Clustering Large Marine Ecosystems by capture data

1. INTRODUCTION

The Johannesburg Plan of Implementation of the World Summit on Sustainable Development (Anonymous, 2002a), noting the Reykjavik Declaration on Responsible Fisheries in the Marine Ecosystem (Anonymous, 2001), set the goal of encouraging the application by 2010 of the ecosystem approach to responsible fisheries (paragraph 29(d) of the Plan). This is an internationally agreed starting point for a new approach to fisheries management and fishery related studies utilizing a multinational, interdisciplinary approach, which integrates information concerning productivity, ecology, fisheries, socio-economic aspects and governance. Since mid-1980s, it has been developed the definition of Large Marine Ecosystems (LMEs) that represented a proposal to give an ecology-based partition of global oceans. The LMEs project called for a more ecologically sensible monitoring of fishery resources, to go beyond the purely biological and socio-economic view of marine resources and improve the awareness of shared resources among countries (Sherman and Alexander, 1986, 1989; Sherman et al., 1990, 1991, 1993).

Initially, 49 LMEs were identified (Sherman and Alexander, 1986) and then an additional 50th was proposed (Bakun et al., 1999). LMEs were defined on the basis of “...consideration of distinct bathymetry, hydrography, productivity, and trophically dependent populations...” (Sherman et al., 1993). This definition is rather broad. For some of the 50 LMEs, not only the ecological aspects but also geopolitical aspects have been considered. In others LMEs, distinct habitats and ecosystems have been put together. For these reasons, and following the publication of numerous papers, books and research results, the list has been expanded and some LMEs subdivided in order to increase their ecological significance, and to expand the coverage of all main shelf areas. The latest list, available at the LME web site managed by NOAA (2002), includes 64 LMEs. However, as the background work for this study initiated in 1999, the paper is based on the 50 LMEs described at that time (see list in Table 1 and map in Appendix 3).

1.1 Overview and scope of the work

The initial purpose of the present work was mainly to made available capture fishery production statistics by LME to scientists carrying out studies on individual LMEs. This encompassed the re-arrangement of the statistics included in the FAO capture database, which are organized into 19 marine fishing areas, and the research of data at the sub-national level needed to disaggregate the national data reported to FAO into defined regions which belong to different LMEs. Preliminary work on the feasibility of re-arranging the FAO capture statistics into the LMEs’ borders was carried out and the congruences and incongruences between the two partitions identified. However, in the course of the work, several difficulties have been encountered both in re-assigning the FAO statistics to LMEs and in the availability of additional data from national sources.

Due to data limitations, it has been possible to assemble data series only for a limited number of years (1990-99) and for a majority but not for all LMEs (43 out of 50; see Table 1) as for seven of them sub-national data were not available to FAO. The data compiled can be requested to the FAO Fishery Information, Data and Statistics Unit (FIDI) by scientists interested in LMEs’ studies but there is no plan to update the catch series by LME.

Table 1. List of the 50 Large Marine Ecosystems (as from Sherman and Duda, 1999)

* LMEs for which, given the unavailability of sub-national capture statistics, data were not compiled.

Although in one of the LME definitions (Sherman et al., 1993) it is mentioned that “...the seaward limit of the LMEs extends beyond the physical outer limit of the shelves to include all or a portion of the continental slopes as well...” the principal characteristics described in studies on single LMEs (e.g. Sherman and Alexander, 1986, 1989; Sherman et al., 1990, 1991, 1993, 1996, 1998; Sherman and Tang, 1999; Kumpf et al., 1999) refer mostly to the marine areas over the continental shelves. Furthermore, it seems that recently this definition has been refined as “Large Marine Ecosystems are regions of ocean space encompassing coastal areas from river basins and estuaries to the seaward boundaries of continental shelves and the outer margins of the major current systems” (Anonymous, 2002b). For these reasons, only capture statistics of species spending most of their life cycles in the shelf areas have been considered in this analysis, thus excluding all species items classified as oceanic for the other study contained in this volume.

As the short period of data availability did not allow a thorough analysis of trends by LMEs, this study manly focuses on the fishery characteristics of LMEs with reference to the major species groups caught in each LME and tries to identify similar patterns among the various LMEs. The ten years series of capture data available for each LME have been grouped on the basis of the ‘International Standard Statistical Classification for Aquatic Animals and Plants’ (ISSCAAP) which has been recently revised (FAO, 2001a,b) and a LME cluster analysis of the similarity of the average total catches of each group for the studied period. This has produced 11 clusters of LMEs which have similar characteristics in their capture profiles.

For each LME, stacked area charts of species groupings’ catches have been also prepared to show the variations along the 10-year period (see charts in Appendix 2) and the differences in trends between the various LMEs belonging to the same cluster have been discussed when appropriate.

2. METHODS

2.1 Re-arrangement of FAO capture statistics by LME and grouping of species items

A data sub-set from the FAO capture database (FAO, 2001c) was created including the 1990-99 catches for all non-oceanic species items. Only capture production of fishes, crustaceans and molluscs were considered, excluding catches of marine mammals, miscellaneous aquatic animals and products, and aquatic plants. Catches of freshwater and diadromous fishes reported as caught in marine waters (e.g. in the Baltic Sea) have also been included. A dataset comprising 867 species items was obtained. The total catches of these species items represent about 90% of the global marine catches as the oceanic species constitute the remaining 10% (see the “Oceanic” study in this volume). This figure is close to a previous estimate of LMEs producing approximately 95% of the world total marine capture production (Sherman, 1994).

In order to re-arrange FAO catch statistics data by LMEs, the following criteria were followed:

• catch data by country/FAO fishing area, including data for Distant Water Fleets (DWFs) when available, were extracted from the dataset and directly re-assigned to the corresponding LME whenever congruent;

• additional data for those LMEs whose boundaries are not coincident with those of the FAO fishing areas have been extracted from regional databases managed by FAO (e.g. GFCM for the Mediterranean and the Black Sea, CECAF for the Eastern Central Atlantic area and ex-ICSEAF for the Southeast Atlantic) and by regional bodies (e.g. NAFO for the Northwest Atlantic and ICES for the Northeast Atlantic) while additional data at the sub-national level had to be retrieved from national yearbooks of fishery statistics and national databases (see Appendix 1 for a complete list of additional sources consulted);

• catches by species have been assigned to LMEs also on the basis of their ranges;

• in the cases in which the total catches from an additional and more detailed national source differed from the total catches previously submitted by the country’s FAO national correspondent, the proportions by species and by sub-national area for each year as reported in the additional source data were applied to the figures included in the FAO database. In this way, the statistics by species items included in the FAO database could have been redistributed on a sub-national level basis and used for building the LME data sets.

