Taxonomic distinctness as a measure of diversity applied over a large scale: the benthos of the Norwegian continental shelf

Authors

  • K. E. ELLINGSEN,

    1. Marine Biodiversity, Department of Biology, University of Oslo, PO Box 1066 Blindern, 0316 Oslo, Norway;and Plymouth Marine Laboratory, Prospect Place, West Hoe, Plymouth, Devon PL1 3DH, UK
    Search for more papers by this author
  • K. R. CLARKE,

    1. Marine Biodiversity, Department of Biology, University of Oslo, PO Box 1066 Blindern, 0316 Oslo, Norway;and Plymouth Marine Laboratory, Prospect Place, West Hoe, Plymouth, Devon PL1 3DH, UK
    Search for more papers by this author
  • P. J. SOMERFIELD,

    1. Marine Biodiversity, Department of Biology, University of Oslo, PO Box 1066 Blindern, 0316 Oslo, Norway;and Plymouth Marine Laboratory, Prospect Place, West Hoe, Plymouth, Devon PL1 3DH, UK
    Search for more papers by this author
  • R. M. WARWICK

    1. Marine Biodiversity, Department of Biology, University of Oslo, PO Box 1066 Blindern, 0316 Oslo, Norway;and Plymouth Marine Laboratory, Prospect Place, West Hoe, Plymouth, Devon PL1 3DH, UK
    Search for more papers by this author

Kari Elsa Ellingsen, Marine Biodiversity, Department of Biology, University of Oslo, PO Box 1066 Blindern, 0316 Oslo, Norway. Fax: +47 22854726; E-mail: k.e.ellingsen@bio.uio.no

Summary

  • 1Data on soft-sediment macrobenthos from the Norwegian continental shelf (56–71°N) was used to examine the use of average taxonomic distinctness (Δ+, a measure of the average degree to which species are related to each other) as a diversity measure.
  • 2Δ+ for all taxa combined decreased with both latitude and depth. In contrast Δ+ for annelids and crustaceans (71% of the fauna) was positively correlated to latitude and depth, whereas molluscs (20·1% of the fauna) were not related to latitude or depth. Depth explained most of the variance in Δ+ for annelids and crustaceans. Neither latitude nor grain size appeared to be important in explaining large-scale patterns in the taxonomic breadth of the three dominant phyla. In contrast latitude rather than depth explained most of the variance in Δ+ of all taxa combined.
  • 3The finding that molluscs showed different patterns of Δ+ than the two other dominant phyla shows that the biodiversity of different phyla responded differently to environmental gradients and that one phylum cannot be used as a surrogate for others in such studies. The most likely explanation for the inconsistency in Δ+ is because the taxonomic classification system also varies between different phyla, and taxonomically related biodiversity measures might be more meaningfully applied to a single phylum than to all taxa combined.
  • 4Δ+ and species richness (S) displayed different patterns of biodiversity. Thus Δ+ is not a surrogate for species richness. Species richness of all taxa combined and annelids were highest in the middle depth range, and lowest in the shallowest and deepest parts of the shelf. Most of the variance in species richness is accounted for by depth, followed by latitude and grain size for all three dominant phyla and all taxa combined.

Introduction

Because of the impracticality of routinely attempting comprehensive surveys of all living species from the whole spectrum of habitat types, ‘surrogacy’ methods will clearly become the norm in biodiversity estimation (Harper & Hawksworth 1994). One important approach to biodiversity estimation is extrapolation from taxon to taxon, focal group to inclusive group, site to site and sample to inventory, across spatial scales (Hammond 1994; Ward et al. 1998). Several surrogates of this kind have been proposed (Féral et al. 2003). These might take the form of a subset of easily censused species from the total biota, but the crucial issue then concerns how valid a measure this surrogate provides of the true total biodiversity, and how the biodiversity of one component of the biota reflects that of others. Some limited success has been achieved in extrapolating biodiversity from one taxon to another (e.g. Mikusinski, Gromadzki & Chylarecki 2001; Ricketts, Daily & Ehrlich 2002) and from a single taxon to a wider range of taxa (e.g. Sahlen & Ekestubbe 2001; Olsgard, Brattegard & Holthe 2003). However, on the Norwegian continental shelf, patterns of benthic biodiversity have been shown to vary between dominant taxonomic groups (Ellingsen 2001, 2002; Ellingsen & Gray 2002).

