Fisheries Research 72 (2005) 301–310
Studying species associations from commercial catch data: a Baltic Sea application Laura Uusitaloa,∗ , Kimmo Vehkalahtib , Sakari Kuikkaa , Pirkko S¨oderkultalahtic a
Department of Biological and Environmental Sciences, University of Helsinki, P.O. Box 65, FIN-00014 Helsinki, Finland b Department of Mathematics and Statistics, University of Helsinki, P.O. Box 54, FIN-00014 Helsinki, Finland c Finnish Game and Fisheries Research Institute, Pukinm¨ aenaukio 4, FIN-00720 Helsinki, Finland Received 22 April 2004; received in revised form 14 October 2004; accepted 19 October 2004
Abstract Finland has 1100 km coast line with relatively strong environmental gradients. Finland has also gathered exceptionally comprehensive data on commercial catches of several fish species along the coast in relatively small units. Multidimensional scaling was applied to these data to test the hypothesis that the effects of the environmental gradients to the abundance and species composition of fish can be seen in the catches. This analysis divided the species into three groups: marine species, salmonid species and coastal species groups, the reproduction biology being the connecting factor between the species in each group. This implies that the reproduction success in a given environment is the main factor affecting the productivity and distribution of species along the coast of Finland. The commercial catch data proved informative in regard to the ecological similarity of the species. © 2004 Elsevier B.V. All rights reserved. Keywords: Species associations; Baltic Sea; Coastal species; Freshwater species; Catch data
1. Introduction The brackish-water ecosystem of the Baltic Sea is unique in the sense that it includes both marine and freshwater fish species, many of which are found in considerable amounts and are subject to commercial fishery. The Baltic Sea fish stocks have adapted to the low-salt conditions of their environment. Cod ∗ Corresponding author. Tel.: +358 9 191 58992; fax: +358 9 191 58257. E-mail address:
[email protected] (L. Uusitalo).
0165-7836/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.fishres.2004.10.005
(Gadus morhua L.), for example, have adapted to the low-salinity conditions in the Baltic Sea: the eggs of the eastern Baltic cod stock float at salinities of 14.5 ± 1.2‰ and those of the western Baltic stock at 20–22‰ (Nissling and Westin, 1997). The freshwater species, on the other hand, have had to adapt to the saline conditions in the Baltic and have developed specific populations in coastal areas with limited salinity (Ojaveer et al., 1981), e.g. the European grayling (Thymallus thymallus (L.)) stocks in the Baltic Sea are genetically different from the closely located river stocks (Koskinen et al., 2002).
302
L. Uusitalo et al. / Fisheries Research 72 (2005) 301–310
In a multi-species fishery like in the Baltic Sea coastal areas, not all species can be managed separately. It is thus beneficial to fishery management and conservation of the stocks to have information on which species react similarly to different pressures of the environment, and which are the critical factors affecting their success in a given environment. In the present analysis we assess the species associations in the Baltic Sea coastal areas of Finland using the commercial catch statistics. Several definitions exist for biological or species associations, which are defined as co-occurring species (Legendre and Legendre, 1978) or as species that have similar reactions to the properties of the environment (Fager and McGowan, 1963). These definitions ignore the taxonomical similarity of species and focus on ecological resemblance, i.e. they take into account the environmental factors behind the productivity of populations. The traditional approaches based on experimental testing and random data sampling are often costly and can give information only on limited geographical and time scales. Moreover, the sampling error may be very high due to limited sampling over time and space. These approaches can thus be complemented with an approach that utilizes existing large datasets, e.g. catch statistics. The ecological results of studies based on such datasets are inevitably broad in nature. They may, however, add valuable information to the more sophisticated, yet smaller-scale analyses of traditional ecological studies. The Finnish catch statistics provide an exceptionally good data for this purpose. The Finnish coast line offers an interesting natural experimental setting that allows the study of species responses in an environment with various varying environmental factors. For example, the salinity changes from 4‰ in Archipelago Sea to 1‰ in the northernmost part of Bothnian Bay. The commercially exploited species occurring along the coast are relatively local, which allows connecting the environmental conditions of reproduction to the abundance of the species in the area. Tagging experiments have shown that, for example, most of the herring (Clupea harengus membras L.) that spawn in the Archipelago Sea also are caught in the same area (Lehtonen et al., 1983). The migrations of pikeperch (Sander lucioperca L.) in the Archipelago Sea were also generally less than 30 km and those of pike (Esox lucius L.) less than 10 km even when tags returned less than 2