The data obtained from different sources and harmonized to the FAO data were used to build 1990-1999 time series for 43 Large Marine Ecosystems. As stated above, for seven LMEs (i.e. 41, 42, 43, 44, 45, 46, and 47) of the Northwest Pacific area, this was not possible either by using the data from the FAO database or by obtaining additional detailed data either at the regional or national level, and therefore they have not been considered in the analysis.

Data by species items were subsequently aggregated into 12 groupings based on the ISSCAAP divisions and groups, as shown in Table 2.

Table 2. Groupings of species considered in the cluster analysis

* Not included in calculations for the cluster analysis

2.2 Cluster analysis

Catches by species groupings were summed up along the ten years period and their percentages in each LME calculated. A cluster analysis, aiming at identifying clusters of LMEs that present similarities in terms of catch composition by species groupings, was performed using the analytical method “partitioning around medoids” or pam, as in the statistical software S-Plus, 2000. The cluster analysis was based on eleven of the groups shown in Table 2 as the ISSCAAP group 39 (‘Marine fishes not identified’) was excluded from the calculations of the percentages used in the cluster analysis. Catches reported in this group may indeed include very different species in different LMEs. However, as the percentage of catches reported as ‘Marine fishes not identified’ is a good inverse indicator of the degree of breakdown by species in which catch statistics are reported from different countries/areas, the percentage of ‘Marine fishes not identified’ on total shelf catches for each LME is shown in each trend charts of Appendix 2.

The pam technique consists of several steps performed by the software, which accepts a matrix of data in which rows (n) are objects (individual LMEs) and columns (p) are variables (ISSCAAP based groupings of species). The algorithm pam computes k representative objects, called medoids, which together determine a clustering. Each object is then assigned to the cluster corresponding to the nearest medoid or, in other words, the function minimizes the sum of the dissimilarities of all objects to their nearest medoid. On the basis of these calculations, a silhouette value s(i) is calculated for each object (LME) as an indication of how well that object has been assigned to a cluster. The value s(i) lies between -1 and +1; objects with a silhouette value close to +1 are well classified, for values around 0 an object lies between two clusters, for values close to -1 objects are not well classified (S-Plus, 2000; for further information on the pam method, see Kaufman and Rousseeuw, 1990).

The outputs of this analysis are a cluster membership list of the LMEs and two types of graphs: a clusplot (Pison et al., 1999) and a silhouette plot (Rousseeuw, 1987). The clusplot is based on the reduction of the multivariate dimensions of the data by principal component analysis (PCA), which yields a first component which accounts for maximal variance, then a second component with maximal variance among all components perpendicular to the first and so on. The clusplot displays objects relative to the first and second principal components and all observations are represented by points in a plot in which the component 1 is plotted on the horizontal axis and component 2 on the vertical one. Around each cluster an ellipse is drawn. The distance between two clusters can be represented as a line connecting the cluster centers (Pison et al., 1999). The silhouette plot consists of a bar graph, in which each object is represented by a bar of length s(i), ranked in decreasing order and showing the objects as visually grouped in clusters. The average silhouette width of the plot (average of the s(i) over all objects in the data set) gives an indication of how well objects have been classified for that given number of clusters. As a rule of thumb, the average silhouette width should be around or higher than 0.25 in order to be able to affirm that a structure in the data has been found.

3. CLUSTERS OF LARGE MARINE ECOSYSTEMS

The clustering procedure was run for different numbers of clusters (from 9 to 13) to identify the number of clusters for which the highest average silhouette would be obtained. For 11 clusters, an average silhouette width of 0.23 was reached, slightly below the 0.25 reference value. For 12 and 13 clusters, the same average silhouette width was obtained, but with increasing number of clusters including single LMEs. Therefore, in the clustering by 12 and 13 groups almost half of the clusters would have been constituted by a single LME. Hence, the pam analysis grouping the 43 LMEs into 11 clusters was considered as the most statistically and ecologically relevant. Memberships of each cluster are listed in Table 3. The clusplot (Figure 1), as generated by the software, represents the LMEs as points included in an ellipse as an indication of cluster membership. The connecting lines representing the distance between clusters have been removed as the clusplot would have been illegible and because the distance between clusters is not relevant for this study.

Table 3. LMEs' cluster membership as results of the pam analysis for 11 clusters

Figure 1. Clusplot for the 11 clusters with individual LMEs identified

Figure 2 shows the silhouette plot of each LME within its cluster. As said above, the silhouette width value is an indicator of how well that object has been assigned to that cluster.

Figure 2. Silhouette plot by LME for 11 clusters

3.1 Discussion by cluster

In this section, each cluster is briefly described evidencing the common characteristics among the LMEs that have led to their classification into the same cluster. Large Marine Ecosystems assigned to the cluster are listed together with the relevant ocean, hemisphere and a general categorization of the climate. A bar chart for each cluster shows the catch percentages of species grouping of LMEs belonging to the same cluster. Charts representing the 1990-99 catch trends by species groupings of each LME are shown in Appendix 2. Information on primary productivity is derived from that produced by the SeaWiFS project (2002), on a model developed by Behrenfeld and Falkowski (1997), as it is presented in the Large Marine Ecosystems web site (NOAA, 2002).

3.1.1 Cluster 1

Figure 3. Cluster 1: catch percentages of species groupings

The first cluster comprised only one LME (Eastern Bering Sea). In Figure 3 is shown the catch percentage of each species grouping (listed in Table 2) for the 1990-99 period. Catches of Gadiformes (ISSCAAP group 32) are predominant in this LME; other groups of some importance are flatfishes, salmons (in group 1X-2X) and crustaceans.

This is an LME characterized by an extreme environment at high latitude, in which temperature, currents and seasonal oscillations influence the productivity. According to SeaWiFS global primary productivity estimates, this LME has been classified as a moderately high productivity ecosystem.

