Risk-based evaluation of the vulnerability of the Skagerrak–Kattegat marine fish community to Swedish fisheries

Abstract

Progress towards ecosystem-based fisheries management calls for useful tools to prioritize actions. To select suitable methods for local circumstances, evaluating approaches used in other jurisdictions can be a cost-effective first step. We tested Productivity Susceptibility Analysis (PSA) to assess the potential vulnerability of the marine fish community in the Skagerrak–Kattegat (Eastern North Sea) to possible interactions with all Swedish fisheries operating in the area. This analysis combines attributes for a species productivity with attributes related to the susceptibility to capture to quantify a single score for vulnerability: high, medium, or low risk. Results indicate that demersal trawl and gillnet fisheries were associated with the highest risk levels if interaction occurs, i.e. having the highest prevalence of species with potentially high vulnerability to the fisheries. Mixed results were seen when comparing the assessment results with available data. The main benefit of utilizing PSA in the area is the comprehensiveness of the assessment, including data-deficient fisheries and species. Drawbacks include potential overestimation of actual risks. Overall, together with available data, PSA in the studied area provides a comprehensive map of potential risks for further actions and may progress a science-based, precautionary management of the area.

by-catch, ecological risk assessment, ecosystem approach to fisheries, Productivity Susceptibility Analysis

Introduction

Approaches for ecosystem-based fisheries management (EBFM) calls for, amongst others, science-based and precautionary management of fisheries (Long et al., 2015). Even if implementation of EBFM is extensively mandated, there is, as of today, no single solution on how to adapt the current management system to conform to EBFM objectives (Trochta et al., 2018). To grasp the complexity of fishing pressure on ecosystems, it is increasingly recognized that the time, expense and data requirements for full quantitative assessments such as those for targeted stocks cannot be met for the suite of by-catch, protected species, and habitats and ecological communities. Therefore, risk-based ecological assessments offer an alternative solution that is a balance between cost and effectiveness (Hobday et al., 2011). These tools trade specificity for generality, based on the relationship between ecological vulnerability and proxies that are widely available and correlated with vulnerability. Today, different approaches to ecological risk assessment (ERA) are utilized for EBFM applications by many fishery management authorities around the world, either qualitative or more quantitative (e.g. Gullestad et al., 2017; Smith et al., 2017; Gaichas et al., 2018).

Compared to the development of traditional stock assessment methods, ERAs have a much shorter history of development. Australia pioneered addressing risks associated with fishery interactions with a diverse range of data-limited species occurring as by-catch or protected species through development of Productivity Susceptibility Analysis (PSA; Stobutzki et al., 2001). In its original design, PSA is an assessment of potential risk to fail management objectives related to by-catch species. It is not assessing the actual impact in terms of actual fishing pressure, since this requires more data such as abundance and mortality from fishing. This information is generally lacking for most non-target species. Instead, the PSA estimates the relative vulnerability between species to a fishery to be able to cost-effectively prioritize further action. Since then, PSA has been further developed and used in, e.g. the ERA framework developed for Commonwealth fisheries in Australia (Hobday et al., 2011). A related version has also been applied as stand-alone assessments to estimate the vulnerability of targeted data-limited fish stocks in the United States (Patrick et al., 2010). This latter application has been questioned by Hordyk and Carruthers (2018) who found that there is perhaps no linear and additive relationship between vulnerability and susceptibility. Furthermore, addressing cumulative assessments of fishing pressure on a species needs further attention (Micheli et al., 2014). It is apparent that refinement of tools for and applications of ERA is an ongoing exercise.

Even if implementing EBFM is mandated in European Union (EU) fisheries (EU, 2013), no ERAs are presently and routinely used by member states’ fisheries management agencies. One impediment to uptake may be that there has been limited research related to applicability in a European context to provide the scientific basis for managers. In other regions of the world, numerous studies have been done on how to address ecological risks of fishing. In terms of PSA, there are today at least 23 scientific publications related to fisheries, covering over 1000 stocks/species, predominantly from the United States, Canada and Australia (Hordyk and Carruthers 2018). The International Council for Exploitation of the Sea (ICES) has, however, expressed interest in the use of PSA for data-limited stocks in European waters in recent years (WKLIFE III, 2013). So far there are only two European studies on marine fish, one on skate stocks in the Celtic Sea (McCully Phillips et al., 2015) and one demersal fish in the Mediterranean (Osio et al., 2015). PSA has also been applied to evaluate risks with incidental catch of marine mammals in Europe (Brown et al., 2013). Even if many European fisheries today are more data rich compared to some Australian fisheries at the time PSA was developed, there are still data gaps. There are no routine assessments of non-commercial fish concerning potential vulnerability from interaction with fisheries, the majority of the fish species in e.g. Swedish coastal waters (Hornborg et al., 2013). Mandatory logbook documentation focuses on landed commercial species, and only a small part of the total catches is monitored by control or scientific observer programmes (Pérez Roda et al., 2019). The monitoring for fisheries regulated under the Common Fisheries Policy varies between fisheries and countries. In, e.g. Sweden, this equals to ∼1% of fishing effort in certain fisheries, predominantly demersal trawling; minor fisheries are not monitored by scientific observers at all (Bergenius et al., 2018).