Most studies of biodiversity in both marine and terrestrial systems deal with species richness patterns. However, the biological diversity of an area is more than the numbers of species present (Harper & Hawksworth 1994). A variety of diversity indices combine the species richness and how evenly the individuals are distributed among the species (see Magurran 2004 for an overview). However, assemblages with the same number of species may comprise species that are closely related to each other taxonomically, or they may be more distantly related.

As an alternative approach to the measurement of biodiversity and for conservation purposes Vane-Wright, Humphries & Williams (1991) proposed to use an index of taxonomic distinctiveness of species based on cladistic classification (see also May 1990). Later,Warwick & Clarke (1995) introduced two new measures of biodiversity, taxonomic diversity (Δ) and taxonomic distinctness (Δ*) based on quantitative data, and subsequently Clarke & Warwick (1998) defined average taxonomic distinctness (Δ+) for presence/absence data. Clarke & Warwick (2001) also proposed a further index, variation in taxonomic distinctness (Λ+). All these measures differ from more conventional diversity indices by incorporating the degree to which the species are taxonomically related to each other. As noted by Warwick & Clarke (1995), fully resolved cladograms are not available for most groups of organisms, and taxonomically based biodiversity measures have therefore been based on the somewhat arbitrary Linnaean classification system.

In marine soft-bottom studies multivariate methods have proven much more sensitive to small changes in faunal composition than species richness and diversity indices (e.g. Gray et al. 1990; Warwick & Clarke 1991). Though not possessing the sensitivity of multivariate methods, indices based on the taxonomic relatedness of species may give additional information to other univariate measures of biodiversity. Benthic communities that have been perturbed are generally kept in an early successional stage with relatively few species, many of which are closely related. In contrast, less perturbed benthic communities in a late successional stage tend to comprise a range of different species belonging to many phyla (Warwick & Clarke 1995). Species richness and average taxonomic distinctness are not conceptually (or mechanistically) related, and there is no reason to suppose that a pair of species, drawn randomly from an undisturbed assemblage with naturally low species richness, is less taxonomically distinct than a pair of species drawn randomly from a naturally rich community.

The majority of marine studies in which indices of taxonomic distinctness have been applied have focused on patterns of biodiversity in relation to anthropogenic perturbations (e.g. Warwick & Clarke 1995, 1998; Hall & Greenstreet 1998; Brown, Clarke & Warwick 2002). Most studies of diversity gradients have been done on small spatial scales, that of alpha diversity, but in recent years there has been an upsurge of interest in large-scale patterns of biodiversity in marine systems. Particularly, the importance of studies of diversity at different spatial scales and patterns of beta diversity have been emphasized (e.g. Clarke & Lidgard 2000; Izsak & Price 2001; Ellingsen & Gray 2002), but yet few published studies of large-scale patterns of taxonomic distinctness are available (but see Price, Keeling & O’Callaghan 1999; Woodd-Walker, Ward & Clarke 2002).

Whether taxonomic distinctness is related to environmental variables such as, for example, sediment type and water depth has rarely been tested. Using a data set on free-living marine nematodes from the coasts of Britain and Chile, Warwick & Clarke (1998) examined the influence of habitat type and diversity on average taxonomic distinctness (Δ+). They concluded that species richness measures of biodiversity are much more strongly affected by habitat type and complexity than Δ+. However, they argued that further studies are necessary before any generalization could be attached to their findings.

Previously, Ellingsen & Gray (2002) studied patterns of biodiversity on the Norwegian continental shelf, focusing on alpha, beta and gamma diversity as well as rarity. Here we extend these analyses using information on how closely the organisms are related to each other. In the present paper we concentrate on Δ+ based on presence/absence of species rather than quantitative data, so that comparisons can be made with data where sampling is less carefully controlled. The objectives of this paper were to: (1) examine whether average taxonomic distinctness is related to latitude, depth and/or sediment type; (2) determine whether one phylum can act as a surrogate for the total fauna or any other phyla; and (3) examine the relationship between average taxonomic distinctness and species richness.

Materials and methods

We utilized a data set of soft-sediment macrobenthos from 101 sites occurring in five large areas along a transect of 15° of latitude on the Norwegian continental shelf (from Ellingsen & Gray 2002). Area 1 was the southernmost area and area 5 the northernmost (Fig. 1). The samples were taken with a 0·1-m2 van Veen grab. At each site five grabs for analyses of macrobenthos were taken, and the samples were sieved on a 1-mm sieve. Water depth ranged from 65 to 434 m, and there was considerable variation in sediment characteristics (silt-clay content: 0–99%; median grain size Mdϕ: 0·67–10·79; total organic matter: 0·5–11·1%, Table 1). Further details of sampling and analyses are given in Ellingsen & Gray (2002).