The ten year trend (see Figure 14 in Appendix 2) shows decreasing catches of all major species groups in recent years with the only exceptions being diadromous fishes and crustaceans.

3.1.2 Cluster 2

Figure 4. Cluster 2: catch percentages of species groupings

The second cluster, adjacent to the LME in cluster 1, is also ‘monotypic’. The Gulf of Alaska is a highly productive ecosystem (SeaWiFS data). It also presents a significant upwelling phenomenon linked to the presence of the counterclockwise gyre of the Alaska Current (NOAA, 2002).

The catch composition of this LME differs from all other LMEs in being characterized by a strong prevalence of the freshwater and diadromous group (Figure 4), this linked to the rich salmon fisheries. Recent researches (Brodeur et al., 1999) have hypothesized changes in the future production of salmons as a consequence of long term shifts in the plankton biomass in the last decades. However, recent catch trends are rather stable (see Figure 15).

3.1.3 Cluster 3

LME no.	LME name	Ocean	Hemisphere	Climate
LME 3	California Current	Pacific	Northern	Temperate
LME 7	Northeast U.S. Continental Shelf	Atlantic	Northern	Temperate
LME 8	Scotian Shelf	Atlantic	Northern	Temperate
LME 9	Newfoundland Shelf	Atlantic	Northern	Subarctic

Figure 5. Cluster 3: catch percentages of species groupings

This cluster groups four of the historically most productive LMEs of the northern hemisphere, three in the Northwest Atlantic and one in the Northeast Pacific. They are all classified as moderately high productivity ecosystems, with the exception of the Northeast U.S. Continental Shelf, which is considered as highly productive and is structurally more complex than the other three, with marked temperature and climate changes, river runoff, estuarine exchanges, tides and complex circulation regimes. For what concerns the California Current ecosystem, this is a transition ecosystem between subtropical and subarctic water masses with an upwelling coastal phenomenon (Bakun, 1993) that determines strong interannual oscillations of the productivity of the ecosystem and, consequently, of the catch levels of different species groups.

The catch composition of this cluster is quite diverse as four species groupings contribute, on average among the four LMEs, at least 10% of the total shelf catches (Figure 5). These groups are: clupeoids (35), Gadiformes (group 32), molluscs (5X) and crustaceans (4X). The trend charts (Figure 16) show the marked decreases of Gadiformes catches in the Atlantic LMEs in the early 1990s up to the cod collapse in 1993-94, while in the same years the Gadiformes catches (mainly of Merluccius products) in the California Current increased and have remained high since then. The LME 7 is characterized by molluscs’ catches (almost 50% of the total) while an increase of crustacean catches in the LMEs 8 and 9 can be noted in recent years although it is not clear if this is due to ecological or to economical reasons (Caddy and Garibaldi, 2000).

3.1.4 Cluster 4

Figure 6. Cluster 4: catch percentages of species groupings

After cluster 6, this is the second cluster for number of LMEs in the present analysis. It is composed by eight LMEs, which, although in different manners, are all enriched by high level of nutrients. This cluster can be subdivided into two main sub-groups: enclosed and semi-enclosed seas (Gulf of California, Baltic Sea and Black Sea), which are strongly influenced by human induced eutrophication, river runoff and/or by a lack of rapid exchange with the adjacent oceans (NOAA, 2002; Kullenberg, 1986; Caddy, 1993) and upwelling ecosystems (two in the Pacific ocean: Humboldt Current and Pacific Central American Coastal, and two in the Atlantic ocean: Canary Current and Guinea Current) that show important upwelling and other seasonal nutrient enrichments (Bernal et al., 1983; Bakun et al., 1999; Bas, 1993; Binet, 1983). The Gulf of Mexico, although it is partially isolated from the Atlantic Ocean and water enters into it from the Yucatan Channel and exits from the Straits of Florida creating the Loop Current which is associated to nutrients flow and upwelling phenomena (Lohrenz et al., 1999), can not be considered as a semi-enclosed sea. Furthermore, this large scale and complex LME is affected by such levels of enriching river runoff (especially from the Mississippi) that large hypoxic areas have been detected in the Gulf in recent years (see Rabalais et al., 1996).

All of these ecosystems are characterized by predominant catches of small-pelagic clupeoids (group 35) that represent over half of the total identified shelf catches in all LMEs (Figure 6). Catch trends (Figure 17), although referring to a limited number of years, show that ups and downs do not occur only in LMEs driven by upwelling regimes but that also enclosed and semi-enclosed LMEs have a high variability in catches.

3.1.5 Cluster 5

Figure 7. Cluster 5: catch percentages of species groupings

The ecosystems in this cluster are distinguished by a very high percentage of crustacean catches (grouping 4x; Figure 7). The second species group in terms of catches is clupeoids in the Southeast U.S. Continental Shelf, flatfishes in the West Greenland Shelf, non-oceanic tunas in the Agulhas Current, and molluscs in the Northern Australian Shelf and Great Barrier Reef. Catch trends in recent years are very diverse and it is difficult to find common elements (see Figure 18).

These ecosystems are characterized by a rather wide range of productivity levels, from low (West Greenland Shelf) and moderate (Southeast U.S. Continental Shelf and Agulhas Current) to moderately-high and high productivity (Great Barrier Reef and Northern Australian Shelf respectively) according to the SeaWiFS estimates. Geographically, with the exception of the West Greenland Shelf and, partially, of the Northern Australian Shelf, these LMEs all lay along the eastern margins of the continents. Nutrient enrichment and mixing are due to different factors: offshore upwelling regime, although not as intense as in the higher latitude regions, in the Southeast U.S. Continental Shelf (Yoder, 1991; NOAA, 2002); tidal effects in the Great Barrier Reef (Brodie, 1999; NOAA, 2002); changes in sea and air temperature in the West Greenland Shelf (Hovgard and Buch, 1990); current-associated in the Agulhas Current (Beckley, 1998); and tidal mixing, monsoons and tropical cyclones in the Northern Australian Shelf (Furnas, 2002).

