Trends in cpue and related changes in spatial distribution of demersal fish species in the Kattegat and Skagerrak, eastern North Sea, between 1981 and 2003

Abstract

We explored the trends in ln-transformed catch per unit effort, defined as average weight (kg) per 1 h trawling, and the spatial distribution of 32 demersal fish species in the Kattegat and Skagerrak using International Bottom Trawl Survey data collected between 1981 and 2003. As in other areas, the biomass of roundfish species such as cod, pollack, hake, and ling drastically decreased during this period most likely owing to fishing pressure. However, other commercially important fish species, e.g. haddock, whiting, and some flatfish, showed a constant or increasing trend during the same period. Non-commercial species showed no or an increasing trend in ln-cpue, by as much as 40 times in hagfish. Furthermore, analyses of the spatial distribution of 14 selected fish species by means of distribution maps of ln-cpue suggested that fish stocks contracted and expanded in response to decrease and increase of the stock biomass, respectively, with some flatfish species (i.e. plaice and flounder) and hagfish representing the exceptions to this general pattern.

Introduction

The last few decades of overexploitation have led to a worldwide decrease, and even collapse, of several fish stocks (e.g. see Myers and Worm, 2003 ). Fishing activities, with removal of target fish species and habitat perturbation, may also lead to changes on all trophic levels ( Jennings and Kaiser, 1998 ). In fact, the global collapse of predatory fish in the northern hemisphere has produced a change in the entire ecosystem, and the species at lower trophic levels (zooplanktivorous pelagic fish and invertebrates) may have steadily increased due to predation release ( Pauly et al ., 1998 ; Myers and Worm, 2003 ).

The effects of changes in target fish abundance on their spatial distribution have been discussed extensively (e.g. Winters and Wheeler, 1985 ; Crecco and Overholtz, 1990 ; Gordoa and Hightower, 1991 ; Marshall and Frank, 1994 ; Swain and Sinclair, 1994 ; Rose and Kulka, 1999 , and references therein). Variations in spatial fish distributions, when not adequately accounted for, can lead to overestimation of the stock biomass and underestimation of the fishing mortality ( Crecco and Overholtz, 1990 ; Rose and Kulka, 1999 ). This phenomenon may lead, in extreme cases, to wrong management decisions and ultimately to the collapse of commercial stocks ( Rose and Kulka, 1999 ). Although the causes for the spatial distribution of a fish population are still poorly understood, density-dependent processes are often suggested as a possible explanation ( Myers and Stokes, 1989 ; MacCall, 1990 ). Alternatively, spatial distributions may also be affected by the degree of stock exploitation ( Winters and Wheeler, 1985 ) and environmental factors ( Corten and van de Kamp, 1996 ; Heessen, 1996 ; Rose and Kulka, 1999 ). Information about the factors affecting spatial distribution of non-commercial species is scarce in the literature (but see Heessen and Daan, 1996 ).

The eastern North Sea including the Kattegat and Skagerrak (ICES Division IIIa) is a transitory area between the central North Sea and the Baltic Sea that has not received as much attention as other regions of the North Sea (see Daan et al ., 1996 ). However, both the inshore and offshore areas of the eastern Skagerrak have undergone high levels of exploitation. The considerable reduction in abundance of large individuals (>30 cm) of several commercial fish species during the last 30 years was likely an effect of high fishing pressure ( Svedäng, 2003 ). Nevertheless, only cod has been extensively investigated in the area ( Svedäng, 2003 ; Svedäng and Bardon, 2003 ).

Factors affecting fish abundance and distribution often operate at regional scales ( Heessen and Daan, 1996 ; Rose and Kulka, 1999 ). Hence, in this paper we investigate the changes in biomass and spatial distribution of both commercial and non-commercial demersal fish species in the open areas of the Kattegat and Skagerrak during the last 23 years.