months after the release were excluded (Lehtonen et al., 1983). Bream (Abramis brama L.) and roach (Rutilus rutilus L.) generally migrated less than 20 km in the same study. Also the flounder (Platichthys flesus L.) and perch (Perca fluviatilis L.) populations are quite local (Aro and Sj¨oblom, 1982, 1983). The recruitment of the stocks is in many cases dependent on environmental factors: e.g. increase in the abundance of roach has been attributed to eutrophication (Lappalainen et al., 2001), year-class strengths of pikeperch and perch have been shown to correlate with temperature of the first summer (Lehtonen and Lappalainen, 1995), and the recruitment of cod in the Baltic Sea is strongly dependent on suitable salinity and oxygen concentrations (Ojaveer et al., 1981). The locality of the fish stocks and the impact of environmental factors are the corner stones of our analysis. The current dataset also allows us to evaluate the usefulness of catch statistics in studying species similarities and in evaluating species productivity. We are aware of no other such studies on a similar geographic scale.
2. Materials and methods 2.1. Description of study area The present study utilizes an extensive dataset that includes all commercial fish catches from the coastal areas of Finland during 1980–2000. The total volume of the Finnish commercial fishery in the ICES subdivisions 29–32 as described by yearly statistics (mean yearly catch during 1980–2000) are: Atlantic cod 1635 t, European flounder 67 t, European whitefish (Coregonus lavaretus L.) 1146 t and pikeperch 217 t (FGFRI, 2001). The coastline of Finland, consisting of the northern gulfs of the Baltic Sea, is 1100 km in length measured as a straight line (Gran¨o et al., 1999). The total length of the shoreline is more than 46 000 km, however, including over 73 000 islands (Gran¨o et al., 1999). Gran¨o et al. (1999) referred to this mosaic-like landscape as an archipelago coast and noted that it has a high geodiversity, a concept similar to biodiversity, with a clear impact on the biodiversity of the area. The Archipelago Sea on the southwestern coast of Finland comprises of over 22 000 islands and has more than 14 000 km
L. Uusitalo et al. / Fisheries Research 72 (2005) 301–310
of coastline (Gran¨o et al., 1999), providing fish with abundant shallow waters. Here, the salinity is also the highest occurring along the coast of Finland; about 5‰. The salinity decreases when moving towards the bays, either northward or eastward, so that at the bottom of the Bothnian Bay salinity is 1‰ and in the Gulf of Finland, near the eastern border of Finland, about 4‰. In the river mouths, salinity is near 0‰. This combination of climatic and oceanic factors is unique in the world. The length of the coastline relative to the area, i.e. shore density, varies along the coast as well. The shore density is highest in the Archipelago Sea, in the western part of the southern coast, and in the Quark area in the middle of the Gulf of Bothnia (Gran¨o et al., 1999). The coastal zone is also narrower in the Gulf of Bothnia than in the Gulf of Finland and has a generally lower shore density (Gran¨o et al., 1999). The northern part of the coast thus offers fewer spawning habitats for species spawning in shallow coastal waters. Temperature is the third important variable factor in the environmental conditions. The duration of the ice-covered season averages 194 days in the northern˚ most part of the Baltic Sea and 60 days in the Aland Archipelago (Sein¨a and Peltola, 1991). These factors together induce remarkable variance in the fish assemblage in different parts of the Baltic. 2.2. The data The coastal fish catch data from Finland span 21 years (1980–2000), including all commercial catches of 19 fish species (FGFRI, 2001). These species include Baltic herring, European sprat (Sprattus sprattus (L.)), cod, flounder, northern pike, vendace (Coregonus albula L.), whitefish, Atlantic salmon (Salmo salar L.), sea-run brown trout (Salmo trutta trutta L.), smelt (Osmerus eperlanus (L.)), bream, ide (Leuciscus idus (L.)), roach, burbot (Lota lota (L.)), perch, pikeperch, European eel (Anguilla anguilla (L.)), rainbow trout (Oncorhynchus mykiss (Walbaum)), and turbot (Scophthalmus maximus (L.)). The methods of data compilation and the reliability of the statistics are described in the Finnish Game and Fisheries Research Institute’s (FGFRI) quality report of the statistics (FGFRI, 2002, 2004). The Finnish Ministry of Agriculture and Forestry maintains registers of fishermen and fishing vessels, and all 3000 fishing units are required to report on their fishing activity to the au-
303
Fig. 1. Northern parts of the Baltic Sea; ICES statistical rectangles marked. The study includes the rectangles between the strong line and Finnish coast.