3.1.6 Cluster 6

Figure 8. Cluster 6: catch percentages of species groupings

This is the cluster with the highest number of LMEs. These nine ecosystems are probably less characterized than others and for this reason they have been grouped together by the clustering routine. Geographically, this cluster groups all tropical ecosystems, with the sole exception of the North Sea, and it includes four out five of the Indian Ocean LMEs. The general greater marine biodiversity of tropical regions is so reflected in catch composition. The main distinguish feature is the high catch percentages for miscellaneous coastal fishes (group 33) and miscellaneous pelagic fishes (group 37). Secondly, catches of herrings, sardines and anchovies (group 35) and of crustaceans (4x) in the nine ecosystems exceed 10% on average. (Figure 8). Most of these ecosystems are characterized by fishing activities mainly concentrated, for different reasons, on the coastal areas and this explain the high percentages of miscellaneous coastal fish catches. Catch trends in the 1990-99 period (Figure 19) are quite diverse and it is difficult to identify a common pattern. However, for most of these ecosystems, with the only exception of the North Sea and the Sulu-Celebes Sea, statistics are reported with a poor species breakdown, as can be deducted by the high percentages of catches included in the “Marine fishes not identified” category (see texts in charts of Figure 19).

Primary production ranges from low (Insular Pacific-Hawaiian and Sulu-Celebes Sea) to high (North Sea, Northeast Brazilian Shelf and Arabian Sea) with the remaining LMEs classified as moderately or moderately-high (South China Sea) productive.

It should be noted that, according to the LMEs web site (NOAA, 2002), the Insular Pacific-Hawaiian LME does not include only the Hawaii, as usually shown in the maps representing the LMEs (e.g. map in Appendix 3), but it extends also to shelf areas of several other Pacific islands. Catch statistics have been considered accordingly. This region is dominated by the equatorial currents system (NOAA, 2002). Fishery production in the Insular Pacific-Hawaiian and Sulu-Celebes Sea LMEs is mostly concentrated in the coastal waters as the islands are usually surrounded by very narrow shelf areas.

The Northeast Brazil Shelf is characterized by high levels of nutrients in the inner part of the shelf (Medeiros et al., 1999). The North Sea includes one of the most diverse coastal regions of the world, with a great variety of habitats (NOAA, 2002). Three of the Indian Ocean ecosystems (Somali Coastal Current, Arabian Sea and Bay of Bengal) are influenced by monsoons. In the Somali Coastal Current and in the Arabian Sea, the Southwest Monsoon from May to October cause seasonal upwelling phenomena that are on the other hand lacking in the Bay of Bengal (information derived, respectively, from Bakun et al., 1998; NOAA, 2002; Dwivedi, 1993). In the Arabian Sea, about 65% of fish landings derive from artisanal fisheries and this would explain the prevalence of coastal species catches but it may also be influenced by the presence of low-oxygen water, which restricts productivity at depths of 200 m and more (Dwivedi and Choubey, 1998; NOAA, 2002). The elongated and narrow shape, semi-enclosed character and circulation patterns of the Red Sea protect the coast from storms and provide habitats for a large number of marine coastal species (Baars, et al., 1998). Different sub-systems within the ecosystem have been identified in the South China Sea (Pauly and Christensen, 1993).

3.1.7 Cluster 7

Figure 9. Cluster 7: catch percentages of species groupings

Also in this cluster, group 35 (clupeoids: herrings, sardines and anchovies) is the most important species group in shelf catches but, unlike for cluster 4, other groups (i.e. mostly coastal fishes but also crustaceans, molluscs and miscellaneous demersal fishes for the Indonesian Seas) also contribute significant capture production (Figure 9). Catch trends have been rather stable in recent years (Figure 20) with moderate increases in total shelf catches if comparing the last year (1999) respect to the first year (1990) of the considered period, with the exception of the Indonesian Seas where catches have been quite steadily increasing.

As for its catch composition, the Mediterranean Sea seems one of the most diverse and stable LME in terms of species groupings, their shares in total catches and trends. Its unusual biodiversity for a temperate sea is confirmed by the fact that the Mediterranean and Black Sea together cover only the 0.8% of the total surface of the oceans but represent about 5.5% of the total world marine fauna (Fredj et al., 1992).

According to the productivity classification by SeaWiFS, the four LMEs in this cluster are moderately-high (Indonesian Seas), moderately (Brazil Current) or low naturally productive ecosystems (Caribbean Sea and Mediterranean Sea) but the productivity of the last two LMEs is increased by nutrient input from rivers, estuaries and human induced activities. These LMEs have in common a composite structure of environmental conditions, with local areas of upwelling, wind-driven currents, high water temperatures at least in some periods of the year, nutrient inputs from rivers or human activities (see studies on the single LMEs: Richards and Bohnsack, 1990, for the Caribbean Sea; Bakun, 1993, for the Brazilian Current; Caddy, 1993, for the Mediterranean Sea; Zijlstra and Baars, 1990, for the Indonesia Seas).

3.1.8 Cluster 8

Figure 10. Cluster 8: catch percentages of species groupings

This single LME cluster includes the Patagonian Shelf, which is characterized by high catches of molluscs, mostly cephalopods, and Gadiformes (Figure 10). Cephalopod fisheries developed in the early 1980s by Distant Water Fleets but, since the early 1990s, also local fleets (i.e. Argentina and Uruguay) are actively targeting these species. Following a drop in 1998, cephalopod catches in this area are still increasing (Figure 21). Instead, catches of Gadiformes, mostly by local fleets, increased continuously since the 1970s but from mid-1990s are declining.

These fisheries take place in one of the most extensive continental shelf of the world. According to the SeaWiFS estimates of global primary productivity, the Patagonian shelf is an area of high productivity and it is influenced by intense western boundary currents and wind- and tide-driven upwelling (Bakun, 1993; NOAA, 2002).

3.1.9 Cluster 9

Figure 11. Cluster 9: catch percentages of species groupings

In this cluster, the six ecosystems have a temperate or subarctic climate and five of them belong to the same oceanic region, the Northeast Atlantic. With the exclusion of the New Zealand Shelf and the Celtic-Biscay Shelf, which are influenced also by warm currents, respectively the South Equatorial and the Gulf Currents, the other ecosystems are categorized as high latitude and extreme environments, in which temperature, currents, tides and seasonal oscillations affect productivity. The same division in two sub-groups applies also to data on primary productivity (SeaWiFS, (2002): the New Zealand Shelf and the Celtic-Biscay Shelf are considered highly productive ecosystems, the Iceland Shelf, the Barents Sea and the Faroe Plateau are moderately highly productive ecosystems, and the East Greenland Shelf is a low productivity ecosystem.

