Factors affecting greenhouse gas emissions in fisheries: evidence from Iceland's demersal fisheries

Abstract

Fishing produces low CO₂ emissions per unit output compared to other animal protein sources. However, emissions from fishing grew by 28% from 1990 to 2011 and fishing currently contributes about 4% of the emissions of world food production. The purpose of this paper is to identify the relationship between various factors and emissions in fisheries. We analyse the development of CO₂ emissions from the Icelandic individual transferable quota regulated fishing fleet from 1997 to 2018. The results show that emissions per unit of catch fell around 40% during this period. The main findings are that overall catches and abundance are by far the most important factors determining emissions, the bigger the catches and the greater the abundance, the smaller the emissions per unit of output. Fuel prices are a distant third factor and technological change has played a minor role in this development. In addition, the importance of different factors affecting emissions, varies between vessels depending on types of fishing gear. The results indicate that building up fish stocks not only increases output but also increases profitability and reduces emissions per unit of output, as long as the fisheries management system preserves incentives for efficient fishing.

carbon emissions, fisheries, Iceland, sustainability

Introduction

Food production is one of the largest contributors to anthropogenic greenhouse gas (GHG) emissions. Its share is estimated around a quarter of all emissions (Steinfeld et al., 2006; Edenhofer, 2015). The production of animal protein is growing, and is estimated to account for around half of all GHG emissions in food production during the last two to three decades (Garnett, 2009; Pelletier and Tyedmers, 2010; Herrero et al., 2016; Webb and Buratini, 2018). Carbon dioxide (CO₂) is the most important GHG, because it contributes around 65% of total GHG emissions from anthropogenic sources (IPCC, 2014). Fishing and aquaculture produce protein with lower emissions per unit of output than almost all land-based animal protein sources (Hilborn et al., 2018; Turrell, 2019; Tsakiridis et al., 2020). Still, fisheries are a significant contributor to global CO₂ emissions, as fossil fuel is its main energy source (Tyedmers, 2004; Driscoll and Tyedmers, 2010; Parker et al., 2015a; Parker et al., 2018). Estimated CO₂ emissions of the world's fishing fleet were in the range of 178–207 million in 2016 (Parker et al., 2018; Greer et al., 2019), which was around 0.5% to the world's CO₂ emissions from human activities and 4% of emissions associated with food production. Emissions from the world's fishing industry grew by 28% from 1990 to 2011 and the emissions per kilo of landed catch increased by 21% during this period on average in all of the world's fisheries (Parker et al., 2018). Therefore, it is obviously important to understand this development and identify the best ways to lower emissions in fisheries.

GHG emissions in fisheries can be reduced in six ways. Firstly, fuel utilization can be improved by better technology. This is a gradual, ongoing process influenced by regulations, technical development, and fuel prices (Abernethy et al., 2010; Parker et al., 2015b). Secondly, by utilizing more of the catch for human consumption instead of using it as feed for aquaculture, as is common for many pelagic species (Saevaldsson and Gunnlaugsson, 2015; Parker et al., 2018). This is because utilizing the catch for human consumption shortens the food chain which leads to lower emissions (Garnett, 2011). Thirdly, by the selection of fishing gear. Generally, passive fishing gear like long lines and gillnet are more fuel efficient (i.e. less fuel per unit of catch) than active gear like trawls (Ziegler and Hansson, 2003; Schau et al., 2009; Avadí and Fréon, 2013) and purse seiner more efficient than most other methods (Thrane, 2004; Tyedmers and Parker, 2012). However, there is high variability in fuel efficiency, especially within some passive fishing gear methods like pots and traps and hooks and lines. Some fishers using these methods are very fuel efficient whereas some are very fuel inefficient (Parker and Tyedmers, 2015). Fourthly, taxing emissions (carbon taxes) should in theory reduce emissions because of lower fishing effort, smaller fishing fleets and more incentives to develop more fuel efficient technologies (Parker et al., 2015b; Waldo and Paulrud, 2017).

Fifthly, the fisheries management system can affect emissions (Waldo et al., 2016). In 1991, a uniform individual transferable quota (ITQ) system was implemented in Iceland, including almost all the country's fisheries (Matthíasson, 2003). Normally, the main rational for managing fisheries by an ITQ system is to promote economic efficiency and cost reduction (Grafton, 1996). However, catch share management systems, like ITQ systems, are well suited to reduce emissions. Emissions are lower in ITQ managed fisheries than compared to other, less flexible management systems governing the same fisheries (Ziegler et al., 2016). ITQ management promotes so-called “autonomous adjustment” of the fishing effort, where fleet size adjusts to an optimal level through trade and consolidation (Hoshino et al., 2020). Waldo et al., (2016) examined CO₂ emissions from Nordic fisheries (i.e. in Denmark, the Faeroes, Iceland, Norway, and Sweden). Their main finding was that fisheries management and optimal harvesting of fish stocks were the most important contributors to lowering emissions. The paper concluded that significant reductions in CO₂ emissions were possible in all of these fisheries except for the Icelandic fisheries because it already had an efficient fisheries management system (i.e. the ITQ system). A similar analysis by Waldo and Paulrud, (2017) concluded that the most important step to reduce GHG emissions in Swedish fisheries was to improve harvest efficiency of the fishing fleet and reduce its size. The paper concluded that by introducing an ITQ system, emissions would be reduced by 30%, because less effort would be undertaken in fishing. Only a part of Swedish fisheries is currently managed by an ITQ system. Merayo et al., (2018) analysed Danish fisheries. An ITQ system was partially introduced in the country's fisheries in 2003 and fully in 2007. They concluded that an ITQ system can be a good solution to reduce overcapacity in fisheries. In addition, it would have positive environmental effects because of lower fishing effort that would reduce fuel consumption and hence lower emissions.

