eco- reached 100%, marine biodiversity would be secured, global catches of predatory nomic predator-prey model. The model shows varying and prey fish would reach levels of 21 and 116 million tons, respectively, and a degrees of management total of 137 million tons of fish would be caught sustainably (Figure 23). effectiveness and assumes GDP growth according to If management took into consideration all potential effects of fishery activities baseline scenario SSP1 on future fishing opportunities, 100% effectiveness would be achieved. Optimum (see section 4 for details on the model). management from an economic perspective would also stipulate the total allowa- (in million tons) ble catch for individual stocks in such a way that it actually regulates and restricts fishery activities. Global catches of predatory fish We come to the conclusion that only a management system that focuses on Global catches relationships in the ecosystem can meet the various needs of a sustainable fishing of prey fish industry: to achieve high catch volumes while increasing ecosystem resilience by protecting marine biodiversity and habitats.

The effectiveness of fishery management is currently estimated at an average of between 50 and 60% (Mora et al. 2009; Watson et al. 2009; Quaas et al. 2016). There is thus considerable scope for development here. Current catch quantities could be maintained at this level of effectiveness. However, the large predatory fish would have to be heavily fished in order to reduce the pressure on smaller prey fish, thus allowing for slightly larger catches. In fact, that is currently common practice. Compared to the model of best-possible management, this would cause a loss of stability in the balance of the ecosystem and shift future fish consumption in the direction of prey fish.

If management effectiveness slipped below the current level, this would be expressed in a sharp reduction in the catches of both predatory and prey fish. This means that ensuring the maximum possible effectiveness of fishery management is critical for maintaining catch yields in the face of simultaneous increases in the global demand for fish.

In the last step, we analyse how the LMEs can help to cover the global requirement for protein. To do this, we use the estimates from the third model (surplus model for 64 LMEs) and compare them with projected regional fish consumption. We use international estimates of future socioeconomic development for the projections, e.g. population trend and economic growth (Shared Socioeconomic Pathways, SSP – see footnote 3).

Summary | 11 In the SSP1 scenario with the lowest assumed population growth, the global fish supply in 2050 will be able to meet approximately 81% of the global requirements of what will then be almost 8.5 billion people. In the SSP3 scenario, where population growth is strongest, only 75% of fish requirements will be met by wild- caught fish by the same date.

It is generally assumed that the huge growth rates in aquaculture seen over the last 30 years were needed to meet the world’s growing appetite for fish. In terms of numbers, half of the world’s fish currently comes from aquaculture production. If the results of our projections are accurate and the volume of fish catches in 2050 could meet around 80% of the world’s requirements, the need for further growth in aquaculture production would abate if the fish was distributed in a more equitable manner.

And the distribution problems are continuing to grow: fish consumption in the regions along the East Asian coast could decline significantly by 2050. Fish is traded globally and prices depend on global demand. If fish prices increase accordingly based on this demand, fish will become unaffordable for a large swathe of the population of LMEs along the East Asian coast. These people would have to switch to affordable alternative sources of protein and fish would be exported at a higher export price.

Key Findings of the Study and WWF Comments

1 According to the projections made by this study, it will on the one hand be possible to fish approximately 112 million tons of fish around the world in 2050 if the current moderate level of fishery management effectiveness remains the same. On the other hand this would risk the health of predatory fish stocks which will be fished too hard. This may cause a perilous destabilisation of the ecosystem. It would appear that marine resources are already close to fully exploited (2010: total catch of 101 million tons), leaving little room to increase catch volumes in the future.

There is only one way to increase global catch quantities that is both relevant and sustainable and thus meets the growing demand: fishery management must be improved significantly worldwide and any decisions made must place far greater emphasis on ecological interactions than has been the case to date. The interactions between predatory and prey fish is one such example. This type of differentiated, economically optimised and fully enforced management system could enable sustainable catches of approximately 137 million tons worldwide in 2050.

12 WWF Comment Improve Management Fishing is exerting considerable pressure on fish stocks and their habitats in all areas of the world’s oceans. WWF is committed to an ecosystem-based fishery management that safeguards the future of marine ecology and the human population. Part of this strategy is not only to conserve vital stocks of large predatory fish but also to protect habitats and endangered species. Total allowable catch limits are set to actually regulate the fishing industry. From the current perspective, this management model would constitute a major improvement in quality and one that is urgently required to make fishing sustainable. Ultimately, it would lead to more fish, which could then be distributed more equitably. Illegal fishing, which accounts for an estimated 30% of the global catch, is evidence of a particularly damaging consequence of poor management. It reflects increased compe- tition and higher demand accompanied by weak controls. The European Union has a particular responsibility to solve this problem. Firstly, EU Member States must be more consistent in implementing the existing regulation against illegal fish imports. Secondly, they must ensure that any fishing activity they conduct in waters outside the EU is both fair and sustainable. Furthermore, EU agreements with third countries must focus on prioritising regional fishing and first and foremost guarantee that local populations are provided with local fish.

2 If the quality of fishery management, at a minimum, stays at its current moderate level, there would be enough wild-caught fish available in 2050 (112 million tons) to theoretically supply each world citizen with 12 kg (per person per year). This roughly matches the average quantity currently recommended by the WHO and a large number of countries.

WWF Comment More ‘Fish Fairness’ WWF considers that the belief that there is enough fish for everyone requires closer scrutiny. Firstly, maintaining the status quo is simply not an option for global fisheries as tolerable limits have already been reached for 58% of stocks and exceeded for 31%, the latter being classified as overfished. In addition, there is currently no fair distribution mechanism for fish that is geared towards real needs. Secondly, the WHO’s recommended intake of fish primarily focuses on the valuable micronutrients rather than on the protein. In many countries, the current demand for fish is well above the average WHO recommendation because the affected areas actually rely on fish for their basic protein supply and very few alternatives are available. In Senegal, 24 kg of fish are consumed per capita per year and fish provides almost half of the animal protein consumed. In Germany and France, per capita consumption of 14 and 32 kg respectively also exceeds the WHO’s recommended 11.7 kg. However, in these countries, fish provides just 7% of the animal protein consumed. Even if we were to abstain completely from fish in northern Europe, we would not suffer from protein deficiency. The situation is very different in poorer regions with high levels of fish consumption.

3 We can assume that developed countries will use the option of importing fish at higher prices when they are confronted with a shortage in fish supply in 2050. Developing countries with abundant fish stocks will then export their fish rather than eat it themselves. Rich countries would thus still be able to afford ‘their’ fish in the future while poorer nations would not. For poor coastal countries, the probability that poverty and hunger will become more widespread within their borders increases.

Summary | 13 In 2050, LMEs in Africa and in Latin America (with the exception of northwest Africa and Peru) and those along the Asian coast will not be able to meet the local demand for wild-caught fish. Neighbouring countries of LMEs in East Asia, West Africa and in western South America could export their fish due to high fish prices and low prices for substitute goods. On the other hand, developed countries with high purchasing power such as Australia and the USA would probably increase their fish imports. Germany, France or South Africa could import fish from other marine regions to offset the major shortfalls that will sometimes occur in their own supply.

WWF Comment I Can Have Your Fish and Eat it Today, Europe imports about a quarter of the world’s total fish catch and represents the largest market for fish and fish products globally. More than half of the fish imported into the EU originates in developing countries. In statistical terms, we in Europe have already eaten all of the fish from our own waters by the middle of any given year. For the remainder of the time, we eat imported fish which is then missing elsewhere as a source of nutrition and/or as the cornerstone of local economic structures. The high demand for imported fish would almost certainly decrease if fish stocks in the European Union’s own waters were once again at healthy levels. We must assume that fish consumption in the Northern hemisphere will have an even more serious impact in the future on the living conditions of those who depend on fish in various ways. Moreover, our analysis of distribution flows clearly suggests that any additional catches will not be used to meet the growing demand in fish-dependent countries. However, the increasing scarcity of resources and unequal distribution of marine fish must not be borne by the poorest countries. This would fuel conflicts and exacerbate instability particularly if the fishing sector is not better regulated.

WWF Conclusion

Our report ‘Fishing for Proteins – Impacts of marine fisheries on global food security to 2050’ identifies the key factors driving a sustainable future fish supply. It also highlights that consistent changes are required in the fishing industry and in its administration to ensure that the worldwide problems of hunger and poverty do not continue well into the future. That would be contrary to the commitments set out in the United Nations plan of action for the future: ending hunger and poverty by 2030 are two of the 17 Sustainable Development Goals (SDGs). To achieve these goals, fishery management, amongst other things, must be improved significantly everywhere. Apart from bad management, fish stocks also suffer from the effects of climate change as well as the pollution and destruction of their habitats. Investment in improved fishery management, in sustainable aquaculture, in the protection of vital marine habitats and in fair trade policies would restore the productivity of our seas and pay off for billions of people in developing countries. Our results clearly show that the world’s growing population must not serve as an excuse for even more reckless exploitation of our seas. In fact, the solution to these problems can be achieved by implementing and enforcing ecosystem-based and sustainable fishery management. In addition, fair access rights and prices must be guaranteed. An increasing supply of sustainably produced, fair trade fish is not merely intended to ease the conscience of Euro- pean consumers; it must also benefit fishermen and fish farmers in developing countries with measurable effects.

The responsibility for this rests with us – not only politically but also as consumers.

14 FISHING FOR PROTEINS Impacts of Marine Fisheries on Global Food Security to 2050. A Global Prognosis

1. Structure of the Report We begin by describing developments in recent years, current fish consumption, and explain facts about fish, food security and the fish supply using a newly developed index of fish dependency. Alongside a global perspective, we also consider individual regions and selected representative countries.

Next, we calculate the quantity of wild-caught fish on a global as well as selected regional basis that will be available in 2050. For modelling, we consider various economic scenarios and levels of management quality, and compare the results with the future demand for fish. In this comparison as well, we consider both the global and the regional persepctive at the level of large marine ecosystems.

We calculate future demand based on shared socioeconomic pathways, (SSPs3) and the regional supply of alternative protein sources – and thereby generate an entirely novel approach for forecasting demand. For calculating potential fish production, we adopt a global predator-prey model. This means that we additionally take into consideration the biological interactions and amplify the current model with an element of ‘ecological realism’.

Finally, we combine regional fish production in the large marine ecosystems (LMEs) with the regional and global demand for fish. In the last section we present the findings from our model on the future of the ocean fishery and its effects on fish consumption, and discuss issues related to the distribution of resources and the challenges for trade.

All calculations and models are based on data from »»Sea Around Us (http://www.seaaroundus.org/) for global fish landings and prices in the large marine ecosystems (LMEs); »»FAO (http://faostat3.fao.org/home/E) for consumption levels and import-and-export prices for protein-rich foods: »»scientific literature on estimates for preference parameters; »»Shared Socioeconomic Pathways’ (SSPs) for income and population scenarios.

Climate Change

This study focuses on the biological and economic effects and on the effects of fishery management quality on future fish catches and consumption levels. However, climate change is also likely to play an important role in the future of fishing (Cheung et al. 2010; Lam et al. 2012; Merino et al. 2012). While ocean warming may increase productivity for some stocks (Kjesbu et al. 2014; Voss et al. 2011), ocean acidification and warming (Voss et al. 2015; Blanchard et al. 2012) generally decrease the productivity of stocks. Against this background of mostly adverse climate change effects on fisheries, estimates for future fish catches may be regarded as somewhat optimistic.

Fishing for Proteins | 15 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry 1. Red BerryDark BerryMedium BerryLight RedDark RedMedium RedLight

11. Grey 2. Orange The term fish consumption is defined here as the amount GreyDark GreyMedium GreyLight OrangeDark OrangeMedium OrangeLight 2. Fish Consumption of fish that is available for consumption in a specific Base Colours 3. Yellow BaseColoursBackground BaseColoursTintedBox Black country: production (excl. non-food uses) plus imports YellowDark YellowMedium YellowLight minus exports plus or minus changes in stocks.4 4. Green GreenDark GreenMedium GreenLight Fish consumption considered here includes pelagic, demersal and other marine 5. Earth fish, freshwater fish, molluscs, crustaceans and cephalopods from marine and EarthDark EarthMedium EarthLight aquaculture production. The data covers the 50 years from 1961 to 2011. Figure 6. Brown 1 shows the increase in global fish consumption, which totalled 130 million tons BrownDark BrownMedium BrownLight in 2011. While Africa, America, Europe and Oceania only show slight increases in 7. Blue consumption over 50 years, fish consumption in Asia strongly increased from the BlueDark BlueMedium BlueLight 1980s on. This increase is mainly driven by China and its expanding fish produc- 8. Aqua tion (mainly aquaculture). AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight Fig. 1 140 World fish consumption 10. Berry 120 BerryDark BerryMedium BerryLight between 1960 and 2010 (in million tons). 100 11. Grey GreyDark GreyMedium GreyLight Source: FAO FishStatJ 80 database 60 Base Colours BaseColoursBackground BaseColoursTintedBox Black Africa 40 America 20 Asia Europe 0 Oceania 1960 1970 1980 1990 2000 2010 World Africa Americas Asia Europe Oceania World The development of the world’s per capita fish consumption in the same period is shown in Figure 2. Within 50 years it more than doubled and reached more than 19 kg per capita in 2013. Per capita fish consumption increased in every continent. However, the absolute amount of fish eaten by one person differs between regions. Africa has the lowest value of fish consumption per person, with 4.5 kg in 1961 increasing to 10.8 kg in 2011. Up to 1990, Europe had the highest level of per capita fish consumption (21.3 kg); since then, Oceania has led the field (26.5 kg in 2011). The steepest increase in fish consumption per person can be observed for Asia, obviously also driven by the strong increase in Chinese aquaculture production (see Fig. 2).

Figure 3 shows differences in current fish consumption levels at country level. Developed countries show the highest (a mean of 26.8 kg in 2013) while low-income food-deficit countries (LIFDCs) have the lowest per capita consumption (a mean of 7.6 kg in 2013). These differences in consumption depend on fish prices Fig. 2 Per capita world fish 30 consumption between 1960 and 2010. 25 (kg/per capita/year). Source: FAO FishStatJ 20 database 15

Africa 10 America Asia 5 Europe Oceania 0 World 1960 1970 1980 1990 2000 2010

Africa Americas Asia Europe Oceania World 16 and availability of fish as well as substitutes, income and socioeconomic factors (FAO 2016).

In terms of world consumption patterns, the share of demersal, pelagic and other marine fish decreased over time while the share of freshwater fish increased (see Figure 4). Again, the strong growth in aquaculture production, especially in China, accounts for this shift and has led to increased consumption of species such as cat- fish, tilapia, pangasius (a freshwater fish), shrimps and bivalves (shellfish such as 1. Red molluscs, crustaceans and cephalopods). The consumption of freshwater species RedDark RedMedium RedLight grew from 1.5 kg per capita to 6.5 kg per capita in the period studied. 2. Orange OrangeDark OrangeMedium OrangeLight

The consumption pattern at continental level reflects the global trend (see Fig.3. Yellow 5). However, the share of consumption of demersal and pelagic fish decreased in YellowDark YellowMedium YellowLight

Asia, America and Europe. One reason for this decrease might be that aquacul4. Green- ture products serve as a cheap alternative to wild catches. The picture in Africa GreenDark GreenMedium GreenLight

looks different. Here, the consumption pattern is relatively constant over time5. Earth with only a slight increase in the consumption of pelagic species. EarthDark EarthMedium EarthLight

6. Brown Northern Europe and North America favour demersal fish while the Mediterra- BrownDark BrownMedium BrownLight

nean and East Asian countries prefer cephalopods. Overall, 74% of the 19.77. kg Blue global per capita fish consumption in 2010 were finfish, 25% or 4.9 kg per capita BlueDark BlueMedium BlueLight were shellfish (FAO 2016). 8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry Fig. 3 BerryDark BerryMedium BerryLight Global per capita 11. Grey fish consumption GreyDark GreyMedium GreyLight (average 2008 to 2010) Base Colours (in kg/year). BaseColoursBackground BaseColoursTintedBox Black Source: FAO 2014

0–2 2–5 5–1 10–20 20–30 30–60 >6

Fig. 4 120 Global fish consumption Pelagic Fish pattern (in million tons). 100 Source: FAO FishStatJ Demersal and other database 80 Marine Fish

Pelagic Fish 60 Freshwater Fish Demersal and other Marine Fish 40 Molluscs, Other

Freshwater Fish Molluscs, Other 20 Crustaceans Crustaceans Cephalopods 0 Cephalopods 1961 1970 1980 1990 2000 2011

Fishing for Proteins | 17 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry BerryDark BerryMedium BerryLight

11. Grey GreyDark GreyMedium GreyLight

Base Colours BaseColoursBackground BaseColoursTintedBox Black

Pelagic Fish

Demersal and other Marine Fish

Freshwater Fish

Molluscs, Other

in Mio. tons Crustaceans

Cephalopods

Africa America 12 16 10 14 12 8 10 6 8 4 6 4 2 2 0 0 1961 1970 1980 1990 2000 2011 1961 1970 1980 1990 2000 2011

Asia Europe 100 18 16 80 14 12 60 10 8 40 6 20 4 2 0 0 1961 1970 1980 1990 2000 2011 1961 1970 1980 1990 2000 2011

Fig. 5 2.1 Fish Consumption in Selected Case Study Countries Fish consumption pattern in the case study continents In addition to the broad overview provided above, we focused on eight countries in (in million tons). Source: FAO FishStatJ order to provide further and more detailed insights. The case study countries were: database 2016 »»France and Germany (Europe), Pelagic Fish Peru and the United States of America (America), Demersal and »» other Marine Fish »»China and Indonesia (Asia), Freshwater Fish Senegal and South Africa (Africa). Molluscs, Other »» Crustaceans The choice was based on the following criteria: (1) each continent (except Oceania) Cephalopods should be represented, (2) developed as well as developing countries should be included, (3) fish and fisheries should play an important role for those countries.

The African countries Senegal and South Africa have the lowest fish consumption overall, starting with 0.06 million tons (Senegal) and 0.1 million tons (South Africa) increasing to roughly 0.3 million tons in both countries. Peru’s fish consumption slightly exceeds African values with 0.14 million tons in 1961 and 0.65 million tons in 2011. In contrast, fish consumption in the USA started much higher in 1961 with 2.5 million tons and increased to 6.8 million tons in 2011. Although Indonesia’s fish consumption was less than 1 million tons in the early 1960s, it has reached values similar to the USA in recent years with a maximum of 6.9 million tons in 2011. The leader in fish consumption is China. It doubled its fish consumption from 3.4 million tons in 1961 to 6.9 million tons in 1984. From the 1980s until today, China’s fish consumption grew at very high rates, leading to 46 million tons in 2011. In comparison, the development of fish consumption was quite moderate in the two European countries Germany and France. Fish consumption levels in Germany stayed relatively constant over time at 0.7 million tons in 1961 and 1.2 million tons in 2011. France experienced a stronger increase, rising from 0.7 million tons in 1961 to 2.2 million tons in 2011.

18 1. Red RedDark RedMedium RedLight

BlueDark BlueMedium BlueLight 2.1. OReradnge ORedDarrangeDark k ORedrangeMediuMedimum ORedLighrangeLight t 8. Aqua AqBlueDaruaDarkk AquaBlueMediuMediumm AquaLighBlueLightt 3.2. YOerallnogwe YOerangeDarllowDarkk YOerangellowMMediuediumm YOerallnowLighgeLightt 9.8. PinAquka PiAqnukDaraDarkk PinkAquaMMediuediumm PinkLighAquaLightt 3.4. YGreeellonw GreeYellownDarDarkk GreenYellowMMediuediumm GreenLighYellowLightt 109. .Pin Berk ry BPiernkDarryDarkk BPinkerrMyMediuediumm BPinkLigherryLight t 5.4. EGreearthn EaGreerthDarnDarkk EGreenarthMediuMediumm EGreenLigharthLightt 1110. GreBeryry GreyDarBerryDarkk GreBeryryMediuMediumm GreBeryryLighLight t 5.6. EBroartwhn BEarowrthDarnDarkk BEarowrthnMediuMediumm BEarowrthnLighLight t Ba11s. eGre Coylours BGreyDaraseColouk rsBackground BGreasyeMediuColourmsTintedBox BlGreacykLigh t 7.6. BBroluewn BrownDark BrownMedium BrownLight Base Colours BaseColoursBackground BaseColoursTintedBox Black 7. Blue

0.5 8 0.4 Africa America 0.5 86 0.3 0.4 64 0.2 0.3 2 0.1 4 0.2 0 20 0.11960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 0 0 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

Asia Europe 50 3 40 50 23 30 40 20 2 30 1 10 20 0 01 101960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

0 0 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

Senegal Fig. 6 Differences between the case study countries can again be found in the per capita Fish consumptionSouth Africa in the fish consumption over time (see Fig. 7). While South Africa’s per capita consump- eight caseSenegal study countries. tion is similar to the African average, Senegalese people consume four to five Peru(in million tons). South Africa times more fish than people in South Africa. Peru and the USA both exceed their Source:USA FAO FishStatJ continent’s average. Also, these two countries show a similar development in per Peru database USAIndonesia capita fish consumption over time. However, Peru shows a high fluctuation. This Senegal China variability is mainly driven by the strong dependence (up to 80%) of the Peru- South Africa Indonesia vian fishery on anchoveta (Engraulis ringens). For example, in the early 1980s, France China Peru anchoveta stocks west of South America drastically decreased – mainly because Germany USA France of an El Niño event (FAO 2016a). This decrease was reflected in the Peruvian per

GermanyIndonesia capita consumption. China France China and Indonesia experienced a strong increase in per capita consumption,

Germany similar to the overall Asian development. Per capita consumption in Germany and Europe is relatively steady over time. Germany lies below the European average. In contrast, France exceeds the European average in terms of fish intake per person and experienced a relatively strong increase over time from 18 kg per person to 35 kg per person.

When comparing consumption patterns at country level, differences between the developing countries (South Africa, Senegal, Indonesia and Peru) and the developed countries (China, France, Germany and the USA) can be noticed (see Fig. 8). In the developing countries, marine fish form the biggest share of consumed fish with pelagic fish clearly dominating. Also, except for Indonesia, the share of freshwater fish consumption is very small. Indonesia is one of the biggest aquaculture producers in the world. The main freshwater species produced are carp, tilapia and gourami; shrimp are also cultivated (FAO 2016b). A similar argument might hold true for Peru, which shows increased consumption of freshwater species and molluscs from 1990 on. Towards the end of the 1980s, Peru initiated aquaculture production of trout, tilapia, shrimps and scallops which developed successfully in subsequent years (FAO 2016c).

Among the developed countries considered in this report, Germany and France

Fishing for Proteins | 19 1. Red RedDark RedMedium RedLight

BlueDark BlueMedium BlueLight 2. Orange 1. Red OrangeDark OrangeMedium OrangeLight RedDark RedMedium RedLight 8. Aqua AquaDark AquaMedium AquaLight 3. Yellow BlueDark BlueMedium BlueLight 2. Orange YellowDark YellowMedium YellowLight OrangeDark OrangeMedium OrangeLight 9. Pink 8. Aqua PinkDark PinkMedium PinkLight 4. Green AquaDark AquaMedium AquaLight 3. Yellow GreenDark GreenMedium GreenLight YellowDark YellowMedium YellowLight 10. Berry 9. Pink BerryDark BerryMedium BerryLight 5. Earth PinkDark PinkMedium PinkLight 4. Green EarthDark EarthMedium EarthLight GreenDark GreenMedium GreenLight 11. Grey 10. Berry GreyDark GreyMedium GreyLight 6. Brown BerryDark BerryMedium BerryLight 5. Earth BrownDark BrownMedium BrownLight EarthDark EarthMedium EarthLight Base Colours 11. Grey BaseColoursBackground BaseColoursTintedBox Black 7. Blue GreyDark GreyMedium GreyLight 6. Brown BrownDark BrownMedium BrownLight Base Colours BaseColoursBackground BaseColoursTintedBox Black 7. Blue

Senegal South Africa Africa Senegal SouthPeru Africa AfricaUSA America Peru USAIndonesia AmericaChina Asia Indonesia ChinaFrance AsiaGermany Europe France Germany 40 Europe 30

30 Africa America 40 3020 20 30 2010 10 20 0 100 10 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

0 0 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

Asia Europe 40 40

30 30 40 40 20 20 30 30 10 10 20 20 0 0 10 1960 1970 1980 1990 2000 2010 10 1960 1970 1980 1990 2000 2010

0 0 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

Fig. 7 have the highest intake of marine fish. However, over time this share has Per capita fish consumption decreased in France in favour of molluscs and crustaceans. In Germany freshwa- in the eight case study ter fish consumption has increased. In the USA, overall consumption has stayed countries and the relatively stable, while, as in France and Germany, the share of marine fish has corresponding continent (black line) decreased while the share of freshwater fish and crustaceans has increased. The (in kg/year). increase in the consumption of freshwater species, crustaceans and molluscs may Source: FAO FishStatJ be driven by imports of aquaculture products, which are cheaper compared to database wild catches. Senegal South Africa China is the world´s leading country in aquaculture production and shows the Peru highest share of freshwater fish and molluscs consumption in recent decades. In USA contrast, its consumption of pelagic and demersal fish is the smallest of all the Indonesia eight case study countries. China France The rapid increase in fish consumption in the developing economies in Asia can Germany be explained by the correlation between increasing fish consumption and increasing wealth: per capita intake of fish increases fastest where wealth and urbanisation are combined and where domestic supply is increasing as well (HLPE 2014).

Overall, the share of marine fish and seafood in world consumption has declined over time, while the share of freshwater fish has shown an increase. However, marine fish still forms the majority of the world´s fish consumption, and some countries, for example South Africa, rely almost entirely on wild catches. Never- theless, although Sub-Saharan Africa currently contributes less than 1% of global aquaculture production, it registers as the fastest-growing aquaculture industry measured by its growth rate (World Resources Institute 2013). African aquaculture production is still very low. There are a number of reasons for this, including difficult market conditions and a focus on smallholder aquaculture, which is very important for local food security but cannot meet the goal of increased fish production at national level (Beveridge et al. 2010).