The marine environment of the New Zealand Shelf is very diverse and includes estuaries, mudflats, mangroves, seagrass and kelp beds, reefs, seamount communities and deep-sea trenches (NOAA, 2002). The Celtic-Biscay Shelf is characterized by strong interdependence of human impact and biological and climate cycles (Koutsikopoulos and Le Cann, 1996). The East Greenland and Iceland LMEs are both characterized by a seasonal ice cover and by marked fluctuations in salinity, temperature and phytoplankton, factors that can contribute to variations of annual catches of cod and small pelagics (Skjoldal et al., 1993). In the Barents Sea, the ice-coverage extends over one third to two thirds of the LME and it varies considerably during the year and inter-annually (NOAA, 2002). The shallow parts of the shelf in the Faroe Plateau are well mixed by extreme tidal currents and no stratification occurs during the summer (NOAA, 2002).

With regard to catch composition, these ecosystems have in common high percentages of miscellaneous pelagic fishes (group 37; Figure 11) which, for the North-East Atlantic areas, are mostly due to peak catches of capelin in 1992-93. In the LMEs 17, 19 and 48 these peaks have a ‘boom and bust’ profile and, in the latest years of the observed period, catches of capelin are markedly decreased (Figure 22). Another fish group that shows relevant catches in all ecosystems of this cluster is group 32 (cods, hakes, haddocks), with the sole exception of the East Greenland LME that has been affected by the cod collapse of the early 1990s. In the other three northernmost Atlantic LMEs, total catches of the whole gadiform group have been rather stable during the 10 years examined (see also, Jakupsstovu and Reinert, 1994; Jacobsen, 1997; Nakken, 1998). In the two temperate ecosystems (i.e. New Zealand and the Celtic-Biscay shelves), the second species group in terms of catches is, respectively, miscellaneous demersal fishes (group 34) and clupeoids (group 35).

3.1.10 Cluster 10

Figure 12. Cluster 10: catch percentages of species groupings

The three ecosystems in this cluster are all western boundary ecosystems. The Norwegian Shelf and the Benguela Current are characterized by a high productivity according to the SeaWiFS classification, whereas the Iberian Coastal LME is considered as moderately productive. The catch composition pattern is dominated by three groups: herrings, sardines and anchovies (group 35), miscellaneous pelagic fishes (group 37) and cods, hakes and haddocks (group 32; Figure 12). Catches of Gadiformes are however very significant, and important for their value, only in the Norwegian Shelf and Benguela Current areas.

The Norwegian Shelf LME has a complex fishery history with concomitant influences of ecological anomalies, high fishery mortality and early implementation of management measures (Blindheim and Skjoldal, 1993; NOAA, 2002). Its high productivity is probably to be linked to the nutrient rich, cold arctic waters that characterize this LME (Furnes and Sundby, 1980). Since the early 1990s there has been a significant increase in Clupea harengus catches (Figure 23) which stock recovered after two decades of very low abundance.

The Iberian Coastal LME’s productivity is climate and upwelling driven. It is characterized by favorable factors for the production of clupeoids and other small pelagic fishes (Wyatt and Perez-Gandaras, 1989). Trends in catches by species groupings have been quite steady in recent years (Figure 23).

In the Benguela Current LME is one of the most strongly wind-driven coastal upwelling systems known and it presents favorable conditions for a rich production of small pelagics of groups 35 and 37 (Bakun, 1993). Harvests are characterized by stock fluctuations according to the variations in the primary and secondary level productivity.

3.1.11 Cluster 11

Figure 13. Cluster 11: catch percentages of species groupings

This single ecosystem cluster includes the Antarctic LME, which is unique both for its geographic and climatic characteristics. It is classified as a low productivity ecosystem, according to the SeaWiFS data, a consequence of the extensive seasonal ice cover and extreme weather conditions. The ecological and biological characteristics of Antarctic marine species are also unique from a food-chain point of view in that it is peculiarly short and based almost entirely on krill, a key species crucial to the sustainability and production of all other fisheries (Chopra and Hansen, 1997).

The Antarctic species most significant for fisheries have been considered as oceanic, either epipelagic or deep-water, and their catch trends are discussed in the “Oceanic” study of this volume. As for catches of shelf species, this LME exhibits a prevalence of miscellaneous demersal catches (group 34) and a much smaller percentage of coastal fishes (group 33; Figure 13), although fitting the Antarctic fishes into the categories of the three miscellaneous groups (i.e. coastal, demersal and pelagic) proved to be rather difficult (FAO, 2001b). Catches of shelf species have been remarkably reduced in the early 1990s (Figure 24).

4. CONCLUSION

The general analysis of cluster composition, common characteristics and catch trends (see Appendix 2) of LMEs in the same cluster presented some unexpected analogies between ecosystems of different marine regions and confirmed similarities between areas in which well known ecological phenomena take place (e.g. upwelling regimes).

As expected, ecosystems with extreme characteristics (i.e. northernmost Pacific and Antarctic LMEs) have peculiar catch patterns and, not presenting similarities with other LMEs, have been included in single clusters. Another cluster that includes only a single LME (Patagonian Shelf), is characterized by predominant catches of cephalopods and Gadiformes.

Three clusters (i.e. 4, 7 and 10) are dominated by catches of clupeoids, but some differences between the three groups of LMEs can be noted. The large marine ecosystems in cluster 4 are highly productive, enriched by nutrients as they are either semi-enclosed seas or have upwelling regimes, with clupeoids representing about 50-70% of the catches in their shelf areas (excluding catches reported as “Marine fishes not identified”). Also LMEs in cluster 10 are highly productive and, in addition to clupeoids, they are characterized by catches of Gadiformes and non-clupeoid small pelagics. In contrast, LMEs in cluster 7 have moderate or low productivity and theirs catch composition is more diverse with several other groups (i.e. coastal fishes, crustaceans, molluscs and miscellaneous demersal fishes) represented by significant catches.