The sixth and final item is to improve fisheries management and allow fish stocks to grow, leading to an increase in catch per unit effort (CPUE) and resulting in a reduction in fuel consumption per unit of catch. This has mostly been studied theoretically. Examples are a paper by Arnason, (2007) where the theory was introduced that higher biomass would lead to higher CPUE and hence lower emissions. Parker et al., (2017) concluded that management systems that lead to higher biomass would lead to lower emissions in Australian lobster fisheries. Extensive theoretical research examined the GHG emissions of Tasmanian southern rock lobster (Jasus edwardsii) fisheries. It concluded that targeting maximum economic yield when managing the fisheries, instead of targeting maximum sustainable yield, resulted in a total GHG emissions reduction of 80%. This was because of bigger biomass of the lobster stock and hence higher CPUE (Farmery et al., 2014). Finally, a recent paper by Hornborg and Smith, (2020) used a simple production model (i.e. Pella–Tomlinson model) to examine how fuel efficiency varied in accordance with the level of exploitation and biomass depletion. The findings were that the fuel use per unit of catch was highly related to fishing effort and stock size. The more effort and hence less stock size, the lower the fuel efficiency. Studies assessing the actual fuel efficiency in fisheries include a paper by Ziegler and Hornborg, (2014). They studied the fuel efficiency of Swedish demersal trawl fisheries from 2002 to 2010. They showed that the difference in fuel efficiency between large and small trawlers was small. The main findings were that the fuel usage was negatively correlated to stock size, with the larger the stock of eastern Baltic cod the smaller the fuel use per landed kilogram. Schau et. al. (2009) analysed energy usage of the Norwegian fishing fleet from 1980 to 2005. They showed that fuel efficiency depended on catch rate (i.e. catch per day), especially for trawlers. Parker and Tyedmers (2015) studied the energy consumption of global fisheries since 1990. They showed a significant variability in fuel consumption per unit of catch, where small pelagic fisheries were the most efficient and crustaceans the least. Among their findings were that fuel efficiency in Europe and Australia has improved since the beginning of the 21st century. They concluded that the most effective improvement in fuel efficiency in fisheries would come by rebuilding fishing stocks where they are overexploited.

This paper contributes to this literature by analysing the GHG emissions of Icelandic demersal fisheries from 1997 to 2018, as well as identifying the factors that affect the emissions. The aim of the paper is to examine what factors affect GHG emissions in Iceland's demersal fisheries. Iceland is one of the largest fishing nations in the world, with annual average catches of about 1.4 million tonnes during the last two decades, and is classified as one of the world's leading demersal fishing nations (FAO, 2017). It is also well managed, with a conservative harvest rule and an effective management system (Elvarsson et al., 2020; Gunnlaugsson et al., 2020). In this research, a theoretical model is developed that describes the production of a wanted good, in this case fish, which also leads to the production of an unwanted good, GHG emissions. The model is based on classic microeconomic theory and the variables that determine the production of the individual firm are prices of inputs and outputs, overall catch limits, and stock size. An empirical model is developed and applied to data for five different fleet segments using both passive and active gears to estimate the relationship between the different exogenous variables and the production of GHG emissions.

Icelandic fisheries

History and management

Fisheries have been the backbone of the Icelandic economy for centuries (Arnason, 1993). Throughout the 20th century and until 2014, fisheries were the country's main source of export earnings. Cod (Gadus morhua) has historically been the most important species in Iceland fisheries. In 2018, cod was 22% of the total catch of the Icelandic fishing fleet. The value of cod landings represented around 45% of the overall value of Icelandic landings and 56% of the value of demersal landings in 2018. The second most important demersal species was haddock (Melanogrammus æglefinus), with 8% of the total catch and 10% of the total demersal catch value. In total, the catch value of Icelandic demersal fisheries was 80% of the landed value of Icelandic vessels in 2018. The remaining value is mostly pelagic species (Statistics Iceland, 2020a).