thorities, using either EU logbooks or monthly coastal reports. At the end of the year, the FGFRI obtains the data and checks, analyzes, and supplements it. The data are also cross-checked with landing statistics and with data collected from the wholesale buyers of fish. Both the data acquisition and compilation of statistics are controlled by EU regulations, e.g. EC 3880/91 (FGFRI, 2002). The total number of observations, aggregated to a monthly level, is more than 13 000. The data consist of catches of the fish species mentioned in kilograms per statistical rectangle (Fig. 1) for each month. Rainbow trout, turbot, and eel were excluded from the analyses, however, because the data for these species were quite sparse. Each of the statistical rectangles cover an area between half a parallel and one meridian; e.g. 60.0–60.5◦ N 21.0–22.0◦ E. We are not aware of other geographically as detailed catch statistics in the Baltic Sea or Atlantic area. A total of 44 rectangles, along or near the coast of Finland, were included in the analyses (Fig. 1). Lying on the coast, many included both water surface and land in their area. The areas of the rectangles varied from 2543 to 3120 km2 , depending on the latitude, and their water surface areas varied from 75 km2 to the entire area of the rectangle. Yearly catches were used to reduce noise, such as random and seasonal variation. Furthermore, the data were transformed from total catches into catches per
304
L. Uusitalo et al. / Fisheries Research 72 (2005) 301–310
unit area of water surface, due to the different water surface areas of the rectangles. The final dataset consisted of yearly catches of 16 fish species from 44 rectangles during 21 years, a total of 924 observations from each species. Total catches were chosen instead of catch per unit effort (CPUE) because of their better comparability. The fish catching methods vary along the coast, and the mesh sizes of the nets have changed during the years as well as the sizes of most gears as a result of technological development in gear materials. This suggests that the catchabilities have increased during the years, and CPUE may thus be a poor estimator of stock sizes across the years. Furthermore, the CPUE might give biased picture of species that are mostly caught as by-catch. We suggest that the total catch is a reasonable proxy for the relative stock size, which is the required level of data accuracy for this type of analysis. 2.3. Statistical methods Ordination methods are widely used in ecology to group species and relate species abundance and environmental data (Fern´andez-Al´aez et al., 2002; Hoeinghaus et al., 2003; Soininen and K¨on¨onen, 2004). Multidimensional scaling (MDS) was chosen among the ordination methods because of its ability to deal with abundance data or other types of data where it is possible that zeros emerge from both ends of the environmental gradient, such as from there being either too high or too low salinity for the species to occur there (Legendre and Legendre, 1998). The yearly variation and possible interannual autocorrelation were removed from the dataset of 21 years by summing the yearly catches in the rectangles for each species separately. The aggregated data thus consisted of summed catches of 16 fish species from 44 rectangles. Since the catch statistics varied remarkably between the rectangles, the data were standardized, by subtracting the means and dividing by the standard deviations of the summed catches of each species. MDS was applied to determine the ordination of the rectangles (Legendre and Legendre, 1998). Classical scaling (Torgerson, 1958) was used to obtain an initial configuration for an iterative least squares scaling (Gower and Hand, 1996). For simplicity, the correlation (subtracted from 1) was chosen as a measure of dissimilarity or distance between the rectangles.