An unexpected finding was a cluster of five ecosystems where the majority (between 30 and 65%) of identified catches on the continental shelf are made of crustacean species. This seems to be the only common feature amongst the LMEs of cluster 5, which are quite diverse in their productivity, climate, and second ranking species group in terms of catches. The remaining clusters are characterized by catches distributed quite evenly amongst the major groups of species (i.e. clusters 3 and 6) or with a slight predominance of miscellaneous pelagic fishes (cluster 9).

However, given the global coverage and the limitations in data availability, this study only aimed at providing basic information on catch composition by LME for future studies on single LMEs and some possible starting points for more in-depth ecologically oriented researches on fishery trends.

5. REFERENCES

Anonymous. 1998. An Ecosytem Strategy for the Assessment and Management of International Coastal Ocean waters. Leaflet issued by IUCN, NOAA and IOC. 8 p.

Anonymous. 2001. Reykjavik Declaration on Responsible Fisheries in the Marine Ecosystem. On-line at: http://www.refisheries2001.org/index.htm. Viewed 06/11/02.

Anonymous. 2002a. World Summit on Sustainable Development - Plan of Implementation. On-line at:

http://www.johannesburgsummit.org/html/documents/summit_docs/plan_final1009.doc. Viewed 13/12/02.

Anonymous. 2002b. Oceans and the World Summit on Sustainable Development. A Large Marine Ecosystems Strategy for the assessment and management of international coastal waters. Leaflet issued by IUCN, NOOA and IOC, 8 p.

Baars, M.A., P.H. Schalk, and M.J.W. Veldhuis. 1998. “Seasonal fluctuations in plankton biomass and productivity in the ecosystems of the Somali Current, Gulf of Aden, and Southern Red Sea”. In: K. Sherman, E.N. Okemwa, and M.J. Ntiba (eds.), Large Marine Ecosystems of the Indian Ocean: assessment, sustainability, and management. 143-174 p. Blackwell Science, Inc., Malden (Mass.), USA.

Bakun, A. 1993. “The California Current, Benguela current, and Southwestern Atlantic Shelf ecosystems: a comparative approach to identifying factors regulating biomass yields”. In: K. Sherman, L.M. Alexander, B.D. Gold (eds.). Large Marine Ecosystems: Stress, Mitigation, and Sustainability. 143-174 p. AAAS Press. Washington DC, USA.

Bakun, A., C. Roy, and S. Lluch-Cota. 1998. “Coastal Upwelling and other processes regulating ecosystem productivity and fish production in the Western Indian Ocean”. In: K. Sherman, E.N. Okemwa, and M.J. Ntiba (eds.), Large Marine Ecosystems of the Indian Ocean: assessment, sustainability, and management. 103-141 p. Blackwell Science, Inc., Malden (Mass.), USA.

Bakun, A., J. Csirke, D. Lluch-Belda, and R. Steer-Ruiz. 1999. “The Pacific Coastal LME”. In: K. Sherman, K. and Q. Tang (eds.). Large marine ecosystems of the Pacific Rim: assessment, sustainability, and management. 268-280 p. Blackwell Science, Inc., Malden (Mass.), USA.

Bas, C. 1993. Long-term variability in the food chains, biomass yields, and oceanography of the Canary Current ecosystem. In: K. Sherman, L.M. Alexander, B.D. Gold (eds.). Large Marine Ecosystems: Stress, Mitigation, and Sustainability. 94-103 p. AAAS Press. Washington DC, USA.

Beckley, L.E. 1998. “The Agulhas Current Ecosystem with particular reference to dispersal of fish larvae”. In: K. Sherman, E.N. Okemwa, and M.J. Ntiba (eds.), Large Marine Ecosystems of the Indian Ocean: assessment, sustainability, and management. 255-276 p. Blackwell Science, Inc., Malden (Mass.), USA.

Behrenfeld, M. and P.G. Falkowski. 1997. “Photosynthetic rates derived from satellite-based chlorophyll concentration”. Limnol. Oceanogr. 42 (1): 1-20.

Bernal, P.A., F.L. Robles, and O. Rojas. 1983. “Variabilidad física y biológica en la región meridional del sistema de corrientes Chile-Perú”. In: G.D. Sharp and J. Csirke (eds.) Proceedings of the Expert Consultation to examine changes in abundance and species of neritic fishe resources. San José, Costa Rica, 28-29 April 1983. FAO Fisheries Report. No. 291, Vol. 3, 683-711 p.

Binet, D. 1983. “Phytoplancton et production primaire des regions côtiérs à upwellings saissoniers dans le Golfe de Guinée”. Océanogr. Trop. 18: 331-355.

Blindheim, J. and H.R. Skjoldal. 1993. “Effects of climate changes on the biomass yield of the Barents Sea, Norwegian Sea, and West Greenland large marine ecosystems”. In: K. Sherman, L.M. Alexander, B.D. Gold (eds.). Large Marine Ecosystems: Stress, Mitigation, and Sustainability. 185-198 p. AAAS Press. Washington DC, USA.

Brodeur, R.D., B.W. Frost, S.R. Hare, R.C. Francis, and W.J. Jr. Ingraham. 1999. “Interannual variations in zooplankton biomass in the Gulf of Alaska, and covariation with California Current zooplankton biomass”. In: K. Sherman, K. and Q. Tang (eds.). Large marine ecosystems of the Pacific Rim: assessment, sustainability, and management. 106-138 p. Blackwell Science, Inc., Malden (Mass.), USA.

Brodie, J. 1999. “Management of the Great Barrier Reef as a Large Marine Ecosystem”. In: K. Sherman, K. and Q. Tang (eds.). Large marine ecosystems of the Pacific Rim: assessment, sustainability, and management. 428-437 p. Blackwell Science, Inc., Malden (Mass.), USA.

Caddy, J.F. 1993. “Contrast between recent fishery trends and evidence for nutrient enrichment in two large marine ecosystems: the Mediterranean and the Black Seas”. In: K. Sherman, L.M. Alexander, B.D. Gold (eds.). Large Marine Ecosystems: Stress, Mitigation, and Sustainability. 137-147 p. AAAS Press. Washington DC, USA.