Iceland secured full control of its fishing resources in 1975 when the Exclusive Economic Zone (EEZ) reached 200 nautical miles. Various effort based management systems were used from 1975 in the demersal fisheries, with very poor results (Runolfsson and Arnason, 2001). Total Allowable Catch (TAC) exceeded the recommendations of scientists and catches exceeded the TAC during most of these years (Danielsson, 1997). By the mid-eighties, the industry was losing significant amounts of money every year and the most important stocks were in decline (Arnason, 1996). Something had to be done. Individual vessel quotas (IVQ) were introduced in most demersal fisheries in 1984. Then, in 1991, a uniform ITQ system was set which included almost all of Iceland's fisheries, making the quota transferable between vessels (Matthíasson, 2003). Normally, the main motive for managing fisheries by an ITQ system is to promote economic efficiency and lower costs (Hannesson, 2004; Arnason, 2008). ITQ systems can also help to improve stock health. This is achieved by curbing overexploitation and creating incentives for the fishing industry to invest in the rebuilding of overexploited fish stocks (Chu, 2009; van Putten et al., 2014). In Iceland, implementing a sustainable TAC proved easier under an ITQ management system than the systems that had proceeded it. This is because catches generally exceeded the TAC until the ITQ system was fully established (Danielsson, 1997; Matthíasson, 2003; ICES, 2019).

Biological management of the demersal stocks

In this study, the 20 most important demersal species are included, as well as the benthic invertebrates as they are partly targeted by the same fleet sectors (Table 1). We also include pelagic beaked redfish (Sebastes mentella) as this species is targeted by traditional trawlers rather than dedicated pelagic vessels, as is the case with other pelagics. These species cover about 97% of the total demersal catch since 1990 and therefore all commercially important demersal species.

Table 1.

Demersal stocks used in this analysis.

A: Name of stock	B: Biomass method	C: Fleet sectors	D: Proportion of catch in codeq.	E: Biomass index scaling factor	F: Codeq
Atlantic cod (Gadus morhua)	Fishable biomass ages 4+ (t)	SMT	46.4%	1.00	1.00
Haddock (Melanogrammus aeglefinus)	Fishable biomass 45 + cm (t)	SMT	12.5%	1.00	1.03
Saithe (Pollachius virens)	Fishable biomass +40 cm (t)	SMT	6.8%	1.00	0.57
Golden redfish (Sebastes norvegicus)	Fishable biomass ages 3+ (t)	T	5.8%	1.00	0.63
Atlantic wolffish (Anarhichas lupus)	Fishable biomass (t)	SMT	1.9%	1.00	0.72
European plaice (Pleuronectes platessa)	Fishable biomass (t)	MT	1.5%	1.00	0.97
Norway lobster (Nephrops norvegicus)	Fishable biomass 6+ (t)	M	0.8%	1.00	2.34
Ling (Molva molva)	Fishable biomass +75 cm (t)	MT	0.7%	1.00	0.55
Tusk (Brosme brosme)	Fishable biomass +40 cm (t)	MT	0.4%	1.00	0.42
Greenland halibut (Reinhardtius hippoglossoides)	CPUE	T	6.9%	70.3	1.85
Northern shrimp (Pandalus borealis)	Survey index	MT	5.1%	2.21	0.97
Beaked redfish (demersal) (Sebastes mentella)	Survey index	T	3.9%	0.632	0.77
Beaked redfish (pelagic) (Sebastes mentella)	Survey index	T	3.7%	80.6	0.77
Lumpsucker (Cyclopterus lumpus)	Female survey index	S	1.1%	1.72	1.10
Angler (Lophius piscatorius)	Survey index	MT	0.6%	6.65	1.84
Lemon sole (Microstomus kitt)	Survey index	MT	0.5%	1.50	1.41
Great silver smelt (Argentina silus)	Survey index	T	0.5%	1.11	0.41
Spotted wolffish (Anarhichas minor)	Survey index	MT	0.3%	0.84	0.85
Atlantic halibut (Hippoglossus hippoglossus)	Survey index	MT	0.3%	3.59	2.63
Iceland scallop (Chlamys islandica)	Survey index	M	0.3%	0.25	0.36

Columns are A: Name of stock. B: Biomass estimation method available, i.e. is it a true biomass estimate or an index that needs to be calibrated to actual stock size. C: What demersal fleet sectors utilize this species (S = small vessel, M = medium, T = trawlers). D: Proportion of total demersal catch in cod equivalents from 1990–2019, this is an indicator of the relative value of the stock in the fisheries during this period. E: Scaling factor to scale the survey indices to arrive at a more realistic true biomass value, see main text for further explanation. F: Average value in cod equivalents since 1991. Sources: Marine and Freshwater Research Institute and the Directorate of Fisheries.

Fishable biomass estimates are available for nine of these species that are about 80% of the total catch volume. Most of the other species only have relative survey biomass indices and one, the Greenlandic halibut (Reinhardtius hippoglossoides), has only commercial CPUE. The indices are usually represented as biomass and mostly calculated based on swept area methods. However, these biomass estimates are only indices, sometimes the indices were lower than the catch in a given year and therefore not realistic. Therefore, they must be scaled for more realistic true biomass estimates. We have, therefore, scaled them so that the lowest ratio of biomass index versus catch was 1.5. This ought to give a more realistic scale for the biomass than using the raw values. We acknowledge that this is a weakness in the analysis but expect the error caused by this to be small as the largest stocks, particularly cod, have true biomass estimates.