Monoculture of fish is increasingly replacing the traditional consumption of small fish species in some low-income countries (FAO COFI 2014). Small pelagic

20 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua Pelagic Fish AquaDark AquaMedium AquaLight

9. Pink Demersal and other PinkDark PinkMedium PinkLight Marine Fish 10. Berry Freshwater Fish BerryDark BerryMedium BerryLight Fig. 8 Pelagic Fish

Fish consumption pattern 11. Grey Demersal and Molluscs, Other GreyDark GreyMedium GreyLight in the eight case study in Mio. tons other Marine Fish countries Crustaceans (in million tons). Base Colours Freshwater Fish Source: FAO FishStatJ BaseColoursBackgr Molluscs,ound B aOtherseColoursTintedBox Black

Cephalopodsdatabase Crustaceans Cephalopods

Senegal South Africa 0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0 0 1961 1970 1980 1990 2000 2011 1961 1970 1980 1990 2000 2011

China Indonesia 50 8

40 6 30 4 20 2 10

0 0 1961 1970 1980 1990 2000 2011 1961 1970 1980 1990 2000 2011

Peru USA 0.8 8

0.6 6

0.4 4

0.2 2

0 0 1961 1970 1980 1990 2000 2011 1961 1970 1980 1990 2000 2011

France Germany 2.5 1.2

2 1 0.8 1.5 0.6 1 0.4 0.5 0.2 0 0 1961 1970 1980 1990 2000 2011 1961 1970 1980 1990 2000 2011

Fishing for Proteins | 21 fish in particular have a unique nutritional composition and are hence of great nutritional importance. Moreover, small fish are more affordable and more easily accessible than larger fish or other animal-based foods and vegetables (Kawarazuka and Bené 2011). Consumption of these should be encouraged and promoted, while the use of small pelagic fish for fish meal and fish oil should be reconsidered (Tacon and Metian 2013). In addition to Germany, pelagic fish play an important role in consumption for all developing countries in the case studies (see Fig. 8). The availability of marine fish in the future is very important, especially for those countries with low aquaculture production.

Conclusion »»Total fish consumption has increased over time, but the share of marine fish and seafood has declined. »»Marine catches still play an important role in fish consumption, some countries rely on wild catches at 100%. »»Increasing fish consumption is driven mainly by Chinese aquaculture. »»Aquaculture products are not a substitution option for all countries.

2.2 Global Importance of Fish as a Protein Source

A unique combination of high-quality protein and vital nutrients make fish an in- valuable food. Fish is not only a source of animal protein – 150 g of fish provides about 50 to 60% of an adult’s daily protein requirements – but also of fatty acids, vitamins and other essential elements such as iodine and selenium, which do not occur in such quantity and diversity in cereals, other crops or meat (Beveridge et al. 2013; Kawarazuka and Béné 2011; WOR2 2013). Dietary diversity and dietary adequacy are high on the agenda in the fight against hunger and malnutrition. Severe poverty is highly correlated with the intake of ‘too many’ staples and too little protein, fat and micronutrients.

In coastal regions of developing countries fish is often the only affordable and relatively easily available source of animal protein. In countries like Sierra Leone, which has a very low overall food security, the share of fish in animal protein is over 50%. Marked differences exist between and within countries and regions in terms of the quantity and variety consumed per capita and the subsequent contribution to nutritional intake. In a global comparison, Africa and Latin America consume relatively little fish (around 10 kg per person per year), whereas per capita consumption in Asia, North America and Europe is above the global average (20 kg) at around 22 kg per year.5 This reflects the factors that affect fish consumption: how available fish is, how expensive it is, whether there are dietary traditions in relation to fish and how developed the country is. Generally speaking, the lower the income the lower the fish consumption. Dietary tradition relates to the fact that countries with a strong fishing tradition due to a long coastline, many fish-rich rivers or islands tend to still consume more fish (FAO 2016).

The WHO recommends on average an annual intake of 11.7 kg fish per person (about 32 g per day or 225 g per week). Averaged over world regions, only Africa and Latin America come close to meeting this reference value. Globally speaking, however, there is an unequal distribution of fish and Northern hemisphere consumes too much fish per capita.

In 2013, fish accounted for 6.7% of all protein consumed and for 17% of the global

22 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

population’s intake of animal protein. In developing countries, this share was10. Ber ry 19.6% and 24.7% in low-income food-deficit countries (LIFDCs) (see Fig. 9). For BerryDark BerryMedium BerryLight

3.1 billion people, fish accounts for 20% of their animal protein; for 4.3 billion11. Grey people, this share is 15% (FAO 2016). Some small island states such as Kiribati, GreyDark GreyMedium GreyLight

Micronesia and the Maldives depend almost exclusively on fish as a proteinBas e Colours source (FAO 2016). The average daily dietary contribution of fish in terms of BaseColoursBackground BaseColoursTintedBox Black calories is about 34 calories per capita. In countries where there is a lack of alternative protein food and where there is a traditional preference for fish (e.g. Senegal), as well as in several small island states such as the ones named above, the daily fish calorie intake reaches 130 calories per capita or more (FAO 2016).

However, based on this data, we may underestimate the importance of fish as a protein and nutrient provider, in particular for countries with low food security and/or poor populations. There are a number of reasons for this.

»»There is a huge variation between and within countries, in particular coastal regions within small island countries depending to a much higher extent on fish as a source of animal protein. When these coastal regions are remote, i.e. far away from major markets and not easily accessible, and have a high prevalence of poverty, substitution possibilities are limited in the short run. Consumption in this case is supply driven.

»»Consumption data is likely to be underestimated in view of the under-recorded contribution of subsistence fisheries and small-scale fisheries in official statistics (FAO 2014; Pauly 2016). Hence, actual fish consumption in developing countriesFish is probablyconsumption higher than official data reports. per capita (kg)

»»EconomicFish dependence contribution onto fish as a source of income plays an important role in coastal areastotal animal in developing protein (%) countries with repercussions on food security.

Fig. 9 30 Total fish consumption and fish contribution to total 25 animal protein. Source: FAOSTAT. 20 Fish consumption per capita (kg) 15 Fish contibution to total animal protein (%) 10 .... WHO recommendation for fish(kg/caput/year) 5

0 Oceania Europe Northern Asia Africa Latin America World America and Caribbian

The dietary pattern of poor people typically has a very strong component of staple food (in particular maize, rice and other cereals), with fish consumption an important factor in helping to correct an imbalanced calorie/protein ratio. Fish often represents an affordable source of animal protein that may not only be cheaper than other animal protein sources, but also part of local and/or traditional recipes. In countries with a long coastline such as Senegal and islands such as Indonesia, fish accounts for, or exceeds, 40% of total animal protein intake (see Tab. 1 next page).

Fishing for Proteins | 23 Prevalence of Fish consumption Fish contribution undernourishment (kg per capita to total animal (% of population) per year) protein (%) 2013 – 2015 2011 2011 Tab. 1 China 9.3 33.5 20.56 Dependence on fish in the diet in the eight case Indonesia 7.6 28.9 54.82 countries in this study. Senegal 10.0 23.5 43.73 Source: FAO South Africa < 5 5.7 5.2 Peru 7.5 22.7 22.28 France < 5 34.8 13.3 Germany < 5 14.2 7.28 USA < 5 21.7 7.37

In Figure 10 we provide a graphical representation of ‘the role of fish in the diet in relation to economic development’. While very low income tends to be associated with hunger, rising income first leads to adequate availability of calories and subsequently to adequate nutritional quality. At low income levels (bottom left) fish contributes to an unbalanced diet: fish tends to be consumed in very small or very large quantities depending on availability. At very high income levels (top right), fish also tends to contribute to an unbalanced, overly protein-rich diet. Economic development is associated with structural change in which the share of people involved in fishing declines in line with increasing development.

Fig. 10 Importance of fish in the diet in relation to economic Strong fish Structural change development of a country dependency as or population. source of income (jobs, foreign exchange) Adequate Adequate Fish contributes to an number of nutrition unbalanced diet calories

hunger Low adaptive Fish contributes to capacity a balanced diet

Economic developement Fish contributes to an unbalanced diet

While in absolute numbers animal protein intake is lower in developing countries than in developed countries, the growing share of animal protein worldwide is driven mainly by developing countries catching up, particularly in Africa and Asia. Splitting up animal protein intake into fish and meat, the contribution of fish to total animal protein intake has been declining slightly since 1990 at the expense of other animal proteins.

In poor countries where fish is a traditional food, increasing income leads to increased meat consumption and greater consumption of more valuable fish

24 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

species, e.g. demersal instead of pelagic fish. Not surprisingly, consumption10. Berofry fish from aquaculture is growing fast. Between 1990 and 2012, fish consumption BerryDark BerryMedium BerryLight from wild sources remained almost the same, but consumption of fish from11 . Grey aquaculture has multiplied by five. Half of fish produced for human consumption GreyDark GreyMedium GreyLight came from aquaculture in 2012, compared to just 5% in 1962 and 37% in 2002Base Colou rs (FAO 2015). BaseColoursBackground BaseColoursTintedBox Black

Just as there is enough food on average to feed everyone on the planet and hunger is a question of distribution, there is also no protein gap considering current world average protein supply.

Fig. 11 120 Total protein consumption in the eight case study countries subdivided into 100 total protein (excluding fish, light blue) and fish protein (dark blue). 80 Fish protein 2009-11 Source: FAOSTAT in gr/caput/day

60 Non-Fish protein 2011 Fish protein in gr/caput/day (2009-11). (in g/caput/day) 40 World average Non-Fish protein protein supply (2011) (in g/caput/day) 20 Nutrition recommendation for total protein intake .... Nutrition recommendation 0 for total protein intake South Senegal Peru USA Indonesia Germany France China (in g/caput/day) Africa Global average protein consumption (g/caput/day) Globally, the average protein supply in 2011 was 79 g per capita per day, while the average protein need was 49.6 g per capita per day. The latter was calculated from the recommended 0.8 g per kg of body weight and the average weight of a person (62 kg) in 2011. Figure 11 shows protein consumption in the eight case study countries. The height of each column represents total protein consumption: the darker blue parts represent fish and the lighter blue parts represent all other protein. Based on this data, all countries except Liberia, Guinea-Bissau, Mozam- bique, Haiti and Madagascar had enough protein available in 2011. Nevertheless, the distribution within and between countries is highly unequal.

2.3 Food Security and Fish

Undernourishment is a major problem worldwide, with one in seven people undernourished and more than one-third of infant mortality attributable to undernutrition. This is especially the case in many developing countries, with the bulk of undernourished people living in rural areas (see Tab. 1). Most of the world’s undernourished people live in South Asia, closely followed by Sub-Saha- ran Africa and East Asia.

In addition to the prevalence of undernourishment, researchers often use several other indicators in order to assess the food security level of a country such as stunting, measured by height-for-age, and wasting, measured by weight-for-age.

Fishing for Proteins | 25 A common multidimensional index used is the Global Hunger Index (GHI, see Tab. 2). To reflect the multidimensional nature of hunger, the GHI combines the following four component indicators into one index:

»»1/3 for undernourishment: the proportion of undernourished people as a percentage of the population, reflecting the share of the population with insufficient caloric intake »»1/6 for child wasting: the proportion of children under the age of five who suffer from wasting, i.e. low weight for their height, reflecting acute undernutrition »»1/6 for child stunting: the proportion of children under the age of five who suffer from stunting, i.e. low height for their age, reflecting chronic undernutrition »»1/3 for child mortality: the mortality rate of children under the age of five, partially reflecting the fatal synergy of inadequate nutrition and unhealthy environments (all standardised based on thresholds set slightly above the highest country-level values observed worldwide for that indicator between 1988 and 2013.6

This indicator emphasises the nutrition situation of children – a vulnerable sub- set of the population for whom a lack of dietary energy, protein or micronutrients (i.e. essential vitamins and minerals) leads to a high risk of illness, poor physical and cognitive development or death. It also combines independently measured indicators to reduce the effects of random measurement errors (GHI 2015).

The GHI categorises countries according to hunger severity into low (values <10), moderate (scores of 10 – 20), serious (20 – 35), alarming (35 – 50) and extremely alarming (>50). In our target countries, Senegal and Indonesia fall into the serious category, South Africa has moderate levels, and all others are in the low category. The index is not calculated for Germany, France and the USA, which are considered to be generally food secure.

Tab.2 1995 2005 2015 (Multidimensional) Global With data from Hunger Index 1995 to 2015 for case study countries. 1993–1997 2003–2007 2010–2016 Source: Welthungerhilfe China 23.2 13.2 8.6 (WHH); International Food Policy Research Institute Indonesia 32.5 26.5 22.1 (IFPRI); and Concern Senegal 36.9 28.5 23.2 Worldwide 2015 South Africa 16.5 21.0 12.4 Peru 25.0 18.8 9.1

Conclusion: »»The relevance of fish for food and nutrition security manifests itself in its value as a protein and micronutrient source. »»Poor people generally consume too few micronutrients and too little protein. »»Globally speaking, there is no protein gap.

26 2.4 Dependence on Fish

In this section we combine the country-specific situation of food (in)security and general health/hunger situations with the socioeconomic value of fish and fisheries to the livelihoods in these countries in order to describe the countries´ dependence on fish.

The following factors determine dependence on fish for food and nutrition:

»» Fish as a food source, i.e. as a source of calories. This is particularly relevant in countries such as Senegal with food insecurity and a high fish intake.

»»Fish as a protein and micronutrient source. This is particularly relevant in countries where the protein share of fish in the diet is high or low. It is also particularly relevant in almost all poor countries that have an imbalanced diet, where people eat too much staple food and not enough micronutrients (e.g. Senegal, Indonesia). The fisheries and aquaculture sector plays, and can continue to play, a particularly prominent role in diversified and healthy diets. While average per capita fish consumption may be low, even small quantities of fish can have a significant positive nutritional impact, given that it is a concentrated source of essential dietary components. Micronutrient deficiencies7 affect hundreds of millions of people, particularly women and children in the developing world. More than 250 million children worldwide are at risk of vitamin A deficiency (leading to blindness), more than 30% of the world’s population are iron deficient, 200 million people have goitre with 20 million suffering from learning difficulties as a result of iodine deficiency and 800,000 child deaths per year are attributable to zinc deficiency.

»»Economic dependence on fish as an income source to buy (healthy) food. This is particularly relevant in countries with high poverty rates and in countries where people working in fisheries are comparably poor. When they lose their jobs in the fish industry, poverty rises and this affects diets in two ways: hunger and quality of the diet. A lack of money to buy food leads first to a less balanced diet, with a tendency to eat less fish and meat, hence less protein and micronutrients and in even more severe cases to a lower intake of calories overall. The severity of this effect depends on the strength of the formal or informal social security systems in place. As an example, if Senegal stopped fishing completely, this would have adverse impacts on the livelihoods of fishermen with widespread hunger as a likely consequence, at least in the short run. This is a typical distribution problem: globally speaking, enough is available, but it is distributed highly unequally.8

Hence, fish contributes to the nutritional security of poor households in developing countries in various ways. These include a consumption pathway where the direct consumption of fish increases intakes of not only calories, but more importantly protein, micronutrients and omega-3 oils, and a cash-income pathway, whereby the fish industry contributes to employment and higher overall food consumption in poor countries (see Fig. 12). Commercialisation, fish processing and small-scale aquaculture offer important livelihood opportunities, particularly for women in developing countries, through their direct involvement in the production, processing and sale of fish. In particular, countries where a large share of the population is dependent on fishing (e.g. Senegal) may suffer from higher hunger rates and political instability if the fishing sector declines. According to estimates by the FAO, a total of 660 to 820 million people are directly or indi-

Fishing for Proteins | 27 rectly dependent on fisheries and aquaculture. The FAO estimates the number of fishermen alone at 54 million, of which 87% live in Asia. In developing countries, the majority of them work in small-scale fisheries with low fish production per person: on average 1.5 tons per year compared to 25 tons per fisherman and year in Europe (FAO 2014).

Fig. 12 Main drivers of Nutrition dependence Economic dependence dependence on fish. Source: Own presentation. Protein and micro- Catch values nutrient source

Availability and Number of people price of substitutes employed

2.5 Fish Dependence Index

Our fish dependence indicator measures the degree of dependence on fish as a source of income and food, in particular as a source of protein. It is based on a composite ranking of the following factors:

»»food security: prevalence of undernourishment for the period from 2011 to 2013 in %, data from the FAO; »»fish consumption: fish share of total animal protein intake in % for 2011, data from the FAO; »»capture production per capita for 2011, data from FAOfishstat; »»gross domestic product (GDP) per capita as a proxy for substitution capacity for 2011, data from the World Bank.9

The indicator measures the short-run dependence. In the long run, the possibilities for compensating through new industries (e.g. building up an aquaculture industry) and new sources of protein (e.g. plant-based sources) have to be considered.

The indicator is similar in some ways to the indicator developed by Allison et al. (2009a, 2009b) and used in Badjeck et al. (2013). The main difference is the more recent data used in this index and the stronger focus on fish as food. Using more recent data (from 2011) is only possible because this index is composed of fewer components. The results of the two indices are in most cases very similar.

Fish as a share of animal protein in food and capture production as a proxy for economic relevance are the main factors and weighted equally. GDP per capita (in USD) and food security (prevalence of undernourishment) are the moderating factors. A high income reduces the dependence created by a high level of the two main factors due to compensation effects. This means that a high income increases possibilities for compensating for a loss of jobs or the loss of fish as a source of food, e.g. through imports and social security. A high level of food security reduces the potentially harmful effects of losing fish as a food source.

28 Figure 13 shows the worldwide distribution of fish dependence. Asian island states and West African coastal countries are most dependent on fish, whereas dependence in Europe is comparably low. Figures 13a to 13d show the effects of the application of one of the four factors: a) catch per capita, b) share of fish in total consumption of animal protein, c) per capita GDP, d) share of undernourishment.

The index reveals that, not surprisingly, countries with a very high share of fish in the diet are highly dependent on fish (see Tab. 3, page 31). One important aspect is that these countries usually also have a comparably large fish industry and are not wealthy or particularly food secure countries. At the same time, the fish protein in high-income food-secure countries is, on average, relatively lower. Nevertheless, poverty and fish dependence do not necessarily go hand in hand (see Tab. 3, left). Among the poorest countries are countries with no access to the ocean (e.g. Ethiopia, Central African Republic).

Germany – with a low catch per person, a low share of fish protein, high income and high food security – has very little dependence on fish. South Africa and United States are also not particularly dependent on fish, even though they have

Fig. 13 Fish dependence scores worldwide. Source: Own presentation.

high medium high medium low no data This world map is a composite ranking of the following factors:

Fig. 13a Per capita catch by country.

high medium high medium low no data

Fishing for Proteins | 29 Fig. 13b Share of fish in total consumption of animal protein.

high (> 20%) medium high (betw. 10 and 20%) medium (betw. 5.6 and 10%) low (< 5.6%) no data

Fig. 13c Per capita GDP in USD.

high (> 26,932) medium high (betw. 11,360 and 26,932) medium (betw. 4,133 and 11,360) low (< 4,133) no data

Fig. 13d Share of undernourishment in population.

high medium high low no data

30 Tab. 3 8 poorest countries Fish 8 countries with the Fish Fish dependence of the depend- highest share of fish in depend- eight poorest countries and ence total animal protein intake ence the eight countries with the highest share of fish protein Liberia medium Sri Lanka very high consumption (own results). Malawi high Bangladesh very high Niger low Solomon Islands very high Central African Republic medium Kiribati very high Mozambique high Micronesia very high Ethiopia low Cambodia very high Guinea very high Sierra Leone very high Togo high Maldives very high

Tab. 4 Per capita GDP Prevalence of Fish share in Catch per Dependence Fish dependence in the (USD) undernourish- animal protein person (tons) score eight case study countries. ment (%) 2011 2011-13 2011 2011 South Africa 12,291 5.0 5.02 0.01 medium Senegal 2,163 12.3 43.73 0.03 very high Peru 10,429 9.6 22.28 0.28 high United States of America 49,804 5.0 7.37 0.02 medium Indonesia 8,438 9.3 54.82 0.02 very high Germany 42,080 5.0 7.28 0.00 low France 37,325 5.0 13.30 0.01 high China 10,286 11.0 20.56 0.01 high

a very large absolute catch and low consumption (see Tab. 4, page 31). The target countries also include Senegal and Indonesia, where many people are involved in the fishing industry, fish is an essential part of the diet and there is a relatively low level of food security.

As an example showing the complexity of fish dependence, Senegal is considered highly dependent on fish.10 According to FAO figures, an estimated one million people are directly or indirectly dependent on fish. Fish accounts for 44% of animal protein intake, but only 12% of total protein. With a consumption of, on average, 24 kg of fish per year per head, Senegal ‘over-consumes’ fish if the WHO recommendation of 11.7 kg is taken as a reference. At 60 g per head per day, protein availability is also above the level needed (the recommended level is 49 g). So, even with a moderate decline in fish intake, there would be no protein gap in Senegal. Nevertheless, 10% of the population is undernourished. In coastal rural areas, the fishery sector is the main source of income (Thiao et al. 2012). However, despite an adequate protein supply, a decline in the fishing sector would likely result in increasing poverty and hunger in Senegal (Lam et al. 2012).

Conclusion: »»Poverty and fish dependence do not necessarily coincide. »»However, poor countries with a comparably large fishing sector have a very high risk of becoming food insecure if fish is lost as an income source.

Fishing for Proteins | 31 2.6 Current Fish Consumption and Fish Supply

Since fish consumption and fish supply differ between regions it is interesting to see which regions are able to meet their demand with their own production and which regions depend on imports to meet their demand for fish. Using population and catch data for 64 LMEs, we first calculate to what extent fish landings from the LMEs meet local fish consumption in 2010 for the population living in the respective neighbouring countries.11 Up to 95% of the annual global fish catch originates from these 64 areas (Sherman et al. 2009). They pose particular challenges for regional, and in some cases, multinational management. We used data from the University of Vancouver’s Sea Around Us project for these calculations (Sea Around Us-database).

Fig. 14 Population, catches and share of local consumption for population in 2010 that was covered by LME catches in 2010 Source: Sea Around Us database/own mapping. The regions depicted in Figure 14 refer to LMEs. Data on catches per LME in Pop. 2010 Scenario (mn) 2010 is taken from the Sea Around Us database. Data on population per country > 50 in 2010 is taken from the IPCC Shared Socioeconomic Pathways Scenario 1. Fish 50–150 consumption per capita and country in 2010 is taken from the FAOStat database. 150–500 (Important: According to the FAO, fish consumption includes all kinds of fish 500–1,000 and seafood products including aquaculture and inland fisheries products while, according to Sea Around Us, the catches only cover marine fish and seafood in Fraction the LMEs. High seas catches are not included in our model. Land-locked coun- 0–80% tries are not included. 80–100% 100–500% > 500% Figure 14 shows how LMEs differ in the extent to which local landings are sufficient to meet local needs in terms of fish consumption. LMEs in which landings Catches (mn tons) are not sufficient to meet local fish demand are indicated in red (coverage 0 to Keine Daten 80%) or yellow (coverage of 80 to 100%). LMEs in which landings are more than 0.01 – 0.60 sufficient to meet local fish demand are indicated in light green (coverage between 0.61 – 1.50 1.51 – 4.00 100% and 500%). Extremely over-supplied LMEs are able to meet much more 4.01 – 8.00 than local demand and are indicated in dark green (coverage greater than 500%). 8.01 –12.86 There is a substantial variation: in very Arctic waters, e.g. the Canadian High Arctic, North Greenland, the Beaufort Sea or the Insular Pacific Hawaiian LME, fish production covers less than 1% of local consumption while the Scotian Shelf, Newfoundland-Labrador Shelf, Icelandic Shelf and Sea and the Faroe Plateau stand out with their massive production, which exceeds local needs and leads to coverage of more than 1,000%.

32 Landings in 9 of the 64 LMEs are much higher than local needs. The north east coast of North America and the northwest coast of Europe, including Iceland and the Faroe Plateau, are characterised by very high catches relative to consumption in the local population. In the remaining 16 LMEs, landings are sufficient to cover local consumption of fish.

Thirty-nine LMEs have landings that are not sufficient to cover local consumption; of these 34 have a fraction of less than 80% and 21 have a fraction below 50%. The weakest LMEs are the areas in the High Arctic, north of Canada and Russia. LMEs around Australia struggle to meet at least half of the local demand.

The ratio of landings to local needs for all LMEs considered is 82%. This means that 82% of the aggregated fish consumption in all LMEs is covered by marine catches from LMEs. So, obviously those catches are not sufficient to satisfy consumption. As mentioned earlier, the catches do not include aquaculture, inland or high seas production. Catches from these sectors probably account for the 18% of fish demand that is not covered by LMEs.

A similar calculation using the WHO recommendation for fish consumption of 11.7 kg per capita per year, instead of the actual fish consumption levels of 2010, lead to a ratio of landings to local needs of 144% for all LMEs considered. This indicates that in 2010 LME catches were sufficient to satisfy the basic needs in terms of fish consumption according to the WHO. However, a redistribution of the resource is required. In this scenario, 39 LMEs are in the under-supplied category while 9 LMEs are in the extremely over-supplied category.

Overall, it seems that LMEs with many bordering countries tend to have a lower potential to meet demand. Examples include the Mediterranean, the Caribbean Sea and the Baltic Sea. In contrast, LMEs with only one or a few coastal countries perform better. Also, LMEs located in the North Atlantic and the East Pacific seem to have a stronger potential to meet local demand. This also holds true for Europe, the east and west coast of the USA and the west coast of Latin and South America. In contrast, LMEs around Africa (North West Africa being the exception) and along the Asian coasts as well as in Australia show a deficit in marine fish production compared to local demand.

In our case studies, only Indonesia and China are facing a clear under-supply in terms of marine catches. However, both countries might find suitable substitutes through their aquaculture production.