Figure 1.

Geographic positions of the 101 sampling sites in five areas on the Norwegian continental shelf (from Ellingsen & Gray 2002).

Table 1.  Summary of depth and sediment characteristics along the Norwegian continental shelf. Mdϕ: median grain size; Silt-clay: fraction of sediment < 0·063 mm (%); TOM: total organic matter (%) (from Ellingsen & Gray 2002)
AreaNo. of sitesDepth (m)MdϕSilt-clay (%)TOM (%)
1 16 65–743·48–3·59 2·5–6·00·8–1·2
2 21 71–1252·30–3·90 0·8–13·80·6–2·1
3 25 93–3561·67–10·79 0·0–99·00·5–11·1
4 19212–4342·97–6·4224·7–96·02·9–10·4
5 20160–3650·67–5·83 5·9–92·41·3–10·2
Total101 65–4340·67–10·79 0·0–99·00·5–11·1

Analyses were done on data pooled over five grabs from each site. In addition to species richness (S) we used average taxonomic distinctness (Δ+), a measure of the average degree to which species in an assemblage are related to each other. Average taxonomic distinctness is defined as: Δ+ = [Σ Σi<j ωij]/[s(s − 1)/2], where s is the number of species present, the double summation is over {i = 1, …s; j = 1, … s, such that i < j}, and ωij is the ‘distinctness weight’ given to the path length linking species i and j in the hierarchical classification (Clarke & Warwick 1998).

The taxonomic levels used in this study are species, genus, family, order, class and phylum, according to the classification described by Howson (1987). Values of Δ+ are based on equal step lengths between these six taxonomic levels. Thus the weighting between taxonomic levels for different species in the same genus is ω = 16·67, for species in different genera, but the same family ω = 33·33, for species in different families, but the same order ω = 50, etc., and ω = 100 for species connected at the highest (taxonomically coarsest) level. The above analyses were undertaken using PRIMER version 5 (Clarke & Gorley 2001). Multiple regressions were performed using Minitab.

Results

taxonomic levels

A total of 807 species from 101 sites was divided among 508 genera, 239 families, 80 orders, 33 classes and 16 phyla. Annelids displayed the highest number of species on the shelf (344 species), followed by crustaceans (229 species) and molluscs (162 species, Table 2). Likewise, annelids dominated generic richness, but molluscs and crustaceans had a higher number of families (74 and 71, respectively) than annelids (48, Table 3). Molluscs had representatives of 23 orders compared with 15 of annelids and only 11 of crustaceans, and molluscs also had the highest number of classes (six, Table 3). The number of species, genera, families, orders and classes were highest in area 3, the southern-central area, and lowest in area 1, the southernmost area, for all taxa combined (Table 3). Annelids and molluscs showed a similar pattern, but crustaceans had highest numbers of all these taxonomic levels in area 5, the northernmost area (Table 3).

Table 2.  The number of species of each phylum and for all phyla combined in the areas and on the total shelf
PhylumArea 1Area 2Area 3Area 4Area 5Total shelf
Annelida 78142226140197344
Crustacea 42 73105 72112229
Mollusca 37 58 98 57 67162
Echinodermata  8 15 22 13 14 36
Sipuncula  2  3  7  4  7  9
Cnidaria  3  4  4  2  3  6
Tunicata  1  3  5  2  1  5
Brachiopoda  0  1  2  3  0  3
Chelicerata  0  1  2  0  1  3
Hemichordata  1  1  2  0  0  2
Phoronida  1  2  1  1  1  2
Priapulida  1  0  1  1  1  2
Echiura  0  1  0  0  0  1
Nemertea  1  1  1  1  1  1
Platyhelminthes  1  1  0  1  0  1
Pogonophora  0  1  1  0  0  1
All phyla176307477297405807
Table 3.  The number of species, genera, families, orders and classes for all taxa combined and the dominant phyla in each area and on the total shelf
TaxaAreaSpeciesGeneraFamilyOrderClass
All taxa11761401003916
23072291385726
34773281747129
42972321345725
54052881485422
Total8075082398033
Annelida1 78 53 2710 1
2142107 3612 2
3226148 4114 2
4140111 3914 2
5197138 4114 2
Total344211 4815 2
Crustacea1 42 36 26 4 1
2 73 55 32 6 2
3105 77 46 9 3
4 72 52 37 7 2
5112 82 53 9 4
Total229141 7111 4
Mollusca1 37 33 2910 3
2 58 47 4214 4
3 98 64 5221 6
4 57 44 3619 5
5 67 46 3518 5
Total162103 7423 6