In a comparatively well-studied area such as the Skagerrak–Kattegat area (Figure 1), a diverse range of Swedish fisheries operates (Table 1). Fully quantitative stock assessments only cover a limited number of species and trends are monitored for only a handful more (in total 39 commercial species comprised of 32 fishes and 7 invertebrates are considered in the annual report on stock status released by the Swedish management authorities; SwAM, 2019), and many of them lack quantitative management objectives (Östman et al., 2016). Most species (i.e. non-commercial fish) have only limited or no regular assessment, primarily through conservation efforts such as the Swedish IUCN Red List of Threatened Species (Hornborg et al., 2013; Swedish Species Initiative, 2019). In fisheries with on-board observers of actual catches, the limited coverage may also not be sufficient to detect trends for rare species (Maxwell and Jennings, 2005). Thus, applying PSA and comparing the outcome with current understanding (such as observer data, and other assessments available) may further the understanding of validity and usefulness of PSA as a tool in a European context and facilitate progress towards implementing EBFM.

Figure 1.

Map of the Skagerrak–Kattegat area.

Table 1.

Swedish fisheries on the Swedish west coast, where an ‘x’ indicates the region in which each fishery takes place.

Fishing segment	Fishery ID	Gear	North Sea	Skagerrak	Kattegat	Min depth (m)	Max depth (m)	Observer coverage	Target speciesa (90% of landing value)
Pelagic active gears	PEL1	Pelagic trawl	x	x	x	0	300	No	Atlantic herring (Clupea harengus) and sprat (Sprattus sprattus)
	PEL2	Demersal trawlb	x			50	100	No	Sandeel (Ammodytes tobianus and A. marinus)
	PEL3	Purse seine	x	x	x	0	150	No	Atlantic mackerel (Scomber scombrus)
	PEL4	Purse seine	x	x	x	0	150	No	Atlantic herring (C. harengus) and sprat (S. sprattus)
Demersal active gears	DEM1	Otter trawl (shrimp and fish)	x	x	x	150	410	Yes	Northern shrimp (Pandalus borealis), Atlantic cod (Gadus morhua) and saithe (Pollachius virens)
	DEM2	Otter trawl (shrimp)		x	x	70	400	Yes	Northern shrimp (P. borealis)
	DEM3	Otter trawl (fish)	x			80	250	Noc	Saithe (P. virens), Atlantic cod (G. morhua), Haddock (Melanogrammus aeglefinus), Pollock (Pollachius pollachius), wolffish (Anarhichas sp.)
	DEM4	Otter trawl Norway lobster and fish		x		45	240	Yes	Norway lobster (Nephrops norvegicus), Atlantic cod (G. morhua), Witch flounder (Glyptocephalus cynoglossus), Haddock (M. aeglefinus), European plaice (Pleuronectes platessa), Saithe (P. virens), Angler fish (Lophius piscatorius), Hake (Merluccius merluccius)
	DEM5	Otter trawl Norway lobster and fish			x	25	75	Yes	Norway lobster (N. norvegicus), Atlantic cod (G. morhua), European plaice (P. platessa), Common sole (Solea solea), Whiting (Merlangius merlangus)
	DEM6	Otter trawl Norway lobster		x	x	30	175	Yes	Norway lobster (N. norvegicus)
Passive gears	PAS1	Creels (Norway lobster)		x	x	35	85	Yes	Norway lobster (N. norvegicus)
	PAS2	Creels (European lobster)		x	x	10	40	No	European lobster (Homarus gammarus)
	PAS3	Creels (Brown crab)		x	x	10	40	No	Brown crab (Cancer pagurus)
	PAS4	Demersal nets		x		10	200	No	Atlantic cod (G. morhua), Common sole (S. solea), Turbot (Scophthalmus maximus)
	PAS5	Demersal nets			x	10	80	Yesd
	PAS6	Pelagic nets		x	x	0	20	No	Atlantic mackerel (S. scombrus), Atlantic herring (C. harengus)
	PAS7	Hook and lines (pelagic)		x	x	0	20	No	Atlantic mackerel (S. scombrus)
	PAS8	Hook and lines (demersal)		x	x	80	200	No	Atlantic cod (G. morhua)
	PAS9	Creels, fyke nets (wrasse)		x	x	0	10	No	Gold-sinny wrasse (Ctenolabrus rupestris), Corkwing wrasse (Symphodus melops), Ballan wrasse (Labrus bergylta)

a

Determined by fishing location, gear, and available quotas.

b

Belongs to the pelagic segment out of vessel tradition (more industrial production).

c

Some observer trips exists from transboundary (Skagerrak and North Sea) fishing trips but the fishery is not within the regular sampling strata.

d

The fishery was introduced in the observer programme in 2019 (no data available for this analysis).

Aim

The aim of the study is to utilize PSA as a screening tool to assess potential risks to the fish community from unintentional catches in Swedish west coast fisheries. This is done by adapting PSA to regional circumstances, assessing the potential vulnerability of fish species to the fisheries operating in the area, and evaluating results in light of current data availability. This is one of the first tests of PSA in a European context, and we discuss management and research implications for progress towards EBFM based on current understanding of risk. This approach also begins the necessary transition from single fishery ERAs to those accounting for multiple fleets.