Caddy, J.F. and L. Garibaldi. 2000. “Apparent changes in the trophic composition of world marine harvests: the perspective from the FAO capture database”. Ocean and Coastal Management. 43 (8-9): 615-655.

Chopra, S. and C. Hansen. 1997. “Deep ecology and the Antarctic marine living resources: lessons for other regimes”. Ocean and Coastal Law Journal. 3(1-2): 117-148.

Dwivedi, S.N. 1993. “Long term variability in the food chains, biomass yield, and oceanography of the Bay of Bengal ecosystem”. In: K. Sherman, L.M. Alexander, B.D. Gold (eds.). Large Marine Ecosystems: Stress, Mitigation, and Sustainability. 43-52 p. AAAS Press. Washington DC, USA.

Dwivedi, S.N. and A.K. Choubey. 1998. “Indian Ocean large marine ecosystems: Need for national and regional framework for conservation and sustainable development”. In: K. Sherman, E.N. Okemwa, and M.J. Ntiba (eds.), Large Marine Ecosystems of the Indian Ocean: assessment, sustainability, and management. 327-333 p. Blackwell Science, Inc., Malden (Mass.), USA.

FAO (Food and Agriculture Organization of the United Nations). 2001a. “Report of the nineteenth session of the Coordinating Working Party on Fishery Statistics”. Nouméa, New Caledonia, 10-13 July 2001. FAO Fisheries Report. No. 656. 91 p.

FAO (Food and Agriculture Organization of the United Nations). 2001b. “Proposal for a revision of the ISSCAAP groups of the Marine Fishes division”. In: FAO Agency Report to the Coordinating Working Party on Fishery Statistics. Nouméa, New Caledonia, 10-13 July 2001, 42-49 p. On-line at ftp://ftp.fao.org/fi/document/cwp/cwp_19/CWP-19-fao.pdf .

FAO (Food and Agriculture Organization of the United Nations). 2001c. Yearbook of Fishery Statistics - Capture Production 1999. Vol. 88/1. FAO, Rome, Italy. Downloadable at: http://www.fao.org/fi/statist/FISOFT/FISHPLUS.asp

Fredj, G., D. Bellan-Santin and M. Meinardi. 1992. “Etat des connaissances sur la faune marine méditerranéenne”. In: D. Bellan (ed.), Spéciation et biogéographie en mer Méditerranée, Bull. Inst. Océanogr. Monaco, no. sp. 9, 133-145 p.

Furnas, M.J. 2002. Land-sea interactions and oceanographic processes affecting the nutrient dynamics and productivity of Australian marine ecosystems. On-line at: http://www.ea.gov.au/coasts/publications/somer/annex1/land-sea.html#HDR7 . Viewed 25/07/02.

Furnes, G.K. and S. Sundby. 1980. “Upwelling and wind induced circulation in Vestfjorden”. In: R. Saetre and M. Mork (eds.), Proceedings from Norwegian Coastal Current Symposium. University of Bergen, Geilo, Norway, 9-12 September 1980.

Hovgard, H. and E. Buch, E. 1990. “Fluctuation in the cod biomass of the West Greenland Sea Ecosystem in relation to climate”. In: K. Sherman, L.M. Alexander, and B.D. Gold (eds.). Large Marine Ecosystems: Patterns, Processes, and Yields. 36-43 p. AAAS Publications Washington DC,.USA.

Jacobsen, O. 1997. “Skilled fishers, bungling economy”. SAMUDRA Report. No. 18, 20-25p.

Jakupsstovu, S.H. and J. Reinert. 1994. “Fluctuations in the Faroe Plateau stock”. In: J. Jakobsson, O.S. Astthorsson, R.J.H. Beverton, B. Bjoernsson, N. Daan, K.T.Frank, J. Meincke, B. Rothschild, S. Sundby, S. Tilseth (eds.). Cod and climate change. Proceedings of a symposium held in Reykjavik, 23-27 August 1993. ICES Marine Science Symposia. Vol. 198, 194-211 p.

Kaufman, L. and P.J. Rousseeuw. 1990. Finding groups in data: An introduction to Cluster Analysis. Wiley, New York.

Koutsikopoulos, C. and B. Le Cann. 1996. “Physical processes and hydrological structures related to the Bay of Biscay anchovy”. Scientia Marina, Vol. 60, suppl. 2, 9-19 p.

Kullenberg, G. 1986. “Long-term changes in the Baltic Ecosystem”. In: K. Sherman and L.M. Alexander (eds.). Variability and Management of Large Marine Ecosystems, AAAS Selected Symposium 99, Westview Press, Boulder, CO, USA.

Kumpf, H., K. Steidinger, and K. Sherman. 1999. The Gulf of Mexico Large Marine Ecosystem: Assessment, Sustainability, and Management. Blackwell Science, Inc., Malden (Mass.), USA. 736 p.

Lohrenz, S.E., D.A. Wiesenburg, R.A. Arnone, and X. Chen. 1999. “What Controls Primary Production in the Gulf of Mexico?” In: H. Kumpf, K. Steidinger, and K. Sherman. 1999. The Gulf of Mexico Large Marine Ecosystem: Assessment, Sustainability, and Management. Blackwell Science, Inc., Malden (Mass.), USA.

Medeiros, C., S.J. Macedos, F.A.N. Feitosa, and M.L. Koenig. 1999. “Hydrography and phytoplankton biomass and abundance of north-east Brazilian waters”. Arch. Fish. Mar. Res./Arch. Fisch. Meeresforsch. Vol. 47, no. 2-3, 133-151 p.

Nakken, O. 1998. “Past, present and future exploitation and management of marine resources in the Barents Sea and adjacent areas”. Fisheries Research. Vol. 37, No. 1-3, 23-25 p.

NOAA (National Oceanic and Atmospheric Administration). 2002. Large marine ecosystems of the world. On line at: http://www.edc.uri.edu/lme/default.htm . Viewed 18/04/2002.

Pauly, D. and V. Christensen. 1993. “Stratified models of Large Marine Ecosystems: a general approach and an application to the South China Sea.”. In: K. Sherman, L.M. Alexander, B.D. Gold (eds.). Large Marine Ecosystems: Stress, Mitigation, and Sustainability. 148-174 p. AAAS Press. Washington DC, USA.