The average number of species per genus on the total shelf was the same for the three dominant phyla and for all taxa combined (1·6 species, Table 4). Annelids had a higher average number of species per family (7·2 species), per order (22·9 species) and class (172 species) than the other dominant groups, while molluscs had the lowest values (2·2, 7 and 27 species, respectively, Table 4). The number of species per class for all taxa combined on the shelf was only 24·5. The average numbers of species per genus, family, order and class were higher at the scale of the total area than at the scale of an area for all taxa combined and for each of the dominant phylum (Table 4).

Table 4.  The number of species (S) per genus (G), family (F), order (O) and class (C) for all taxa combined and the dominant phyla in each area and on the total shelf
TaxaAreaS/GS/FS/OS/C
All taxa11·31·8 4·5 11·0
21·32·2 5·4 11·8
31·52·7 6·7 16·5
41·32·2 5·2 11·9
51·42·7 7·5 18·4
Total1·63·410·1 24·5
Annelida11·52·9 7·8 78·0
21·33·911·8 71·0
31·55·516·1113·0
41·33·610·0 70·0
51·44·814·1 98·5
Total1·67·222·9172·0
Crustacea11·21·610·5 42·0
21·32·312·2 36·5
31·42·311·7 35·0
41·42·010·3 36·0
51·42·112·4 28·0
Total1·63·220·8 57·3
Mollusca11·11·3 3·7 12·3
21·21·4 4·1 14·5
31·51·9 4·7 16·3
41·31·6 3·0 11·4
51·51·9 3·7 13·4
Total1·62·2 7·0 27·0

The proportion of species in each of the dominant phyla changed along the shelf, although these changes were small (Fig. 2). Annelids had the highest proportion of species in area 5 and the lowest in area 1. Conversely, the proportion of mollusc species was highest in area 1 and lowest in area 5, whereas crustaceans had highest proportion of species in area 5 and lowest in area 3.

Figure 2.

Proportion (%) of the total number of species on the shelf and in each area (1–5) from each of the dominant groups: Annelida (black bars), Crustacea (cross-hatched bars) and Mollusca (white bars).

taxonomic distinctness and environmental variables

The mean value of the average taxonomic distinctness (Δ+) on the total shelf varied between the dominant phyla and was highest for molluscs (69·1), followed by annelids (63·5) and crustaceans (60·1) (Table 5). As expected, mean Δ+ for all taxa combined (87·0) was higher than for each of the dominant phyla. There was a decrease in mean Δ+ from area 1 to area 5 for all taxa combined (Table 5). Conversely, for annelids and crustaceans mean Δ+ increased from area 1 to area 4, but area 5 had lower values (Table 5). Molluscs had highest mean Δ+ in area 1, followed by area 4. At the smallest scale examined (i.e. a site) crustaceans displayed the widest range of Δ+ (47·3–72·2), followed by molluscs (56·7–76·7), and annelids (61·1–65·5) (Fig. 3b–d).

Table 5.  Mean number of species (S) and mean taxonomic distinctness (Δ+) per site for all taxa combined and the three dominant phyla in each area and on the total shelf
TaxaAreaSΔ+
All taxa1 65·8888·78
2101·9587·37
3 83·2487·01
4 71·9586·73
5105·7585·58
Total 86·7187·05
Annelida1 31·8162·09
2 54·2963·28
3 44·1663·36
4 40·2664·19
5 62·3064·16
Total 47·1763·46
Crustacea1 12·0657·20
2 19·0557·59
3 12·0860·00
4 10·3263·87
5 20·2561·34
Total 14·8160·05
Mollusca1 11·8171·57
2 16·8166·61
3 18·6467·45
4 14·7471·30
5 14·6569·73
Total 15·6569·10
Figure 3.

Relationship between taxonomic distinctness (Δ+) and latitude (°N) for: (a) all taxa, r = −0·49, P < 0·001; (b) Annelida, r = 0·63, P < 0·001; (c) Crustacea, r = 0·42, P < 0·001; (d) Mollusca, r = 0·09, NS.