Material and methods

Productivity Susceptibility Analysis

The PSA tool estimates the relative and potential vulnerability of a species to a fishery depending on the species’ productivity (life history attributes such as maximum age and size) in combination with the susceptibility to the fishery (depending on fishing attributes such as mesh size and discard mortality) based on qualitative scoring (Hobday et al., 2011). The latest PSA method available and used in Australia was used as a starting point (ERM, 2017), with some minor adaptations. The same life history parameters for productivity attributes were used, but cut-off values for low–medium–high productivity were adapted to regional circumstances by following the method in Hobday et al. (2011), i.e. determined from analysis of the distribution of attribute values for the species assessed (Table 2). These values differed from those found in Australia (Hobday et al., 2011) and Central America (Micheli et al., 2014) based on the different fish community composition, where, e.g. maximum size and age at maturity were most similar to Micheli et al. (2014) and maximum age most similar to Hobday et al. (2011). One change was made to reproductive strategy in the form of adding sex change as a low productivity attribute based on the evidence of risk (e.g. Kindsvater et al., 2017). Full list of values for productivity attributes used for the different species is found in Supplementary Table S1.

Table 2.

Cut-offs for low–medium–high productivity risk scores used in this study.

Productivity attribute	Low productivity (high risk = 3)	Medium productivity (medium risk= 2)	High productivity (low risk= 1)
Age at maturity (years)	>4	3–4	<3
Maximum age (years)	>20	10–20	<10
Fecundity (eggs per year)	<1 000	1000–40 000	>40 000
Maximum size (cm)	>70	30–70	<30
Size at maturity (cm)	>40	20–40	<20
Trophic level	>3.9	3.5–3.9	<3.5
Reproductive strategy	Live bearer, sex change	Demersal egg layer	Broadcast spawner

Not all fish species on the Swedish west coast are caught in all fisheries. However, the PSA here took a fully precautionary approach as an initial screening and assessed potential vulnerability of the whole fish community to all fisheries. This includes all marine fish species (145 species) that have been recorded as occurring on the Swedish west coast (ICES area 3a) based on the following databases and criteria:

• resident and reproducing marine fish extracted from the Swedish Taxonomic Database;

• caught in the International Bottom Trawl Survey between 1972 and 2018 (ICES SD 20-23);

• recorded in on-board observer data between 1996 and 2018 (ICES SD 20, 21 or 23).

The list included 117 fish species without specific monitoring for stock assessment-related purpose in Sweden today (but a few may be monitored elsewhere, such as anchovy Engraulis encrasicolus). Some of the species have quantitative stock assessments and might thus be excluded from the semi-quantitative PSA as suggested under the latest guidelines for risk management in Australia (ERM, 2017). There were, however, benefits to including them all to estimate potential risk for all fish in the area to discard pressure from all fisheries, including in data-deficient circumstances (fisheries without observer data). This also allows risks identified to be evaluated against quantitative data when available (further detailed below).

Productivity

The productivity attributes were primarily extracted from the Swedish Species Initiative (Swedish Species Initiative, 2019); if species attributes were missing in this source, data from FishBase (2019) or Heessen et al. (2015) were used. If ranges or different estimates were provided for an attribute, the score associated with highest risk was used to be precautionary.

Susceptibility

Susceptibility attributes for Swedish commercial fishing effort on the Swedish west coast (Figure 1) were collated from mandatory fishery logbook records, gear regulations, literature, and fisheries experts. The fishing activities are highly varied in effort and fishing pattern. As such, the fleets operated can be sorted in many ways, e.g. separated by gears, size of vessels, activity level. Here, the activities were separated based on a combination of target species and gear into 19 individual fisheries (Bergenius et al., 2018), each one with differences in susceptibility attributes (Table 1). The 19 fisheries were grouped into three fishing segments (pelagic active gears, demersal active gears, and passive gears), and the fish community vulnerability to each one of them was assessed separately.

All fish species were assumed to be able to be caught by the different segments, but the extent to which the species was susceptible depended on a combination of species and fishery attributes. This is different to earlier approaches that have mainly used data on fish actually caught as by-catch in the fishery. The different method chosen here is motivated by the need for a precautionary approach due to the lack of or sparse observer data and implies both that results are theoretical but risk-based until data or further analysis can support other evidence. Availability and depth range of the species were based on the same sources as for productivity attributes. The cut-offs for low–medium–high susceptibility were only slightly modified compared to ERM (2017; Table 3):

Table 3.

Susceptibility attributes and cut-offs for low–medium–high risk used in this study.

Attribute	High susceptibility (risk = 3)	Medium susceptibility (risk = 2)	Low susceptibility (risk= 1)
Availability	Endemic/sub-population	North hemisphere	Global
Encounterability	>66% depth overlap	33–66% depth overlap	<33% depth overlap
Selectivity	Fishery- and species-dependent	Fishery- and species-dependent	Fishery- and species-dependent
PCM	Fishery- and species-dependent	Fishery- and species-dependent	Fishery- and species-dependent

For fisheries-specific information, see Supplementary Table S2.

• Availability, same categorization as in Australia but the medium risk description was changed to northern hemisphere instead of southern hemisphere.

• Post-capture mortality (PCM), i.e. potential mortality risk of a species from either being landed or discarded in a fishery, was based on expert opinion due to the lack of data for all fisheries and species. This was guided by general experience from research conducted at the Swedish University of Agricultural Sciences related to the introduction of the EU landing obligation (ICES, 2016). Species that was defined as target species in a fishery (defined as contributing to 90% of landing value; Table 1) were attributed high risk in that fishery, and small pelagic fishes are given highest potential mortality risk score in all gears from their low survival rate when discarded; the rest of the species were attributed a low-to-high mortality depending on fishery (influenced by, e.g. depth).