Material and methods

Trends in biomass

We analysed the trends in cpue of 32 demersal fish species collected in the Kattegat and Skagerrak ( Figure 1 ) during the winter International Bottom Trawl Survey (IBTS) between 1981 and 2003. The IBTS has been carried out in the area since 1979 in January/February, and from 1991 also in September, each year by the Swedish National Board of Fisheries onboard the RV “Argos”. The main aim of the survey is to provide annual estimates of recruitment for several commercial species, but it is also a reliable source of stock abundance data that agree with stock assessment estimations ( Heessen, 1996 ). We used winter data because of the longer time-series. Moreover, in winter most individuals recruited the previous spawning season have already reached a size caught by the survey gear ( Heessen and Daan, 1996 ). Catch rates (cpue) were calculated in weight (kg) per 1 h of trawling, including for each species only those hauls whose depths are considered suitable in the literature. Cpue data were ln-transformed [(ln (cpue + 1)] to satisfy the assumption of normality and homogenous variance. According to the IBTS protocol (see for details ICES, 1992 ), the sampling area was stratified by ICES rectangle of 0.5° latitude and 1° longitude. Haul duration was 30 min at 4 knots except in adverse weather conditions. Between 23 and 49 trawl hauls were performed each year, with a larger number of hauls in the past decade ( Table 1 ). The trawl employed was a standard GOV (Grande Ouverture Verticale) with a 16 mm mesh size in the codend. In this paper the terms cpue and biomass are used as synonyms.

Table 1

Summary information for Swedish IBTS trawl hauls in the Kattegat and Skagerrak between 1981 and 2003.

Year	Period	No. of hauls	Bottom depth (m)
1981	03–19 Feb	32	22–183
1982	09–18 Feb	23	30–232
1983	31 Jan–16 Feb	35	20–210
1984	06–21 Feb	35	21–135
1985	04–21 Feb	32	21–150
1986	03–19 Feb	42	21–248
1987	02–19 Feb	49	22–300
1988	01–18 Feb	39	22–265
1989	06–23 Feb	43	18–234
1990	05–21 Feb	45	18–235
1991	12–28 Feb	38	19–265
1992	03–20 Feb	43	18–258
1993	08–25 Feb	45	18–258
1994	31 Jan–17 Feb	48	18–259
1995	30 Jan–16 Feb	48	19–228
1996	29 Jan–15 Feb	48	18–257
1997	27 Jan–13 Feb	46	19–250
1998	26 Jan–12 Feb	45	19–256
1999	25 Jan–11 Feb	46	18–256
2000	24 Jan–10 Feb	46	19–253
2001	22 Jan–08 Feb	45	18–233
2002	21 Jan–07 Feb	45	19–237
2003	27 Jan–12 Feb	46	19–257

Figure 1

Depth profile of the Kattegat and Skagerrak.

Spatial distribution

We analysed the change in spatial distribution of 14 fish species. The species were chosen to represent three different groups, including five commercial roundfish (cod, pollack, saithe, haddock, and whiting), five flatfish (plaice, flounder, common sole, lemon sole, and long rough dab), and four non-target fish species (poor cod, four-bearded rockling, lumpsucker, and hagfish). The spatial distribution of each species was qualitatively examined by means of the Natural Neighbor interpolation method. This method does not extrapolate values beyond the range of data and it is particularly suitable for data sets containing dense data in some areas and sparse data in other areas (see Sambridge et al ., 1995 for details). For the sake of consistency, the ratio (maximum range divided by the minimum range) and angle (preferred orientation of the major axis) of anisotropy were fixed at 1 and 0, respectively, for all maps. This means that we did not assume any preferred orientation in the data. Two distribution maps were created for each species, corresponding to the highest and lowest level of ln-cpue observed in the time-series. We used ln-cpue values in order to be consistent with the estimated temporal trends and to render maps more comprehensive. The period 1981–1985 was avoided for this purpose because of the low number of hauls ( Table 1 ). Natural Neighbor interpolation was performed using Surfer 8 (2002) computer software.

Results

Trends in biomass

The 32 demersal fish species displayed large fluctuations in ln-cpue in the Kattegat and Skagerrak during the period 1981–2003. Commercial roundfish species, such as cod, saithe, hake, and ling, showed a generally decreasing trend in ln-cpue, sometimes reaching very low values during the latest years, as in the case of pollack ( Figure 2a ). However, only for cod and pollack were the negative trends significant ( Table 2 ). On the other hand, whiting exhibited a remarkably stable ln-cpue, whereas haddock and saithe showed large fluctuations without a clear trend ( Figure 2a ). Only the ln-cpue of Norway pout increased significantly during the study period ( Table 2 ).

Table 2

The 32 fish species analysed in the study. Coefficients of determination and significance values (linear correlation) of ln-cupe temporal trends are also shown. n.s.: not significant, p > 0.05.