Some stocks do not have biomass estimates available for the entire period, e.g. biomass indices are only available for deep-water species since 2000. In these cases, so as not to have the biomass as zero, the biomass estimates from the closest year available was used. The species biomass was weighed using their relative unit price to the price of cod to get an indicator on the value of the fish stocks in the sea. This relative price, known as cod equivalent (codeq), is widely used in aggregation in Iceland due to the dominance of the cod stock in the nation's fisheries (Gunnlaugsson et al., 2018). The cod equivalent value refers to the value of the species compared to cod, hence, the value of one tonne of cod is always one codeq. Haddock, with a value of 1.03 (Table 1), is therefore slightly more valuable than cod, 1.03 tonne of cod are equal in value to 1 tonne of haddock. Codeq values were first calculated in 1991 and the values we use are the average for each species since then. The same cod equivalent value is used for all years.

The total biomass of all demersal species was 2.3 million tonnes in 1990 and 2.0 million tonnes cod equivalents, as seen in Figure 1. Both indices declined after that and were low until around 2000 when they began increasing again to the current level of about 2.8 million tonnes biomass and 2.4 million tonnes cod equivalents. This change is almost solely due to fluctuations in the biomass of cod. Biomass of other species measured in cod equivalents was remarkably stable through the period at about 1 million tonnes except with a dip between 1993 and 2003.

Figure 1.

Size of the demersal fish stocks (fishable biomass) in the Icelandic EEZ in weight (thin whole line) and in cod equivalents (bold whole line). Also shown are the amount of other species than cod in cod equivalents (thick dashed line) and the biomass of cod (thin dashed line). Note that as the cod equivalent of cod is 1 then the stock size of cod is the same in weight and cod equivalents. Sources: Marine and Freshwater Research Institute and the Directorate of Fisheries.

In the 1980s, there was considerable overfishing in Iceland's fisheries as catches were usually much higher than levels recommended by scientists (Danielsson, 1997). Fishing stocks were therefore overfished and some depleted. Gradually, catches have been getting closer to the advice by the Icelandic Marine and Freshwater Research Institute, which the Ministry of fisheries bases its TAC on. In 1995, an important milestone was reached in the management of the Icelandic cod stock. That year, a harvest control rule (HCR) was introduced. HCRs have gradually been introduced to more species, ensuring that the catch is close to the advised TAC. This has lowered the fishing mortality and the stock (especially the cod stock) has been rebuilt as Figure 1 shows. According to the rule, the TAC in year t is:

$$\begin{eqnarray*} TA{C_t} = \frac{{\left( {TA{C_{t - 1}} + H*B{4^ + }} \right)}}{2},\,\, \end{eqnarray*}$$

(1)

where B4⁺ is the estimated biomass of cod that is at least 4-years old (i.e. the fishable stock). H is the harvest rate tuning parameter. It was 0.25 from 1995 to 2010 when it was reduced to 0.20 and has not changed since that year (Elvarsson et al., 2020; MFRI, 2020). Seven other species are now also managed with harvest control rules and new species are regularly being added.

Oil consumption in the demersal fisheries

Table 2 shows the estimated oil consumption by Iceland's demersal fisheries. The table shows the average oil consumption per landed kilo in cod equivalent kilos of all demersal fisheries. In addition, the table shows each of the five fleet classes examined in this paper. The table shows the average of the first four years and the last four years for the time period covered by this study. The average reduction was around 40% over the study period. It was remarkably similar for all of the five fleet classes. The greatest reduction was observed in boats larger than 200 gross registered tonnage (GRT) and fresh fish trawlers. It was almost 50% for both fleet classes. The lowest reduction was among freezer trawlers, around 36%. The table indicates that the boat category (i.e. the three boat classes in this study) is more efficient than the trawlers; as their fuel usage is considerably lower per unit of catch.

Table 2.

Estimated oil consumption (kilogram oil per kilo of catch in cod equivalent kilos) of all Iceland's demersal fisheries (weighted average) and in the five fleet classes.

Bs	1997–2000	2015–2018	Change
Weighted average of all demersal fisheries	1.47	0.89	−39.5%
Boats < 10 GRT	1.21	0.71	−41.4%
Boats 10–200 GRT	1.15	0.63	−45.1%
Boats > 200 GRT	1.16	0.58	−49.5%
Fresh fish trawlers	2.03	1.03	−49.2%
Freezer trawlers	1.51	0.97	−35.7%

Source: Statistics Iceland.