• Encounterability was based on the species depth distribution in relation to depth intervals for the fisheries based on 5 and 95%-percentile for all hauls between year 2000 and 2018 in the on-board observer database (Table 1). For fisheries without observer programme, expert opinion on normal depth range was used instead. Furthermore, all species were categorized as benthic, semi-pelagic or pelagic (Supplementary Table S2). Overrides to depth overlap between species and fishery were then used for benthic species in pelagic fisheries (e.g. susceptibility low for a benthic species in a pelagic fishery regardless of depth overlap), and likewise for a pelagic species to demersal fisheries. Semi-pelagic species had no overrides but only based on depth overlap, likewise for pelagic fish to pelagic fisheries, and demersal fish to demersal fisheries.

• Selectivity attributes for the different fisheries were based on expert opinion from local experience on selectivity attributes of the gears used in the fisheries from, e.g. industry-research collaboration on developing more selective gears (species and size) in the area (see, e.g. Nilsson et al., 2018) and data from on-board observers (mainly trawl fisheries). Usually, there is a size component for this attribute, e.g. mesh size relative to length at maturity. Since some gears have a complex sorting system (combinations of mesh size, panels, grids, escape openings), the size component was considered to be overly complex for the intended purpose. Here, we chose to sort the species into the same groups as for PCM and attributed low–medium–high risk of capture in different gears based on local expert opinion guided by the collective experience described above.

Full documentation on cut-off values is found in Supplementary Table S2.

Vulnerability and risk level

Once productivity and selectivity attributes were defined, a combined score of potential vulnerability V for each species and fishery was calculated to estimate low–medium–high risk for a species with the encounter, following the method developed by Stobutzki et al. (2001). This is done through calculating the Euclidian distance from the origin of an x–y scatter plot:

$$V=\sqrt{P^2+S^2},$$

(1)

where productivity (P) is calculated as the average risk score for all seven productivity attributes and susceptibility (S) is the geometric mean for all four susceptibility attributes [availability (A) × encounterability (E) × selectivity (S) × PCM]^1/4. The different approach to combining individual attributes of productivity (P) and susceptibility (S) into a single P or S score reflects the assumption that each of the P attributes is an equally likely estimate of overall productivity (hence averaging), while for S, the geometric mean is used because low score for any one attribute (e.g. A, or S, E, or PCM) lowers the overall risk score considerably (Hobday et al., 2011). The vulnerability scores are then translated into risk scores through dividing the potential scores for V into equal thirds; >3.18 equals to high risk, between 3.18 and 2.64 medium risk, and lower than 2.64 low risk (Figure 2; for further information see Hobday et al., 2011).

Figure 2.

PSA plot with the contours illustrating cut-offs for low–medium–high-risk areas for species. The colour scale represents the V-value, with blue areas the lowest risk and the red the highest.

Vulnerability of each species to the different fisheries was analysed on an aggregate level in terms of (i) mean potential vulnerability for all species per fishing segment and separate fisheries and (ii) composition of potentially low–medium–high-risk species in separate fisheries.

Risks in relation to available data

The outcome of the PSA was compared with existing data to evaluate the estimated vulnerability in relation to what is known about the species and fisheries interaction in the area. This was done by examining the available observer data per fishery (aggregated volume of discarded catch over the last 10 years, 2008–2018) through comparing (i) the vulnerability to a fishery with observed discard volume in the same fisheries and (ii) the relative discard pressure of different fisheries relative to the estimated vulnerability. Lastly, to determine actual discard pressure on fish species in the area, biomass estimates are needed. Since this is lacking for most species evaluated here, the outcome of the Swedish IUCN Red List was used as a proxy (ArtDatabanken, 2015). The Swedish IUCN Red List of Threatened Species is updated every 5 years and is primarily based on abundance trends from available sources, such as survey data. The PSA results were examined in terms of the estimated vulnerability score for red-listed fish species to the different fisheries, including available observer data on discard volumes. The analysis of results relative to the Swedish IUCN Red List also included evaluation of the PSA methodology, such as mean productivity scores and species threat status.

Results

Productivity

Of the 145 assessed species, 46% had estimates for all seven productivity attributes whereas 4% had only 1–2 known attributes for productivity (i.e. high-risk value for productivity based on data deficiency, such as Striped seasnail Liparis liparis). The average value was 6 out of 7 attributes known. This implies that the productivity scores can in general be seen as robust, i.e. most of the species had most of the attributes known. Seventeen percent of the fish community was categorized as red-listed nationally [i.e. belonging to the categories near threatened (NT), vulnerable (VU), endangered (EN), critically endangered (CR), regionally extinct (RE), or data deficient (DD)]. These species generally had all productivity attributes known, only a few missed one or two. The average productivity scores indicated lower productivity and higher risk for the threatened species (VU, EN, CR), 2.54 ± 0.33 compared to 2.09 ± 0.43 for the whole fish community, thus supporting the productivity-based vulnerability in terms of increased risk level.

Vulnerability

Potential risk to species on segment and fisheries level

The potential risk the three fishing segments’ pose to the fish community differed in terms of average vulnerability scores for the whole fish community. The demersal trawl segment was associated with the highest risk in terms of mean vulnerability score of the fish community (2.97 ± 0.08), followed by pelagic (2.88 ± 0.03), whereas passive fisheries had the lowest (2.82 ± 0.21). Passive fisheries also had the highest variability due to the diversity of fishing methods and practices.

The separate fisheries differed in terms of proportion of species having high vulnerability to the fishery if interaction occurs (Figure 3). Demersal gillnets posed the greatest risk, with higher risk levels than for the demersal trawl fisheries. The fishery with the least amount of high-risk interactions in the number of species with high vulnerability was the targeting of live wrasses with pots and creels (PAS9); 44% of the fish community could, however, be at medium risk by the fishery. For full list of fisheries and species, see Supplementary Table S3.