Group	Scientific name	Common name	r 2	p
Commercial roundfish	1 Gadus morhua	Cod	−0.57	<0.001
	2 Melanogrammus aeglefinus	Haddock	0.18	n.s.
	3 Merlangius merlangus	Whiting	0.15	n.s.
	4 Merluccius merluccius	Hake	−0.11	n.s.
	5 Molva molva	Ling	−0.08	n.s.
	6 Pollachius pollachius	Pollack	−0.58	<0.001
	7 Pollachius virens	Saithe	0.06	n.s.
	8 Trisopterus esmarkii	Norway pout	0.24	<0.05
Flatfish	9 Arnoglossus laterna	Scaldfish	0.52	<0.001
	10 Glyptocephalus cynoglossus	Witch	0.26	<0.05
	11 Hippoglossoides platessoides	Long rough dab	0.56	<0.001
	12 Limanda limanda	Dab	0.66	<0.001
	13 Microstomus kitt	Lemon sole	0.08	n.s.
	14 Platichthys flesus	Flounder	−0.01	n.s.
	15 Pleuronectes platessa	Plaice	0.34	<0.01
	16 Scophthalmus maximus	Turbot	−0.02	n.s.
	17 Scophthalmus rhombus	Brill	0.41	<0.001
	18 Solea solea	Common sole	0.18	<0.05
Non-target species	19 Anarhichas lupus	Wolffish	−0.07	n.s.
	20 Callionymus lyra	Dragonet	0.25	<0.05
	21 Callionymus maculatus	Spotted dragonet	0.41	<0.01
	22 Cyclopterus lumpus	Lumpsucker	−0.05	n.s.
	23 Eutrigla gurnardus	Grey gurnard	0.29	<0.01
	24 Lophius piscatorius	Angler	0.07	n.s.
	25 Lumpenus lampretaeformis	Snake blenny	0.44	<0.001
	26 Lycodes vahlii	Vahl's eelpout	0.26	<0.05
	27 Myoxocephalus scorpius	Bullrout	0.63	<0.001
	28 Myxine glutinosa	Hagfish	0.58	<0.001
	29 Raja radiata	Starry ray	−0.29	<0.01
	30 Rhinonemus cimbrius	Four-bearded rockling	0.02	n.s.
	31 Trachinus draco	Greater weever	−0.01	n.s.
	32 Trisopterus minutus	Poor cod	−0.46	<0.001

Figure 2

Trends in ln-cpue of 32 fish species collected during the first quarter IBTS performed in the Kattegat and Skagerrak between 1981 and 2003. Lines represent 2-year moving averages. Vertical bars represent upper 95% confidential intervals.

In contrast, the ln-cpue of flatfish species displayed a constant or increasing general trend ( Figure 2b ). A significant positive trend was observed in plaice, dab, long rough dab, common sole, scaldfish, witch, and brill ( Table 2 ). Conversely, flounder ln-cpue was fairly stable, and those of turbot and lemon sole fluctuated without a clear trend.

Among non-target fish species, the positive trend in the ln-cpue of hagfish, snake blenny, and spotted dragonet is noteworthy ( Figure 2c ). Grey gurnard, dragonet, bullrout, and Vahl's eelpout also significantly increased ( Table 2 ). Poor cod and starry ray were the only two non-target species whose ln-cpue significantly decreased during the study period ( Table 2 ). Some species (wolffish, angler, greater weever) showed large interannual fluctuations in ln-cpue without any trend, not surprising considering that they were caught in very small quantities.

Spatial distribution

Although there were slight differences among years in the spatial location of IBTS hauls, the eastern Kattegat was evenly covered every year. No hauls were made in the western part of the central Kattegat. Moreover, the western and central regions of the Skagerrak had a spatially irregular coverage ( Figure 3a–c ). This uneven spatial coverage needs to be kept in mind when interpreting species distributions. In particular, the interpolation between the southwestern Kattegat and the western Skagerrak, where there are no hauls, is a mere artefact produced by the software used and should not be considered reliable. Therefore, in order to avoid misinterpretations of the maps, the locations of the hauls are also indicated on Figure 3 .

Figure 3

Spatial distribution of 14 fish species at the highest (left map) and lowest (right map) values of ln-cpue observed during the IBTS survey in the Kattegat and Skagerrak between 1981 and 2003. Circles represent haul positions. Scale bars represent ln-cpue. The Natural Neighbor interpolation method was used to create the distribution maps (see text for details).

Analysis of the fish distribution showed that commercial roundfish tended to be aggregated at a low level of biomass and dispersed over a wider geographic range at high biomasses ( Figure 3a ). This pattern was particularly evident for pollack and saithe that, at low ln-cpue, were aggregated in restricted areas of the central Skagerrak. Haddock were mostly in the Skagerrak and southern Kattegat at low ln-cpue and over the whole studied area at high levels of ln-cpue. Cod seemed to be spread over the whole Kattegat and to have a more patchy distribution at high and low levels of biomass, respectively ( Figure 3a ). For whiting the contrast between low and high ln-cpue was small, which might explain the lack of any clear difference in spatial distribution.