Material and methods

Theory

We assume that fishing is an economic activity, driven by the goal of profit maximization. Our behavioural model is therefore based on the standard microeconomic model of the profit maximizing firm. We assume a production technology involving a vector of wanted outputs, y, and a vector of unwanted outputs, e, produced using a vector of inputs, x. Since the analysis in this paper deals with fisheries, we assume that production is affected by natural conditions, z. We write the production set as

$$\begin{eqnarray*} g\left( {y,x,z,e} \right) \le 0. \end{eqnarray*}$$

(2)

We assume the function g() is continuous and differentiable. We assume that the total quantity of the fleet is regulated by a total allowable catch limit, |$\bar y$|⁠. Producers face a vector of prices, p, for outputs and a vector of input prices, w. We assume the unwanted output is a pure externality and is not priced. Therefore, the firm solves the following optimization problem

$$\begin{eqnarray*} &&{\max _{y,x}}\left\{ {p'y - w'x} \right\}\\ &&\rm{S.t.}\\ && g\left( {y,x,z,e,\bar y} \right) \le 0,\\ && z = {z_0},\\ && y \le \bar y, \end{eqnarray*}$$

(3)

where z₀ is the state of the natural conditions. This results in a set of output supply functions and input demand functions conditional on natural conditions

$$\begin{eqnarray*} x\left( {p,w|{z_0},\bar y} \right)\hbox{ and }y\left( {p,w|{z_0},\bar y} \right). \end{eqnarray*}$$

(4)

Substituting (4) into (2) gives the indirect production function of the firm, expressing the profit maximizing, technically feasible production mixes of inputs and outputs.

$$\begin{eqnarray*} {\rm{\Gamma \,\,}}\left( {p,w,{z_0},e,\bar y} \right) = \,\,g\,\,\left( {y\left( {p,w|{z_0},\bar y} \right),x\left( {p,w|{z_0},\bar y} \right),{z_0},e} \right) = \,\,0\\ \end{eqnarray*}$$

(5)

Under the assumption that the firm is profit maximizing, it is efficient, and the expression holds with equality. This expression gives us the latent relationship between output and input prices and the production of the unwanted good. This relationship can be used to derive the effects of changes in prices of inputs and outputs on the production of the unwanted good.

It is assumed that the production of the unwanted output does not impact the production of the wanted output. This implies that the production of the unwanted output is separable from production of the wanted output, operating in a dimension of its own. Under the assumption that an inverse function of |${\rm{\Gamma }}( {p,\,\,w,\,\,{z_0},\,\,e,\,\,\bar y} )$| with respect to unwanted good, e, exists, the following expression can be obtained for the unwanted output

$$\begin{eqnarray*} {\rm{\Gamma }}_e^{ - 1}\,\,\left( {p,w,{z_0},\bar y} \right) = \,\,e, \end{eqnarray*}$$

(6)

where the subscript, e, to the function indicates the dimension of the unwanted output. Expression (6) gives us the profit maximizing production of the unwanted output under any array of input and output prices, natural conditions, and overall catch limits. It can be regarded as the implicit supply function of the unwanted good.

Data

Data on costs, prices, and quantities of inputs and outputs was obtained from Statistics Iceland. It contains yearly data on several different segments of the Icelandic fishing fleet. This analysis focuses on the five fleet segments that operate in the demersal fishery, three using passive gear (mostly hand line, longline, and gillnets) and two using active gear (different types of trawlers). The data covers the period 1997–2018. Yearly data on carbon emissions was collected from the Icelandic Environment Agency. This data is based on fuel consumption and is collected as a part of Iceland's reporting to the United Nations Framework Convention on Climate Change through annual National Inventory Reports. Yearly data on stock sizes was collected from the Icelandic Marine and Freshwater Research Institute. It includes data on the 20 commercially exploited species of demersal fish reported in Table 1. An aggregate abundance index and an aggregate output index were constructed to simplify the analysis. It weighs together different species in cod equivalents, as described above. Table 3 shows the descriptive statistics of the variables used in this study.

Table 3.

Descriptive statistics of the variables studied in this analysis.

Variable	Average	Standard deviation	Unit
Oil price	40.61	22.72	Icelandic krona per liter
Output price	169.78	57.66	Icelandic krona per kilo
Composite input pricea	6 102	1316.81	Index
Stock size	1.83	0.35	Million tonnes
Output quantity
Boats < 10 GRT	41.96	20.08	Thousand tonnes
Boats 10–200 GRT	107.99	29.47	Thousand tonnes
Boats > 200 GRT	93.39	31.65	Thousand tonnes
Fresh fish trawlers	88.35	17.03	Thousand tonnes
Freezer trawlers	210.96	41.88	Thousand tonnes
Emissions
Boats < 10 GRT	41.35	28.67	Thousand tonnes CO2 eq
Boats 10–200 GRT	96.72	42.92	Thousand tonnes CO2 eq
Boats > 200 GRT	81.02	26.08	Thousand tonnes CO2 eq
Fresh fish trawlers	125.50	34.32	Thousand tonnes CO2 eq
Freezer trawlers	279.15	68.43	Thousand tonnes CO2 eq

Sources: Statistics Iceland, Icelandic Environment Agency, and Marine and Freshwater Research Institute.

a

This is a measure of quantity of other inputs in constant prices.