Figure 3.

Fish community potential vulnerability to different fisheries in terms of proportion of species being at low–medium–high risk to the practice.

Potential risk to species and recorded catches

For fisheries with observer data, the discard volume of different species varied between fisheries (Table 4). Note that these volumes are not comprehensive since it only illustrates discarded catch from on-board observer trips and is not raised to be an estimate of total discard pressure for the whole fishery. Atlantic cod Gadus morhua was found to have a high or medium vulnerability to all but one fishery (PAS9; Supplementary Table S3). The species was also recorded as discarded in all fisheries with observer data. On a fishery level, Atlantic cod dominates fish discards in Norway lobster creels (PAS1), but in terms of total discard volume of the species, otter trawling for Norway lobster and fish in the Skagerrak area (DEM4) dominates by far. Another common species discarded, common dab Limanda limanda, is not targeted in any fishery. This species was assessed as having a medium vulnerability to five fisheries and low to the rest—even if the total magnitude of discard is equivalent to Atlantic cod. Even if most species included in the PSA were found in observer data records, some vulnerability scores may be false positives since the species have no record of being caught in any of the observed fisheries, e.g. striped seasnail, rudderfish Centrolophus niger.

Table 4.

Top discarded species^a in tonnes for fisheries with observer data with vulnerability scores found in this study colour coded (red = high risk; yellow = medium risk; green = low risk).

Species	Common name	Fishery						Total
		PAS1	DEM1	DEM2	DEM4	DEM5	DEM6
Gadus morhua	Atlantic cod	59	307b	8	2623b	402b	215	3 614
Limanda limanda	Common dab	2	3	1	282	515	2416	3 218
Trisopterus esmarkii	Norway pout		2472	166	20	3	6	2 666
Pleuronectes platessa	European plaice	0.1	5	3	669b	447b	704	1 830
Hippoglossoides platessoides	American plaice	0.1	86	34	520	130	1020	1 789
Micromesistius poutassou	Blue whiting		1472	13	3			1 489
Merlangius merlangus	Whiting	3	155	22	466	487b	297	1 430
Pollachius virens	Saithe	0.3	701b	0.4	311b	2	5	1 020
Amblyraja radiata	Starry ray		123	2	556	19	20	719
Glyptocephalus cynoglossus	Witch flounder		429	11	165b	25	71	701
Melanogrammus aeglefinus	Haddock		135	6	372b	66	67	646
Platichthys flesus	European flounder	0.2			44	359	211	614
Argentina silus	Greater argentine		514	10				524
Clupea harengus	Atlantic herring		336	7	72	32	24	471
Squalus acanthias	Picked dogfish		13		292	18	4	326
Cyclopterus lumpus	Lumpfish		286		13	6		305
Eutrigla gurnardus	Grey gurnard		2	1	62	130	106	301
Merluccius merluccius	European hake	0.1	10	3	99b	45	99	256
Microstomus kitt	Lemon sole	0.6	1	1	67	25	63	158
Chimaera monstrosa	Rabbit fish		134	2	20			156
Trachinus draco	Greater weaver		0.1		1	73	19	93
Coryphaenoides rupestris	Roundnose grenadier		90	0.6	1.9			93
Etmopterus spinax	Velvet belly		86	2	2			90
Solea solea	Common sole	0.2			1	16b	35	53

a

Species with over 50 tonnes of estimated total discards from observer trips from all fisheries between years 2008 and 2015.

b

Targeted species in the fishery.

Comparing the estimated vulnerability scores with available discard data identified both support for the findings and the need for further investigations (Table 4). Vulnerability and discard levels followed the same pattern in terms of highest discard volume and vulnerability for Grey gurnard Eutrigla gurnardus, Lemon sole Microstomus kitt, and Greater argentine Argentina silus; the highest risk score was attributed to the fishery with the highest discard volume. In contrast, the lowest vulnerability score for some species, such as common dab L. limanda and American plaice Hippoglossoides platessoides, was seen for fisheries with the highest discard volumes of the species.

Vulnerability and status of the species

The fish community comprise 25 fish species that are red-listed, six of which are also targeted species (Table 5). Some of them also belong to the most discarded fish by volume in fisheries with observer data (Table 4). Most fisheries pose a potential high risk to these species, with the exception of a few passive gears. Only three species (Lumpfish Cyclopterus lumpus, Fourbeard rockling Enchelyopus cimbrius, and Atlantic halibut Hippoglossus hippoglossus) were assessed as not having high vulnerability to any of the fisheries, but they have still often medium vulnerability to most.

Table 5.

Red-listed fish species (ArtDatabanken, 2015) occurring in the studied area and potential vulnerability to different fisheries (high risk = red, medium risk = yellow, low risk = green), with percentage of total discard by fishery listed for those with observer effort (aggregated between years 2008 and 2018).