Flatfish, on the other hand, did not show clear distribution patterns. Plaice and flounder did not show any differences in the degree of aggregation/dispersion at different levels of biomass ( Figure 3b ). The distribution of common sole was little changed in the years of lowest and highest ln-cpue, with two major aggregations, one in the central Kattegat and one in the southern Skagerrak. However, at high ln-cpue, the dense concentrations were more extensive. Lemon sole was aggregated in some areas of the Kattegat at low values of ln-cpue and distributed over the whole Kattegat and in the Skagerrak at high ln-cpue, whereas long rough dab was distributed over a somewhat broader area at high levels of biomass.

Spatial distribution of the non-target fish species resembled that of the commercial roundfish species. Poor cod were present in the Skagerrak at low ln-cpue and additionally in the Kattegat at high ln-cpue ( Figure 3c ). Four-bearded rockling and lumpsucker expanded over the whole study area at high biomasses. Hagfish, on the other hand, seemed to have a constant geographic range, being concentrated in small regions of the eastern Skagerrak at both high and low levels of biomass.

Discussion

Trends in biomass

Overall, the results based on fish ln-cpue for the Kattegat and Skagerrak are similar to those in other areas of the North Atlantic, i.e. a steady decrease of several commercially important roundfish (cod, pollack, hake, ling). Since the non-target fish species did not show any decrease in ln-cpue (except starry ray and poor cod), the decrease in roundfish is probably attributable to fishing pressure, even though we cannot rule out other possibilities, such as environmental change. However, two roundfish species, haddock and saithe, showed large fluctuations in ln-cpue without any clear trend. Similarly, whiting showed no trend in biomass during the period analysed. The results agree with stock assessment values for cod, haddock, and saithe evaluated for the whole North Sea, whereas the stable whiting biomass in the Kattegat and Skagerrak is different from stock assessment estimates ( ICES, 2003 ). It must be kept in mind that our analysis covered only a limited area of the North Sea, so comparisons could be misleading and, although the ln-cpue was relatively constant in the area, the average size of whiting has decreased (JH, unpublished data). The discrepancy between stock assessment values and our ln-cpue estimates suggests that in order to better understand the dynamics of fish stocks, biomass estimation should be performed at regional and local scales and implemented using as many data sets as possible.

Flatfish species displayed no or positive trends in ln-cpue. This agrees with the findings of Heessen and Daan (1996) in different areas of the North Sea, and could be due to the different ecology and life history traits of flatfish compared with roundfish. For example, dab and long rough dab grow rapidly and mature at relatively low size compared with most commercial demersal fish ( Heessen and Daan, 1996 ; Jennings and Kaiser, 1998 ). Therefore, balancing the higher mortality of the older and bigger individuals, those species are less vulnerable to intensive exploitation than roundfish ( Jennings and Kaiser, 1998 ). On the other hand, there are indications of increased invertebrate productivity in coastal waters of the North Sea over the last two decades. This could have increased the food supply for juvenile flatfish and, thus, positively affected flatfish populations ( Rijnsdorp and van Leewen, 1996 ). However, the IBTS is carried out in open waters where, even in a small system such as the Kattegat and Skagerrak, the increase in invertebrates as an effect of eutrophication could be minor compared with that in coastal zones.

Non-commercial species, which are neither a target nor a common bycatch of commercial fisheries (e.g. those with a vermiform body shape), showed also no or a positive trend in ln-cpue during the period analysed. An extreme example is the hagfish, whose ln-cpue increased almost 40 times during the last decade. This species is a scavenger ( Britton and Morton, 1994 ), feeding mostly upon dead or dying carcasses of other organisms on the sea bed. Therefore, one possible explanation of its expansion could be the increased amount of discarding from commercial vessels ( Jennings and Kaiser, 1998 ; Martini, 1998 ). This mechanism could also in part explain the increased biomass of the flatfish acknowledged to benefit from organisms damaged by fishing activity ( Millner and Whiting, 1996 ; Rijnsdorp and van Leewen, 1996 ; Jennings and Kaiser, 1998 ). The decrease in abundance of several top-predatory fish species (e.g. roundfish, see above) could represent an alternative explanation for the increase in abundance of several flatfish and non-target species; through predation release ( Pauly et al ., 1998 ; Myers and Worm, 2003 ). However, predator–prey coupling is difficult to demonstrate especially in open high-diversity systems and in exploited areas ( Jennings and Kaiser, 1998 ), and the cascading impact on other parts of the ecosystem should be evaluated in order to provide evidence for this hypothesis.