The dimensionality of the ordination was assessed primarily by evaluating the ordination diagrams to find meaningful biological interpretations. In addition, a statistical goodness-of-fit measure, stress, was applied. Stress is the normalized value of the residual sum of squares, which is the criterion value of the least squares scaling. Smaller stress indicates a better fit in the statistical sense. The aim was to find a reasonable number of dimensions to display the information in sufficient precision. To find a meaningful grouping of the rectangles, cluster analysis was conducted with the same distance matrix that was used with the MDS. Several initial groupings were achieved with the k-medoid method (Kaufman and Rousseeuw, 1990), and the final solution of four groups was found with the single linkage method. The clusters were visualized together with the MDS dimensions. The ordination of the fish species was performed using the complete dataset of 924 observations and 16 fish species. The data were standardized by subtracting the means and dividing by the standard deviations of the catches of each species. To examine the stability of the species associations, the data were divided into two sets: the 1980s and the 1990s (the latter included the year 2000). MDS was applied to the entire dataset and the two time slices, using the same methods and dissimilarity measure as with the rectangles. Other measures, such as Bray-Curtis, were tried but the correlation (subtracted from 1) appeared to be a better choice for these data. The dimensionalities of the ordinations were assessed by evaluating the ordination diagrams, and by comparing the stress values of the scalings in different dimensions. The ordinations were made comparable with each other using Procrustes analysis (Cox and Cox, 1994) by matching the ordinations of the two time slices to the ordination of the entire dataset. Groups among the fish species were found by hierarchical clustering, since the distance matrix of 16 species was too small for k-medoid clustering. Ward’s method, an agglomerative algorithm minimizing the intragroup variance, was applied with the correlation (subtracted from 1) as the dissimilarity measure. The clustering was visualized as a tree graph (dendrogram). The statistical analyses were carried out with SURVO MM software (Mustonen, 2001).
L. Uusitalo et al. / Fisheries Research 72 (2005) 301–310
305
3. Results For interpretation, the ordination of the rectangles appeared to be sufficient in two dimensions, with a stress value of 23.4%. A further dimension did not bring up any substantial information. Together with the cluster analysis, the ordination revealed four distinct groups among the rectangles (Figs. 2 and 3), each group comprising of rectangles of different nature. The groups were named Pelagic, Archipelago & Southern coast, Northern coast, and Bay bottom. The pelagic group consisted of rectangles that had no or very little coastline or archipelago. Archipelago & Southern coast group consisted of rectangles along the southern part of ˚ the coast of Finland, and the Aland Archipelago. The Northern coast group included the rectangles along the northern part of the coast, except for the three northernmost coastal rectangles that formed the Bay bottom group. The ordination of the species was also found satisfactory for interpretation with two dimensions (stress 23.2%). The ordination diagram (Fig. 4) shows the associations of the commercially exploited fish species in the Baltic Sea. The freshwater species (pikeperch, bream, ide, perch, burbot, pike, and roach) cluster near each other. Cod and sprat were found near each other, as were herring and flounder, which were between
Fig. 2. Least squares MDS ordination diagram of the aggregated rectangle data, with different symbols indicating the groups revealed by the cluster analysis of the same data.
Fig. 3. Map of groups revealed by the cluster analysis of the aggregated rectangle data.
the freshwater species and the cod/sprat group. The salmonid species (smelt, vendace, trout and salmon) clustered near each other as well. Hierarchical clustering (Fig. 5) revealed three biologically compatible groups: the freshwater species, together with the marine species herring and flounder, formed one group, the salmonid species (smelt, whitefish, vendace, trout, and salmon) formed another, and cod and sprat formed the third group. These associations, revealed by ordination
Fig. 4. Least squares MDS ordination diagram of the species data.