Empirical model

We assume a Cobb Douglas production technology (see e.g. Green 2008). This model is chosen as it is one of the the simplest empirical models consistent with economic theory and does therefore not require large datasets for identification. We assume that the production of the unwanted output could be related to either inputs or outputs. With one wanted output, one unwanted output, n inputs and one natural condition, expression (6) for the dimension of the unwanted output can be written as

$$\begin{eqnarray*} {{\rm{\Gamma }}_e}\,\,\left( {p,w,{z_0},e,\bar y} \right) &=& y\left( {p,w|{z_0},\bar y} \right)z_0^\vartheta {\bar y^\eta }{e^\lambda }\\ &&-\,\, \prod {\left( {{x_i}\left( {p,w|{z_0},\bar y} \right)} \right)^{{\alpha _i}}}z_0^\theta {\bar y^\delta }\,\,{e^\gamma } = 0,\\ y\left( {p,w|{z_0},\bar y} \right)\,\,{e^\lambda } &=& \prod {\left( {{x_i}\left( {p,w|{z_0},\bar y} \right)} \right)^{{\alpha _i}}}z_0^\theta {e^\gamma },\\ e &=& {\left( {\frac{{y\left( {p,w|{z_0}} \right)z_0^\vartheta {{\bar y}^\eta }}}{{\prod {{\left( {{x_i}\left( {p,w|{z_0}} \right)} \right)}^{{\alpha _i}}}z_0^\theta }}} \right)^{\frac{1}{{\gamma - \lambda }}}}\\ &=& {\rm{\Gamma }}_e^{ - 1}\,\,\left( {p,w,{z_0},\bar y} \right). \end{eqnarray*}$$

(7)

This expression leads to an empirical model that is simplified as

$$\begin{eqnarray*} {\rm{ln\,\,}}\left( {{e_j}} \right) &=& {\beta _{j,0}}\,\, + \sum {\beta _{j,i}}{\rm{ln}}\left( {{w_i}} \right) + {\beta _{j,n + 1}}{\rm{ln}}\left( p \right) + {\beta _{j,n + 2}}{\rm{ln}}\left( {{z_0}} \right)\\ &&+\, {\beta _{j,n + 3}}{\rm{ln}}\left( {{{\bar y}_j}} \right), \end{eqnarray*}$$

(8)

where j denotes fleet segment and |${\beta _{i,\,\,j}}$| are sums of the parameters of the production technology for variable i in fleet segment j. The estimated parameters represent the elasticities of emissions supply with respect to prices, stock size, and catch limits, e.g. |$\frac{{\delta {\rm{ln}}( {{e_j}} )}}{{\delta {\rm{ln}}( {{z_0}} )}} = \frac{{\delta {e_j}}}{{{e_j}}}\,\,\,\,{( {\frac{{\delta {z_0}}}{{{z_0}}}} )^{ - 1}} = {\beta _{j,\,\,n + 2}}\,\,$|⁠. Note that all fleet segments are assumed to meet the same price incentives and natural conditions but catch limits vary, since overall catch limits for fleet segments can change by quota trade between fleet segments. Homogeneity of degree zero is imposed on the model by |$\mathop \sum \limits_{i\,\, = \,\,1}^{n + 1} {\beta _{j,\,\,i}} = \,\,0$|⁠.

An empirically estimable version of expression (8) is

$$\begin{eqnarray*} {\rm{ln\,\,}}\left( {{e_j}} \right) &=& {\beta _{j,0}}\,\, + \sum {\beta _{j,i}}{\rm{ln}}\left( {{w_i}} \right) + {\beta _{j,n + 1}}{\rm{ln}}\left( p \right) + {\beta _{j,n + 2}}{\rm{ln}}\left( {{z_0}} \right)\\ &&+\,\,{\beta _{j,n + 3}}{\rm{ln}}\left( {{{\bar y}_j}} \right) + {\varepsilon _j}, \end{eqnarray*}$$

(9)

where |${\varepsilon _j}$| is an error term. The unwanted output in this analysis is carbon dioxide emissions and the natural condition is abundance of fish, as measured by estimated stock size. Together the error terms can be stacked in the vector ε. Given the time series properties of the data, we assume that the error term has the following distribution

$$\begin{eqnarray*} \varepsilon \sim N\left( {0,{\rm{\Sigma }}} \right), \end{eqnarray*}$$

(10)

where ε has an AR(1) structure. Furthermore, since more than one type of vessel is included in the model, contemporaneous correlations are assumed in Σ. The model is therefore estimated using non-linear, seemingly unrelated regression.