The IBTS has been carried out in the Kattegat and Skagerrak in a standard manner (unvaried sampling design, same fishing procedure and gear used) since 1979. Therefore, in our study the factors affecting fish catchability and cpue data in commercial fisheries (such as different fishing power, non-random trawling, diurnal and seasonal variation in fish distribution; Garrod, 1964 ; Gulland, 1964 ) can be ruled out. However, other possible sources of change in the availability of fish to the survey, such as age-dependent behaviour ( Swain et al ., 1994 ), cannot be excluded. Furthermore, the IBTS covers only the open sea areas, thus missing the inshore part of fish distributions.

Spatial distribution

Our spatial analysis suggests a general trend for all commercial roundfish species: a tendency to be aggregated and dispersed at low and high levels of biomass, respectively. Although we tried to be as consistent as possible in the analysis, the Natural Neighbor interpolation method used is merely a qualitative method and provides a visual representation of a phenomenon that needs to be confirmed using quantitative statistical tools. Moreover, the spatially uneven coverage of the survey and the interpolation between the southwestern Kattegat and the western Skagerrak needs to be kept in mind in the interpretation of the maps. The fact that fish populations seem to aggregate at low levels of biomass and disperse at high levels has previously been demonstrated for pelagic shoaling fish (reviewed by Winters and Wheeler, 1985 ) and for target demersal fish stocks such as haddock ( Crecco and Overholtz, 1990 ; Marshall and Frank, 1994 ) and cod ( Swain and Sinclair, 1994 ; Rose and Kulka, 1999 ). In our study we have shown a similar aggregation/dispersion phenomenon in non-target fish species. The only exception was that of hagfish, which were always restricted to a small area of the Skagerrak in spite of a huge interannual change in ln-cpue. This finding is not surprising considering that hagfish are highly site-specific, requiring specific substrata and high salinity ( Martini, 1998 ). Although the contraction and expansion behaviour of marine fish stocks is considered to occur in response to density-dependent habitat selection mechanisms ( MacCall, 1990 ), the factors originally triggering these processes are still poorly understood. However, food requirement and predation avoidance possibly drive the observed pattern ( MacCall, 1990 ). Another explanation could be related to fish spawning behaviour. In fact, as the data were collected in February, i.e. at the beginning of the spawning season for several species, the observed changes in spatial distribution might be driven by density-dependent behaviour related to spawning or by shifts in spawning time. An alternative reason could be associated with environmental change ( Gordoa and Hightower, 1991 ; Rose and Kulka, 1999 ).

Flatfish, on the other hand, seem not to follow the same general pattern. Long rough dab, common sole, and lemon sole showed a higher and lower level of aggregation at low and high ln-cpue, respectively. For plaice and flounder, conversely, this pattern was not evident. Swain and Morin (1996) found that the range of distribution of American plaice was not related to changes in abundance in the Gulf of St Lawrence. This possibly shows a generally different behaviour of flatfish compared with roundfish. An alternative hypothesis is that plaice and flounder have never reached, during our time-series, biomass levels high enough for density-dependent effects to be triggered.

The implications of spatial redistribution of fish stocks in response to density-dependent factors are huge. Variations in fish spatial distribution can lead to overestimating the stock size and underestimating fishing mortality ( Crecco and Overholtz, 1990 ; Rose and Kulka, 1999 ), with resulting risks for overexploitation and stock collapse. In fact, Rose and Kulka (1999) showed that the crash of the northern cod stock off Newfoundland was the result of misinterpretation of fishery cpue influenced by hyper-aggregation. Furthermore, they stressed that cpue from the fishery should not be used as an index of demersal fish abundance without prior knowledge of the spatial characteristics of the population and of the fishing vessels (see also Harley et al ., 2001 ; Salthaug and Aanes, 2003 ). Therefore, the change in spatial distribution of fish stocks in the Kattegat and Skagerrak area presented in this paper is a phenomenon particularly important not only from an ecological point of view but also from a management perspective.

We thank the staff at the Institute of Marine Research in Lysekil and the crew of the RV “Argos” who contributed to data collection, and to Julia Blanchard, Henk Heessen, and Verena Trenkel for helpful comments and suggestions on an early draft of the manuscript.