Conclusion:

»»Catches vary strongly at LME level. »»In approximately two-thirds of all LMEs, the total demand for fish in 2010 could not be covered by local marine catches. »»Marine catches can cover 82% of global demand for fish. »»Excess catches are mainly reported for LMEs in the North Atlantic and East Pacific.

Fishing for Proteins | 33 We are interested in future global marine catches 3. The Bio-economic Model and effects on consumption levels, specifically focusing on the year 2050. Fish markets are globalised to a very large extent (Smith et al. 2010; Asche et al. 2015). Global markets allocate worldwide catches so that they equal total demand for both human consumption and animal feed. Global catches are determined by the fishing effort applied and the productivity of the fish stocks. Thus, an assess- ment of potential future catches requires a bio-economic modelling approach that combines the ecological approach describing productivity of fish stocks with the economic part describing the economic incentives to exerting fishing effort and markets that allocate fish catches to different consumers. The literature suggests that the efficiency of fishery management plays a central role in this respect (Costello et al. 2008; Quaas et al. 2016).

The model is designed to tell us how the overall size of the fish catches changes under different economic and fishery management scenarios and how the total global catch is allocated in terms of regional catches and regional consumption quantities. To address these questions, we use a nested modelling approach. Here we briefly sketch the modelling approach; details are given in the technical appendix.12

3.1 The Model Approach

Questions 1 and 2 are addressed in a global model that separates predatory and prey fisheries using Lotka-Volterra stock dynamics. With this model approach we take biological interaction into account. The model assumes that the change in biomass over time depends on the natural growth of a stock, the interaction between predator and prey and fishing activities. The predator has a negative impact on the prey biomass, meaning that if the predator stock increases, the prey stock will decrease because the predator feeds on prey. On the other hand, if the prey stock increases, the predator stock will also increase. Hence, the prey stock has a positive impact on the predator’s biomass. Fishing activities are influenced by the demand parameter and reduce the change of a stock over time. Available studies investigating a similar research question at the global level have so far not taken into account such ecological interactions and lumped all fish into one aggregate surplus production model (World Bank 2009), or considered several biologically independent stocks (Quaas et al. 2016; Costello et al. 2016).

Furthermore, we look into the interactions of predatory and prey fisheries on the global fish markets, as well as into the interaction between fish and non-fish protein food, by means of a stylised consumer demand system, where different types of fish and non-fish protein food are imperfect substitutes (Anderson 1985; Quaas and Requate 2013). Besides predatory fish and forage fish, we consider protein-rich non-fish food, including beans, dairy products, eggs, lentils, peas, maize, meat, nuts and rice.

Questions 1 and 3 are addressed in a regionalised model where each large marine ecosystem hosts one individual fish stock. In this model we abstract from ecological interactions between the stocks and consider a generalised Schaefer surplus production model for each LME, but include economic interactions that are mediated by the global fish market. Thus, we do not differentiate between predator and prey but assume that all fish in one LME can be seen as one stock.

34 The change of biomass over time is determined by the natural growth of a stock and the fishing activity. Again, fish production is affected by the LME-specific demand parameters for fish goods. As fish is a traded commodity on a worldwide market, demand in one region of the world will affect the production in another region: a higher world-market price makes it more attractive to exert fishing effort. In our regional demand model we consider regional consumption of three commodities at the LME level: domestically produced fish, imported fish and protein-rich non-fish substitution goods, again including beans, dairy products, eggs, lentils, peas, maize, meat, nuts and rice.

3.2 Data and Estimation of Model Parameters

We use data from three main sources for the estimating parameters of the bio-economic fishery models:

»»Sea Around Us (www.seaaroundus.org), »»FAO database FAOStat (http://faostat3.fao.org/home/E) and »»FAO Fisheries and Aquaculture Department’s database FishStatJ (http://www.fao.org/fishery/statistics/software/fishstatj/en).

Fig. 15 Data on catches and landed values of fish is taken from the Sea Around Us data- Large marine ecosystems base. From this database we use time series of catches and landed values from considered in this study. 1950 to 2010 for 64 LMEs (see Figure 15).

,Ä64 Ä ,Ä64 ,66 Ä ,Ä58 Ä ,Ä19 ,20 ,Ä57 Ä56 ,54 ,Ä55 ,Ä18 , ,Ä59 ,Ä21 Ä63 ,Ä60 Ä23 ,Ä1 , ,Ä22 , ,Ä53 ,Ä2 ,Ä52 ,Ä65 ,Ä9 ,Ä24 Ä8 Ä , ,Ä25 ,62 ,Ä51 ,Ä7 ,Ä48 ,Ä50 ,Ä3 ,Ä26 ,Ä6 ,Ä47 ,Ä49 ,Ä10 ,Ä4 ,Ä5 ,Ä27 ,Ä33 Ä Ä32 Ä11 ,12 , Ä , ,Ä34 ,35 ,Ä36 ,Ä17 ,Ä28 ,Ä37 ,Ä31 ,Ä16 ,Ä38 ,Ä39 ,Ä45 ,Ä40 ,Ä29 Ä ,Ä13 Ä15 ,30 , ,Ä44 ,Ä41 ,Ä43 ,Ä46 ,Ä42 ,Ä14

,Ä61

1. East Bering Sea 15. South Brazil Shelf 28. Guinea Current 42. Southeast Australian Shelf 55. Beaufort Sea 2. Gulf of Alaska 16. East Brazil Shelf 29. Benguela Current 43. Southwest Australian Shelf 56. East Siberian Sea 3. California Current 17. North Brazil Shelf 30. Agulhas Current 44. West-Central Australian Shelf 57. Laptev Sea 4. Gulf of California 18. Canadian Eastern Arctic - 31. Somali Coastal Current 45. Northwest Australian Shelf 58. Kara Sea 5. Gulf of Mexico West Greenland 32. Arabian Sea 46. New Zealand Shelf 59. Iceland Shelf and Sea 6. Southeast U.S. Continental Shelf 19. Greenland Sea 33. Red Sea 47. East China Sea 60. Faroe Plateau 7. Northeast U.S. Continental Shelf 20. Barents Sea 34. Bay of Bengal 48. Yellow Sea 61. Antarctic 8. Scotian Shelf 21. Norwegian Sea 35. Gulf of Thailand 49. Kuroshio Current 62. Black Sea 9. Newfoundland-Labrador Shelf 22. North Sea 36. South China Sea 50. Sea of Japan/East Sea 63. Hudson Bay Complex 10. Insular Pacific-Hawaiian 23. Baltic Sea 37. Sulu-Celebes Sea 51. Oyashio Current 64. Central Arctic Ocean 11. Pacific Central-American 24. Celtic-Biscay Shelf 38. Indonesian Sea 52. Sea of Okhotsk 65. Aleutian Islands 12. Caribbean Sea 25. Iberian Coastal 39. North Australian Shelf 53. West Bering Sea 66. Canadian High Arctic- 13. Humboldt Current 26. Mediterranean 40. Northeast Australian Shelf 54. Northern Bering- North Greenland 14. Patagonian Shelf 27. Canary Current 41. East-Central Australian Shelf Chukchi Seas

Fishing for Proteins | 35 The Sea Around Us database includes differentiated data on catches, landings and estimated discards of large-scale, small-scale, subsistence and recreational fishery. The data in the Sea Around Us database combines reported values from the FAO and additional unreported data estimated by Sea Around Us. Additional information on missing data was collected for this estimate (Pauly and Zeller 2015). The main sources were governmental websites and publications, statistical agencies responsible for the fishing industry, international research organisations such as the FAO, the International Council for the Exploration of the Seas (ICES) or regional fisheries management organisations such as the Northwest Atlantic Fisheries Organization (NAFO) as well as academic literature. Based on this information, anchor-points in time are derived and used for interpolation. A linear interpolation between anchor-points is used to reconstruct commercial catches. Either population trends or trends in the number of fishers over time are used to interpolate between anchor points for non-commercial catches from subsistence and recreational fisheries. The reconstructed catches are then combined with the reported FAO data.

The Sea Around Us catch data is disaggregated into catches by species, functional groups and size groups of 0-30, 30-89 and >89 cm average length. We use this information to group the catches into predator and prey categories. We take a size-based approach and consider large fish to be most likely predatory and small fish to be most likely forage fish. Specifically, we consider all fish greater than 90 cm to be predatory while fish smaller than 90 cm and invertebrates are considered to be prey. In the global predator-prey model and the global demand model, the data is aggregated per year over all 64 LMEs, leading to 61 observations each (= number of years from 1950 to 2010) for predator and prey catches. In the fish supply model at LME-level we use total catches per year aggregated over size groups. In this model the data has been aggregated by LME and year. In total, we use a dataset with 3,904 observations.

Figure 16 shows the development in total catches and catches from 1950 to 2010. The amount of total catches peaks in 1996 at 123 million tons. Since then, catches are decreasing. This is in line with Pauly and Zeller (2016) who also calculated the global peak in catches in 1996 at a level of 129 million tons. In contrast to the global data of Pauly and Zeller (2016), we do not include high seas catches, which results in the difference of 6 million tons. In addition, some small island states, such as Wallis and Futuna Islands (France), Saint Helena (UK) or Nauru, are not included because their geographical position does not lie within the area of an LME.

The biological parameters are estimated using the Catch-MSY method developed by Martell and Froese (Martell and Froese 2013). This method allows biological parameters, such as biomass based on catch data, to be estimated. It requires time series of catch data and prior ranges for the parameter values as well as possible ranges of stock sizes from the initial and the final period.

After specifying initial stock sizes and limits for the final stock size, a parameter set is randomly drawn from the prior parameter distribution. Then, the underlying fish supply model is used to calculate the biomass corresponding to the level of harvest given the parameter set. If this biomass is in a reasonable range, the parameter set is stored. In our analysis, we repeat this procedure 10,000,000 times for each LME. We use samples of 1,000 randomly picked accepted parameter values in our model computations to compute mean estimates and confidence intervals. Thus, all results reported below are based on averages and standard deviations obtained from 1,000 separate model runs.

36 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry BerryDark BerryMedium BerryLight

11. Grey GreyDark GreyMedium GreyLight

Base Colours BaseColoursBackground BaseColoursTintedBox Black

Fig. 16 150 Development of global total catches and predatory/prey catches from 1950 to 2010. (in million tons) Source: SeaAroundUs 100

Total Prey Predator 50

0 1940 1960 1980 2000 2020

Total Prey Predator We use the above-mentioned approach by Martell and Froese with a Schaefer surplus production model for each LME and for one global stock of aggregated fish. In the global predator-prey model we extend the approach by Martell and Froese and determine parameter values for a Lotka-Volterra predator-prey model (Hannesson 1983). In each case, initial parameter sets are randomly drawn from a uniform distribution to be tested. Economic theory predicts a positive relationship between fish stock biomass and market supply of fish (or no relationship at all in the case of a pure schooling fishery), and thus a negative relationship between stock biomass and the fish price. We use price data in each run for a tested parameter13 set to check whether this requirement is met. Biological parameters that do not pass this test are rejected. Otherwise, we use the resulting information on the relationship between price and stock biomass to obtain an estimate for economic parameter values.

3.3 Global and Regional Demand Systems

To quantify the demand systems, we use data on fish prices from Sea Around Us. The data on landed values is used to derive prices. The ex-vessel fish prices, used to calculate the landed values, are derived using two approaches. Local ex-vessel prices, converted to US dollars, form the starting point. These are combined with ex-vessel prices calculated from reported landed values and catches.

Real prices are determined by deflation using the consumer price index of 2005 (Sumaila et al. 2015). We use this price data for the calculation of production values in the global demand model. Figure 17 shows the development of ex-vessel prices over time for predator and prey fish. Since prey also include valuable invertebrates such as shrimp, lobster or sea urchins, the prey price does not deviate much from the predator price.

In addition to fish production quantities and fish prices, the global demand model also requires data on total expenditures and consumption levels for the three commodities: predatory fish, prey fish and non-fish protein-rich substitution goods. Total national expenditures are calculated from production, export and import values, while national consumption is calculated from production, export and import quantities (see Fig. 19).

Data on substitution goods is taken from the FAO Statistics division, with a

Fishing for Proteins | 37 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry BerryDark BerryMedium BerryLight

11. Grey GreyDark GreyMedium GreyLight

Base Colours BaseColoursBackground BaseColoursTintedBox Black

1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow 2600 YellowDark YellowMedium YellowLight

2400 4. Green GreenDark GreenMedium GreenLight 2200 5. Earth 2000 EarthDark EarthMedium EarthLight 1800 6. Brown 1600 BrownDark BrownMedium BrownLight 1400 7. Blue BlueDark BlueMedium BlueLight 1200 8. Aqua 1000 AquaDark AquaMedium AquaLight 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 20059. Pink ’10 PinkDark PinkMedium PinkLight

10. Berry Fig. 17 BerryDark BerryMedium BerryLight Prey time series running from 1961 to 2013. Collection of the data is restricted by the Global real ex-vessel price Predator availability of both trade and production data and the length of the corresponding11. Grey per year in 2005 time series. Considering all these limitations, the following commodities are GreyDark GreyMedium GreyLight (in USD per ton). Source: Sea Around Us/ included in the group of substitution goods: beans, dairy products, eggs, lentils,Base Colou rs 14 BaseColoursBackground BaseColoursTintedBox Black Own graphic peas, maize, meat, nuts and rice.

The global demand model reflects total global production (quantity in tons), Prey global export price (per ton in current USD) and global export value (in current Predator USD) for the above-mentioned commodities. In contrast to this, the demand model at LME-level does not require data on predatory fish and prey fish, but on domestically produced fish and imported fish.

We use the FAO database to provide data on total national domestic production (quantity in tons), exports (quantity in tons, value in current USD) and imports (quantity in tons, value in current USD) from 1976 to 2010. The FAO FishstatJ database does not contain information on production values for fish goods hence we calculate the production value as the product of production quantities and export prices. Export (import) prices for fish as well as for substitution goods are calculated by dividing export (import) values by the corresponding quantities.

Sea Around Us only provides deflated landed values (real prices) of predatory and forage fish and is hence not comparable to the nominal FAO price data for non- fish substitution goods. For this reason, data on global export values and global export quantities of fish is also taken from the FAO FishStatJ database for the period from 1976 to 2010. Global nominal export prices per ton are calculated using this data. The FAO FishstatJ database does not differentiate between forage and predatory fish.Prey For this reason, we were not able to calculate the price for these two types ofPredator fish separately. Instead, a common global export price per year for Substitution Goods

Fig. 18 250 Global expenditure per year (in billion USD 200 from 1976 to 2010). Source: FAO FishstatJ/ 150 Own graphic 100 Predator Prey 50 Substitution- goods 0 1976 1981 1986 1991 1996 2001 2006 ’10

38 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry BerryDark BerryMedium BerryLight

11. Grey GreyDark GreyMedium GreyLight

Base Colours BaseColoursBackground BaseColoursTintedBox Black

Fig. 19 120 7 Predator and Prey Production Substitution Goods Production 3 3000 Global production per year (in million tons from 1976 to 2010. 100 2500 Data from FAOSTAT,Sea Around Us/Own graphic 80 2000 Predator Prey 60 1500 Substitution- goods 40 1000 Exponentiell Substitution Goods 20 500

0 0 1975 1980 1985 1990 1995 2000 2005 2010

both types of fish is calculated. Furthermore, both wild-caught fish and fish from aquaculture are included in the FishstatJ database. Since we calculate a global price for all Preyfish commodities in this database, this price does not differ between wild-caught Predatorfish and fish from aquaculture either. Hence, the information entered into our databaseSubstitution from Goodsthese two sources relates to: Exponentiell (Substituion Goods) »»total global marine catches of predatory and forage fish (in tons), »»the global export value of predatory and forage fish (in current USD), »»the global export price for fish (per ton in current USD).

Finally, total national expenditures are calculated as the sum of production and import values minus export value.

National consumption is calculated as the sum of domestic production and imports minus exports. For some observations, the resulting value for consumption is negative. In these cases, we set the negative values equal to zero.

We assume one global consumer who has preferences regarding the quantities of each commodity she consumes. We further assume that the consumer prefers to substitute fish with another species of fish rather than with non-fish substitute goods.

To calculate global demand for a) predatory fish, b) forage fish and c) protein-rich non-fish substitute goods, we use the following yearly input data at global level: export prices, production quantities, total expenditures for all three commodities and a parameter that expresses the above-mentioned preferences of our consumer. With the given information on predatory and forage fish and substitute goods, we estimate the demand parameters for each commodity for each year from 1976 to 2010. The demand parameters indicate the estimated share each commodity has in the consumption of protein-rich food.

For the projections of global fish demand in 2050, we then calculate the mean of these demand parameters over time per LME and commodity. To match catch data from Sea Around Us at LME level with FAO country data, we estimated each country’s share in a particular LME based on the spatial overlap of the country’s coastal waters with the LME. Some countries included in the FAO data do not

Fishing for Proteins | 39 have coastal waters in any LME, either because they are landlocked or because there is no LME defined in their coastal waters. We remove countries without access from our dataset.15

Some countries have coastal waters in more than one LME. For these countries we assumed that the fraction of trade and production associated with an LME is equal to the fraction of the country’s area of coastal waters in the corresponding LME. The shares are calculated using GIS software and information on the area of exclusive economic zones (EEZs) and LMEs.

Estimating regional consumption requires the aggregation of all input data (domestic production quantities, etc.) at LME level. With regard to the regions, we group input data by the 64 LMEs provided by Sea Around Us. In each LME, the consumer has preferences regarding the quantities of each commodity she consumes. We assume that each LME’s consumer prefers to substitute fish with another species of fish rather than with a non-fish substitute good and that she differentiates between imported and domestically produced fish.

To calculate regional demand for a) imported fish, b) domestically produced fish and c) protein-rich non-fish substitute goods, we use the following yearly regional level input data: export prices, import prices, domestic production quantities, imported production quantities, total expenditures for all three commodities and two parameters that express the above-mentioned preferences per LME.

Using this information, we estimate the demand parameters for each good and each year from 1976 to 2010. The demand parameters specify the estimated share of each of the three goods in the total consumption of protein-rich foods. To project the demand for fish in 2050, we calculate the average value per LME and per good from the time series of the demand parameters.

3.4 Scenarios: Socioeconomic Pathways and Fishery Management

Fish consumption crucially depends on income that consumers in different parts of the world spend on fish and non-fish protein-rich food. It also depends on population numbers.

With regard to income and population numbers, we base our scenarios on GDP development data from the International Institute for Applied Systems Analysis (IIASA) quantification of the Intergovernmental Panel on Climate Change (IPCC): so called ‘shared socioeconomic pathways’ (SSPs) describing world futures in the 21st century.16 Five SSPs describe five scenarios of global future societal development. These SSPs are one component of the IPCC scenarios integrating future changes in climate and society to investigate climate impacts and options for mitigation and adaptation (O’Neill et al. 2015). Among those, SSP1 is deemed to describe sustainable development. We use SSP1 for our baseline scenario.

Figure 20 shows the projection of GDP in each of the scenarios. In order to cover the range of future GDP we also used scenarios SSP3 (minimum GDP scenario) and SSP5 (maximum GDP scenario).

For the base case of the SSP1 scenario, global GDP increases by a factor of 3.757. The income elasticity of demand for food is the parameter that determines what

40 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry BerryDark BerryMedium BerryLight

11. Grey GreyDark GreyMedium GreyLight

Base Colours BaseColoursBackground BaseColoursTintedBox Black

1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

Fig. 20 350 5. Earth Development of global EarthDark EarthMedium EarthLight GDP from 2010 to 2050 300 6. Brown for IPCC scenarios SSP1, BrownDark BrownMedium BrownLight SSP3 and SSP5. 250 7. Blue (in trillion USD) BlueDark BlueMedium BlueLight 200 SSP1 8. Aqua AquaDark AquaMedium AquaLight SSP3 150 SSP5 9. Pink PinkDark PinkMedium PinkLight 100 10. Berry 50 BerryDark BerryMedium BerryLight

11. Grey 0 GreyDark GreyMedium GreyLight 2010 2020 2030 2040 2050 Base Colours BaseColoursBackground BaseColoursTintedBox Black SSP1 SSP2 SSP3 SSP4 SSP5 fraction of additional income will be spent on food by 2050. The review by Cireira and Masset (2010) indicates that, while the income elasticity of fish demand is close to 1, the best global estimate for the income elasticity of food demand is 0.48.

For the base case, we assume that global expenditures for fish and non-fish protein-rich food increase by a factor of 0.48 x 3.757. In addition, we consider a very conservative scenario (SSP3) where food expenditures increase by a factor of 0.48 x 2.758 and no further technical progress is made in fishing technology; and a high-pressure scenario (SSP5) where global food demand increases by a factor of 4.534 and income elasticity is 1, which may be adequate for fish as well (Cireira and Masset 2010).

The corresponding population development is depicted in Figure 21. To calculate future fish demand, we will use the scenarios SSP1 and SSP3 in order to cover the range of possible population numbers in 2050. While SSP1 assumes the smallest population level in 2050 (8.5 billion) with a population increase factor of 1.23, SPP3 refers to the biggest population level in 2050, namely 9.95 billion with a population increase factor of 1.45.

With regard to the supply of non-fish protein-rich food, we estimate the trend between 1976 and 2010, which was characterised by an annual growth rate of 2.09%. In all scenarios we assume that this growth rate can be sustained until SSP1 SSP2 SSP3 SSP4 SSP5

Fig. 21 11 Development of the global population from 2010 to 2050 in the IPCC scenarios 10 SSP1, SSP3 and SSP5. (in billion people) 9 SSP1

SSP3 8 SSP5

6 2010 2020 2030 2040 2050

Fishing for Proteins | 41 2050. The economic parameters estimate gives an estimate of the improvement in fishing technology and corresponding reduction in fishing cost by 2.4% per year for predator fisheries and 1.1% per year for prey fisheries on average. This is in line with previous findings in the literature (Squires and Vestergaard 2013). We also assume that this trend will persist until 2050, except in our most conservative scenario in which we assume that there will be no further technical progress.

Fishery management may have a very important effect on future stock development and catches (Froese and Proelß 2010; Quaas et al. 2016; Costello et al. 2016). We consider different scenarios with respect to management effectiveness. One such scenario is that all fisheries will be managed according to the maximum sustainable yield, i.e. in such a way that the long-term catches in tons are maxim- ised (Froese and Proelß 2010).

Furthermore, we consider different scenarios where fishing effort is managed by means of total allowable catches and effort regulations (Grafton et al. 2005). In our modelling approach, we follow Quaas et al. (2016) and conceptualise management effectiveness as the fraction of external costs of fishing that are internalised in the fishermen’s decisions with regard to their catch effort. Such external costs arise if individual fishermen do not fully take into account the effects of fishing on future fishing opportunities. Economically optimal management would set the total allowable catch (TAC) so that 100% of external costs of fishing are taken into account and fisheries are regulated. No management at all would correspond to open access conditions and 0% of the external costs of fishing would be taken into account. We quantify management effectiveness based on Mora et al. (2009). We consider the case of perfect management (management effectiveness at 100%) and eight cases of imperfect management (management effectiveness from 20% to 90%, respectively). We neglect costs of management, e.g. for monitoring and enforcement, throughout the analysis.

42 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink 4. Results and Discussion The following section presents model output regard- PinkDark PinkMedium PinkLight ing fish supply and fish demand in 2050 based10. B errony the global demand model and the global predator- BerryDark BerryMedium BerryLight

prey model. 11. Grey GreyDark GreyMedium GreyLight

We assume maximum sustainable yield management for all fisheries. This allowsBase Colours us to answer the question of the extent to which the fish stocks in the global oceans BaseColoursBackground BaseColoursTintedBox Black could contribute to the supply of protein for the world population in 2050. Based on the global predator-prey model, the global surplus production model and aggregated results from the regional model, we present estimates of the maximum sustainable yield (MSY) that the global fish stocks could supply. Figure 22 shows the estimates for the three models; Table 5 shows the corresponding figures:

Fig. 22 250 Estimate of maximum sustainable yield that the global fish stocks could 200 supply according to different models. (in million tons) 150 Global MSY estimate

100

0 Predator-Prey Surplus production LME surplus model type production

The first bar shows the global catch quantity for a predator-prey model; the second and third bars show the maximum catch for Schaefer surplus models. In the second model, one global stock is assumed for the entire sea; the third model represents the total maximum sustainable yields from 64 LMEs each of which has one stock. In this third model, we analyse the potential contribution of the LMEs to meet global and regional needs for fish protein.

The diagram indicates that the global MSY is 112 million tons in the Schaefer surplus model. We calculate that total global catches reached 101 million tons in 2010.17 This means that marine resources are already almost fully exploited, which does not leave much space for an increase in catches in the future.

Mean catch Standard deviation (million tons) (million tons) Tab. 5 Yield-oriented, global predator-prey 160 91 Mean catches in 2050 according to three model Global surplus production specifications. (global Schaefer model) 112 1 Aggregate of regional surplus production model (64 LMEs) 111 3

Fishing for Proteins | 43 1. Red RedDark RedMedium RedLight

The globalBlueDar MSYk estimateBl ueMediuform the predatory-preyBlueLight 2. O ramodelnge is much larger – almost OrangeDark OrangeMedium OrangeLight Senegal 8. Aqua 160 million tons. If the management strategy focuses exclusively on yield, the South Africa total globalAquaDar catchk could thereforeAquaMedium be substantiallyAquaLight 3. Yello whigher. A growing global YellowDark YellowMedium YellowLight Peru 9. Pink population coupled with a rising demand for protein-rich food like fish could USA PinkDark PinkMedium PinkLight 4. Green justify a management strategy that is focused on maximising bioGreenDarmass.k However,GreenMediu m GreenLight 10. Berry in doing this, all other conservations objectives of a sustainable fishing industry Indonesia BerryDark BerryMedium BerryLight 5. Earth (ecological effects, ecosystem-related effects, socioeconomic consequences) would China EarthDark EarthMedium EarthLight 11. Grey have to be disregarded. A high maximum catch also comes with much higher GreyDark GreyMedium GreyLight 6. Brown France uncertainty. This is in line with studies showing that the stabilityBrownDar kof ecosystemsBrownMediu m BrownLight Germany Base Colours declines sharply if predatory populations are disproportionately reduced (Britten BaseColoursBackground BaseColoursTintedBox Black 7. Blue et al. 2014; Essington et al. 2015).