A simplified version of expression (9) was also estimated. It assumes that only natural conditions matter to the production of the carbon dioxide emissions and that the emissions are proportional to overall catch. Under this assumption, expression (9) simplifies to

$$\begin{eqnarray*} {\rm{ln\,\,}}\left( {{e_j}} \right) &=& {\beta _{j,0}}\,\, + {\beta _{j,n + 2}}{\rm{ln}}\left( {{z_0}} \right) + {\rm{ln}}\left( {{{\bar y}_j}} \right) + {\varepsilon _j},\\ {\rm{ln\,\,}}\left( {\frac{{{e_j}}}{{{{\bar y}_j}}}} \right) &=& {\beta _{j,0}}\,\, + {\beta _{j,n + 2}}{\rm{ln}}\left( {{z_0}} \right) + {\varepsilon _j}. \end{eqnarray*}$$

(11)

The same AR (1) error structure is assumed for expression (11) as for expression (9). Technological development is likely to affect emissions. A version of expressions (9) and (11) with trend is therefore estimated to account for technological change during the study period.

Results

Figures 2a–f show the relationship between emissions per unit of catch limit as a function of stock size according to expression (12) for each fleet segment and the weighted average for all fleet segments.

Figure 2. a-f

The relationship between emissions per unit of catch and stock for each fleet segment and weighted average for all fleet segments. Sources: Statistics Iceland, Icelandic Environment Agency, and the Marine and Freshwater Research Institute.

Figures 2a–f show a very close relationship between stock size and emissions per unit of catch. It shows a different level of emissions between passive gear (figures 2a–c) and active gear technology (figures 2d–e) and different sensitivity to stock size.

Comparison of adjusted R² statistics supported that expression (9) could be estimated based on a common model for two categories of vessels, those with passive (mostly hand line, longline, and gillnets) and active (bottom trawls) gear. Table 4 shows the adjusted R² for the four models considered; simple (expression 11), simple with trend, implicit emissions supply (expression 9), and implicit emissions supply with trend. The results in Table 4 show the explanatory power of the simple model is lower that the implicit supply model for all fleet segments except for fresh fish trawlers. This indicates that this fleet segment is insensitive to price changes. Furthermore, the table shows that the trends add very little to the explanatory power of the models.

Table 4.

Fit statistics (adjusted R²) for the estimated models for each fleet segment.

	Simple model		Supply model
Type	No trend	Trend	No trend	Trend
<10 GRT	0.72	0.72	0.99	0.99
10–200 GRT	0.74	0.74	0.94	0.94
>200 GRT	0.80	0.79	0.82	0.83
Fresh fish trawlers	0.79	0.79	0.74	0.72
Freezer trawlers	0.57	0.56	0.78	0.78

Table 5 shows the estimated emissions supply elasticities for stock size, quota, oil price and fish price, and the relative trend. An elasticity is the relative change in emissions given a relative change in an independent variable (e.g. percentage change in emissions per 1% change in input). The results show that emissions are very responsive to stock size and that the elasticity of emissions with respect to stock size for passive gear is generally smaller than for active gear. None of the trend parameters are statistically significant and all of them are small. This indicates a small role for technology in reducing emissions during this period. Both gear categories show a clear relationship between total quota and emissions, but the emissions elasticity of quota is much closer to the one for passive gear than active, indicating that active gear emissions grow more slowly with increased output. The results show that the passive gear fleet segments are more sensitive to price changes than the active gear segments. Fuel price elasticities are negative, as expected. It is worth noting that the fuel price elasticities are much smaller than the stock size elasticities and that this difference is greater for active compared to passive gear. Output price (fish price) elasticities are positive for both segments, as expected, but larger for passive than active gear. The autocorrelation parameters are in the range 0.3–0.4 and are statistically significant in all cases, supporting the AR (1) definition of the error term.

Table 5.

Estimated elasticities of emissions supply with respect to stock size, quota, oil price, and fish price and the relative trend (expressions 9 and 11), with t-values in parenthesis.

		Simple model		Supply model
Elasticities	Gear class	No trend	Trend	No trend	Trend
Stock size	Passivea	−1.40	−1.20	−0.94	−0.97
		(−8.37)	(−2.99)	(−7.94)	(−3.45)
	Activeb	−1.50	−1.52	−1.34	−0.95
		(−8.81)	(−3.94)	(−6.82)	(−2.20)
Trend	Passive		−0.0062		0.0009
			(−0.56)		(0.07)
	Active		0.0065		−0.0216
			(0.06)		(−1.21)
Total quota	Passive	1.00	1.00	0.95	0.96
				(56.29)	(57.81)
	Active	1.00	1.00	0.87	0.87
				(21.54)	(20.56)
Fuel price	Passive			−0.44	−0.44
				(−8.44)	(−8.03)
	Active			−0.18	−0.13
				(−2.15)	(−1.53)
Fish price	Passive			0.51	0.50
				(5.77)	(4.27)
	Active			0.31	0.44
				(2.25)	(2.44)

a

Mostly hand line, longline, and gillnets

b

Different types of bottom trawls

Discussion

From 1997 to 2018, the average CO₂ emissions per unit of catch were reduced by approximately 40% in Iceland's demersal fisheries. The results indicate that this was mostly due to larger demersal fish stocks, resulting in higher TACs and increases in CPUE. The emissions supply elasticity of stock size was –0.95 for active and –0.97 for passive fishing gear, based on the supply model with a trend. This means that a 1% increase in stock size reduces fuel consumption by slightly less than 1% for both active and passive fishing. The other estimated models showed even larger reductions in emissions relative to stock size. The emissions supply elasticity of total quota was estimated to be 0.96 for passive fishing gear and only 0.87 for active. This indicates that emissions do not increase proportionally to output, and thus the more you catch the less you emit per unit. If a harvest rule is in place, as in Iceland where TAC depends on stock size, an increased stock size leads to reductions in emissions per unit of output, both through increased CPUE and increased output. Therefore, sustainable management of fish stocks is by far the most important way to reduce GHG emissions. Our findings are therefore in accordance with the recent paper and predictive models by Hornborg and Smith, (2020) since fuel usage per unit of catch was highly related to stock size.