Common name	Species	IUCN (2015)	PEL1	PEL2	PEL3	PEL4	DEM1	DEM2	DEM3	DEM4	DEM5	DEM6	PAS1	PAS2	PAS3	PAS4	PAS5	PAS6	PAS7	PAS8	PAS9
Norway bullhead	Micrenophtys lilljeborgii	DD
Lumpfish	Cyclopterus lumpus	NT					94	0		4	2	0	0
Fourbeard rockling	Enchelyopus cimbrius						39	0		41	8	10	2
Sailray	Rajella lintea						0	0		0	0	0	0
Norway redfish	Sebastes viviparus						31	0		12	8	49	0
Sea lamprey	Petromyzon marinus						100	0		0	0		0
Velvet belly	Etmopterus spinax	VU					96	2		2	0	0	0
Atlantic cod	Gadus morhua						9a	0	a	73a	11a	6	2			a	a			a
Tope shark	Galeorhinus galeus						0	0		0	0	0	0
Haddock	Melanogrammus aeglefinus						21	1	a	58a	10	10	0
Whiting	Merlangius merlangus						11	2		33	34a	21	0
European hake	Merluccius merluccius						4	1		39	18a	39	0
Greenland shark	Somniosus microcephalus						0	0		0	0	0	0
Starry ray	Amblyraja radiata	EN					17	0		77	3	3	0
Atlantic wolffish	Anarhichas lupus						0	0	a	22	10	23	44
Rabbit fish	Chimaera monstrosa						17	0		77	3	3	0
Atlantic halibut	Hippoglossus hippoglossus						10	0		81	9	1	0
Ling	Molva molva						11	0		44	10	4	31
Thornback ray	Raja clavata						11	0		23	61	4	0
Pollack	Pollachius pollachius	CR					69	0	a	31	0	0	0
Picked dogfish	Squalus acanthias						4	0		89	6	1	0
Porbeagle	Lamna nasus						0	0		0	0	0	0
European eel	Anguilla anguilla						47	0		0	16	12	24
Roundnose grenadier	Coryphaneoides rupestris						97	1		2	0	0	0
Blue skate	Dipturus batis	RE					38	0		62	0	0	0

a

Targeted species in the fishery.

Available discard data reveals that two threatened species have not been recorded during observer effort the past 10 years (Porbeagle Lamna nasus and Greenland shark Somniosus microcephalus). A few threatened species that are also commercial, such as Atlantic wolffish Anarhichas lupus and Atlantic halibut H. hippoglossus, are landed in some fisheries if caught in commercial size and thus are not present in discards. As an example, Atlantic wolffish is found in discard data from three demersal trawl fisheries: DEM4, DEM5, and DEM6 (high vulnerability; Table 5). The reason for not being present in discards from the other two demersal fisheries (DEM1 and DEM2, also high vulnerability) may be because the species is landed. Even if the highest volumes discarded of wolffish are from the passive gear (PAS1), actual fishing pressure is probably the highest in DEM3 (which lacks dedicated observer effort), since wolffish belongs to the targeted species in this fishery.

Swedish fisheries were further assessed as posing potential high risk to most fish species considered as not applicable for the Swedish IUCN Red List, i.e. not covered by Swedish conservation assessments (Supplementary Table S3).

Discussion

PSA as a tool for the studied area

This study has utilized PSA to comprehensively illustrate current understanding of risk for the Swedish west coast fish community from Swedish fisheries and compared the outcome with available data to quantitatively address them. The approach to risk evaluation utilized here, i.e. assuming that the whole fish community may encounter all fisheries operating in the area, has both benefits and drawbacks. The main benefit is the systematic screening of potential vulnerability to all fisheries for all fish species, including fisheries without discard data. Compared to current practise, performing an inclusive risk assessment across fisheries regardless of data availability provides insight that is currently unavailable for fishery managers in the area. A limitation of the PSA is that presence in an area underpins an assumption that fishing interactions are possible, and so the risk levels found here should be interpreted with caution when documentation of interaction between a species and particular fishery is lacking; some fish species may in reality never have been caught in a fishery. If interaction is documented to occur, a PSA provides an estimate of vulnerability, which is known to be higher compared to fully quantitative stock assessments (Zhou et al., 2016). This bias results from the precautionary treatment of uncertainty, and the limited number of categories for scoring. The contributors to the risk score need to be investigated as part of a management response, beginning with the highest risk species (ERM, 2017). If risks cannot be dismissed, species that currently lack monitoring or management should be prioritized. This builds the knowledge base necessary to progress towards EBFM implementation in a cost-effective and precautionary manner. The study performed here is thereby of importance to the initiated development of an EBFM strategy for the studied area (SwAM, 2016), where benefits and drawbacks of utilizing PSA compared to other tools need to be balanced against available methods, data, management objectives, and monitoring costs. PSAs represent one way of identifying vulnerable ecosystem elements (e.g. by-catch or protected species) and could be part of a fit-for-purpose toolbox for the region, which must be jointly developed by researchers and managers.

Previous knowledge of fish vulnerability from Swedish west coast fisheries beyond current quantitative stock assessments for commercial species is centred on limited sets of individual fisheries or commercial species. For example, several studies have reconstructed historical data to illustrate decline in the abundance of some commercial species over longer time-series (e.g. Cardinale et al., 2010; Cardinale et al., 2015). Others have compared discard composition and amount for different fisheries over time (Bergenius et al., 2018), for different fishing methods for the same species (e.g. Hornborg et al., 2017), or differences in discard pressure between different demersal trawling practises (e.g. Hornborg et al., 2013). This analysis thus advances earlier research on fish vulnerability through studying potential pressure from fisheries, including those with no quantitative data (e.g. the pelagic segment, most fisheries with passive gears) and on non-commercial species that have never been evaluated before. This is important since most of the species that are found in the area and included in this analysis are not landed or recorded. In addition, on-board observer data of fish discard levels are mostly non-existent beyond demersal trawl fisheries, and highly limited in terms of coverage relative to fishing effort. The number of species classified as high risk to the fisheries is consistent with observed declines of many commercial fish stocks (e.g. the historical baseline studies described above; Hornborg et al., 2013). The potential for species with recorded depletions to be used as sentinels for similarly vulnerable species identified in the PSA, but without monitoring data, merits further investigation.