306
L. Uusitalo et al. / Fisheries Research 72 (2005) 301–310
Fig. 5. Dendrogram of the hierarchical cluster analysis of the species data, indicating three groups of species.
and clustering, correspond closely to fish biologists’ views of species similarity. The fact that biologically meaningful groups emerged in the analysis based on mere catch statistics supports the view that good catch statistics data can be sufficient for the analysis of biological associations of species. To study the stability of the species associations, the data were divided into two sets: the 1980s and 1990s. The cutoff point is well justified by the fact that the cod catches decreased in the study area since the mid-1980s, and the cod fishery virtually ended by 1990. Thus the two periods represent different types of ecosystem state – one with cod present both in the ecosystem and fishery and the other with cod virtually absent from both. The ordinations were performed in two dimensions to compare the results with each other and the entire dataset. The stress values were 20.0% for the 1980s and 22.2% for the 1990s. The ordination diagrams (Fig. 6) show results consistent with the entire dataset suggesting that collapse of the cod stocks did not alter the ecosystem so much that it would have strongly affected the reciprocal associations of the other species. 4. Discussion The present study suggests that catch statistics can be used as the basis of successful study of species as-
Fig. 6. Least squares MDS ordination diagrams of the species data in two time slices, the 1980s (above), and the 1990s.
sociations, even without fishing effort data that would support the use of relative biomass estimates. This is an important finding, since catch statistics are often available even though there were no fish abundance data produced by a random sampling procedure. A wide variety of habitat types (high variance in the environmental factors) as well as high variance in the catches in the study area are probably needed, however, to obtain data that are truly informative of species associations and to avoid associations that are artefacts of a specific environmental type. It may be important that the asso-
L. Uusitalo et al. / Fisheries Research 72 (2005) 301–310
ciations found reflect the ecological associations, since if these are only artefacts of the specific environmental type, they may not hold if the environmental conditions change. The coastal areas of Finland are ideal for this type of large-scale analysis of the effects of environmental factors. The associations found in this study (Figs. 4 and 5) appear to reflect quite faithfully the biological similarities between the species. The high variance of environmental factors in the coastal waters of Finland induces widespread variance in the total catches, enabling information to be gathered on species associations. The factor that apparently explains the grouping is the reproduction biology. Cod and sprat, which form one group together (Fig. 5), are pelagic/benthopelagic species that also spawn pelagically. Sprat spawn in the Baltic Sea at salinities of 5‰ and more, at temperatures of 4–14 ◦ C in the deeps and their associated slopes, from the west˚ ern shores of the Baltic to the Aland Sea and the Gulf of Bothnia (Ojaveer et al., 1981). Baltic cod eggs need salinities of 10‰ or more to develop, and thus their spawning is limited to areas at least 60–80 m deep (Ojaveer et al., 1981). The second group consists of salmonid species that mainly reproduce in rivers or in the sea area on hard sand or gravel bottoms and prefer cool waters. Searun brown trout and salmon reproduce in rivers and a high proportion of recruitment is based on rearing systems (ICES, 2003a). Sea-spawning whitefish and vendace spawn on open-sea reefs and occur mainly in the northernmost parts of the Baltic Sea, where the salinity is low (Ojaveer et al., 1981). Smelt also have migratory forms that ascend rivers and spawn on sandy or gravelly bottoms. In the third group, the connecting factor appears to be dependency on shallow coastal waters in reproduction. The pikeperch, roach, bream, pike, burbot, perch, and ide are freshwater species, and their reproduction is limited to river mouths, inlets, and coastal waters on which their life cycle is dependent (Ojaveer et al., 1981). Flounder and herring were found in the same group with these species. Even though they are marine species, their reproduction is strongly dependent on the coastal zone as well: The flounder stocks of the Finnish coast spawn in the littoral and coastal zone in depth of 2–20 m (Sandman, 1906) in contrary to the
307