Pison, G., A. Struyf, and P.J. Rousseeuw. 1999. “Displaying a clustering with CLUSPLOT”. Computational statistics & data analysis, 30(4): 381-392.

Rabalais, N.N., W.J. Jr. Wiseman, R.E. Turner, D. Justic, B.K. Sen-Gupta and Q. Dortch. 1996. “Nutrient changes in the Mississippi River and system responses on the adjacent continental shelf”. Estuaries. Vol. 19 (2B): 386-407.

Richards, W.J. and J.A. Bohnsack. 1990. “The Caribbean Sea: a Large Marine Ecosystem in crisis”. In: K. Sherman, L.M. Alexander, and B.D. Gold (eds.). Large Marine Ecosystems: Patterns, Processes, and Yields. 44-53 p. AAAS Publications Washington DC,.USA.

Rousseeuw, P.J. 1987. “Silhouettes: A graphical aid to the interpretation and validation of cluster analysis”. J. Comput. Appl. Math. 20: 53-65.

SeaWiFS (Sea-viewing Wide Field-of-view Sensor) Project. 2002. On-line at: http://seawifs.gsfc.nasa.gov/SEAWIFS.html .

Sherman, K. 1994. “Sustainability, biomass yields, and health of coastal ecosystems: an ecological perspective”. Mar. Ecol. Prog. Ser. 112: 277-301.

Sherman, K. and L.M. Alexander (eds.). 1986. Variability and Management of Large Marine Ecosystems. AAAS Selected Symposium 99, Westview Press, Boulder, CO, USA. 319 p.

Sherman, K. and L.M. Alexander (eds.). 1989. Biomass Yields and Geography of Large Marine Ecosystems. AAAS Selected Symposium 111. Westview Press, Boulder, CO, USA. 493 p.

Sherman, K. and A.M. Duda. 1999. “Large marine ecosystems: an emerging paradigm for fishery sustainability”. Fisheries. 24(12): 15-26.

Sherman, K. and Q. Tang (eds). 1999. Large Marine Ecosystems of the Pacific Rim: Assessment, Sustainability, and Management. Blackwell Science, Inc., Malden (Mass.), USA. 465 p.

Sherman, K., L.M. Alexander, and B.D. Gold (eds.). 1990 (2nd printing 1992). Large Marine Ecosystems: Patterns, Processes, and Yields. AAAS Publications Washington DC, USA. 242 p.

Sherman, K., L.M. Alexander, and B.D. Gold (eds.). 1991. Food Chains, Yields, Models, and Management of Large Marine Ecosystems. Westview Press, Inc. Boulder, CO. 320 p.

Sherman, K., L.M. Alexander, B.D. Gold (eds.). 1993. Large Marine Ecosystems: Stress, Mitigation, and Sustainability. AAAS Press. Washington DC, USA. 376 p.

Sherman, K., N.A. Jaworski, and T.J. Smayda (eds). 1996. The Northeast Shelf Ecosystem: Assessment, Sustainability, and Management. Blackwell Science, Inc., Malden (Mass.), USA. 564 p.

Sherman, K., E.N. Okemwa, and M.J. Ntiba (eds). 1998. Large Marine Ecosystems of the Indian Ocean: Assessment, Sustainability, and Management. Blackwell Science, Inc., Malden (Mass.), USA. 394 p.

Skjoldal, H.R., T.T. Noji, J. Giske, J.H. Fossaa, J. Blindheim, and S. Sundby, S. 1993. Mare cognitum. Science plan for research on marine ecology of the Nordic Seas (Greenland, Norwegian, Iceland Seas) 1993-2000. A regional GLOBEC program with contributions also to WOCE and JGOFS. Inst. Mar. Res., Bergen, Norway, 162 p.

S-PLUS. 2000. Professional Edition for Windows, Release 3 (August 2000). MathSoft, Inc., Seattle, Washington.

Wyatt, T. and G. Perez-Gandaras. 1989. “Biomass Changes in the Iberian Ecosystem”. In: K. Sherman and L.M. Alexander (eds.). Biomass Yields and Geography of Large Marine Ecosystems. AAAS Selected Symposium 111. Westview Press, Boulder, CO, USA.

Yoder, J.A. 1991. “Warm-temperate food chains of the Southeast Shelf ecosystem”. In: K. Sherman, L.M. Alexander, and B.D. Gold (eds.). Food Chains, Yields, Models, and Management of Large Marine Ecosystems. 49-66 p. Westview Press, Inc. Boulder, CO, USA.

Zijlstra, J.J. and M.A. Baars. 1990. “Productivity and fisheries potential of the Banda Sea ecosystem”. In: K. Sherman, L.M. Alexander, and B.D. Gold (eds.). Large Marine Ecosystems: Patterns, Processes, and Yields. 54-65 p. AAAS Publications Washington DC,.USA.

APPENDIX 1. - Additional sources

Table 4 lists the sources from which additional capture statistics have been extracted to complement the FAO capture database in building on the LME data series. The seven LMEs (41 to 47) for which, given the unavailability of sub-national data, it was not possible to prepare data series by LME are excluded. The LMEs without any additional sources are those congruent with the FAO fishing areas and to which FAO statistics were assigned directly.

Table 4. Additional data sources used to complement the FAO database

APPENDIX 2. - Capture trends (1990-1999) of each LME by cluster

Figure 14. Cluster 1: capture trends of LME 1

Figure 15. Cluster 2: capture trends of LME 2

Figure 16. Cluster 3: capture trends of LMEs 3-7-8-9

Figure 17: Cluster 4: capture trends of LMEs 4-5-13-22-26-27-28-50

Figure 18: Cluster 5: capture trends of LMEs 6-10-30-38-39

Figure 19: Cluster 6: capture trends of LMEs 11-16-21-31-32-33-34-35-36

Figure 20: Cluster 7: capture trends of LMEs 12-15-25-37

Figure 21: Cluster 8: capture trends of LME 14

Figure 22: Cluster 9: capture trends of LMEs 17-18-19-23-40-48

Figure 23: Cluster 10: capture trends of LMEs 20-24-29

Figure 24: Cluster 11: capture trends of LME 49

APPENDIX 3. - Map of the 50 LMEs

Map of the 50 LMEs (modified from Anonymous, 1998).