Disparities found in this study between discard volumes and vulnerability score could either be caused by (i) absence of information on catch volumes for most species/fishery combinations, and thus lack of inclusion in the PSA; (ii) species- and fisheries-specific differences in behaviour, substrate preferences, spatial patchiness, seasonality that do limit catch volume; (iii) historical depletion such that catch is presently low; (iv) observer data collection effort (e.g. differences in time or place of sampling); or (v) that a fishery with high potential risk to a species but low discard volumes is either effective in avoiding the species or landing a larger proportion of the catch (i.e. have quota for the species or catch larger individuals). The actual pressure on commercial species for the different fisheries is influenced by quota availability and abundance, which both vary between years. Without data, it is thus difficult to establish which is the most important potential cause for disparity between vulnerability score and discard volume. What is known, in general terms, is that the pelagic segment dominates in catch volume in Swedish fisheries (over 70% of catch volume of all Swedish fisheries in 2015; Bergenius et al., 2018). These fisheries (PEL1–4) are considered to have low discard ratios, but this is difficult verify due to challenges in monitoring high volumes. Demersal trawl fisheries (26% of landings in 2015) are monitored by scientific observers (1% of effort) and also have the highest by-catch ratio of fish per landing (Bergenius et al., 2018). The passive gear segment contributes with a small landing volume but is very diverse and dominates in terms of number of vessels (∼300 vessels on the Swedish west coast); only one fishery has observer data (creelers for Norway lobster). Monitoring of discards has recently begun for one more passive gear (demersal gillnet). It would be interesting to follow up this observer effort in the future, supported by the findings in this study.

The initial PSAs developed and performed in Australia covered a wide range of taxa and fishing practises (Smith et al., 2007). As expected, targeted species were generally found to be associated with the highest risk, with between 1 (one Danish seine fishery and one trawl fishery) and 159 (otter trawl fishery) high-risk species per fishery. The proportion of vulnerable species depends on the number and composition of species assessed, so results are not easily compared to this study. Other studies using PSA in other regions [following the PSA method of Hobday et al. (2011)] have been restricted to smaller samples of fisheries, i.e. focused on a specific segment. As an example, Micheli et al. (2014) performed a study on small-scale fisheries with passive gears in Mexico. They identified high risk for 22.2% of the 81 species examined and further increased to 38.3% when incorporating cumulative risk from all fisheries. Set gillnets were found to have the potentially greatest impact in the study from Mexico (22.2% of species at high risk) whereas trap fishing had very low (≤1.2%). These findings are supported by this study, where the generally higher proportion of species at high risk for passive gears found here (ranging between 3 and 68%) is likely a combination of fishing pattern (e.g. depth) and species targeted (other taxa included and potential differences in life histories). Furthermore, species identified to be at high risk in passive gear fisheries has been shown to have low productivity, such as elasmobranchs and marine mammals (Micheli et al., 2014). In this study, several species had a very low susceptibility score in, e.g. the passive gear segment (except gillnets) but were scored as having a high vulnerability due to low productivity (e.g. Greenland shark S. microcephalus and Blue ling Molva dypterygia). When further examined, some high-risk species can even be fully excluded from many of the passive gear fisheries since there is no depth overlap (fisheries performed shallower than depth distribution of the species, seen for, e.g. the fishery-species combination of PAS9 and Blue ling).

When utilizing PSA in a management context, it is therefore crucial to establish the attributes most influencing the overall vulnerability score for each species-fishery combination. The Australian Ecological Risk Management guide describes Residual Risk Guidelines for the analysis of PSA results, where risk may be reduced (ERM, 2017). High and medium risk species are evaluated against potential reasons for high risk, such as missing attributes or low productivity despite low susceptibility. This evaluation should be done in partnership with fishery managers, scientists, and fishers and is an important next stage for the findings in this study. It will help eliminate out false positives (species that do not interact with the fishery), prioritize data analysis of current monitoring, prioritize expanded monitoring or commercial catch data, and develop mitigation strategies to reduce risks, all of which contributes to an EBFM approach. Since individual species strategies may become overwhelming (e.g. 100 individual species plans), effort to develop mitigation for groups of like species will be advantageous. For example, single measures that modify an area fished may reduce risk for large number of vulnerable species. These measures can also be tested in a simulation using the PSA (e.g. Leadbitter, 2013).

Methodological challenges

ERAs are in general sensitive to methodological choices (Piet et al., 2017). One example of relevance to this study is cut-off values used for the categorization of low–medium–high productivity, which should be tuned to the region or species groups, allowing for resolution in relative risk. These values for age and size at maturity in this study, based on the species composition in the region, thus differed between this study and values used in Australia (ERM, 2017) and in central America (Micheli et al., 2014). The differences seen between regions are most likely ecosystem dependent such as water temperature, nutrient levels but are also influenced by the species composition, i.e. number and features of the species included. The Australian cut-off values are based on southern hemisphere species from a broad range of habitats ranging from the Antarctic to the tropics (Hobday et al., 2011) whereas this study and Micheli et al. (2014) included species from a narrower ecosystem range, i.e. the Swedish west coast and Baja California in Mexico, respectively. Choosing species composition thus affects cut-off values for risk levels and should therefore be guided by the purpose of the assessment—whether it is to estimate relative risk levels of all fish species in a range of ecosystems or in a narrower range, such as within a temperate or tropical system. Since the PSA outcome is a measure of relative risk, the cut-off values should be based on the set of species found in the fishery range since (i) ecosystem dynamics differ between areas, (ii) a PSA outcome is a measure of relative not absolute risk, and (iii) results are intended to inform priority management measures.