more southern stocks that spawn in the deeps. Baltic herring are, as well, coastal spawners that deposit their eggs on the seabed in shallow littoral waters (Haegele and Schweigert, 1985), and Urho and Hild´en (1990) demonstrated that herring productivity is very dependent on the presence of shallow areas. Baltic herring also live and show good growth in the northernmost parts of the Baltic Sea, where the salinity is very close to zero (Parmanne, 1990), suggesting that they are very well adapted to the low-salinity conditions of the Baltic Sea. To our knowledge, no other studies are available that would demonstrate this similarity of herring with freshwater species. This group is here called the ‘coastal species group’. The finding that grouping could be explained by reproduction biology is in good accordance with Werner’s (2002) results that the key factors affecting fish species’ success in certain environments are often related to reproduction, since egg and larval stages are often more vulnerable to environmental perturbations than are adults. The ecological requirements of egg and larvae and the environmental factors at the spawning and nursery sites thus together largely determine the success of a species, and species that succeed in similar environments form associations. The lack of recreational catches in the data is a drawback, since they form a fairly important part of the total catches of some fish species, especially pike, perch, and pikeperch. The type of scaling used, however, takes into account the fact that only certain components of the catches are included to the input data. As long as the proportions of commercial and recreational catches are not too area-specific, they cannot give rise to the current results. This is supported by the fact that even rectangles very near each other (where the ratios between commercial and recreational catches are probably similar) were classified to separate groups (Figs. 2 and 3). The high variance in cod recruitment in the Baltic may also have affected the results. There was good recruitment of cod in the late 1970s, followed by collapse of the cod stock in the early 1990s (ICES, 2003b). In the study area, the cod catches were higher than 10 kg/ha for most of the 1980s, and more than 30 kg/ha in 1984. By the early 1990s, however, the catches declined to near zero and did not recover during the study period. Under these circumstances, the cod catches do not necessarily reflect properly the abundance of the stock.
308
L. Uusitalo et al. / Fisheries Research 72 (2005) 301–310
Fig. 7. Composition of the total fish catches other than herring in each rectangle. The species groups as identified by the hierarchical cluster analysis.
The catch statistics furnished abundant information on the species’ relationships with the environmental variables, as is clearly shown by analysis of the rectangles (Figs. 2 and 3). The fish catch data alone divided the statistical rectangles into four ecologically compatible groups, enabling us to assume that environmental characteristics largely govern the community structure. The varying nature of the coast of Finland is also illustrated by the composition of the catches in different parts of the coast. The herring is the most abundant fish species throughout the coast and comprises 39–99% of the yearly average catches in every rectangle. The composition of the remaining catch varies considerably along the coast, however (Fig. 7). The species groups in Fig. 7 are those identified in the hierarchical clustering analysis (Fig. 5). In the northernmost part of the Baltic, salmonid species comprise the majority of the non-herring catch. In the southern Gulf of Bothnia and Gulf of Finland pelagic areas, the marine species sprat and cod dominate this fraction. Coastal species (the coastal species group of Fig. 5, excluding herring) are
caught in each rectangle that meets the coastline, but they dominate the non-herring catch only in the southwestern rectangles, where there is especially high shore density. This variability in the catch composition along the coast suggests that the environmental factors along the Baltic seacoast of Finland are variable enough to support this type of analysis. Paszkowski and Tonn (2000) showed that in physically similar lakes the fish and bird assemblages were quite similar as well. This phenomenon, called community concordance, suggests that the communities respond similarly but independently of each other to similar environmental factors. Another related concept is coherence, in which different ecosystems behave similarly through time (Magnuson et al., 1990). The former basically implies community dependence on longterm environmental factors, whereas the latter implies dependence on short-term factors. Both of these processes probably have an impact here, but as the time slice analysis showed (Fig. 6), no short-term factors were present that would have masked the community concordance effect. In dealing with species associations, it must be noted that while reacting to the properties of the physical environment, species frequently found together also affect each other’s environmental conditions. They may thus have evolved adaptations to each other’s presence as well. Kupschus and Tremain (2001) showed, however, that in the estuarine area of Indian River Lagoon, central Florida, physical environmental variables were of primary importance to the fish assemblages, whereas biological interactions were of minor importance.