Next, we analyse how global catches depend on the management effectiveness for the three scenarios in the global bio-economic predator-prey model. Figure 23 shows the global catches in 2050 for the demand scenario based on the reference scenario with GDP increase from SSP1 and an income elasticity of food demand of 0.48. 0,5 8

0,4 In a scenario where there is perfect6 management (100% effectiveness), global 0,3 catches of predatory fish and prey fish reach levels of 21 and 116 million tons, re- 4 0,2 spectively. Together, this totals 137 million tons, significantly above current yields (Figure 23). The current level2 of fishery management effectiveness, however, 0,1 averages out at about 50-60% (Mora et al. 2009; Watson et al. 2009; Quaas et al. 0 2016). If it remains at this level,0 global catch yields in 2050 would be only slightly 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 above the current level. Current management does not sufficiently reflect on the species´ interactions. Thus, predatory fish would be heavily fished, which would alleviate predatory pressure on forage fish, enabling slightly higher total catches. Fig. 23 Compared to a scenario of perfect management, fish consumption would gradual- Global fish catches accord- ly shift towards smaller and smaller species. 50 ing to global bio-economic 3 predator-prey model based 40 on varying degrees of A decrease in management effectiveness to levels below the current status leads 2 30management effectiveness to a strong decrease in catches for both predatory and forage fish. Thus, in line (reference scenario, income 20 with the finding of Quaas et al. (2016), achieving a sufficiently high degree of growth from SSP1). management effectiveness is 1essential for sustaining fish catches while global fish 10 (in million tons) demand continues to increase. 0 0 1960 1970 Global catches1980 1990 2000 2010 1960 1970 1980 1990 2000 2010 of predatory fish For the sake of comparison, we consider the high-pressure scenario with GDP

Global catches growth based on SSP5 and a unit income elasticity of fish demand. Results are of prey fish shown in Figure 24. If management was perfect, global catches of predatory fish

30 140

25 120 100 20 80 15 60 10 40

5 20

0 0 20 30 40 50 60 70 80 90 100 20 30 40 50 60 70 80 90 100 Management effectiveness (%) Management effectiveness (%)

44 1. Red RedDark RedMedium RedLight

0,5 8

0,4 6 0,3 4 0,2 2 0,1 0 0 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

50 3 40 2 30 20 1 10

0 0 1960 1970 1980 1990 2000 2010 1960 1970 1980 1990 2000 2010

1. Red RedDark RedMedium RedLight

BlueDark BlueMedium BlueLight 2. Orange 30 140 OrangeDark OrangeMedium OrangeLight Senegal 8. Aqua South Africa 25 AquaDark AquaMedium120 AquaLight 3. Yellow YellowDark YellowMedium YellowLight Peru 9. Pink 100 20 USA PinkDark PinkMedium PinkLight 4. Green 80 GreenDark GreenMedium GreenLight 15 10. Berry Indonesia BerryDark BerryMedium 60 BerryLight 5. Earth China EarthDark EarthMedium EarthLight 10 11. Grey 40 GreyDark GreyMedium GreyLight 6. Brown France 5 BrownDark BrownMedium BrownLight Germany Base Colours 20 BaseColoursBackground BaseColoursTintedBox Black 7. Blue 0 0 20 30 40 50 60 70 80 90 100 20 30 40 50 60 70 80 90 100 Management effectiveness (%) Management effectiveness (%)

Fig. 24 and prey fish could be sustained at similar levels as in the reference scenario. Global fish catches accord- However, with decreasing management effectiveness, the catches decrease much ing to global bio-economic more strongly than in the baseline scenario. 0,5predator-prey model based 8 on varying degrees of 0,4 6 management effectiveness Results for the most conservative scenario with demand growth derived from 0,3 (high-pressure scenario, SSP3, and assuming that there will be no further technical progress in fishing, 4 0,2income growth from SSP5, are shown in Figure 25. In this case, fish catches could be sustained even if unit income elasticity of management did not improve2 compared to current levels. However, the as- 0,1 food demand). sumptions made here are not very realistic: in particular, the trend of improving 0 (in million tons) 0 1960 1970 1980 fishing1990 technology2000 2010is likely to1960 continue 1970 in the coming1980 decades,1990 which2000 would lead2010 to strongly increased fishing pressure. Nevertheless, this scenario shows that Fig. 25 economic driving forces, in particular increasing demand and technical progress Expected global fish catches in fishing technology, are central factors influencing the future fate of fisheries. according to a global bio-economic predator-prey In the last step we analyse how the LMEs within the global ocean can contribute 50 model based on varying 3 to meeting the needs of fish protein intake around the globe. To do this, we use 40 degrees of management effectiveness (low-pressure the MSY estimates for the different2 LMEs and compare them with regional fish 30 scenario, income growth consumption, based on FAO data for fish consumption in 2010 and the popu- 20 from SSP3, no technical lation in 2050 assumed from the two scenarios used (SSP1 and SSP3). These progress in fishing). 1 scenarios refer to the extremes in population development with SSP1 referring 10 (in million tons) to the smallest projected population in 2050 and SSP3 referring to the largest 0 0 1960 1970 Global catches1980 projected1990 population2000 2010 among the1960 five scenarios.1970 1980 1990 2000 2010 of predatory fish Global catches The results are depicted in Figure 26 and Figure 27. The LMEs are again col- of prey fish our-coded according to their potential to meet local needs. The red and yellow

30 140

25 120 100 20 80 15 60 10 40

5 20

0 0 20 30 40 50 60 70 80 90 100 20 30 40 50 60 70 80 90 100 Management effectiveness (%) Management effectiveness (%)

Fishing for Proteins | 45 Fig. 26 Projected MSY catches (in million tons), population size (in millions) and share of local needs (in %) that could potentially be met by LME in 2050 under ideal conditions and population development according to SSP1 scenario.

Fig. 27 Projected MSY catches (in million tons), population size (in millions) and share of local needs (in %) that could potentially be met by LME in 2050 under ideal conditions and population development according to SSP3 scenario. dots indicate that the LME is not able to meet local needs – not even according Pop. 2010 Scenario (mn) to the MSY management scenario assumed here. The green dots indicate that the > 50 LME is able to meet more than local needs with MSY management. 50–150 150–500 Compared to 2010, the results are quite similar. In 2050, 38 LMEs in the SPP1 500–1,000 scenario and 37 LMEs in the SSP3 scenario are not able to meet the needs of the >1,000 local population. The extremes of 2010 can also be found in the 2050 scenarios. In very Arctic waters, e.g. the Canadian High Arctic, North Greenland or Beaufort Fraction Sea or the Insular Pacific Hawaiian LME, fish production meets less than 1% of 0–80% their demand while the Scotian Shelf, Newfoundland-Labrador Shelf, Icelandic 80–100% Shelf and Sea and the Faroe Plateau stand out with their massive overproduction, 100–500% > 500% which leads to coverage of more than 1,000%.

Catches (mn tons) However, a few LMEs change their category. The North Sea and the Sea of Japan No Data will probably be able to meet local needs in 2050. In contrast, the Arabian Sea 0.01 – 0.60 and the California Current are likely to not be able to meet local needs in 2050. 0.61 – 1.50 1.51 – 4.00 4.01 – 8.00 Regarding the world’s potential to meet population needs in terms of fish, the 8.01 –13.15 scenarios differ. In the SSP1 scenario with the lower population in 2050, the

46 Fig. 28 Net import and net export of fish per LME in 2050 (in million tons).

Net exporter (mn ton) >5,0 5.0–1.0 1.0–0.5 <0.5 Net Importer (mn ton) world’s fish supply can meet 81% of the world’s fish needs. In the SSP3 scenario >5.0 with the higher population, only 75% of the world’s fish needs can be met by 5.0–1.0 the world’s fish supply. Similar to the situation in 2010, the missing 19% and 1.0–0.5 25%, respectively, are probably supplied by fish from aquaculture, high seas and <0.5 inland production.

In the SSP1 scenario with the slow population development, 28 LMEs experience a decrease of more than 10% in their share of MSY catches and local needs. The average decrease in these LMEs is 25%. In 20 LMEs the share increases by more than 10% with an average increase of 37%. Overall, decreases and increases seem to level out since the world’s share only decreases by 1%.

In the SSP3 scenario with the strong population development, the share reduces by more than 10% in 19 LMEs. The average decrease in these 19 LMEs is 35%. In contrast, 28 LMEs experience an increase in their share by more than 10%. On average, the increase is 43%. However, since the world’s share shows a clear decrease from 82% in 2010 to 75% in 2050, it seems that although more LMEs will probably face a stronger deficit in fish supply in the SSP1 scenario, this total deficit will be bigger in the scenario with the higher population.

Overall, Figure 26 and Figure 27 clearly show that the world’s future fish needs will probably not be met by marine catches alone. Aquaculture will be required as well, with the caveat that some aquaculture production uses wild captured fish as well (Essington et al. 2015).

In order to find out which LMEs are likely to export fish products and which LMEs are likely to depend on imports in the future, we calculate net import and export quantities per LME in 2050 from estimates of the demand model at LME-level. The distribution is shown in Figure 28.

Figure 29 and Figure 30 show the absolute deviation of fish consumption between 2010 and 2050 for all LMEs. The LMEs in Figure 29 experience a decrease in fish consumption in 2050 compared to 2010 while Figure shows the LMEs where fish consumption will increase.

According to Figure 29, waters along the East Asian coast will experience the biggest decrease in fish consumption by 2050 although these areas struggle to

Fishing for Proteins | 47 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

meet local demand with local supply (see Figure 26 and Figure 27). Since LMEs10. Berry BerryDark BerryMedium BerryLight interact on a global market, price levels for fish and substitution goods will influ- ence people´s decision on whether they will consume fish or turn to a substitute11. Grey GreyDark GreyMedium GreyLight instead. If fish prices are sufficiently high, fish will become an unaffordable good Base Colours for a substantial share of the population in the LMEs along the East Asian coast. BaseColoursBackground BaseColoursTintedBox Black These people will turn to the more affordable substitution goods and fish is going to be exported for the higher export price. This might explain the leading position of East Asian LMEs among the net exporters.

Fig. 29 Oyashio Current LMEs with decreasing fish Canadian Eastern Arctic-West consumption between East Siberian Sea 2010 and 2050 (in million tons). Aleutian Islands Laptev Sea Humboldt Current California Current Sea of Okhotsk Pacific Central-American Arabian Sea Black Sea Kara Sea Barents Sea Beaufort Sea Iberian Coastal Caribbean Sea Baltic Sea Gulf of Alaska Northern Bering - Chukchi East Bering Sea Sulu-Celebes Sea North Sea Sea of Japan / East Sea Guinea Current Mediterranean Sea Kuroshio Current Bay of Bengal East China Sea Yellow Sea South China Sea

0 5 10 15 20 25

48 1. Red RedDark RedMedium RedLight

2. Orange OrangeDark OrangeMedium OrangeLight

3. Yellow YellowDark YellowMedium YellowLight

4. Green GreenDark GreenMedium GreenLight

5. Earth EarthDark EarthMedium EarthLight

6. Brown BrownDark BrownMedium BrownLight

7. Blue BlueDark BlueMedium BlueLight

8. Aqua AquaDark AquaMedium AquaLight

9. Pink PinkDark PinkMedium PinkLight

10. Berry BerryDark BerryMedium BerryLight

11. Grey GreyDark GreyMedium GreyLight

Base Colours BaseColoursBackground BaseColoursTintedBox Black

Fig. 30 Southeast Australian Shelf LMEs with increasing Northwest Australian Shelf fish consumption between Indonesian Sea 2010 and 2050 Gulf of Thailand (in million tons). East Central Australian Shelf West-Central Australian Shelf Benguela Current North Brazil Shelf Gulf of California Canary Current Agulhas Current Newfoundland-Labrador Shelf Northeast U.S. Continental Shelf West Bering Sea Celtic-Biscay Shelf Southeast U.S. Continental Shelf Greenland Sea Southwest Australian Shelf South Brazil Shelf Gulf of Mexico Somali Coastal Current North Australian Shelf Patagonian Shelf Canadian High Arctic – North Greenland Scotian Shelf Northeast Australian Shelf-Great Barrier Insular Pacific-Hawaiian Faroe Plateau Iceland Shelf and Sea Red Sea Hudson Bay Complex New Zealand Shelf East Brazil Shelf Norwegian Sea 0 5 10 15

Fishing for Proteins | 49 Appendix

Tab. 6 National recommended intakes for fish (based on the WHO’s recommended standards).

National recommended intake Recommended Source quantity (g/week) United Kingdom 2 portions (140 g each) per week, one of which should be oily 280 Food Standards Agency (2010)18 Australia/New Zealand 2-3 servings (150g each) 375 Food Standards Australia New Zealand (2013)20 Canada At least 150g each week 150 Health Canada (2011)21 Austria 1-2 portions per week (total 150g) 150 WHO (2003)22 Germany 1-2 portions per week 150 Georgia 12,8-15g fish per day 97 WHO (2003)23 Ukraine 20g fish per day 140 WHO (2003)24 Estonia 2-3 servings per week (50g each) 150 WHO (2003)25 United States of America 8 oz per week 226 http://bit.ly/1nhRps6 Italy 100-240g per week 170 http://bit.ly/294BDQm France 100-200g per week 150 http://bit.ly/29AcfCm Ireland 2x per week 200 http://bit.ly/29Anq8D Norway 2-3x per week 250 http://bit.ly/29KT48J Denmark 2-3x per week 350 (explicit) http://bit.ly/29xPV69 Sweden 2-3x per week 250 http://bit.ly/29AVhkg Iceland 2-3x per week 250 http://bit.ly/29T6jU8 Eastern Mediterranean (Cyprus, Lebanon, Turkey, Greece, Jordan, Syria, Israel, Palestine, Egypt, Libya) 2x per week 180 http://bit.ly/29t25Cn Malaysia 2-3x per week (200-300g/week) 250 http://bit.ly/29T6leL Sri Lanka 2-3x per week (fatty fish) 250 http://bit.ly/29t2F30 Barbados 2-3x per week 250 http://bit.ly/1TbViHR Mexico 2x per week 200 http://bit.ly/29M12LC Argentina 2-3x per week ( 75-100 g each) 244 http://bit.ly/1OLY18D

Total: 31 national recommendations Ø = 204,25 204.25 x 52 = 10.6 kg/capita x year

19) to 25): from Thurstan et al. (2013)