The PSA method has also been criticized in terms of the basic assumptions on linear and additive relationship between attributes. This has been found to cause uncertainties for primarily medium risk values, i.e. being less correlated with actual risk compared to high and low values (Hordyk and Carruthers, 2018). Potential discrepancies can possibly be mitigated through Residual Risk Analysis (ERM, 2017), but it also marks some challenges with utilizing relative risk and not absolute risk to a species, identified early on in the PSA development (Smith et al., 2007). The different approaches to combining individual attributes for productivity and susceptibility to a single score may, however, merit from further investigations. For susceptibility, an additive susceptibility approach was initially used by the PSA developers but was found to cause too many false positives when, in reality, a species could not even be caught in the fishery (Smith et al., 2007). The multiplicative approach to scoring susceptibility recognizes that if any of the four attributes is low, so is overall susceptibility. With regard to the additive productivity attributes, some adaptations of the PSA weigh potential differences based on the relative importance of life history traits as estimates of overall productivity. Duffy and Griffiths (2019) examined redundancies in a related PSA approach used in North America, an approach that uses more attributes than the PSA method used here. They found a non-linear relationship between most productivity attributes. This indicates that they are independently important to the overall productivity score, and averaging is appropriate. However, some attributes (e.g. age and length of maturity) were strongly correlated and could be removed. These relationships can be region and taxa specific, as another study on threatened fish showed that age at maturity had the strongest correlation with threat status, even if size at maturity and maximum age also strongly correlate (Hornborg et al., 2013). A consistent set of attributes, as argued by Hobday et al. (2011), preserves comparability between studies, until the relationship between different life history parameters and risk level can be tuned to specific attributes.

Future research effort could also involve further refinement of the susceptibility attributes used, as they may have high influence on risk level (Hordyk and Carruthers, 2018). One component is encounterability, which could use an attribute based on “swept area estimates” (Pope et al., 2000) that combine differences in both depth range, effort, and area affected through using fishery characteristics such as trawl hours and attraction area of fishing gears. This approach is reflected in another semi-quantitative risk-based tool, Sustainability Assessment for Fishing Effect (SAFE), that is used in some Australian assessments of fish species where data allow (ERM, 2017). This tool is less precautionary than PSA and closer to stock assessment results and also requires more data (Zhou et al., 2016). This method was not chosen for this study since (i) lack of data is the main concern for most of the fisheries assessed here and (ii) we wanted to utilize the most precautionary assessment as a first risk evaluation where SAFE requires many assumptions and may thus risk false negatives. The selectivity component of susceptibility could also be refined quantitatively by developing selectivity fishery-specific estimates using analyses based on published selectivity studies. Furthermore, different fisheries operate in different areas and seasons, which imply different pressure related to, e.g. targeting on spawning grounds. All these factors affecting susceptibility may either be mitigated through standardized Residual Risk Analysis (ERM, 2017) or require method development and by this, most likely more data. However, with more data, more quantitative approaches are available; the intended purpose of PSA is as a systematic screening of risks where data are deficient for many species of interest.

Conclusions and recommendations

PSA offers a holistic view of the potential risk landscape for species in a fishery. This information and subsequent risk reduction efforts support progress towards EBFM for Swedish fisheries on the west coast of Sweden. Depending on management needs, PSA method refinements may be called for. Future risk-based evaluations could also include existing applications, such as cumulative assessments (e.g. Zhou et al., 2019). This should also include fisheries from other countries operating in the area. Based on findings in this study, reccomendations to research and fishery managers are that:

• Initiate a Residual Risk Analysis to rule out false postives. Research or monitoring projects should be considered if current observer efforts are not sufficient to detect trends for a species because it is very rare or historically depleted; a lack of available records does not necessarily imply that the species is not caught and not at risk. Reducing the uncertainty around productivity attributes can decrease risk levels, and so missing life history data could be collected by enhancing existing surveys and on-board observer programmes (Gilman et al., 2017).

• For high-risk species that are commercially targeted and have quantitative stock assessments and poor stock status in the area (e.g. Atlantic cod), results may be used to guide further data collection needs on discards from the high-risk fisheries that are currently without scientific observer effort.

• For resident, non-commercial fish species with high potential vulnerability to fisheries where they are recorded as being caught, risk-reducing strategies should be prioritized. This can be done by investigating if trends may be detected in the limited but existing monitoring data, and if discard mitigation could be achieved either through gear modifications or spatial/temporal measures.

• For wide-ranging species, implementing improved monitoring and mitigation measures requires decisions by national management agencies with regard to their responsibility to minimize risks resulting from their national fleets.

• Lastly, implementing EBFM requires attention to more ecosystem components than fish, future risk assessments should include other by-catch taxa, such as benthic invertebrates, marine mammals and birds occurring in the area, and extend to cover habitats and ecological communities.

Supplementary data

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

Data availability

All data behind the study is found as Supplementary material and sources described in Material and methods.

Funding

This work was funded by the Swedish Research Council Formas (mobility grant 2016-00455) and by the Swedish Agency for Marine and Water Management (grant ID 1861-2019).