5. Conclusion The groupings found in this study relate to biological features of the stocks that are connected with the productivity of the stocks in a given environment. Another approach is to examine the variance over time with respect to some environmental factors, but we argue that the variance over space is at least as important, because sustainability of the stocks requires that the species will be able to reproduce in their native environments. Of the target species of Finnish professional fishermen, only salmon, herring, cod, and sprat are managed through a Total Allowable Catch (TAC) policy; the remaining species are managed using technical measures
L. Uusitalo et al. / Fisheries Research 72 (2005) 301–310
such as mesh size restrictions, closure areas, and closure periods, and large proportions of them are subject to mixed-stock fishery. Mixed-stock fishery is risky to stocks that are usually low in abundance when the target stocks are at high levels, because the high biomass may cause the fishing pressure to rise too high for rare by-catch species. In such cases, management is more sustainable if it is focused on biologically similar stocks in the given environment. Knowledge of species associations in the Baltic may also prove useful in helping to determine the early warning signs of population depletion in species that are less intensively monitored. If the unit catches of an evenly exploited stock decrease, the change may be due to environmental factors that probably also affect species having similar critical biological requirements in that specific environment, and precautionary management measures may be called for here. Further analysis should examine the selectivity features of each fishery and group of stocks to estimate the impact of management efforts on different species in more detail.
Acknowledgements This work was partly financed by the Finnish Biological Interactions Graduate School. Professor Hannu Lehtonen, Samu M¨antyniemi, and Dr. Jyrki Lappalainen provided comments and ideas. Dr. Antti Lappalainen, Tapani Pakarinen, and Esko Araj¨arvi gave invaluable help with some of the graphics. We wish to thank two anonymous reviewers for helpful comments.
References Aro, E., Sj¨oblom, V., 1982. Stock assessment of flounder off the coast of Finland in 1975–1981. ICES CM 1982/J: 25, 15 pp. Aro, E., Sj¨oblom, V., 1983. The migration of flounder in the northern Baltic Sea ICES CM 1983/J: 26, 12 pp. Cox, T.F., Cox, M.A.A., 1994. Multidimensional Scaling. Chapman & Hall, London, p. 213. Fager, E.W., McGowan, J.A., 1963. Zooplankton species groups in the Northern Pacific. Science 140, 453–460. Fern´andez-Al´aez, C., de Soto, J., Fern´andez-Al´aez, M., Garc´ıaCriado, F., 2002. Spatial structure of the caddisfly (Insecta Trichoptera) communities in a river basin in NW Spain affected by coal mining. Hydrobiologia 487, 193–205. FGFRI, 2001. Finnish Fishery Time Series. Finnish Game and Fisheries Research Institute, p. 112.
309
FGFRI, 2002. Professional Marine Fishery 2001. Finnish Game and Fisheries Research Institute. SVT – Official Statistics of Finland – Agriculture, Forestry and Fishery 2002:56, 52 pp. FGFRI, 2004. Quality description of commercial marine fishery data. http://www.rktl.fi/english/statistics/fishing/commercial marine fishery/quality description.html. Gower, J.C., Hand, D.J., 1996. Biplots. Chapman & Hall, London, 277 pp. Gran¨o, O., Roto, M., Laurila, L., 1999. Environment and Land Use in the Shore Zone of the Coast of Finland. Publicationes Istituti Geographici Universitatis Turkuensis, 76 pp. Haegele, C.W., Schweigert, J.F., 1985. Distribution and characteristics of herring spawning grounds and description of spawning behaviour. Can. J. Fish. Aquat. Sci. 42 (Suppl. 