Average taxonomic distinctness (Δ+) for all taxa combined decreased with latitude (r = –0·49, P < 0·001, n = 101, Fig. 3a). This indicates that there was a gradient in taxonomic distinctness for the total fauna, and species in the northern part of the study area were more closely related to each other than those at the southernmost sites. Conversely, although annelids were the most common phylum, Δ+ increased with latitude (r = 0·63, P < 0·001, n = 101, Fig. 3b). Likewise, crustaceans were positively related to latitude, but molluscs showed no such relationship (Fig. 3c,d). These results show that patterns of change in Δ+ for any dominant phylum cannot act as a surrogate for the total fauna.

Water depth along the shelf increased with latitude (rs = 0·81, Table 6). However, Δ+ of annelids and crustaceans showed a stronger positive relationship to depth than to latitude (Fig. 4b,c). Thus, annelid and crustacean species in the shallowest parts of the study area were more closely related to each other than those at the deepest sites. In contrast Δ+ for all taxa combined decreased more strongly with latitude than depth (Fig. 4a). Δ+ for molluscs was not related to depth (Fig. 4d). Again patterns of change in Δ+ are not consistent between phyla.

Table 6.  Spearman rank correlations (rs) between environmental variables, with significant (P < 0·05) coefficients in bold (n for all correlations = 101)
 LatitudeLongitudeDepthSilt-clayMdϕSkIσIKG
  1. Silt-clay, fraction of sediment < 0·063 mm (%); Mdφ, median grain size; SkI, skewness; σI, sorting; KG, kurtosis; TOM, total organic matter (%) (from Ellingsen & Gray 2002).

Longitude   0·70       
Depth   0·81   0·54      
Silt-clay   0·61   0·62   0·80     
Mdϕ   0·21   0·26   0·58   0·77    
SkI0·37−0·34−0·330·17−0·16   
σI   0·70   0·54   0·78   0·79   0·43−0·25  
KG−0·47−0·37−0·50−0·43−0·34   0·53−0·45 
TOM   0·70   0·56  0·86   0·94   0·70−0·24   0·82−0·48
Figure 4.

Relationship between taxonomic distinctness (Δ+) and depth (m) for: (a) all taxa, r = −0·29, P < 0·001; (b) Annelida, r = 0·72, P < 0·001; (c) Crustacea, r = 0·65, P < 0·001; (d) Mollusca, r = 0·13, NS.

Most of the variance in Δ+ for annelids and crustaceans is explained by depth (Table 7a). Although marginally significant, neither latitude nor grain size appeared to be important in explaining large-scale patterns in Δ+ of annelid, crustacean or mollusc assemblages. In contrast latitude rather than depth explained most of the variance in Δ+ of all taxa combined (Table 7a). Note that the total variance in Δ+ for all taxa explained by this analysis is less than for annelids or crustaceans.

Table 7.  (a) Multiple regression of average taxonomic distinctness (Δ+) performed separately for annelids, crustaceans, molluscs and all taxa combined, with latitude, depth and sediment grain size as predictor variables using data over all areas (101 sites). Explanatory variables are given in the order in which they contribute, i.e. using type I sums of squares. The two columns being the percentage variance explained and the F-value for testing the significance of that contribution. All F-values are on (1,97) d.f., i.e. the F-values should be compared with F1,97(5%) = 4·0. (b) Multiple regressions as in (a) but for response variable species richness (S). These use both linear and quadratic terms in the relationship with depth, because linear fits alone were not adequate (see Fig. 6); quadratic fits were not required for latitude or grain size. The F-values for relationships to depth are therefore on (2,96) d.f., i.e. to be compared with F2,96(5%) = 3·1
 Predictor variable% Variance explainedF statistic
(a) Δ+
AnnelidaDepth54·5126·2
Latitude 1·8  4·1
Grain 1·7  4·0
CrustaceaDepth42·7 75·5
Grain 2·1  3·7
Latitude 0·4  0·8
MolluscaDepth 1·7  1·7
Grain 0·03  0·03
Latitude 0·01  0·01
All taxaLatitude23·2 35·0
Grain 7·5 11·4
Depth 5·1  7·7
(b) S
AnnelidaDepth and depth232·0 54·7
Latitude 9·6 16·4
Grain 2·2  3·7
CrustaceaLatitude13·7 17·8
Depth and depth211·9 15·4
Grain 0·5  0·6
MolluscaDepth and depth216·7 23·0
Latitude11·8 16·2
Grain 1·5  2·1
All taxaDepth and depth229·0 42·3
Latitude 4·9  2·9
Grain 0·1  0·2

species richness and environmental variables

At the scale of areas, there was no clear relationship between mean species richness or total species richness and latitude for all taxa combined and each of the dominant phyla (Tables 3, 5). At the smallest scale examined (a site) the numbers of species of molluscs and crustaceans were not related to latitude, whereas annelids and all taxa combined showed positive but weak relationships with latitude (Fig. 5a–d).