50 Fish Supply Model Fish Supply Model a.Fish Global Supply Predator Model- Prey Model

a.WeFish Global assume Supply Predator a Model Lotka- -PreyVolterra Model type of predator-prey model (Hannesson 1983) where refers to the biomass of the predatory Wea.Fishspecies Global assume Supply and Predator a Model Lotkato the- -Prey Volterrabiomass Model type of the of preypredator species.-prey Changes model (Hannesson in biomass 1983) over timewhere ( t and refers t) areto the defined biomass as of the predatory Fish Supply Model 𝑥𝑥 speciesWe a. Global assume and Predator a Lotkato the--Prey Volterrabiomass Model type of the of preypredator species.-prey Changes model (Hannesson in biomass 1983) over timewhere ( t and refers t) areto the defined biomass as of the predatory a. Global Predator-Prey Model 𝑥𝑥 speciesWe assume and a𝑦𝑦 Lotkato the- Volterrabiomass type of the of preypredator species.-prey Changes model (Hannesson in biomass 1983) over time where (𝑥𝑥 ̇ t and refers 𝑦𝑦̇ t) areto the defined biomass as of the predatory WeFish assume Supply a LotkaModel-Volterra type of predator-prey model (Hannesson 1983) where 𝑥𝑥 refers to the biomass of the predatory species and 𝑦𝑦 to the biomass of the prey species. Changes in biomass2 over time (𝑥𝑥̇ t and 𝑦𝑦̇ t) are defined as speciesa. Global and Predator to the -biomassPrey Model of the prey species.𝑡𝑡 Changes𝑥𝑥 𝑡𝑡 in𝑥𝑥 biomass𝑡𝑡 𝑡𝑡 𝑡𝑡over 𝑡𝑡time ( t𝑥𝑥 and t) are defined as 𝑦𝑦 𝑥𝑥̇ = 𝑟𝑟 𝑥𝑥 − 𝑘𝑘 𝑥𝑥22 + 𝑎𝑎𝑎𝑎 𝑦𝑦 − 𝐻𝐻 𝑥𝑥̇ 𝑦𝑦̇ We assume a Lotka-Volterra type of predator-prey𝑡𝑡 model𝑥𝑥𝑦𝑦 𝑡𝑡𝑡𝑡 (Hannesson𝑥𝑥𝑦𝑦 𝑡𝑡𝑡𝑡 𝑡𝑡𝑡𝑡 1983)𝑡𝑡𝑡𝑡 𝑡𝑡 where 𝑥𝑥 refers to the biomass of the predatory 𝑦𝑦 𝑥𝑥𝑦𝑦̇̇ = 𝑟𝑟 𝑥𝑥𝑦𝑦 − 𝑘𝑘 𝑥𝑥𝑦𝑦2 +−𝑎𝑎𝑏𝑏𝑎𝑎𝑏𝑏 𝑦𝑦 − 𝐻𝐻𝐿𝐿 𝑥𝑥̇ 𝑦𝑦̇ Here, and 𝑦𝑦 describe the stock size in year and denote2 the intrinsic growth𝑥𝑥̇ rates,𝑦𝑦̇ kx and ky capture density- species and to the biomass of the prey species.𝑡𝑡 Changes𝑥𝑥𝑦𝑦 𝑡𝑡𝑡𝑡 in𝑥𝑥𝑦𝑦 biomass𝑡𝑡𝑡𝑡 𝑡𝑡𝑡𝑡 𝑡𝑡𝑡𝑡over 𝑡𝑡time ( t and t) are defined as 𝑥𝑥𝑦𝑦̇̇ = 𝑟𝑟 𝑥𝑥𝑦𝑦 − 𝑘𝑘 𝑥𝑥𝑦𝑦22 +−𝑎𝑎𝑏𝑏𝑎𝑎𝑏𝑏 𝑦𝑦 − 𝐻𝐻𝐿𝐿 Here, dependence and for describepredator theand stock prey sizespecies, in year respectively, and and denote and the intrinsicdenote thegrowth interaction𝑥𝑥 rates, k parameters.x and ky capture An increasedensity- in the 𝑡𝑡 𝑡𝑡 𝑡𝑡𝑥𝑥 𝑥𝑥𝑦𝑦 𝑡𝑡𝑡𝑡𝑦𝑦 𝑥𝑥𝑦𝑦 𝑡𝑡2𝑡𝑡 𝑡𝑡𝑡𝑡 𝑡𝑡𝑡𝑡 𝑡𝑡 𝑥𝑥𝑦𝑦̇̇𝑡𝑡 = 𝑟𝑟𝑥𝑥𝑥𝑥𝑦𝑦𝑡𝑡 − 𝑘𝑘𝑥𝑥𝑥𝑥𝑦𝑦𝑡𝑡2 +−𝑎𝑎𝑏𝑏𝑎𝑎𝑏𝑏𝑡𝑡𝑦𝑦𝑡𝑡 − 𝐻𝐻𝐿𝐿𝑡𝑡 Here, dependence 𝑥𝑥 and for𝑦𝑦𝑦𝑦 describepredator theand stock prey sizespecies, in year respectively, 𝑡𝑡, 𝑟𝑟 and 𝑟𝑟 and denote and the intrinsicdenote thegrowth interaction𝑥𝑥̇ rates,𝑦𝑦̇ k parameters.x and ky capture An increasedensity- in the biomass 𝑡𝑡 of prey𝑡𝑡 has a positive impact on the development𝑥𝑥̇ 𝑡𝑡𝑥𝑥= 𝑟𝑟𝑦𝑦𝑥𝑥𝑡𝑡𝑦𝑦− of𝑘𝑘𝑦𝑦 the𝑥𝑥 𝑡𝑡2 + predator’s𝑎𝑎𝑎𝑎 𝑡𝑡𝑦𝑦𝑡𝑡 − 𝐻𝐻 biomass𝑡𝑡 which is why the interaction term 𝑥𝑥 𝑦𝑦 𝑡𝑡𝑦𝑦, ̇𝑟𝑟𝑡𝑡 = 𝑟𝑟𝑦𝑦𝑦𝑦𝑟𝑟𝑡𝑡 − 𝑘𝑘𝑦𝑦𝑎𝑎𝑦𝑦𝑡𝑡 − 𝑏𝑏𝑏𝑏𝑏𝑏𝑡𝑡𝑦𝑦𝑡𝑡 − 𝐿𝐿𝑡𝑡 biomassHere,isdependence positive. andof However,prey for describepredatorhas aan positive increase theand stock prey impact in sizespecies, the on inbiomass theyear respectively, development 𝑦𝑦 ̇ of= andthe𝑟𝑟 𝑦𝑦 predator −and denote of𝑘𝑘 the𝑦𝑦 2and has −predator’sthe𝑏𝑏 𝑏𝑏a intrinsicdenote 𝑦𝑦negative− 𝐿𝐿 biomass thegrowth impact interaction rates,which on the kis parameters.x whyanddevelopment kthey capture interaction An of increasedensity the term prey’s- in the Here, 𝑡𝑡 and 𝑡𝑡 describe the stock size in year 𝑥𝑥 and 𝑦𝑦 denote the intrinsic growth rates, kx and ky capture density- 𝑡𝑡 𝑡𝑡 𝑥𝑥 𝑦𝑦 𝑡𝑡, 𝑟𝑟𝑡𝑡 𝑥𝑥 𝑟𝑟𝑡𝑡 𝑥𝑥𝑎𝑎 𝑡𝑡 𝑏𝑏𝑡𝑡 𝑡𝑡 𝑡𝑡 𝑎𝑎𝑥𝑥 𝑦𝑦 biomassisdependence positive. ofwhich However,prey for is predatorhas why aan thepositive increase and prey’s prey impact interaction in species, the on biomass the respectively,term development𝑥𝑥 ̇of= the𝑟𝑟 𝑥𝑥 ispredator −andnegative. of𝑘𝑘 the𝑥𝑥 2and has +predator’s 𝑎𝑎H 𝑎𝑎at denote and𝑦𝑦negative− L𝐻𝐻 t biomassdenote the impact interaction harvest which on the is levelsparameters. whydevelopment thefor predatorinteraction An of increase the and term prey’s pr eyin the dependence 𝑡𝑡 for𝑡𝑡 predator and prey species, respectively,𝑥𝑥 𝑦𝑦 and and denote the interaction parameters. An increase in the𝑡𝑡 𝑡𝑡 biomass𝑥𝑥𝑡𝑡 of prey𝑦𝑦𝑡𝑡 has a positive impact on the development𝑡𝑡,𝑦𝑦𝑟𝑟̇𝑡𝑡𝑥𝑥 = 𝑟𝑟𝑦𝑦𝑦𝑦𝑟𝑟𝑡𝑡𝑦𝑦− of 𝑘𝑘 𝑦𝑦the𝑎𝑎𝑦𝑦𝑡𝑡 −predator’s𝑏𝑏𝑏𝑏𝑏𝑏𝑡𝑡𝑦𝑦𝑡𝑡 − 𝐿𝐿 biomass𝑡𝑡 which is why the interaction term 𝑎𝑎𝑥𝑥 𝑦𝑦 biomassisspeciesHere, positive.𝑥𝑥 respectively. which and However,𝑦𝑦 is describe why anThus, the increase theprey’s the stock change interaction in sizethe inbiomassin biomassyear term 𝑡𝑡, 𝑟𝑟of isandthe determined is predator𝑟𝑟 negative. denote has bythe H thea t intrinsic andnegative biological Lt denote growth impact growth harvestrates, on ofthe k xlevelsthe anddevelopment stock, k fory capture predator minus ofdensity catches, the and prey’s -prey plus biomass of prey has a positive impact on the development𝑡𝑡 𝑡𝑡 of the𝑎𝑎 predator’s𝑏𝑏 biomass which is why the interaction term 𝑎𝑎𝑥𝑥𝑡𝑡𝑦𝑦𝑡𝑡 biomassisspecies positive. respectively. which However, is why anThus, the increase prey’s the change interaction in the inbiomass biomass term of𝑏𝑏𝑥𝑥 isthe𝑦𝑦 determined ispredator negative. has by H theat andnegative biological Lt denote impact growth harvest on ofthe levelsthe development stock, for predator minus of catches, the and prey’s prey plus isordependence positive.minus the However, interaction for predator an term.increase and prey in species,the biomass respectively, of the𝑡𝑡 𝑡𝑡 predator and 𝑎𝑎 and has 𝑏𝑏 a denote negative the impact interaction on the parameters. development An increaseof the prey’s in the 𝑡𝑡 𝑡𝑡 𝑡𝑡 𝑡𝑡 𝑏𝑏𝑥𝑥𝑥𝑥 𝑦𝑦 𝑦𝑦 𝑎𝑎𝑥𝑥𝑡𝑡𝑦𝑦𝑡𝑡 biomassWeorspeciesbiomass minus assume𝑥𝑥 respectively. whichthe of generalisedpreyinteraction𝑦𝑦 is haswhy Thus, athe positiveterm. Schaefer prey’s the changeimpact interaction harvest onin biomasstheproduction term development𝑡𝑡, 𝑟𝑟 is determinedfunctions, is𝑟𝑟 negative. of the predator’sby H thet and biological L tbiomass denote growth harvestwhich of is levelsthe why stock, the for interactionpredator minus catches, and term pr ey plus𝑎𝑎 𝑥𝑥 𝑦𝑦 biomass which is why the prey’s interaction term 𝑏𝑏𝑥𝑥𝑡𝑡𝑦𝑦𝑡𝑡 is negative. Ht and Lt denote harvest levels for predator and prey orspeciesWeis minus positive.assume respectively. the However, generalisedinteraction Thus, an term. Schaeferincrease the change inharvest the in biomass biomassproduction of is 𝑡𝑡the determined𝑡𝑡functions, predator𝑎𝑎 has by 𝑏𝑏 thea negative biological impact growth on ofthe the development stock, minus of catches,the prey’s plus species respectively. Thus, the change in biomass is𝑦𝑦 determined by the biological growth of the stock, minus catches, plus𝑡𝑡 𝑡𝑡 orWe minus assume the generalisedinteraction term. Schaefer harvest production𝑏𝑏𝑥𝑥𝑡𝑡 𝑡𝑡functions, 𝑎𝑎𝑥𝑥 𝑦𝑦 orbiomass minus the which interaction is why the term. prey’s interaction term 𝑏𝑏𝑥𝑥 𝑦𝑦 is negative. H t and Lt denote harvest levels for predator and prey We assume generalised Schaefer harvest production functions, Χ 𝑥𝑥 Wespecies assume respectively. generalised Thus, Schaefer the change harvest in biomassproduction is𝑡𝑡 determinedfunctions,𝑡𝑡 𝑡𝑡 𝑡𝑡 by𝑥𝑥 𝑥𝑥the biological growth of the stock, minus catches, plus or minus the interaction term. 𝑏𝑏𝑥𝑥 𝑦𝑦 𝐻𝐻 = 𝑞𝑞𝑞𝑞ΧΧ𝑦𝑦𝑥𝑥𝐸𝐸 𝑡𝑡𝑡𝑡 𝑡𝑡𝑡𝑡 𝑦𝑦𝑥𝑥𝑦𝑦𝑥𝑥 for We predator assume and generalised prey species Schaefer respectively. harvest productionHere, and functions, 𝐻𝐻𝐿𝐿 denote== 𝑑𝑑𝑞𝑞𝑑𝑑𝑞𝑞ΧΧ 𝑦𝑦𝑥𝑥catchability𝐸𝐸 coefficients and and  denote the stock 𝑡𝑡𝑡𝑡 𝑡𝑡𝑡𝑡 𝑦𝑦𝑥𝑥𝑦𝑦𝑥𝑥  x y 𝐻𝐻𝐿𝐿 == 𝑑𝑑𝑞𝑞𝑑𝑑𝑞𝑞ΧΧ𝑦𝑦𝑥𝑥𝐸𝐸  elasticitiesfor predator of and output, prey whichspecies are respectively. allowed to differ Here, from and one. 𝑡𝑡 𝑡𝑡 denoteExt and𝑡𝑡Χ 𝑥𝑥 catchabilityE𝑦𝑦𝑥𝑥yt𝑦𝑦𝑥𝑥 are effort coefficients levels directed and at predatory and y and denote prey the fish stock 𝐻𝐻 = 𝑞𝑞𝑞𝑞𝑡𝑡Χ𝑦𝑦 𝐸𝐸 x for predator and prey species respectively. Here, 𝑐𝑐 and 𝑑𝑑𝐿𝐿 𝑡𝑡denote= 𝑑𝑑𝑑𝑑𝑡𝑡 catchability𝐸𝐸𝑥𝑥𝑥𝑥 coefficients and  and  denote the stock respectively.elasticities of output, which are allowed to differ from one.𝐻𝐻𝑡𝑡 E=xt 𝑞𝑞and𝑞𝑞𝑡𝑡Χ𝑦𝑦 𝐸𝐸E𝑦𝑦yt𝑦𝑦 are effort levels directed at xpredatoryy and prey fish 𝐿𝐿𝑡𝑡 = 𝑑𝑑𝑑𝑑𝑡𝑡 𝐸𝐸𝑦𝑦𝑦𝑦 elasticitiesfor predator of and output, prey whichspecies are respectively. allowed to differ Here, from 𝑐𝑐 and one. 𝑑𝑑 denoteExt andΧ 𝑥𝑥 catchabilityEyt are effort coefficients levels directed and at predatory and  and denote prey the fish stock Assumingrespectively.for predator that and the prey marginal species effort respectively. costs are Here,constant and for 𝐿𝐿both 𝑡𝑡denote= 𝑑𝑑fisheries,𝑑𝑑 𝑡𝑡 catchability𝐸𝐸𝑥𝑥𝑥𝑥 and allowing coefficients for a andtrend  ofx declining and y costsdenote due the to stock 𝑐𝑐 𝑑𝑑𝐻𝐻 = 𝑞𝑞𝑞𝑞Χ𝑦𝑦 𝐸𝐸 x y technicalrespectively.Assumingelasticities progress thatof output, the marginal(at which rates areveffortx and allowed costs vy) , fishingtoare differ constant costs from canforone. bothbe E xtwritten fisheries,and E asyt are and effort allowing levels for directed a trend at of predatory declining andcosts prey due fish to elasticities of output, which are allowed to differ from𝑐𝑐 one.𝑑𝑑 𝑡𝑡 Ext and𝑡𝑡 E𝑦𝑦yt𝑦𝑦 are effort levels directed at predatory and prey fish Assumingrespectively.for predator that and the prey marginal species effort respectively. costs are constantHere, and for 𝐿𝐿both denote= 𝑑𝑑fisheries,𝑑𝑑 catchability𝐸𝐸 and allowing coefficients for a andtrend  of decliningand  denotecosts due the tostock respectively.technical progress (at rates vx and vy) , fishing costs𝑐𝑐 can𝑑𝑑 be written as x y Assumingelasticities that of output,the marginal which effortare allowed costs areto differ constant from for one. both Ext fisheries,and Eyt are and effort allowing levels for directed a trend at of predatory declining and costs prey due fish to Assumingtechnical progress that the marginal(at rates veffortx and costs vy) , fishingare constant costs canfor bothbe written fisheries, as and allowing for a trend of declining costs due to 𝑐𝑐 𝑑𝑑 𝑥𝑥 technicalrespectively. progress (at rates vx and vy) , fishing costs can be written as −Χ technical progress (at rates vx and vy) , fishing costs𝑥𝑥 𝑡𝑡 can𝑡𝑡 be written𝑥𝑥 as𝑥𝑥 𝑡𝑡 𝑡𝑡 Assuming that the marginal effort costs are constant𝐶𝐶 (𝐻𝐻 , 𝑥𝑥 for) =both𝑒𝑒𝑒𝑒𝑒𝑒 fisheries,(𝑐𝑐 − 𝑣𝑣 𝑡𝑡 )and𝑥𝑥−ΧΧ 𝑥𝑥allowing𝑥𝑥𝐻𝐻 for a trend of declining costs due to 𝑥𝑥𝑦𝑦 𝑡𝑡 𝑡𝑡 𝑦𝑦𝑥𝑥 𝑥𝑥𝑦𝑦 𝑡𝑡 𝑡𝑡 −Χ𝑥𝑥𝑥𝑥 As technical discussed progress in the (atmain rates text, v xthe and biological vy) , fishing parameters𝐶𝐶 costs(𝐻𝐻𝐿𝐿 ,, 𝑦𝑦𝑥𝑥can) = arebe𝑒𝑒𝑒𝑒 writtenestimated𝑒𝑒(𝑐𝑐 − as𝑣𝑣 𝑡𝑡 )using𝑥𝑥𝑦𝑦−Χ 𝐻𝐻 𝐿𝐿the Catch-MSY method developed by Martell 𝑥𝑥𝑦𝑦 𝑡𝑡 𝑡𝑡 𝑦𝑦𝑥𝑥 𝑥𝑥𝑦𝑦 𝑡𝑡 𝑡𝑡 𝐶𝐶 (𝐻𝐻𝐿𝐿 ,,𝑦𝑦𝑥𝑥 ) = 𝑒𝑒𝑒𝑒𝑒𝑒(𝑐𝑐 − 𝑣𝑣 𝑡𝑡)𝑥𝑥𝑦𝑦−ΧΧ𝑥𝑥𝑥𝑥𝐻𝐻𝐿𝐿 As discussed in the main text, the biological parameters are estimated( )using−Χ𝑥𝑥 the Catch-MSY method developed by Martell and Froese (Martell and Froese 2013). This method𝑥𝑥𝑦𝑦 𝑡𝑡 allows𝑡𝑡 biological𝑦𝑦𝑥𝑥 𝑥𝑥𝑦𝑦 parameters𝑡𝑡 𝑡𝑡 based on catch data to be estimated. It 𝐶𝐶𝑥𝑥(𝐻𝐻𝐿𝐿𝑡𝑡,,𝑦𝑦𝑥𝑥𝑡𝑡) = 𝑒𝑒𝑒𝑒𝑒𝑒(𝑐𝑐𝑥𝑥 − 𝑣𝑣𝑥𝑥𝑡𝑡)𝑥𝑥𝑦𝑦𝑡𝑡−Χ𝑥𝑥𝐻𝐻𝐿𝐿𝑡𝑡 As discussed in the main text, the biological parameters are estimated using𝑥𝑥 the Catch-MSY method developed by Martell requiresand Froese time (Martell series andof catch Froese data 2013). and prior This ranges method𝐶𝐶𝑦𝑦(𝐻𝐻 𝑡𝑡for, 𝑥𝑥allows𝑡𝑡 the) = parameter𝑒𝑒 biological𝑒𝑒𝑒𝑒(𝑐𝑐𝑦𝑦 − 𝑣𝑣 𝑦𝑦 valuesparameters𝑡𝑡)𝑥𝑥𝑡𝑡−Χ 𝐻𝐻as𝑡𝑡 well based as possible on catch ranges data to of be stock estimated. sizes from It 𝐶𝐶𝑦𝑦(𝐿𝐿𝑡𝑡, 𝑦𝑦𝑡𝑡) = 𝑒𝑒𝑒𝑒𝑒𝑒(𝑐𝑐𝑦𝑦 − 𝑣𝑣𝑦𝑦𝑡𝑡)𝑦𝑦𝑡𝑡−Χ𝑥𝑥 𝐿𝐿𝑡𝑡 therequiresandAs discussed initialFroese time and (Martell seriinfinal thees period. mainandof catch Froese text, After data the specifying 2013). andbiological prior This initial ranges methodparameters𝐶𝐶 𝑥𝑥stock(𝐿𝐿 for𝑡𝑡, 𝑦𝑦allows 𝑡𝑡sizes)the= are 𝑒𝑒parameter 𝑒𝑒biological andestimated𝑒𝑒(𝑐𝑐 𝑥𝑥limits− 𝑣𝑣 𝑥𝑥 values parameters𝑡𝑡for )using𝑦𝑦 the as𝐿𝐿thefinal𝑡𝑡 well Catch basedstock as -possible MSY size,on catch methoda parameter ranges data developed to of be stockset estimated. is sizesrabyndomly Martell from It As discussed in the main text, the biological parameters are estimated using𝑡𝑡 𝑥𝑥 the Catch-MSY method developed by Martell and Froese (Martell and Froese 2013). This method𝐶𝐶 (𝐻𝐻 ,allows𝑥𝑥 ) = 𝑒𝑒 biological𝑒𝑒𝑒𝑒(𝑐𝑐 − 𝑣𝑣 parameters𝑡𝑡)𝑥𝑥 −Χ 𝐻𝐻 based on catch data to be estimated. It anddrawntherequires initialFroese from time and the(Martell serifinal pries orperiod. andofparameter catch Froese After data specifying distribution.2013). and prior This initial rangesThen,method 𝑦𝑦stock the 𝑡𝑡for allows 𝑡𝑡sizesunderlyingthe parameter biologicaland 𝑦𝑦limits fish 𝑦𝑦 values parameterssupplyfor 𝑡𝑡the as modelfinal𝑡𝑡 well basedstock isas used possible size,on tocatch a calculate parameter ranges data to theof be stockset biomass estimated. is sizesrandomly from It requiresAs discussed time seri in esthe of main catch text, data the and biological prior ranges parameters𝐶𝐶 (𝐿𝐿 for, 𝑦𝑦 the) = are parameter𝑒𝑒𝑒𝑒 estimated𝑒𝑒(𝑐𝑐 − 𝑣𝑣 values𝑡𝑡 )using𝑦𝑦 𝐿𝐿asthe well Catch as -possibleMSY method ranges developed of stock bysizes Martell from requirescorrespondingdrawnthe initial from time and the serifinal to pri esthe orperiod. of parameterlevel catch Afterof dataharvest specifyingdistribution. and given prior initial the rangesThen, parameter stock the for sizesunderlyingthe set.parameter and If thislimits fish biomass values supplyfor the as model finalis wellin stocka isreasonableas used possible size, to a calculate parameter range,ranges thetheof stockset parameterbiomass is sizesrandomly setfrom is theand initial Froese and (Martell final period. and Froese After specifying 2013). This initial method stock allows sizes biologicaland limits parametersfor the final basedstock size,on catch a parameter data to be set estimated. is randomly It stored.thecorrespondingdrawn initial from In andour the analysis,final to pri the orperiod. parameterlevel we After ofrep harvesteat specifyingdistribution. this given procedure initial the Then, parameter stock 10,000,000 the sizesunderlying set. and times If thislimits fish for biomass supplyforeach the LME. model finalis in Westocka isreasonable useused size, samples to a calculate parameter range, of 1,000 thethe set parameterbiomass randomly is randomly setpicked is drawnrequires from time the seri priores parameterof catch data distribution. and prior Then,ranges the for underlyingthe parameter fish valuessupply asmodel well isas used possible to calculate ranges ofthe stock biomass sizes from acceptedstored.drawncorresponding from In parameterour the analysis, to pri theor valuesparameterlevel we ofrep in harvest eatour distribution. thismodel given procedure computations the Then, parameter 10,000,000 the underlying to computeset. times If this fish meanfor biomass supply each estimates LME. model is in Wea isandreasonable useused confidence samples to calculate range, ofintervals. 1,000 thethe parameterbiomass randomly Thus, all set picked is correspondingthe initial and tofinal the period. level of After harvest specifying given initialthe parameter stock sizes set. and If thislimits biomass for the finalis in stocka reasonable size, a parameter range, the set parameter is randomly set is resultsacceptedstored.corresponding reportedIn parameterour analysis, to below the valueslevel arewe ofbasedrep in harvest eatour onthismodel averagesgiven procedure computations the andparameter 10,000,000 standard to computeset. deviationstimes If this meanfor biomass each obtained estimates LME. is in from Wea andreasonable use1,000 confidence samples separate range, ofintervals. 1,000model the parameter randomlyruns Thus, each. all set picked In is stored.drawn Infrom our the analysis, prior parameter we repeat distribution. this procedure Then, 10,000,000 the underlying times fish for supply each LME. model We is useduse samples to calculate of 1,000 the biomass randomly picked resultsacceptedstored.the global reportedIn parameter ourpredator analysis, below-prey values arewe model, basedrep in eatour we onthismodel extend averages procedure computations the andapproach 10,000,000 standard to by compute Martell deviationstimes meanforand each obtainedFroese estimates LME. (2013) from We and use1,000 and confidence samples determine separate ofintervals. 1,000modelparameter randomlyruns Thus, values each. all picked Infor resultsacceptedthecorresponding Lotkaglobal reported -parameter Volterrapredator tobelow predatorthe-prey values levelare model, based - ofpreyin harvestour wemodel. onmodel extend averages given In computations eachthe the andapproach parameter case, sta ndardinitial to by computeset. Martell parameterdeviations If this meanand biomass sets obtainedFroese estimates to is be in(2013) from testeda reasonableand 1,000 and confidenceare determine randomlyseparate range, intervals. modelparameterthedrawn parameter runs fromThus, values each.a all uniformset Inforis acceptedstored. In parameter our analysis, values we repin oureat modelthis procedure computations 10,000,000 to compute times meanfor each estimates LME. We and use confidence samples of intervals. 1,000 randomly Thus, all picked distribution.resultsthe Lotkaglobal reported- Volterrapredator Biological below predator-prey parametersare model, based-prey wemodel. on are extend averages accepted In eachthe andapproach case,if final sta ndard initialbiomasses by Martell parameterdeviations fall and between sets obtainedFroese to be a(2013) from minimumtested 1,000 and are anddetermine randomlyseparate two-thirds modelparameterdrawn of their runsfrom valuesequilibrium each.a uniform Infor resultsaccepted reported parameter below values are based in our on model averages computations and standard to compute deviations mean obtained estimates from and 1,000 confidence separate intervals. model runsThus, each. all In valuedistribution.the Lotkaglobal without- Volterrapredator Biological fishing. predator-prey parameters model,-prey wemodel. are extend accepted In eachthe approach case,if final initialbiomasses by Martellparameter fall and between sets Froese to be a(2013) minimumtested and are anddetermine randomly two-thirds parameterdrawn of their from valuesequilibrium a uniform for theresults global reported predator below-prey are model, based we on extend averages the approachand standard by Martell deviations and obtainedFroese (2013) from 1,000 and determine separate model parameter runs each.values In for Economicvaluedistribution.the Lotka without- Volterratheory Biological fishing. predicts predator parameters a- preypositive model. are relationship accepted In each case,ifbet finalween initialbiomasses fish parameter stock fall biomass between sets to and be a minimummarkettested aresupply and randomly two of fish-thirds drawn(or ofno their fromrelationship equilibrium a uniform at thethe Lotka global-Volterra predator predator-prey model,-prey model.we extend In each the approach case, initial by parameterMartell and sets Froese to be (2013) tested and are determine randomly parameter drawn from values a uniform for Economicdistribution.allvalue in the without case theory Biological fishing.of apredicts pure parameters schooling a positive fishery),are relationship accepted and thus ifbet final weena negative biomasses fish stock relationship fall biomass between between and a minimummarket stock supply biomassand two of fish- thirdsand (or the ofno fishtheir relationship price. equilibrium We at distribution.the Lotka-Volterra Biological predator parameters-prey model. are accepted In each ifcase, final initialbiomasses parameter fall between sets to be a minimumtested are and randomly two-thirds drawn of fromtheir equilibriuma uniform valueuseallEconomic in price the without case data theory fishing. ofin aeachpredicts pure run schooling afor positive a tested fishery), relationship parameter and thus bet set weena to negative ch fisheck stockwhether relationship biomass this requirementbetween and market stock is supply met.biomass Specifically, of fish and (or the no wefish relationship assume price. We that at valuedistribution. without Biologicalfishing. parameters are accepted if final biomasses fall between a minimum and two-thirds of their equilibrium Economictheuseall in openprice the case accessdata theory ofin aconditionseachpredicts pure run schooling afor positivepHt a =tested C fishery),x (Hrelationshipt ,parameter xt) andand pthus Ltbet set= weenCa to ynegative(L cht, fishyeckt) should stockwhether relationship biomasshold this for requirementbetween theand period market stock between is supply met.biomass Specifically, of1976 fish and and (or the no2000 wefish relationship assume (Quaasprice. We thatet at Economicvalue without theory fishing. predicts a positive relationship between fish stock biomass and market supply of fish (or no relationship at theal.alluse in2012). openprice the case access dataWe use ofin aconditionseach observedpure run schooling for ppricesHt a =tested C fishery), xfrom(Ht ,parameter xSeat) andand Around pthusLt set= CaUs to ynegative(L chandt, yeckt) theshould whether relationship stock hold estimates this for requirementbetween the periodfrom stock the between is test met.biomass runs Specifically, 1976 in and the and the Martell/Froese 2000 wefish assume (Quaasprice. We thatet allEconomic in the case theory of a predictspure schooling a positive fishery), relationship and thus bet weena negative fish stock relationship biomass between and market stock supply biomass of fish and (or the no fishrelationship price. We at procedureal.usethe 2012). openprice accessdataWe to estimate use in conditionseach observed arun log for - plinearisedpricesHt a =tested C xfrom(H t,OLSparameter xSeat) and regression Around pLt set= CUs toy (Lof chandt, theyeckt) the shouldopen whether stock access hold estimates this for conditions requirement the periodfrom usingthe between is test met. the runs costSpecifically, 1976 in functions the and Martell/Froese 2000 we and assume (Quaas including thatet useall inprice the datacase in of each a pure run schooling for a tested fishery), parameter and thus set ato negative check whether relationship this requirementbetween stock is met.biomass Specifically, and the fishwe assumeprice. We that theprocedureal. 2012). timeopen trends accessWe to estimate use below. conditions observed aWe log accept - plinearisedpricesHt = C a xfrom (Hparameter t,OLS xSeat) and regression Around setpLt =if CitUs ygives (Lof andt, theyt ) non the shouldopen -stocknegative access hold estimates forestimat conditions the periodfromes for usingthe betweenboth test the X runsx costand 1976 in Xfunctions ythe .and Parameter Martell/Froese 2000 and (Quaas includingsets that et theuse open price access data in conditions each run forpHt a= testedCx(Ht, parameterxt) and pLt set= C toy(L cht, yeckt) should whether hold this for requirement the period betweenis met. Specifically, 1976 and 2000we assume (Quaas that et theprocedureal.do 2012). nottime pass trends We to this estimate use below. test observed are aWe logrejected. accept- linearisedprices aOtherwise, from parameter OLS Sea regression Around we set use if itUs the gives of and resultingthe nonthe open -stocknegative information access estimates estimat conditions on from esthe for relationshipusingthe both test the X runsx costand between in Xfunctions ythe. Parameter Martell/Froese price and and includingsets stock that al.the 2012). open We access use conditionsobserved pricespHt = C fromx(Ht, Seaxt) and Around pLt = CUsy(L andt, yt )the should stock hold estimates for the periodfrom the between test runs 1976 in theand Martell/Froese 2000 (Quaas et theprocedure time trends to estimate below. aWe log accept-linearised a parameter OLS regression set if it gives of the non open-negative access estimat conditionses for using both the Xx costand Xfunctionsy. Parameter and includingsets that procedurebiomassdoal. not 2012). pass to to Weobtain this estimate use test an observed are estimate a logrejected.-linearised prices for Otherwise,economic from OLS Sea regressionparameter Aroundwe use Usthe ofvalues. and resultingthe theopen stock information access estimates conditions on from the relationshiptheusing test the runs cost between in functionsthe Martell/Froese price and and including stock the biomassdo nottime pass trendsto obtain this below. test an are estimate We rejected. accept for aOtherwise,economic parameter parameter we set use if it the gives values. resulting non -negative information estimat on esthe for relationship both Xx and between Xy. Parameter price and sets stock that theprocedure time trends to estimate below. We a log accept-linearised a parameter OLS regression set if it gives of the non open-negative access estimat conditionses for using both the Xx costand functionsXy. Parameter and including sets that The biomassdo not means pass to obtain andthis standardtest an are estimate rejected. deviations for Otherwise,economic for the 1,000parameter we use parameter the values. resulting sets usedinformation in the computationson the relationship are given between in the price table and below stock dothe not time pass trends this testbelow. are We rejected. accept Otherwise, a parameter we set use if it the gives resulting non-negative information estimat on esthe for relationship both Xx and between Xy. Parameter price and sets stock that The biomass means to obtainand standard an estimate deviations for economic for the 1,000parameter parameter values. sets used in the computations are given in the table below biomassdo not pass torx obtain this test anry estimateare rejected.kx for economic Otherwise,ky parameter wea use the values. resultingb informationcx oncy the relationshipvx betweenvy price xand stocky The means and standard deviations for the 1,000 parameter sets used in the computations are given in the table below Mean biomass 1.44rtox obtain 2.24rany estimate 0.044kx for economic0.0096ky parameter 0.0046a values. 0.014b 49.14cx 24.27cy 0.024vx 0.011vy 0.24 x 0.32y The meansrx and standardry devikxations fork ythe 1,000a parameterb sets usedcx in the computationscy vx are givenvy in the table below StdMeanThe means 0.561.44 and standard0.742.24 devi0.0230.044ations for0.00620.0096 the 1,000 0.00200.0046 parameter 0.00770.014 sets used 10.4549.14 in the computations7.8224.27 0.0050.024 are given0.0030.011 in the table0.200.24x below0.200.32 y StdMean 0.561.44rx 0.742.24ry 0.0230.044kx 0.00620.0096ky 0.00200.0046a 0.00770.014b 10.4549.14cx 7.8224.27cy 0.0050.024vx 0.0030.011vy 0.200.24 x 0.200.32y The meansrx and standardry devikx ations forky the 1,000a parameterb sets usedcx in the computationscy vx are givenvy in the table below y StdMean 0.561.44 0.742.24 0.0230.044 0.00620.0096 0.00200.0046 0.00770.014 10.4549.14 7.8224.27 0.0050.024 0.0030.011 0.200.24x 0.200.32 Meanb. Fish Supply1.44rx Model2.24ry at LME0.044kx- Level 0.0096ky 0.0046a 0.014b 49.14cx c24.27y v0.024x v0.011y 0.24 0.32y Std 0.56 0.74 0.023 0.0062 0.0020 0.0077 10.45 7.82 0.005 0.003 0.20x 0.20 Stdb. InMean theFish total Supply0.56 1.44fish supply Model0.742.24 model, at LME0.023 0.044the-Level change 0.00620.0096 of biomass 0.00200.0046 over time0.00770.014 is defined 10.4549.14 as 24.277.82 0.0240.005 0.0110.003 0.240.20 0.320.20