1), 39–55. Hoeinghaus, D.J., Layman, C.A., Arrington, D.A., Winemiller, K.O., 2003. Spatiotemporal variation in fish assemblage structure in tropical floodplain creeks. Environ. Biol. Fish. 67 (4), 379– 387. ICES, 2003a. Report of the Baltic salmon and trout assessment group. ICES Advisory Committee on Fishery Management. ICES CM 2003/ACFM:20, 262 pp. ICES, 2003b. Report of the Baltic Fisheries Assessment Working Group, ICES Headquarters, 7–16 April 2003. ICES CM 2003/ACFM:21, 522 pp. Kaufman, L., Rousseeuw, P.J., 1990. Finding Groups in Data: An Introduction to Cluster Analysis. Wiley, New York, 342 pp. Koskinen, M.T., Nilsson, J., Veseleov, A., Potutkin, A.G., Ranta, E., Primmer, C.R., 2002. Microsatellite data resolve phylogeographic patterns in European grayling, Thymallus thymallus, Salmonidae. Heredity 88, 391–401. Kupschus, S., Tremain, D., 2001. Associations between fish assemblages and environmental factors in nearshore habitats of a subtropical estuary. J. Fish Biol. 58, 1383–1403. Lappalainen, A., Rask, M., Koponen, H., Vesala, S., 2001. Relative abundance, diet and growth of perch (Perca fluviatilis) and roach (Rutilus rutilus) at Tv¨arminne, Northern Baltic Sea, in 1975 and 1997: responses to eutrophication? Bor. Environ. Res. 6, 107–118. Legendre, L., Legendre, P., 1978. Associations. In: Sournia, A. (Ed.), Phytoplankton Manual. Monographs on Oceanographic Methodology, vol. 6. UNESCO, pp. 261–272. Legendre, P., Legendre, L., 1998. Numerical Ecology, 2nd English edition. Elsevier, Amsterdam, 837 pp. Lehtonen, H., B¨ohling, P., Hild´en, M. 1983. Saaristomeren pohjoisosan kalavarat. Riista- ja kalatalouden tutkimuslaitos, kalantutkimusosasto. Monistettuja julkaisuja 9 (Fish resources of the northern part of the Archipelago Sea. Finnish Game and Fisheries Research Institute. In Finnish). ISBN 951-9092-22-6. Lehtonen, H., Lappalainen, J. 1995. The effects of climate on the year-class variations of certain freshwater fish species. In: Beamish, R.J. (Ed.), Climate Change and Northern Fish Populations. Can. Spec. Publ. Fish. Aquat. Sci. 121, 37–44. Magnuson, J.J., Benson, B.J., Kraz, T.K., 1990. Temporal coherence in the limnology of a suite of lakes in Wisconsin, USA. Freshw. Biol. 23, 145–159. Nissling, A., Westin, L., 1997. Salinity requirements for successful spawning of Baltic and Belt Sea cod and the potential for cod
310
L. Uusitalo et al. / Fisheries Research 72 (2005) 301–310
stock interactions in the Baltic Sea. Mar. Ecol. Progr. Ser. 152 (1–3), 261–271. Ojaveer, E., Lindroth, A., Bagge, O., Lehtonen, H., Toivonen, J., 1981. Fish and fisheries. In: Voipio, A. (Ed.), The Baltic Sea. Elsevier Oceanography Series, vol. 30. Elsevier, Amsterdam, ISBN: 0-444-41884-9. Parmanne, R., 1990. Growth, morphological variation and migrations of herring (Clupea harengus L.) in the northern Baltic Sea. Finnish Fish. Res. 10, 1–48. Paszkowski, C.A., Tonn, W.M., 2000. Community concordance between the fish and aquatic birds of lakes in northern Alberta, Canada: the relative importance of environmental and biotic factors. Freshw. Biol. 43, 421–437. Sandman, J.A., 1906. Kurzer Bericht u¨ ber in Finnland ausgef¨uhrte Unter suchungen u¨ ber den Flunder, den Steinbutt und den Kabeljau. Rapp. P. -v. R´eun. Cons. Int. Explor. Mer. 5, 37– 44.
Soininen, J., K¨on¨onen, K., 2004. Comparative study of monitoring south-Finnish rivers and streams using macroinvertebrate and benthic diatom community structure. Aquat. Ecol. 38 (1), 63–75. Sein¨a, A., Peltola, J., 1991. Duration of the ice season and statistics of fast ice thickness along the Finnish coast 1961–1990. Finnish Mar. Res., 258. Torgerson, W.S., 1958. Theory and Methods of Scaling. Wiley, London. Urho, L., Hild´en, M., 1990. Distribution patterns of Baltic herring larvae, Clupea harengus L., in the coastal waters off Helsinki, Finland. J. Plankton Res. 12 (1), 41–54. Werner, R.G., 2002. Habitat requirements. In: Fuiman, L.A., Werner, R.G. (Eds.), Fishery Science: The Unique Contributions of Early Life Stages. Blackwell Science, Cornwall, ISBN: 0-632-056614. Mustonen, S., 2001. New Windows version of Survo. http://www.survo.fi/mm/english.html.