This study shows that effects of fuel prices are a distant second contributor to reduced emissions in Iceland's demersal fisheries. The emissions supply elasticity with respect to oil price was estimated to be only –0.13 for active gear and –0.44 for passive gear. These findings are in line with previous results examining Nordic fisheries (i.e. Waldo et al., 2016), and a study by Ziegler and Hornborg, (2014) that analysed Swedish fisheries, whose findings were that there was only a weak inverse relationship between fuel price and the fuel used per unit of landings. Therefore, emission taxes on oil will have some, but not substantial, effects on reducing emissions. Other findings are that the contributions of prices (both input and output) were much smaller and the contribution of technological change was negligible. This was because the trend variable, supposed to capture technological development, was surprisingly small, indicating a minor role for technology in reducing emissions. This is contrary to the findings examining Australian and European fisheries since the beginning of the 21st century, where reduction in fuel consumption was achieved by improved technology (Parker and Tyedmers, 2015). Several factors might contribute to our findings. There has been underinvestment in new vessels in Icelandic fisheries for the last three decades (Saevaldsson and Gunnlaugsson, 2015). An indication of this is that in 1997, the average age of all Icelandic fishing vessels was 20 years (Gunnlaugsson and Saevaldsson, 2016). In 2018, this average was 31 years (Statistics Iceland, 2020b). This may indicate that there is an opportunity for further reductions in fuel consumption and hence reductions in emissions in Iceland's fisheries, with newer more fuel-efficient vessels and better engine design and increased use of alternative fuels.

The findings in this paper show the outcomes of a well-managed and established ITQ system. The companies operating within this system maximize their profits given their overall quota limit. The rebuilding of demersal fish stocks, under this management system, has led to a considerable reductions in GHG emissions. The results presented in this paper add to the findings of Waldo et al., (2016), who examined emissions from Nordic fisheries (i.e. in Denmark, the Faeroes, Iceland, Norway, and Sweden). Their main findings were that fisheries management and optimal harvesting of fish stocks was the most important contributor to reducing emissions. The paper concluded that significant reductions in CO₂ emissions were possible in all of these fisheries except for the Icelandic fisheries because it already has a well-managed fisheries system (i.e. the ITQ system). This paper shows, however, that even more emissions reduction were possible within ITQ managed fisheries. Iceland had reduced fishing effort before this study's data range, which started in 1997. Therefore, the so-called “autonomous adjustment” had already occurred, reducing GHG emissions before this research began. So, our results show that even more reductions in emissions are possible within ITQ managed fisheries if management of the fishing stocks is optimized, and the stocks are rebuilt and managed sustainably.

A recent estimate indicates that around 35% of the world's fish stocks are overfished (FAO, 2020) and only 32% of the world's fisheries are in good biological condition (Costello et al., 2016). The lessons from Iceland's fisheries show that the effort to rebuild the demersal of fish stocks during the past decades has also led to a very large drop in GHG emissions. This indicates that efficient and sustainable utilization of fish resources leads to added environmental benefits. It is undisputed that overharvesting leads to reduced profitability. Our results indicate that rebuilding of the world's fish stocks and sustainable utilization of fish stocks would not only improve food availability and profitability of fisheries, but also be the best and most efficient method to reduce emissions in global fisheries. However, if the world is to reap the full benefit of reduced emissions in fisheries by rebuilding of fish stocks, it is necessary to have a fisheries management system with strong efficiency incentives. Otherwise larger fish stocks may lead to more fishing effort or investment in capacity or technology that may in the end cause increased emissions.

Conclusion

Iceland has experienced a very significant reduction in GHG emissions in its demersal fisheries during the last two decades. From 1997 to 2018, the average emissions per unit of catch reduced by around 40% in Iceland's demersal fisheries. The findings presented in this article show that rebuilding of fish stocks has been the most important contributing factor towards reduced GHG emissions in Iceland's demersal fisheries during this period. Other variables examined (i.e. improved technology, oil price, and fish price) had a much smaller impact. Iceland has an established ITQ system. Companies operating under ITQ systems have a strong incentive to use all their inputs efficiently to maximize profits given their allocated fishing quota. This study shows that rebuilding of fishing stocks when fisheries are managed by a system with efficiency incentives may lead to significant reductions in GHG emissions.

Data availability statement

The data underlying this article will be shared on reasonable request to the corresponding author.