b.InStd theFish total Supply 0.56fish supply Model0.74 model, at LME 0.023the-Level change 0.0062 of biomass 0.0020 over time0.0077 is defined 10.45 as 7.82 0.005 0.003 0.20 0.20 b.In theFish total Supply fish supply Model model, at LME the-Level change of biomass over time is defined as b. Fish Supply Model at LME-Level In the total fish supply model, the change of biomass over time is defined2 as Inb. the Fish total Supply fish supply Model model, at LME the-Level change of biomass𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 over𝑙𝑙𝑙𝑙𝑙𝑙 𝑡𝑡time 𝑙𝑙𝑙𝑙𝑙𝑙is defined𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 as𝑙𝑙𝑙𝑙𝑙𝑙 ,𝑡𝑡 𝑥𝑥̇ = 𝑟𝑟 𝑥𝑥 − 𝑘𝑘 𝑥𝑥 − 𝐻𝐻 x lme,t describes the stock size in the large marine ecosystem lme in year2 , lme describes the intrinsic growth rate of the In the total fish supply model, the change of biomass𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 over𝑙𝑙𝑙𝑙𝑙𝑙 𝑡𝑡 time 𝑙𝑙𝑙𝑙𝑙𝑙is defined𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 as𝑙𝑙𝑙𝑙𝑙𝑙 ,𝑡𝑡 𝑥𝑥̇ = 𝑟𝑟 𝑥𝑥 − 𝑘𝑘 𝑥𝑥2 − 𝐻𝐻 xstock,lme,t describes klme is a measurethe stock of size density in the dependence large marine𝑙𝑙𝑙𝑙𝑙𝑙 and ,ecosystem𝑡𝑡 Hlmt,t𝑙𝑙𝑙𝑙𝑙𝑙 describes𝑡𝑡 lme𝑙𝑙𝑙𝑙𝑙𝑙 in yearthe𝑙𝑙𝑙𝑙𝑙𝑙, 𝑡𝑡catches , lme𝑙𝑙𝑙𝑙𝑙𝑙 describes,𝑡𝑡 from the LMEthe intrinsic in year growtht (Clark rate 1991). of the Thus, 𝑥𝑥̇ = 𝑟𝑟 𝑥𝑥 − 𝑘𝑘 𝑥𝑥2 −𝑡𝑡 𝐻𝐻𝑟𝑟 xstock,thelme,t change describes klme is o af biomass measurethe stock is of size the density biologicalin the dependence large growth marine𝑙𝑙𝑙𝑙𝑙𝑙 of and , ecosystem𝑡𝑡a stock Hlmt,t𝑙𝑙𝑙𝑙𝑙𝑙 describesminus𝑡𝑡 lme𝑙𝑙𝑙𝑙𝑙𝑙 the in year the𝑙𝑙𝑙𝑙𝑙𝑙2catches, 𝑡𝑡catches , lme 𝑙𝑙𝑙𝑙𝑙𝑙taken describes,𝑡𝑡 from by the the LME thefishing intrinsic in yearindustry. growtht (Clark In arate 1991). similar of the Thus, way as 𝑥𝑥̇𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = 𝑟𝑟𝑙𝑙𝑙𝑙𝑙𝑙𝑥𝑥𝑡𝑡 − 𝑘𝑘𝑙𝑙𝑙𝑙𝑙𝑙𝑥𝑥𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 −𝑡𝑡 𝐻𝐻𝑟𝑟 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 xstock,theforlme,t thechange describes kgloballme is o a fpredator biomass measurethe stock- preyis of size the density model, biologicalin the dependence welarge assume growth marine𝑥𝑥̇ of aand ecosystemfishinga= stock H𝑟𝑟lmt,t cost 𝑥𝑥describesminus− lme function𝑘𝑘 the in𝑥𝑥 year the2catches catches −, 𝐻𝐻lme taken describes from by the the LME thefishing intrinsic in yearindustry. growtht (Clark In arate 1991). similar of the Thus, way as xlme,t describes the stock size in the large marine𝑙𝑙𝑙𝑙𝑙𝑙 ecosystem,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙 𝑡𝑡 lme𝑙𝑙𝑙𝑙𝑙𝑙 in year𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑡𝑡, 𝑟𝑟lme𝑙𝑙𝑙𝑙𝑙𝑙 describes,𝑡𝑡 the intrinsic growthFishing rate for of Proteins the | 51 stock, thefor thechange kgloballme is o a fpredator biomassmeasure- preyis of the density model, biological dependence we assume growth𝑥𝑥 ̇ of aand fishinga =stock H𝑟𝑟lmt,t cost describes𝑥𝑥minus− function 𝑘𝑘 the𝑥𝑥 thecatches catches− 𝐻𝐻 taken from by the the LME fishing in yearindustry. t (Clark In a 1991). similar Thus, way as stock,xlme,t describesklme is a measure the stock of size density in the dependence large marine and ecosystem Hlmt,t describes lme in yearthe catches , lme describes from the theLME intrinsic in year growth t (Clark rate 1991). of the Thus, the change of biomass is the biological growth of a stock minus the catches𝑡𝑡 𝑟𝑟 taken by the fishing industry. In a similar way as the forstock, thechange global klme iso f predator abiomass measure- preyis ofthe densitymodel, biological dependencewe assume growth of aand fishinga stock Hlmt,t cost minusdescribes function the thecatches catches𝑡𝑡 𝑟𝑟 taken from by the the LME fishing in year industry. t (Clark In a1991). similar Thus, way as forthe the change global o predatorf biomass-prey is the model, biological we assume growth ofa fishinga stock cost minus function the catches 𝑡𝑡 𝑟𝑟 taken−Χ𝑙𝑙𝑙𝑙𝑙𝑙 by the fishing industry. In a similar way as for the global predator-prey model, we 𝑙𝑙𝑙𝑙𝑙𝑙assume𝑙𝑙𝑙𝑙𝑙𝑙 ,𝑡𝑡a fishing𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 cost function𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 All for parameter the global values predator differ-prey for model, the 64 we LMEs.𝐶𝐶 assume(𝐻𝐻 We a,use𝑥𝑥 fishing the) =same cost𝑒𝑒𝑒𝑒𝑒𝑒 function (approach𝑐𝑐 − 𝑣𝑣 as𝑡𝑡) for𝑥𝑥−Χ the𝑙𝑙𝑙𝑙𝑙𝑙𝐻𝐻 global predator-prey model for the 𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝐶𝐶 (𝐻𝐻 , 𝑥𝑥 ) = 𝑒𝑒𝑒𝑒𝑒𝑒(𝑐𝑐 − 𝑣𝑣 𝑡𝑡)𝑥𝑥−Χ𝑙𝑙𝑙𝑙𝑙𝑙𝐻𝐻 regionalisedAll parameter model, values which differ yieldsfor the 64,000 64 LMEs.𝑙𝑙𝑙𝑙𝑙𝑙 parameter We𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 use𝑙𝑙𝑙𝑙𝑙𝑙 sets the,𝑡𝑡 samefor rlme approach, 𝑙𝑙𝑙𝑙𝑙𝑙klme, clme𝑙𝑙𝑙𝑙𝑙𝑙 ,as vlme for, andthe globalX𝑙𝑙𝑙𝑙𝑙𝑙lme,.𝑡𝑡 predator-prey model for the 𝐶𝐶 (𝐻𝐻 , 𝑥𝑥 ) = 𝑒𝑒𝑒𝑒𝑒𝑒(𝑐𝑐 − 𝑣𝑣 𝑡𝑡)𝑥𝑥𝑙𝑙𝑙𝑙𝑙𝑙−Χ𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝐻𝐻 All parameter values differ for the 64 LMEs. We use the same approach as for the𝑙𝑙𝑙𝑙𝑙𝑙 global predator-prey model for the regionalised model, which yields 64,000𝑙𝑙𝑙𝑙𝑙𝑙 parameter𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙 sets,𝑡𝑡 for rlme, 𝑙𝑙𝑙𝑙𝑙𝑙klme, clme𝑙𝑙𝑙𝑙𝑙𝑙, vlme𝑙𝑙𝑙𝑙𝑙𝑙−, Χand,𝑡𝑡 X𝑙𝑙𝑙𝑙𝑙𝑙lme,.𝑡𝑡 𝐶𝐶𝑙𝑙𝑙𝑙𝑙𝑙(𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡, 𝑥𝑥𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) = 𝑒𝑒𝑒𝑒𝑒𝑒(𝑐𝑐𝑙𝑙𝑙𝑙𝑙𝑙 − 𝑣𝑣𝑙𝑙𝑙𝑙𝑙𝑙𝑡𝑡)𝑥𝑥𝑙𝑙𝑙𝑙𝑙𝑙−Χ𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 regionalisedAll parameter model, values which differ yieldsfor the 64,000 64 LMEs.𝐶𝐶 parameter(𝐻𝐻 We ,use𝑥𝑥 sets the) =samefor𝑒𝑒 𝑒𝑒r𝑒𝑒lme (approach𝑐𝑐, klme−, c𝑣𝑣lme ,as v𝑡𝑡)lme for𝑥𝑥, andthe𝐻𝐻 globalXlme. predator-prey model for the All parameter values differ for the 64 LMEs.𝐶𝐶𝑙𝑙𝑙𝑙𝑙𝑙(𝐻𝐻 We𝑙𝑙𝑙𝑙𝑙𝑙, 𝑡𝑡use, 𝑥𝑥𝑙𝑙𝑙𝑙𝑙𝑙 the,𝑡𝑡) =same𝑒𝑒𝑒𝑒𝑒𝑒 (approach𝑐𝑐𝑙𝑙𝑙𝑙𝑙𝑙 − 𝑣𝑣𝑙𝑙𝑙𝑙𝑙𝑙 as𝑡𝑡) 𝑥𝑥for𝑙𝑙𝑙𝑙𝑙𝑙 the,𝑡𝑡 𝐻𝐻 global𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 predator-prey model for the regionalisedAll parameter model, values which differ yields for the 64,000 64 LMEs. parameter We use sets the forsame rlme approach, klme, clme ,as vlme for, andthe globalXlme. predator-prey model for the regionalised model, which yields 64,000 parameter sets for rlme, klme, clme, vlme, and Xlme. regionalised model, which yields 64,000 parameter sets for rlme, klme, clme, vlme, and Xlme.

Fish Supply Model a. Global Predator-Prey Model We assume a Lotka-Volterra type of predator-prey model (Hannesson 1983) where refers to the biomass of the predatory species and to the biomass of the prey species. Changes in biomass over time ( t and t) are defined as 𝑥𝑥 𝑦𝑦 𝑥𝑥̇ 𝑦𝑦̇ 2 𝑡𝑡 𝑥𝑥 𝑡𝑡 𝑥𝑥 𝑡𝑡 𝑡𝑡 𝑡𝑡 𝑡𝑡 𝑥𝑥̇ = 𝑟𝑟 𝑥𝑥 − 𝑘𝑘 𝑥𝑥 2 + 𝑎𝑎𝑎𝑎 𝑦𝑦 − 𝐻𝐻 𝑦𝑦̇𝑡𝑡 = 𝑟𝑟𝑦𝑦𝑦𝑦𝑡𝑡 − 𝑘𝑘𝑦𝑦𝑦𝑦𝑡𝑡 − 𝑏𝑏𝑏𝑏𝑡𝑡𝑦𝑦𝑡𝑡 − 𝐿𝐿𝑡𝑡 Here, and describe the stock size in year and denote the intrinsic growth rates, kx and ky capture density-

dependence𝑡𝑡 for𝑡𝑡 predator and prey species, respectively,𝑥𝑥 𝑦𝑦 and and denote the interaction parameters. An increase in the biomass𝑥𝑥 of prey𝑦𝑦 has a positive impact on the development𝑡𝑡, 𝑟𝑟 𝑟𝑟 of the predator’s biomass which is why the interaction term is positive. However, an increase in the biomass of the predator𝑎𝑎 has𝑏𝑏 a negative impact on the development of the prey’s 𝑎𝑎𝑥𝑥𝑡𝑡𝑦𝑦𝑡𝑡 biomass which is why the prey’s interaction term is negative. Ht and Lt denote harvest levels for predator and prey

species respectively. Thus, the change in biomass is𝑡𝑡 determined𝑡𝑡 by the biological growth of the stock, minus catches, plus or minus the interaction term. 𝑏𝑏𝑥𝑥 𝑦𝑦 We assume generalised Schaefer harvest production functions,

Χ𝑥𝑥 𝑡𝑡 𝑡𝑡 𝑥𝑥𝑥𝑥 𝐻𝐻 = 𝑞𝑞𝑞𝑞Χ𝑦𝑦 𝐸𝐸 𝐿𝐿𝑡𝑡 = 𝑑𝑑𝑑𝑑𝑡𝑡 𝐸𝐸𝑦𝑦𝑦𝑦  for predator and prey species respectively. Here, and denote catchability coefficients and  x and y denote the stock elasticities of output, which are allowed to differ from one. Ext and Eyt are effort levels directed at predatory and prey fish respectively. 𝑐𝑐 𝑑𝑑 Assuming that the marginal effort costs are constant for both fisheries, and allowing for a trend of declining costs due to technical progress (at rates vx and vy) , fishing costs can be written as

−Χ𝑥𝑥 𝑥𝑥 𝑡𝑡 𝑡𝑡 𝑥𝑥 𝑥𝑥 𝑡𝑡 𝑡𝑡 𝐶𝐶 (𝐻𝐻 , 𝑥𝑥 ) = 𝑒𝑒𝑒𝑒𝑒𝑒(𝑐𝑐 − 𝑣𝑣 𝑡𝑡)𝑥𝑥 −Χ𝑥𝑥𝐻𝐻 𝑦𝑦 𝑡𝑡 𝑡𝑡 𝑦𝑦 𝑦𝑦 𝑡𝑡 𝑡𝑡 As discussed in the main text, the biological parameters𝐶𝐶 (𝐿𝐿 , 𝑦𝑦 ) = are𝑒𝑒𝑒𝑒 estimated𝑒𝑒(𝑐𝑐 − 𝑣𝑣 𝑡𝑡 )using𝑦𝑦 𝐿𝐿the Catch-MSY method developed by Martell and Froese (Martell and Froese 2013). This method allows biological parameters based on catch data to be estimated. It requires time series of catch data and prior ranges for the parameter values as well as possible ranges of stock sizes from the initial and final period. After specifying initial stock sizes and limits for the final stock size, a parameter set is randomly drawn from the prior parameter distribution. Then, the underlying fish supply model is used to calculate the biomass corresponding to the level of harvest given the parameter set. If this biomass is in a reasonable range, the parameter set is stored. In our analysis, we repeat this procedure 10,000,000 times for each LME. We use samples of 1,000 randomly picked accepted parameter values in our model computations to compute mean estimates and confidence intervals. Thus, all results reported below are based on averages and standard deviations obtained from 1,000 separate model runs each. In the global predator-prey model, we extend the approach by Martell and Froese (2013) and determine parameter values for the Lotka-Volterra predator-prey model. In each case, initial parameter sets to be tested are randomly drawn from a uniform distribution. Biological parameters are accepted if final biomasses fall between a minimum and two-thirds of their equilibrium value without fishing. Economic theory predicts a positive relationship between fish stock biomass and market supply of fish (or no relationship at all in the case of a pure schooling fishery), and thus a negative relationship between stock biomass and the fish price. We use price data in each run for a tested parameter set to check whether this requirement is met. Specifically, we assume that the open access conditions pHt = Cx(Ht, xt) and pLt = Cy(Lt, yt) should hold for the period between 1976 and 2000 (Quaas et al. 2012). We use observed prices from Sea Around Us and the stock estimates from the test runs in the Martell/Froese procedure to estimate a log-linearised OLS regression of the open access conditions using the cost functions and including the time trends below. We accept a parameter set if it gives non-negative estimates for both Xx and Xy. Parameter sets that do not pass this test are rejected. Otherwise, we use the resulting information on the relationship between price and stock biomass to obtain an estimate for economic parameter values.

The means and standard deviations for the 1,000 parameter sets used in the computations are given in the table below  rx ry kx ky a b cx cy vx vy  x y Mean 1.44 2.24 0.044 0.0096 0.0046 0.014 49.14 24.27 0.024 0.011 0.24 0.32 Std 0.56 0.74 0.023 0.0062 0.0020 0.0077 10.45 7.82 0.005 0.003 0.20 0.20

b. Fish Supply Model at LME-Level In the total fish supply model, the change of biomass over time is defined as

2 𝑥𝑥̇𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = 𝑟𝑟𝑙𝑙𝑙𝑙𝑙𝑙𝑥𝑥𝑡𝑡 − 𝑘𝑘𝑙𝑙𝑙𝑙𝑙𝑙𝑥𝑥𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 − 𝐻𝐻𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 xlme,t describes the stock size in the large marine ecosystem lme in year , lme describes the intrinsic growth rate of the stock, klme is a measure of density dependence and Hlmt,t describes the catches from the LME in year t (Clark 1991). Thus, the change of biomass is the biological growth of a stock minus the catches𝑡𝑡 𝑟𝑟 taken by the fishing industry. In a similar way as for the global predator-prey model, we assume a fishing cost function

−Χ𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 All parameter values differ for the 64 LMEs.𝐶𝐶 (𝐻𝐻 We ,use𝑥𝑥 the) =same𝑒𝑒𝑒𝑒𝑒𝑒 (approach𝑐𝑐 − 𝑣𝑣 as𝑡𝑡) for𝑥𝑥 the𝐻𝐻 global predator-prey model for the regionalised model, which yields 64,000 parameter sets for rlme, klme, clme, vlme, and Xlme.

Demand Model Demand Model Demand Model a. Global Demand Model a.Demand Global Model Demand Model In our global demand model, we consider one representative consumer who has preferences over consumption of three Ina. ourGlobal global Demand demand Model model, we consider one representative consumer who has preferences over consumption of three typesIn our ofglobal protein demand-rich food, model, namely we consider non-fish, one protein representative-rich food items consumer (quantity who C hastt ), high preferences-trophic-level over predatoryconsumption fish of(qu threeantity types of protein-rich food, namely non-fish, protein-rich food items (quantity Ct ), high-trophic-level predatory fish (quantity Htt), and low-trophic-level forage fish (quantity Ltt). types of protein-rich food, namely non-fish, protein-rich food items (quantity Ct ), high-trophic-level predatory fish (quantity Ht), and low-trophic-level forage fish (quantity Lt). Preferences over these goods as well as numeraire consumption Xtt is described by the utility function: PreferencesHt), and low- trophicover these-level goods forage as fish well (quantity as numeraire Lt). consumption X is described by the utility function: Preferences over these goods as well as numeraire consumption Xt is described by the utility function:

Preferences over these goods as well as numeraire consumption Xt is described by the utility function:

𝐸𝐸𝑡𝑡𝜎𝜎 𝜎𝜎𝐸𝐸−𝑡𝑡𝜎𝜎1 𝑡𝑡 𝑡𝑡 𝜎𝜎−1 𝑡𝑡 𝑡𝑡 𝑈𝑈𝑡𝑡 = 𝑁𝑁𝑡𝑡 +𝜎𝜎𝐸𝐸−𝜎𝜎1 𝑙𝑙𝑙𝑙(𝑉𝑉𝑡𝑡) 𝑈𝑈 = 𝑁𝑁 +𝜎𝜎−1 𝑙𝑙𝑙𝑙(𝑉𝑉 ) with Ett being total expenditures of w for protein𝑡𝑡 -rich𝑡𝑡 food in year𝑡𝑡 t, Ntt being numeraire consumption, and Vtt being a sub- with Et being total expenditures of w for protein𝑈𝑈 -=rich𝑁𝑁 food+ in𝑙𝑙𝑙𝑙 year(𝑉𝑉 ) t, Nt being numeraire consumption, and Vt being a sub- utility index for protein food consumption, given by (Quaas and Requate 2013; Quaas et al. 2016). utilitywith E indext being for total protein expenditures food consumption, of w for protein given- richby (Quaas food in andyear Requate t, Nt being 2013; numeraire Quaas consumption,et al. 2016). and Vt being a sub-

utility index for protein food consumption, given by (Quaas and Requate 2013; Quaas et al. 2016).

𝜎𝜎 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−𝜎𝜎1 𝜎𝜎𝜎𝜎−1 𝜎𝜎𝜎𝜎−1 𝜎𝜎𝜎𝜎−1 𝜎𝜎𝜎𝜎−1 𝑡𝑡 𝐻𝐻 𝐿𝐿 𝐻𝐻 𝐿𝐿 𝑉𝑉 = [(1 − 𝜂𝜂 −𝜂𝜂 )𝐶𝐶𝑡𝑡𝜎𝜎𝜎𝜎−1 + 𝜂𝜂 𝐻𝐻𝑡𝑡𝜎𝜎𝜎𝜎−1 + 𝜂𝜂 𝐿𝐿𝜎𝜎𝑡𝑡 𝜎𝜎−1]𝜎𝜎−1 𝑉𝑉𝑡𝑡 = [(1 − 𝜂𝜂𝐻𝐻−𝜂𝜂𝐿𝐿)𝐶𝐶𝑡𝑡 𝜎𝜎 + 𝜂𝜂𝐻𝐻𝐻𝐻𝑡𝑡 𝜎𝜎 + 𝜂𝜂𝐿𝐿𝐿𝐿𝑡𝑡𝜎𝜎 ] 𝑡𝑡 𝐻𝐻 𝐿𝐿 𝑡𝑡 𝐻𝐻 𝑡𝑡 𝐿𝐿 𝑡𝑡 Here, σ reflects the elasticity of substitution𝑉𝑉 = between[(1 − 𝜂𝜂 −different𝜂𝜂 )𝐶𝐶 types+ 𝜂𝜂 𝐻𝐻of food.+ 𝜂𝜂 Following𝐿𝐿 ] Quaas et al. (2016), we assume σ = Here, σ reflects the elasticity of substitution between different types of food. Following Quaas et al. (2016), we assume σ = 1.7.Here, The σ reflects further preferencethe elasticity parameters of substitution ηH and between ηL are differentestimated types using of pricefood. and Following quantity Quaas data fromet al. Sea(2016), Around we assumeUs and theσ = 1.7. The further preference parameters ηH and ηL are estimated using price and quantity data from Sea Around Us and the FAO. Using the yearly prices of non-fish protein-rich food, PCt, predatory fish, PHt and forage fish, PLt, maximisation of the 1.7. The further preference parameters ηH and ηL are estimated using price and quantity data from Sea Around Us and the FAO. Using the yearly prices of non-fish protein-rich food, PCt, predatory fish, PHt and forage fish, PLt, maximisation of the utility with respect to consumption of protein-rich food leads to the following inverse demand functions FAO.utility withUsing respect the yearly to consumption prices of non of- fishprotein protein-rich-rich food food, leads P Ctto, predatorythe following fish, inverse PHt and demand forage fish,functions PLt, maximisation of the

utility with respect to consumption of protein-rich food leads to the following inverse demand functions

1 − 𝑡𝑡 𝜎𝜎1 𝐸𝐸 − 𝐶𝐶𝐶𝐶 𝑡𝑡 𝐻𝐻 𝐿𝐿 𝑡𝑡 𝜎𝜎1 𝑃𝑃 = 𝐸𝐸𝑡𝑡 (1 − 𝜂𝜂 − 𝜂𝜂 ) 𝐶𝐶 − 𝐶𝐶𝐶𝐶 𝑉𝑉𝑡𝑡𝑡𝑡 𝐻𝐻 𝐿𝐿 𝑡𝑡 𝜎𝜎 𝑃𝑃𝐶𝐶𝐶𝐶 = 𝐸𝐸𝑉𝑉𝑡𝑡 (1 − 𝜂𝜂𝐻𝐻 − 𝜂𝜂𝐿𝐿) 𝐶𝐶 𝑉𝑉 1 𝑡𝑡 𝑃𝑃 = 𝑡𝑡 (1 − 𝜂𝜂 − 𝜂𝜂− ) 𝐶𝐶 𝑡𝑡 𝜎𝜎1 𝑉𝑉 𝐸𝐸 − 𝐻𝐻𝐻𝐻 𝑡𝑡 𝐻𝐻 𝑡𝑡 𝜎𝜎1 𝑃𝑃 = 𝐸𝐸𝑡𝑡 𝜂𝜂 𝐻𝐻− 𝐻𝐻𝐻𝐻 𝑉𝑉𝑡𝑡𝑡𝑡 𝐻𝐻 𝑡𝑡 𝜎𝜎 𝑃𝑃𝐻𝐻𝐻𝐻 = 𝑉𝑉𝐸𝐸𝑡𝑡 𝜂𝜂𝐻𝐻 𝐻𝐻 𝑉𝑉 𝑡𝑡1 𝑃𝑃 = 𝑡𝑡 𝜂𝜂 𝐻𝐻− 𝑡𝑡 −𝜎𝜎1 𝑉𝑉𝐸𝐸 − 𝐿𝐿𝐿𝐿 𝑡𝑡 𝐿𝐿 𝑡𝑡 𝜎𝜎1 𝑃𝑃 = 𝐸𝐸𝑡𝑡 𝜂𝜂 𝐿𝐿 𝐿𝐿𝐿𝐿 𝑡𝑡𝑡𝑡 𝐿𝐿 −𝑡𝑡 𝜎𝜎 𝑃𝑃 = 𝐸𝐸𝑉𝑉𝑡𝑡 𝜂𝜂 𝐿𝐿 𝑃𝑃𝐿𝐿𝐿𝐿 = 𝑉𝑉 𝜂𝜂𝐿𝐿 𝐿𝐿𝑡𝑡 from which we estimate the demand parameters ηHt, ηLt, using𝑡𝑡 data on Htt, Ltt, Ctt, PCt, PHt, and PLt for the period 1976 to 2010. from which we estimate the demand parameters ηHt, ηLt, using𝑉𝑉 data on Ht, Lt, Ct, PCt, PHt, and PLt for the period 1976 to 2010. This leads to fromThis leadswhich towe estimate the demand parameters ηHt, ηLt, using data on Ht, Lt, Ct, PCt, PHt, and PLt for the period 1976 to 2010. This leads to ηH 0.1554 ηH 0.1554 ηL 0.3675 ηH 0.1554 ηL 0.3675

ηL 0.3675 For expenditures Ett we use the scenarios described in section 2.3. For the consumption of non-fish protein-rich food, we For expenditures Et we use the scenarios described in section 2.3. For the consumption of non-fish protein-rich food, we assume that the past trend over the period 1976 to 2010 will continue, with an exponential growth rate of 2.09% per year. assumeFor expenditures that the past Et we trend use overthe scenarios the period described 1976 to 2010 in section will continue, 2.3. For withthe consumptionan exponential of growthnon-fish rate protein of 2.09%-rich food,per year. we

assume that the past trend over the period 1976 to 2010 will continue, with an exponential growth rate of 2.09% per year. b. Demand Model at LME Level Fb.or Demand the modelling Model of at regional LME Level demand we grouped countries at large marine ecosystem (LME) level. We assume that there Fb.or Demand the modelling Model of at regional LME Level demand we grouped countries at large marine ecosystem (LME) level. We assume that there is a representative consumer for each large marine ecosystem who consumes protein-rich food in each year t, which is isFor a therepresentative modelling of consumer regional demandfor each welarge grouped marine countries ecosystem at largewho consumesmarine ecosystem protein-rich (LME) food level. in each We yearassume t, which that thereis composedis a representative of a quantity consumer Clme,tlme,t offor non each-fish, large protein marine-rich ecosystem food items, who quantity consumes Flme,tlme,t of protein fish food-rich-items. food inPreferences each year tare, which described is composed of a quantity Clme,t of non-fish, protein-rich food items, quantity Flme,t of fish food-items. Preferences are described by the utility function composedby the utility of function a quantity Clme,t of non-fish, protein-rich food items, quantity Flme,t of fish food-items. Preferences are described by the utility function 𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝜎𝜎 𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑈𝑈𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = 𝑁𝑁𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 + 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝜎𝜎 ln(𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝑈𝑈 = 𝑁𝑁 + 𝐸𝐸𝜎𝜎 − 1 𝜎𝜎 ln(𝑉𝑉 ) 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝜎𝜎 − 1 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 As in the global model, Elme,tlme,t describes the total𝑈𝑈 expenditures= 𝑁𝑁 + for proteinln-(rich𝑉𝑉 food) in year t, Nlme,tlme,t is numeraire consumption, As in the global model, Elme,t describes the total expenditures for𝜎𝜎 protein− 1 -rich food in year t, Nlme,t is numeraire consumption, and Vlme,tlme,t is a sub-utility index for protein food consumption: As52 in the global model, Elme,t describes the total expenditures for protein-rich food in year t, Nlme,t is numeraire consumption, and Vlme,t is a sub-utility index for protein food consumption:

and Vlme,t is a sub-utility index for protein food consumption:

𝜎𝜎 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−𝜎𝜎1 𝑑𝑑 𝑖𝑖 𝜎𝜎𝜎𝜎−1 𝑑𝑑 𝑑𝑑 𝜎𝜎𝜎𝜎−1 𝑖𝑖 𝑖𝑖 𝜎𝜎𝜎𝜎−1 𝜎𝜎−1 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝜎𝜎 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝜎𝜎−𝜎𝜎1 𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,,𝑡𝑡 𝜎𝜎𝜎𝜎−1 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,,𝑡𝑡 𝜎𝜎𝜎𝜎−1 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = [(1 − 𝜂𝜂 − 𝜂𝜂 ) 𝐶𝐶 + 𝜂𝜂 (𝐹𝐹 ) + 𝜂𝜂 (𝐹𝐹 ) ]𝜎𝜎−1 𝑉𝑉 = [(1 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,𝑡𝑡 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,𝑡𝑡) 𝐶𝐶𝑙𝑙𝑙𝑙𝑙𝑙𝜎𝜎 + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,𝑡𝑡) 𝜎𝜎 + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,𝑡𝑡) 𝜎𝜎 ] 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = [(1 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝐶𝐶𝑙𝑙𝑙𝑙𝑙𝑙 + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) ] Here, σ reflects the elasticity of substitution between fish and non-fish protein-rich food. Using the Armington (1969) Here, σ reflects1 the elasticity of substitution between fish and non-fish protein-rich food. Using the Armington (1969) assumption1 allows a distinction to be made between domestically produced and imported fish. Again we assume that the Here, σ reflects the elasticity of substitution between fish and non-fish protein-rich food. Using the Armington (1969) d assumption1 allows a distinction to be made between domestically produced and imported fish. Again we assume that the elasticityassumption of demand allows a is distinction 1.7 according to be to made Asche between et al. (1996) domestically and Quaas produced and Requate and imported (2013). fish. The Again demand we assumeparameters that ηthedlme,Flme,F elasticityii of demand is 1.7 according to Asche et al. (1996) and Quaas and Requate (2013). The demand parameters η lme,F and η lme,Flme,F measure relative preference for domestic and imported fish. Using the yearly prices of non-fish protein-rich foodd elasticityi of demand is 1.7 according to Asche et al. (1996) and Quaas and Requate (2013). The demand parameters η lme,F and η lme,F measure relative preference for domesticd and imported fish. Using the yearlyii prices of non-fish protein-rich food PC,lme,ti, domestically produced fish food items P F,lme,t and imported fish food items P F,lme,t, utility maximisation leads to the and η lme,F measure relative preference for domesticd and imported fish. Using the yearlyi prices of non-fish protein-rich food PC,lme,t, domestically produced fish food items P F,lme,t and imported fish food items P F,lme,t, utility maximisation leads to the following inverse demand functions: d i PfollowingC,lme,t, domestically inverse demand produced functions: fish food items P F,lme,t and imported fish food items P F,lme,t, utility maximisation leads to the

following inverse demand functions:

1 − 𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑑𝑑 𝑖𝑖 −𝜎𝜎1 𝐸𝐸 − 𝐶𝐶,,𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝜎𝜎1 ,,𝑡𝑡 𝑃𝑃 = 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (1 − 𝜂𝜂 − 𝜂𝜂 ) 𝐶𝐶 𝐶𝐶,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙−𝜎𝜎 ,𝑡𝑡 𝑃𝑃 = 𝐸𝐸𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (1 − 𝜂𝜂 − 𝜂𝜂 ) 𝐶𝐶 𝐶𝐶,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑃𝑃 = 𝑉𝑉 (1 − 𝜂𝜂 −1𝜂𝜂 ) 𝐶𝐶 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 − 𝑑𝑑 𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑑𝑑 𝜎𝜎1 𝑑𝑑 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 − 𝐹𝐹𝑑𝑑,,𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,,𝑡𝑡 𝜎𝜎1 𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,,𝑡𝑡 𝑃𝑃 = 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 (𝐹𝐹 ) 𝜂𝜂 𝑃𝑃𝐹𝐹𝑑𝑑,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,𝑡𝑡)−𝜎𝜎 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙𝑑𝑑 ,𝑡𝑡 𝐹𝐹,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑃𝑃 = 𝑉𝑉 𝐹𝐹 1 𝜂𝜂 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 ( )− 𝑖𝑖 𝐸𝐸𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑖𝑖 𝜎𝜎1 𝑖𝑖 𝐹𝐹𝑖𝑖,,𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,,𝑡𝑡 −𝜎𝜎 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,,𝑡𝑡 𝑃𝑃 = 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹 ) 1 𝜂𝜂 𝐹𝐹𝑖𝑖,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,𝑡𝑡 − 𝑙𝑙𝑙𝑙𝑙𝑙𝑖𝑖 ,𝑡𝑡 𝑃𝑃 = 𝑉𝑉𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹 ) 𝜎𝜎 𝜂𝜂 𝑃𝑃𝐹𝐹,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = 𝑉𝑉 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 With the given information on prices and quantities of domestically𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 produced and imported fish and substitution goods for With the given information on prices and quantities of domestically𝑉𝑉 d produced iiand imported fish and substitution goods for theWith period the given 1976 information to 2011, we on estimate prices and the quantitiespreference of parameters domestically η dproducedlme,Flme,F and η iandlme,Flme,F .imported fish and substitution goods for the period 1976 to 2011, we estimate the preference parameters η lme,F and η lme,F . For food consumption expenditures, we use the SSP1 scenario ond income growthi and an income elasticity of food demand theFor periodfood consumption 1976 to 2011, expenditures, we estimate we the use preference the SSP1 parameters scenario on η lme,Fincome and growthη lme,F . and an income elasticity of food demand of 0.48 (Cireira and Masset 2010). For the consumption of non-fish protein-rich food, we determine linear trends for each ofFor 0.48 food (Cireira consumption and Masset expenditures, 2010). For we the use consumption the SSP1 scenario of non-fish on incomeprotein -growthrich food, and we an determine income elasticity linear trends of food for demand each LME based on past observations for the period 1976 to 2010 and assume that they will continue until 2050. LMEof 0.48 based (Cireira on pastand Massetobservations 2010). for For the the period consumption 1976 to 2010of non and-fish assume protein that-rich they food, will we continue determine until linear 2050. trends for each LME based on past observations for the period 1976 to 2010 and assume that they will continue until 2050.

1 The Armington assumption is a standard assumption of computable equilibrium models and implies that consumers are assumed to 1 The Armington assumption is a standard assumption of computable equilibrium models and implies that consumers are assumed to differentiate1 between goods based on origin, that is whether the good is produced domestically or imported. differentiate The Armington between assumption goods based is a onstandard origin, thatassumption is whether of the computable good is produced equilibrium domestically models andor imported. implies that consumers are assumed to differentiate between goods based on origin, that is whether the good is produced domestically or imported. Demand Model Demanda. Global Model Demand Model a.In Globalour global Demand demand Model model, we consider one representative consumer who has preferences over consumption of three Intypes our ofglobal protein demand-rich food, model, namely we consider non-fish, one protein representative-rich food items consumer (quantity who C hast ), highpreferences-trophic-level over predatoryconsumption fish of(qu threeantity typesHt), and of proteinlow-trophic-rich- levelfood, foragenamely fish non (quantity-fish, protein Lt). -rich food items (quantity Ct ), high-trophic-level predatory fish (quantity HPreferencest), and low- trophicover these-level goods forage as fish well (quantity as numeraire Lt). consumption Xt is described by the utility function: Preferences over these goods as well as numeraire consumption Xt is described by the utility function:

𝐸𝐸𝑡𝑡𝜎𝜎 𝜎𝜎−1 𝑈𝑈𝑡𝑡 = 𝑁𝑁𝑡𝑡 +𝐸𝐸𝑡𝑡𝜎𝜎 𝑙𝑙𝑙𝑙(𝑉𝑉𝑡𝑡) 𝜎𝜎−1 with Et being total expenditures of w for protein𝑈𝑈𝑡𝑡 -=rich𝑁𝑁𝑡𝑡 food+ in𝑙𝑙𝑙𝑙 year(𝑉𝑉𝑡𝑡) t, Nt being numeraire consumption, and Vt being a sub- withutility E indext being for total protein expenditures food consumption, of w for protein given- richby (Quaas food in andyear Requate t, Nt being 2013; numeraire Quaas consumption,et al. 2016). and Vt being a sub- utility index for protein food consumption, given by (Quaas and Requate 2013; Quaas et al. 2016).

𝜎𝜎 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎 𝜎𝜎 𝜎𝜎 𝜎𝜎 𝑉𝑉𝑡𝑡 = [(1 − 𝜂𝜂𝐻𝐻−𝜂𝜂𝐿𝐿)𝐶𝐶𝑡𝑡𝜎𝜎−1 + 𝜂𝜂𝐻𝐻𝐻𝐻𝑡𝑡𝜎𝜎−1 + 𝜂𝜂𝐿𝐿𝐿𝐿𝜎𝜎𝑡𝑡−1 ]𝜎𝜎−1 𝜎𝜎 𝜎𝜎 𝜎𝜎 𝑉𝑉𝑡𝑡 = [(1 − 𝜂𝜂𝐻𝐻−𝜂𝜂𝐿𝐿)𝐶𝐶𝑡𝑡 + 𝜂𝜂𝐻𝐻𝐻𝐻𝑡𝑡 + 𝜂𝜂𝐿𝐿𝐿𝐿𝑡𝑡 ] Here, σ reflects the elasticity of substitution between different types of food. Following Quaas et al. (2016), we assume σ = Here,1.7. The σ reflects further thepreference elasticity parameters of substitution ηH and between ηL are differentestimated types using of pricefood. andFollowing quantity Quaas data fromet al. Sea(2016), Around we assume Us and theσ = 1.7.FAO. The Using further the preferenceyearly prices parameters of non-fish η proteinH and η-Lrich are food,estimated PCt, predatory using price fish, and PHt quantity and forage data fish, from P SeaLt, maximisation Around Us and of the the FAO.utility Usingwith respect the yearly to consumption prices of non of-fish protein protein-rich-rich food food, leads P Ctto, predatorythe following fish, inverse PHt and demand forage fish,functions PLt, maximisation of the utility with respect to consumption of protein-rich food leads to the following inverse demand functions

1 𝑡𝑡 −𝜎𝜎 𝐶𝐶𝐶𝐶 𝐸𝐸 𝐻𝐻 𝐿𝐿 𝑡𝑡 1 𝑃𝑃 = 𝑡𝑡 (1 − 𝜂𝜂 − 𝜂𝜂 ) 𝐶𝐶−𝜎𝜎 𝐶𝐶𝐶𝐶 𝐸𝐸𝑉𝑉 𝐻𝐻 𝐿𝐿 𝑡𝑡 𝑃𝑃 = 𝑡𝑡 (1 − 𝜂𝜂 − 𝜂𝜂 1) 𝐶𝐶 𝑉𝑉 𝑡𝑡 −𝜎𝜎 𝐻𝐻𝐻𝐻 𝐸𝐸 𝐻𝐻 𝑡𝑡 1 𝑃𝑃 = 𝑡𝑡 𝜂𝜂 𝐻𝐻−𝜎𝜎 𝐻𝐻𝐻𝐻 𝐸𝐸𝑉𝑉 𝐻𝐻 𝑡𝑡 𝑃𝑃 = 𝑡𝑡 𝜂𝜂 𝐻𝐻 1 𝑉𝑉 𝑡𝑡 −𝜎𝜎 𝐿𝐿𝐿𝐿 𝐸𝐸 𝐿𝐿 1 𝑃𝑃 = 𝜂𝜂 𝐿𝐿−𝑡𝑡 𝐸𝐸𝑉𝑉𝑡𝑡 𝜎𝜎 𝑃𝑃𝐿𝐿𝐿𝐿 = 𝜂𝜂𝐿𝐿 𝐿𝐿𝑡𝑡 from which we estimate the demand parameters ηHt, ηLt, using𝑡𝑡 data on Ht, Lt, Ct, PCt, PHt, and PLt for the period 1976 to 2010. Demand Model 𝑉𝑉 This leads to froma. Global which Demand we estimate Model the demand parameters ηHt, ηLt, using data on Ht, Lt, Ct, PCt, PHt, and PLt for the period 1976 to 2010. ThisIn our leads global to demand model, we consider one representativeηH 0.1554 consumer who has preferences over consumption of three ηHL 0.15540.3675 types of protein-rich food, namely non-fish, protein-rich food items (quantity Ct ), high-trophic-level predatory fish (quantity ηL 0.3675 Ht), and low-trophic-level forage fish (quantity Lt). For expenditures Et we use the scenarios described in section 2.3. For the consumption of non-fish protein-rich food, we Preferences over these goods as well as numeraire consumption Xt is described by the utility function: For assume expenditures that the past Et we trend use overthe scenarios the period described 1976 to 2010 in section will continue, 2.3. For withthe consumptionan exponential of growthnon-fish rate protein of 2.09%-rich food, per year. we assume that the past trend over the period 1976 to 2010 will continue, with an exponential growth rate of 2.09% per year. b. Demand Model at LME Level b.For Demand the modelling Model of at regional LME Level demand we grouped countries𝐸𝐸𝑡𝑡𝜎𝜎 at large marine ecosystem (LME) level. We assume that there 𝜎𝜎−1 F isor a therepresentative modelling of consumer regional demandfor each welarge grouped marine𝑈𝑈𝑡𝑡 = 𝑁𝑁 countries𝑡𝑡 ecosystem+ 𝑙𝑙𝑙𝑙 (at𝑉𝑉𝑡𝑡 )largewho consumesmarine ecosystem protein-rich (LME) food level. in each We yearassume t, which that thereis iswithcomposed a representativeEt being of totala quantity expenditures consumer Clme,t offor nonof each w- fish,for large protein protein marine-rich-rich ecosystemfood food in items, year who tquantity, N tconsumes being F lme,tnumeraire of protein fish food -consumption,rich-items. food in Preferences each and yearVt being tare, which adescribed sub is- composedutilityby the index utility of for function a proteinquantity food Clme,t consumption, of non-fish, protein given by-rich (Quaas food items, and Requate quantity 2013; Flme,t of Quaas fish food et al.-items. 2016). Preferences are described by the utility function 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝜎𝜎 𝑈𝑈𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = 𝑁𝑁𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 + ln(𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝜎𝜎 𝜎𝜎−1 𝐸𝐸𝜎𝜎𝑙𝑙𝑙𝑙𝑙𝑙−,𝑡𝑡1 𝜎𝜎𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−1 𝑈𝑈𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = 𝑁𝑁𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝜎𝜎+ ln𝜎𝜎 (𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡)𝜎𝜎 As in the global model, Elme,t describes the𝑉𝑉𝑡𝑡 = total[(1 −expenditures𝜂𝜂𝐻𝐻−𝜂𝜂𝐿𝐿)𝐶𝐶𝑡𝑡 for+𝜎𝜎 𝜂𝜂protein−𝐻𝐻𝐻𝐻1 𝑡𝑡 -rich+ 𝜂𝜂𝐿𝐿 food𝐿𝐿𝑡𝑡 ]in year t, Nlme,t is numeraire consumption,

Asand in V thelme,t globalis a sub model,-utility Eindexlme,t describes for protein the food total consumption: expenditures for protein-rich food in year t, Nlme,t is numeraire consumption, andHere, V lme,tσ reflects is a sub the-utility elasticity index of for substitution protein food between consumption: different types of food. Following Quaas et al. (2016), we assume σ = 1.7. The further preference parameters ηH and ηL are estimated using price and quantity data from Sea Around Us and the FAO. Using the yearly prices of non-fish protein-rich food, PCt, predatory fish, PHt and forage fish, P𝜎𝜎Lt , maximisation of the 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−1 utility with respect to consumption of protein𝑑𝑑 -rich𝑖𝑖 food leads𝜎𝜎 to 𝑑𝑑the following𝑑𝑑 𝜎𝜎 inverse𝑖𝑖 demand𝑖𝑖 functions𝜎𝜎 𝜎𝜎 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = [(1 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝐶𝐶𝑙𝑙𝑙𝑙𝑙𝑙𝜎𝜎−1 + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡)𝜎𝜎−1 + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡)𝜎𝜎−1 ]𝜎𝜎−1 𝑑𝑑 𝑖𝑖 𝜎𝜎 𝑑𝑑 𝑑𝑑 𝜎𝜎 𝑖𝑖 𝑖𝑖 𝜎𝜎 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = [(1 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝐶𝐶𝑙𝑙𝑙𝑙𝑙𝑙 + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) ] Here, σ reflects the elasticity of substitution between fish and non-fish protein-rich food. Using the Armington (1969) 1 1 − Here,assumption σ reflects allows the elasticitya distinction of substitution to be made between between fish domestically𝐸𝐸𝑡𝑡 and non-fish produced protein𝜎𝜎 and-rich imported food. Using fish. the Again Armington we assume (1969) that the 1 𝐶𝐶𝐶𝐶 𝐻𝐻 𝐿𝐿 𝑡𝑡 d assumptionelasticity of demand allows a is distinction 1.7 according to be to made Asche between et 𝑃𝑃al. (1996)= domestically𝑡𝑡 ( 1and− 𝜂𝜂 Quaas− produced𝜂𝜂 ) and 𝐶𝐶 Requate and imported (2013). fish. The Again demand we assumeparameters that ηthelme,F i 𝑉𝑉 d elasticityand η lme,F of measure demand relative is 1.7 according preference to forAsche domestic et al. (1996)and imported and Quaas fish.1 Usingand Requate the yearly (2013). prices The of nondemand-fish proteinparameters-rich foodη lme,F i d 𝑡𝑡 −𝜎𝜎 i PC,lme,t, domestically produced fish food items P F,lme,t and imported fish food items P F,lme,t, utility maximisation leads to the and η lme,F measure relative preference for domestic and𝐻𝐻 𝐻𝐻imported𝐸𝐸 𝐻𝐻 fish.𝑡𝑡 Using the yearly prices of non-fish protein-rich food d 𝑃𝑃 = 𝑡𝑡 𝜂𝜂 𝐻𝐻 i PfollowingC,lme,t, domestically inverse demand produced functions: fish food items P F,lme,t and imported𝑉𝑉 fish food items P F,lme,t, utility maximisation leads to the 1 following inverse demand functions: − 𝐸𝐸𝑡𝑡 𝜎𝜎 𝐿𝐿𝐿𝐿 𝐿𝐿 𝑡𝑡 𝑃𝑃 = 𝑡𝑡 𝜂𝜂 𝐿𝐿 𝑉𝑉 1 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑑𝑑 𝑖𝑖 −𝜎𝜎 from which we estimate the demand parameters𝐶𝐶,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 ηHt𝐸𝐸, ηLt, using data𝑙𝑙𝑙𝑙𝑙𝑙, 𝑡𝑡on H𝑙𝑙𝑙𝑙𝑙𝑙t, L,𝑡𝑡t, Ct, 1PCt , PHt, and PLt for the period 1976 to 2010. 𝑃𝑃 = (1 − 𝜂𝜂 − 𝜂𝜂 ) 𝐶𝐶𝑙𝑙𝑙𝑙𝑙𝑙− ,𝑡𝑡 This leads to 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 𝑑𝑑 𝑖𝑖 𝜎𝜎 𝐶𝐶,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑃𝑃 = 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (1 − 𝜂𝜂 − 1𝜂𝜂 ) 𝐶𝐶 ηH 𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙0.1554,𝑡𝑡 − 𝑑𝑑 𝐸𝐸 𝑑𝑑 𝜎𝜎 𝑑𝑑 η𝐹𝐹L, 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 0.3675𝑙𝑙𝑙𝑙𝑙𝑙 ,𝑡𝑡 1 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑃𝑃𝑑𝑑 = 𝑙𝑙𝑙𝑙𝑙𝑙,,𝑡𝑡 (𝐹𝐹𝑑𝑑 )−𝜎𝜎 𝜂𝜂𝑑𝑑 𝐹𝐹,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝐸𝐸𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑃𝑃 = 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹 ) 1 𝜂𝜂 − For expenditures Et we use the scenarios described𝑖𝑖 in section𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 2.3.𝑖𝑖 For𝜎𝜎 the𝑖𝑖 consumption of non-fish protein-rich food, we 𝐹𝐹,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝐸𝐸 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 1 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑃𝑃 = (𝐹𝐹 )− 𝜂𝜂 assume that the past trend over the period 1976𝑖𝑖 to 2010𝐸𝐸𝑉𝑉 𝑙𝑙𝑙𝑙𝑙𝑙will, 𝑡𝑡continue,𝑖𝑖 𝜎𝜎with𝑖𝑖 an exponential growth rate of 2.09% per year. 𝑃𝑃𝐹𝐹,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 With the given information on prices and quantities of domestically𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 produced and imported fish and substitution goods for b. Demand Model at LME Level d i Withthe period the given 1976 information to 2011, we on estimate prices and the quantitiespreference of parameters domestically η producedlme,F and η andlme,F imported. fish and substitution goods for For the modelling of regional demand we grouped countries at larged marine ecosystemi (LME) level. We assume that there theFor periodfood consumption 1976 to 2011, expenditures, we estimate we the use preference the SSP1 parameters scenario on η lme,Fincome and growthη lme,F . and an income elasticity of food demand is a representative consumer for each large marine ecosystem who consumes protein-rich food in each year t, which is Forof 0.48 food (Cireira consumption and Masset expenditures, 2010). For we the use consumption the SSP1 scenario of non- fishon incomeprotein -growthrich food, and we an determine income elasticity linear trends of food for demand each composed of a quantity Clme,t of non-fish, protein-rich food items, quantity Flme,t of fish food-items. Preferences are described ofLME 0.48 based (Cireira on pastand Massetobservations 2010). for For the the period consumption 1976 to 2010of non and-fish assume protein -thatrich theyfood, will we continue determine until linear 2050. trends for each by the utility function LME based on past observations for the period 1976 to 2010 and assume that they will continue until 2050.

1 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 The Armington assumption is a standard assumption 𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 of computable𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝐸𝐸 equilibrium 𝜎𝜎 𝑙𝑙𝑙𝑙𝑙𝑙 models,𝑡𝑡 and implies that consumers are assumed to 1 𝑈𝑈 = 𝑁𝑁 + ln(𝑉𝑉 ) differentiate The Armington between assumption goods based is a onstandard origin, thatassumption is whether of the computable good is 𝜎𝜎produced −equilibrium1 domestically models andor imported. implies that consumers are assumed to differentiateAs in the global between model, goods E basedlme,t describes on origin, thethat totalis whether expenditures the good isfor produced protein -domesticallyrich food in or year imported. t, Nlme,t is numeraire consumption, and Vlme,t is a sub-utility index for protein food consumption:

𝜎𝜎 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−1 𝜎𝜎−1 𝑑𝑑 𝑖𝑖 𝜎𝜎 𝑑𝑑 𝑑𝑑 𝜎𝜎 𝑖𝑖 𝑖𝑖 𝜎𝜎 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = [(1 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝐶𝐶𝑙𝑙𝑙𝑙𝑙𝑙 + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) + 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) ]

Here, σ reflects the elasticity of substitution between fish and non-fish protein-rich food. Using the Armington (1969) assumption1 allows a distinction to be made between domestically produced and imported fish. Again we assume that the d elasticity of demand is 1.7 according to Asche et al. (1996) and Quaas and Requate (2013). The demand parameters η lme,F i and η lme,F measure relative preference for domestic and imported fish. Using the yearly prices of non-fish protein-rich food d i PC,lme,t, domestically produced fish food items P F,lme,t and imported fish food items P F,lme,t, utility maximisation leads to the following inverse demand functions:

1 − 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑑𝑑 𝑖𝑖 𝜎𝜎 𝑃𝑃𝐶𝐶,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = (1 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 − 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝐶𝐶𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 1 − 𝑑𝑑 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑑𝑑 𝜎𝜎 𝑑𝑑 𝑃𝑃𝐹𝐹,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 1 − 𝑖𝑖 𝐸𝐸𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑖𝑖 𝜎𝜎 𝑖𝑖 𝑃𝑃𝐹𝐹,𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 = (𝐹𝐹𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡) 𝜂𝜂𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 𝑉𝑉𝑙𝑙𝑙𝑙𝑙𝑙,𝑡𝑡 With the given information on prices and quantities of domestically produced and imported fish and substitution goods for d i the period 1976 to 2011, we estimate the preference parameters η lme,F and η lme,F . For food consumption expenditures, we use the SSP1 scenario on income growth and an income elasticity of food demand of 0.48 (Cireira and Masset 2010). For the consumption of non-fish protein-rich food, we determine linear trends for each LME based on past observations for the period 1976 to 2010 and assume that they will continue until 2050.

1 The Armington assumption is a standard assumption of computable equilibrium models and implies that consumers are assumed to differentiate between goods based on origin, that is whether the good is produced domestically or imported.

Fishing for Proteins | 53 List of Large Marine Ecosystems (according to Sea Around Us)

Agulhas Current Greenland Sea Northern Bering - Chukchi Seas Aleutian Islands Guinea Current Northwest Australian Shelf Antarctic Gulf of Alaska Norwegian Sea Arabian Sea Gulf of California Oyashio Current Baltic Sea Gulf of Mexico Pacific Central-American Coastal Barents Sea Gulf of Thailand Patagonian Shelf Bay of Bengal Hudson Bay Complex Red Sea Beaufort Sea Humboldt Current Scotian Shelf Benguela Current Iberian Coastal Sea of Japan / East Sea Black Sea Iceland Shelf and Sea Sea of Okhotsk California Current Indonesian Sea Somali Coastal Current Canadian Eastern Arctic -West Insular Pacific-Hawaiian South Brazil Shelf Greenland Kara Sea South China Sea Canadian High Arctic - North Greenland Kuroshio Current Southeast Australian Shelf Canary Current Laptev Sea Southeast U.S. Continental Shelf Caribbean Sea Mediterranean Sea Southwest Australian Shelf Celtic-Biscay Shelf New Zealand Shelf Sulu-Celebes Sea Central Arctic Ocean (no data available) Newfoundland-Labrador Shelf West Bering Sea East Bering Sea North Australian Shelf West-Central Australian Shelf East Brazil Shelf North Brazil Shelf Yellow Sea East China Sea North Sea East Siberian Sea Northeast Australian Shelf-Great Barrier East-Central Australian Shelf Reef Faroe Plateau Northeast U.S. Continental Shelf

List of Protein-Rich Non-Fish Food Substitute Goods (FAOstat Database 2016)

Almonds, shelled Groundnuts, shelled Meat, turkey Bambara beans Hazelnuts, shelled Milk, skimmed, cow Beans, dry Kola nuts Milk, skimmed, dried Beans, green Lard Milk, whole, condensed Brazil nuts, shelled Lentils Milk, whole, dried Broad beans, horse beans, dry Maize Milk, whole, evaporated Butter, cow’s milk Maize, green Milk, whole, fresh, cow Cashew nuts, shelled Meat, cattle Nuts, not elsewhere included Cashew nuts, with shell Meat, chicken Nuts, prepared (exc. groundnuts) Cheese, sheep’s milk Meat, duck Peas, dry Cheese, whole cow’s milk Meat, game Peas, green Chestnuts Meat, goat Rice – total (rice milled equivalent) Chick peas Meat, goose and guinea fowl Soybeans Coconuts Meat, horse Walnuts, shelled Cream, fresh Meat, not elsewhere included Walnuts, with shell Eggs, hen, in shell Meat, pig Whey, condensed Eggs, other bird, in shell Meat, rabbit Whey, dry Ghee, buffalo milk Meat, sheep Yoghurt, concentrated or not

54 List of Figures and Tables

Fig. 1: World fish consumption. 16 Fig. 2: Per capita world fish consumption. 16 Fig. 3: Per capita fish consumption. 17 Fig. 4: World fish consumption pattern. 17 Fig. 5: Fish consumption pattern of the case study continents. 18 Fig. 6: Fish consumption in the eight case study countries. 19 Fig. 7: Per capita fish consumption in the eight case study countries and the corresponding continent. 20 Fig. 8: Fish consumption pattern in the eight case study countries. 21 Fig. 9: Total fish consumption and fish contribution to total animal protein. 23 Fig. 10: Fish in the diets in relation to economic development of a country or population. 24 Fig. 11/Z1: Total protein supply and fish in total protein. 25 Fig. 12: Main drivers of fish dependency. 28 Fig. 13/Z2: Fish dependency scores worldwide. 29 Fig. 13a: Per capita catch worldwide. 29 Fig. 13b: Share of fish in total consumption of animal protein. 30 Fig. 13c: Per capita GDP in USD. 30 Fig. 13d: Share of undernourishment in population. 30 Fig. 14/Z3: Population, catches and fraction of local consumption for population in 2010 that was covered by LME catches in 2010). 32 Fig. 15: Large Marine Ecosystems considered in this study. 35 Fig. 16: Development of global total catches, predatory and prey fish catches from 1950 to 2010 37 Fig. 17: Global ex-vessel price per year in 2005 USD per ton. 38 Fig. 18: Global Expenditure per Year in billion USD from 1976 to 2010. 38 Fig. 19: Global production per year in million tons from 1976 to 2010. 39 Fig. 20: Development of Global GDP from 2010 to 2050 for the SSP scenarios of the IPCC. 41 Fig. 21: Development of Global Population from 2010 to 2050 for the SSP scenarios of the IPCC. 41 Fig. 22/Z4: Estimates of global maximum sustainable yield that the global fish stocks could supply from three different model specifications. 43 Fig. 23/Z5: Global fish catches according to global bio-economic predator-prey model for varying degrees of management effectiveness for the reference scenario with income growth from SSP1. 44 Fig. 24: Global fish catches according to global bio-economic predator-prey model for varying degrees of management effectiveness for the high-pressure scenario with income growth from SSP5 and unit income elasticity of food demand. 45 Fig. 25. Expected global fish catches according to a global bio-economic predator-prey model for varying degrees of management effectiveness for the low-pressure scenario with income growth from SSP3 and no technical progress in fishing. 45 Fig. 26: Projected MSY Catches, population size and fraction of local needs that could potentially be covered by LME in 2050 under ideal conditions and population development according to SSP1 scenario. 46 Fig. 27: Projected MSY Catches, population size and fraction of local needs that could potentially be covered by LME in 2050 under ideal conditions and population development according to SSP3 scenario. 46 Fig. 28: Net Import and Net Export of fish per LME in 2050 in million tons. 47 Fig. 29: LMEs with decreasing fish consumption between 2010 and 2050. 48 Fig. 30: LMEs with increasing fish consumption between 2010 and 2050. 49

Tab. 1: Dependence on fish in the diet in the eight case countries in this study. Source: FAO. 24 Tab. 2: (Multidimensional) Global Hunger Index 1995 to 2015 for selected target countries. 26 Tab. 3: Fish dependence of the eight poorest countries and the eight countries with the highest share of fish protein consumption (own results). 31 Tab. 4: Fish dependence in the eight target countries in this study. 31 Tab. 5: Mean catches in 2050 according to three model specifications. 43 Tab. 6: National recommended intakes for fish 50

Fishing for Proteins | 55 Footnotes

2) WHO technical report series 916: Diet, Nutrition and the Prevention of Chronic Diseases – Report of a Joint WHO/FAO Expert Consultation 2002. 2) Recommendation issued by the German Nutrition Society (DGE): 1 to 2 portions of fish per week. Recommended portion size is approx. 150 g = 225 g per week or 11.7 kg per capita per year. https://www.dge.de/ernaehrungspraxis/ vollwertige-ernaehrung/10-regeln-der-dge/ 3) SSPs were developed by the climate change research community (e.g. IPCC) in order to simplify the integrated analysis of the future effects of climate change. They lead to prognoses for population development and economic development, especially for the following elements: 1. population by age, gender and educational status; 2. urbanisation; and 3. Economic development (GDP). In addition to these basic elements, there are other hypothetical scenarios, among others, for 4. energy supply and use; 5. land use; 6. greenhouse gas emissions and air pollution; 7. average global radiative forcing and temperature changes; as well as 8. mitigation costs. 4) http://www.fao.org/fishery/cwp/search/en 5) See also section on fish consumption. 6) Standardised according to the limit values which, between 1988 and 2013, were slightly above the highest country values of the relevant indicator measured worldwide. 7) Among the most common micronutrient deficits, fish has the greatest potential to help alleviate vitamin A, iron and iodine deficits. This is particularly true for small species consumed whole with heads and bones, which can be an excellent source of many essential minerals such as iodine, selenium, zinc, iron, calcium, phosphorus and potassium, as well as vitamins such as A and D and several vitamins from the B group (Kawarazuka and Béné 2011). In addition, fish is usually low in saturated fats, carbohydrates and cholesterol with a few exceptions for selected species 8) A similar conclusion, albeit with a different intention, was reached by Thurstan and Roberts (2014). 9) Measured in USD purchasing power equivalent, i.e. correcting for exchange rate fluctuations using a hypothetical exchange rate to achieve equal purchasing power for a fixed basket of goods. This measure is often used in international comparisons in order to minimise the effects of (short-term) fluctuations in exchange rates when comparing the poverty status in several countries, for example. 10) As in the Allison et al. (2009a, 2009b) indicator. 11) For a full list of LMEs see appendix. The Antarctic and the Central Arctic Ocean are excluded due to a lack of data. 12) The full sets of parameter values, computation results and programming codes are available electronically as supplementary material. For the numerical calculation we employ the interior-point algorithm of the Knitro (version 9.1) optimisation software (Byrd et al. 1999; 2006). All programming codes were implemented in AMPL and are available as supporting material. 13) Details on the resulting accepted parameter values for the global predator-prey model are given in the technical appendix. 14) See appendix for a detailed list of substitution goods. Note: ‘other’ refers to the FAOstat specification ‘not elsewhere included’. 15) The following countries have been removed: Armenia, Austria, Azerbaijan, Belarus, Bermuda, Bhutan, Bolivia (Plurinational State), Botswana, Burkina Faso, Burundi, Central African Republic, Chad, Cook Islands, Czech Republic, Czechoslovakia, Ethiopia, Ethiopia PDR, Republic of Fiji, French Polynesia, Guam, Hungary, Kazakhstan, Kiribati, Kyrgyzstan, Lao People’s Democratic Republic, Lesotho, Luxembourg, Former Yugoslav Republic of Macedonia, Malawi, Mali, Marshall Islands, Mauritius, Federal States of Micronesia, Republic of Moldova, Mongolia, Nepal, Niger, Northern Mariana Islands, Palau, Palestine (Occupied Territories), Paraguay, Rwanda, Samoa, Serbia, Serbia and Montenegro, Slovakia, Swaziland, Switzerland, Tonga, Turkmenistan, Tuvalu, Uganda, Uzbekistan, Vanuatu, Yugoslavia SFR, Zambia, Zimbabwe. 16) https://tntcat.iiasa.ac.at/SspDb/dsd?Action=htmlpage&page=welcome 17) Note that the data underlying this study only considers marine catches in LMEs. High seas catches, aquaculture production and inland fisheries are not included.

56 References

Allison, E.H. et al. 2012. Rights-based fisheries governance: from fishing rights to human rights. Fish and Fisheries, 13: 14-29. DOI: 10.1111/j.1467-2979.2011.00405.x. Allison, E.H. et al. 2009a. Vulnerability of national economies to the impacts of climate change on fisheries. Fish and Fisheries. 10(2), 173-196 DOI: 10.1111/j.1467-2979.2008.00310.x. Allison, E.H. et al. 2009b. Climate change and fisheries: a comparative analysis of the relative vulnerability of 132 countries. Fish and Fisheries, 10: 173-196. Anderson, J.L. 1985. Market Interactions between Aquaculture and the Common-Property Commercial Fishery. Marine Resource Economics 2(1): 1-24. Crane, Russak & Company, Inc. Armington, P. 1969. A Theory of Demand for Products Distinguished by Place of Production. International Monetary Fund Staff Papers, XVI, 159-78. Asche, F. et al. 2015. Fair Enough? Food Security and the International Trade of Seafood. World Development, 67: 151-160. DOI: 10.1016/j.worlddev.2014.10.013.

Badjeck, M-C. et al. 2013. The vulnerability of fishing-dependent economies to disasters. FAO Fisheries and Aquaculture Circular No. 1081 FIPI/C1081 (En) Béné, C. et al. 2015. Feeding 9 billion by 2050 – Putting fish back on the menu. Food Security - The Science, Sociology and Economics of Food Production and Access to Food, 7: 261-274. Beveridge, M. 2013. Meeting the food and nutrition needs of the poor: the role of fish and the opportunities and challenges emerging from the rise of aquaculture. Journal of Fish Biology, 83 (4): 1067-1084. Beveridge, M., Phillips, M., Dugan, P. and Brummet, R. 2010. Barriers to aquaculture development as a pathway to poverty alleviation and food security. In: Advancing the Aquaculture Agenda: Workshop Proceedings, OECD, 345-359. Blanchard, J.L. et al. 2012. Potential consequences of climate change for primary production and fish production in large marine ecosystems. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 367, 2979-2989. BMEL (Bundesministerium für Ernährung und Landwirtschaft) 2015. Welternährung verstehen. Fakten und Hintergründe. Branch, T.A. 2008. Not all fisheries will be collapsed in 2048. Marine Policy 32:38-39. Branch, T.A., Jensen, O.P., Ricard, D., Ye, Y., Hilborn, R. 2010. Contrasting Global Trends in Marine Fishery Status Obtained from Catches and from Stock Assessments, Conservation Biology, 25 (4), 777-786. Britten, G.L. et al. 2014. Predator decline leads to decreased stability in a coastal fish community. Ecology Letters. DOI: 10.1111/ele.12354. Byrd, R., Hribar, M. and Nocedal, J. 1999. An interior point method for large scale nonlinear programming. SIAM Journal of Optimization 9 (4), 877-900. Byrd, R., Nocedal, J. and Waltz, R. 2006. Knitro: An integrated package for nonlinear optimization. In: di Pillo, G., Roma, M. (eds.), Large-Scale Nonlinear Optimization. Springer, p. 35.

Cheung, W.W.L. et al. 2010. Large-scale redistribution of maximum fisheries catch potential in the global ocean under climate change. Global Change Biology 16, 1: 24-35. Christensen, V. et al. 2014. A century of fish biomass decline in the ocean. Marine Ecology Progress Series, 512: 155-166. Cireira, X. and Masset, E. 2010. Income distribution trends and future food demand. Philosophical Transactions of the Royal Society B 365: 2821-2834. DOI: 10.1098/rstb.2010.0164. Clark, C.W. 1991. Mathematical Bioeconomics. 2nd Edition. John Wiley & Sons. New Jersey. Clark, C.W., Munro, G.R. and Sumaila U.R. 2005. Subsidies, buybacks, and sustainable fisheries. Journal of Environmental Economics and Management, 50(1), 47-58. Costello, C. et al. 2016. Global fishery futures under contrasting management regimes. Proceedings of the National Academy of Sciences of the United States of America. DOI: 10.1073/pnas.1520420113. Costello, C. et al. 2012. Status and Solutions for the World’s Unassessed Fisheries. Science, 338: 517-520. DOI: 10.1126/science.1223389. Costello, C., Gaines, S.D. and Lynham, J. 2008. Can Catch Shares Prevent Fisheries Collapse? Science, 321(5896): 1678-1681. DOI: 10.1126/science.1159478.

Essington, T.E. et al. 2015. Fishing amplifies forage fish population collapses. Proceedings of the National Academy of Sciences of the United States of America 112(21), 6648-6652. Evans, L.T. 1998. Feeding the ten billion – Plant and population growth. Cambridge University Press, 264 p.

FAO Committee on Fisheries 2014. Report of the fourteenth session of the Sub-Committee on Fish Trade. Bergen, Norway, 24-28 February 2014. FAO 2014. The State of World Fisheries and Aquaculture: Opportunities and Challenges. Food and Agriculture Organization of the United Nations. Rome.

Fishing for Proteins | 57 FAO 2016. The State of World Fisheries and Aquaculture: Opportunities and Challenges. Food and Agriculture Organization of the United Nations. Rome. FAO Fisheries and Aquaculture Department. Fisheries and aquaculture software. FishStatJ - software for fishery statistical time series. Rome. Updated 23 June 2015. [Cited 22 January 2016]. http://www.fao.org/fishery/statistics/software/fishstatj/en FAO Statistical database, http://faostat3.fao.org/home/E FAO 2016a. Species Fact Sheet Engraulis ringens. http://www.fao.org/fishery/species/2917/en FAO 2016b. Fishery and Aquaculture Country Profiles Indonesia. http://www.fao.org/fishery/facp/IDN/en FAO 2016c. Fishery and Aquaculture Country Profiles Peru. http://www.fao.org/fishery/facp/PER/en Fenichel, E.P. et al. 2016. Measuring the value of groundwater and other forms of natural capital Proceedings of the National Academy of Sciences of the United States of America, 113 (9) 2382-2387; published ahead of print 8 February 2016. DOI: 10.1073/pnas.1513779113 Frid, C.L.J. and Paramor, O.A.L. 2012. Food for Thought. Feeding the world: what role for fisheries? ICES Journal of Marine Science, 69: 145-150. DOI: 10.1093/icesjms/fsr207. Froese, R. and Proelss, A. 2010. Rebuilding fish stocks no later than 2015: will Europe meet the deadline? Fish and Fisheries, 11: 194-202. DOI: 10.1111/j.1467-2979.2009.00349.x.

Grafton, R.Q. et al. 2005. Incentive-based approaches to sustainable fisheries. Canadian Journal of Fisheries and Aquatic Sciences 63, 699-710. Grafton, R.Q., Kompas, T., Hilborn, R. (2007): Economics of Overexploitation Revisited.In: Science 318, 1601.

Hannesson, R. 1983. Optimal Harvesting of Economically Interdependent Fish Species. Journal of Environmental Economics and Management, 10: 329-345. HLPE, 2014. Sustainable fisheries and aquaculture for food security and nutrition. A report by the High Level Panel of Experts on Food Security and Nutrition of the Committee on World Food Security, Rome 2014.

Jennings, S. et al. 2016. Aquatic food security: insights into challenges and solutions from an analysis of interactions between fisheries, aquaculture, food safety, human health, fish and human welfare, economy and environment. Fish and Fisheries. DOI: 10.1111/faf.12152.

Kawarazuka, N. and Béné, C. 2011. The potential role of small fish species in improving micronutrient deficiencies in developing countries: building evidence, 14: 1927-1938. DOI: 10.1017/S1368980011000814. Kelleher, K. 2009. The Sunken Billions. The Economic Justification for Fisheries Reform. Sustainability Criteria for Fisheries Subsides: The Latin American Context. 29-30 July 2009, Guayaquil, Ecuador. The World Bank. Kjesbu O. S. et al. 2014. Synergies between climate and management for Atlantic cod fisheries at high latitudes. Proceedings of the National Academy of Sciences of the United States of America 111 (9), 3478-3483.

Lam, V.W.Y., Cheung, W.W.L., Swartz, W. and Sumaila, U.R. 2012. Climate change impacts on fisheries in West Africa: implications for economic, food and nutritional security. African Journal of Marine Science 34(1):103-117.

Martell, S. and Froese, R. 2013. A simple method for estimating MSY from catch and resilience. Fish and Fisheries, 14:504-514. Merino, G. et al. 2012. Can marine fisheries and aquaculture meet fish demand from a growing human population in a changing climate? Global Environmental Change, 22(4): 795-806. DOI: 10.1016/j.gloenvcha.2012.03.003. McClanahan, T., Allison, E.H. and Cinner, J.E. 2015. Managing fisheries for human and food security. Fish and Fisheries, 16: 78-103. DOI: 10.1111/faf.12045. Mora, C., Myers, R.A., Coll, M., Libralato, S., Pitcher, T.J., Sumaila, R.U., Zeller, D., Watson, R., Gaston, K.J. and Worm, B. 2009. Management Effectiveness of the World’s Marine Fisheries. PLOS Biology, 7(6): 1-11. MRI (Max Rubner-Institut) 2008. Nationale Verzehrsstudie II. Ergebnisbericht, Teil 2. Karlsruhe. https://www.mri.bund.de/ fileadmin/MRI/Institute/EV/NVSII_Abschlussbericht_Teil_2.pdf

O’Neill, B.C. et al. 2015. The roads ahead: Narratives for shared socioeconomic pathways describing world futures in the 21st century, Global Environmental Change. DOI: 10.1016/j.gloenvcha.2015.01.004.

Pauly, D. et al. 2002. Towards sustainability in world fisheries. Nature, 418, 689-695. Pauly, D., Watson, R., Alder, J. 2005. Global trends in world fisheries: impacts on marine ecosystems and food security. Philosophical Transactions of the Royal Society B, 360: 5-12. DOI: 10.1098/rstb.2004.1574. Pauly, D. and Zeller, D. 2015. Catch reconstruction: concepts, methods and data sources. Online Publication. Sea Around Us (www.seaaroundus.org). University of British Columbia. Pauly, D. and Zeller, D. 2016. Catch reconstructions reveal that global marine fisheries catches are higher than reported and declining. Nature Communications, 10244. DOI: 10.1038/ncomms10244.

Quaas, M.F. et al. 2012. Fishing industry borrows from natural capital at high shadow interest rates. Ecological Economics 82, 45-52.

58 Quaas, M.F. and Requate, T., 2013: Sushi or fish fingers? Seafood diversity, collapsing fish stocks and multi-species fishery management. Scandinavian Journal of Economics, 115(2): 381-422. Quaas, M.F. et al. 2016. It’s the economy, stupid. Projecting the fate of fish populations using ecological-economic modeling. Global Change Biology, 22(1): 1365-2486. DOI: 10.1111/gcb.13060.

Sea Around Us, www.seaaroundus.org. University of British Columbia. Sherman, K. and Hempel, G. (Editors) 2009. The UNEP Large Marine Ecosystem Report: A perspective on changing conditions in LMEs of the world’s Regional Seas. UNEP Regional Seas Report and Studies No. 182. United Nations Environment Programme. Nairobi, Kenya. Smith, M.D. et al. 2010. Sustainability and Global Seafood. Science, 327: 784-786. DOI: 10.1126/science.1185345. Squires, D. and Vestergaard, N. 2013. Technical Change and the Commons. Review of Economics and Statistics, 95(5), 1769-1787. Sumaila, U.R. et al. 2012. Benefits of Rebuilding Global Marine Fisheries Outweigh Costs. PLOS one, 7(7). DOI: 10.1371/journal.pone.0040542. Sumaila, U.R., Teh, L., Zeller, D. and Pauly, D. 2015. The global ex-vessel fish price database. In: D. Pauly and D. Zeller (eds.). Catch reconstructions: concepts, methods and data sources. Online Publication. Sea Around Us (www.seaaroundus. com). University of British Columbia.

Tacon, M. and Metian, M. 2013. Fish Matters: Importance of Aquatic Foods in Human Nutrition and Global Food Supply. Reviews in Fisheries Science, 21(1):22-38. DOI: 10.1080/10641262.2012.753405. Thiao, D. et al. 2012. Economic dimension of the collapse of the ‘false cod’ Epinephelus aeneus in a context of ineffective management of the small-scale fisheries in Senegal. African Journal of Marine Science 34 (3): 305-311. Thilstedt, S.H. et al. 2016. Sustaining healthy diets: The role of capture fisheries and aquaculture for improving nutrition in the post-2015 era. Food Policy, Vol 61, 126–131. Thurstan, R.H. and Roberts, C.M. 2014. The past and future of fish consumption: can supplies meet healthy eating recommendations? Marine Pollution Bulletin, 89(1-2): 5-11. DOI: 10.1016/j.marpolbul.2014.09.016. Tilman et al. 2002. Agricultural sustainability and intensive production practices. Nature 418, 671-677

United Nations, Department of Economic and Social Affairs, Population Division (2015). World Population Prospects: The 2015 Revision, http://esa.un.org/unpd/wpp/

Voss, R., Quaas, M.F., Schmidt, J.O., Kapaun, U. 2015. Ocean acidification may aggravate social-ecological trade-offs in coastal fisheries. PLoS ONE 10(3): e0120376. Voss, R. et al. 2011. Temperature change and Baltic sprat: from observations to ecological-economic modelling. ICES Journal of Marine Science, 68(6), 1244-1256.

Watson, R., Gaston, K. J., Worm, B. 2009. Management effectiveness of the world’s marine fisheries. PLoS Biology 7 (6), e1000131. World Bank 2008. The Sunken Billions. The Economic Justification for Fisheries Reform. Profish. World Bank 2015. The Sunken Billions revisited. World Bank: GDP. http://data.worldbank.org/indicator/NY.GDP.MKTP.CD World Resources Institute, 2013. Creating a Sustainable Food Future. A menu of solutions to sustainably feed more than 9 billion people by 2050. World Resources Report 2013-14:Interim Findings, http://www.wri.org/sites/default/files/wri13_report_4c_wrr_online.pdf Worm, B. et al. 2006. Impacts of biodiversity loss on ocean ecosystem services. Science 314, 787-790. Worm, B. et al. 2009. Rebuilding Global Fisheries. Science 325 (5940), 578-585. WWF 2015. Das große Wegschmeißen. Vom Acker bis zum Verbraucher: Ausmaß und Umwelteffekte der Lebensmittelverschwendung in Deutschland.

Acknowledgement

WWF would like to thank the following people who were very helpful in providing critical feedback and valuable information in the preparation of this study: Rolf Willmann (former fisheries expert with the FAO) Edward H. Allison (University of Washington, Seattle) Birgit Meade (Agricultural Economist, USA) Rashid Sumaila (Fishery Economics Research Unit, University of British Columbia, Vancouver) Mark Prein and Anneli Ehlers (Deutsche Gesellschaft für Internationale Zusammenarbeit, GIZ)

Fishing for Proteins | 59 © Copyright of WWF International ® Trademark of WWF International • Printed on 100 % recycled paper • Effective 10/16 CLED 00% 1 RECY (0)30 311 777 0 (0)30 311 777 199 (0)30 311 +49 +49 WWF Deutschland 18 Reinhardtstr. Berlin | GermanyD-10117 Tel.: Fax:

future in which humans live in harmony with nature. www.de | [email protected] Why we are here stop the degradationTo of the planet’s natural environment and to build a