Figure 5.

Relationship between sample species richness (S) and latitude (°N) for: (a) all taxa, r = 0·30, P = 0·002; (b) Annelida, r = 0·45, P < 0·001; (c) Crustacea, r = 0·25, P = 0·01; (d) Mollusca, r = −0·02, NS.

Species richness of the total fauna and the annelids were highest in the middle depth range, and lowest in the shallowest and in the deepest parts of the shelf (Fig. 6a,b). Crustaceans and molluscs showed no clear relation with depth (Fig. 6c,d).

Figure 6.

Relationship (quadratic regression) between sample species richness (S) and depth (m) for: (a) all taxa, R2 = 0·29, P < 0·001; (b) Annelida, R2 = 0·32, P < 0·001; (c) Crustacea, R2 = 0·12, P = 0·002; (d) Mollusca, R2 = 0·17, P < 0·001.

Most of the variance in species richness is accounted for by depth, followed by latitude and grain size for annelids, molluscs and all phyla combined, but latitude explained most of the variance in species richness of crustaceans (Table 7b). Grain size appeared to be unimportant in explaining large-scale patterns in the species richness of the individual or combined phyla. Latitude was a significant factor for the individual phyla, but not for all phyla combined.

There was no clear relationship between Δ+ and the mean number of species at the scale of the areas as well as the total number of species in an area (Tables 3, 5). At the smallest scale, Δ+ and the number of species on the total shelf was not related for molluscs (r = 0·02, P = 0·813, n = 101) and crustaceans (r = –0·1, P = 0·407, n = 101), whereas annelids showed a weak positive relationship (r = 0·26, P = 0·007, n = 101) and all taxa combined showed a weak negative relationship (r =–0·27, P = 0·006, n = 101). Thus Δ+ is clearly not a surrogate for species richness and vice versa.

Discussion

Whereas most marine studies of patterns of taxonomic distinctness have involved a mixture of systematically diverse groups (e.g. macrobenthos: Warwick & Clarke 1995; Clarke & Warwick 1998), or a specific group (e.g. nematodes: Warwick & Clarke 1998; asteroids: Price et al. 1999; copepods: Woodd-Walker et al. 2002; fish: Hall & Greenstreet 1998; Rogers, Clarke & Reynolds 1999; corals: Brown et al. 2002), we investigated differences in patterns of average taxonomic distinctness (Δ+) between the three dominant macrobenthic phyla: the annelids, crustaceans and molluscs, and all taxa combined.

taxonomic groups and environmental variables

Along the Norwegian continental shelf mean Δ+ at 101 sites varied between the dominant phyla (Table 5). Likewise, in a study of soft-sediment biodiversity in the subtropical Hong Kong waters Shin & Ellingsen (2004) found that mean Δ+ was highest for polychaetes (62), followed by molluscs (60) and crustaceans (55). In their study molluscs displayed the widest range of Δ+ (33–83), whereas Δ+ for polychaetes were only ranging from 58 to 65. The finding that polychaetes displayed the smallest range in both a subtropical and temperate area, shows that different taxa may have different expected ranges of Δ+ either as a result of their internal taxonomical hierarchy or different responses to environmental gradients. In the present study annelids dominated species and generic richness, but molluscs had the highest number of families, orders and classes (Table 3).

In this study Δ+ for annelids and crustaceans were positively related to latitude, while molluscs showed no such relationship (Fig. 3b–d). However, latitude is not regarded as the major driver of large-scale patterns of biodiversity, but it may be linked with important environmental factors (e.g. Rohde 1992). In a study of Atlantic asteroids Price et al. (1999) found no relationship between Δ* (taxonomic distinctness based on quantitative species data) and geographical area or depth, suggesting an absence of latitudinal and coastal/deep sea trends. However, in a large ‘ocean-scale’ study of copepods from the Atlantic Woodd-Walker et al. (2002) found abrupt changes in taxonomic distinctness (Δ+) around 40°N and 40°S. In the tropics and subtropics copepod communities were characterized by high stable taxonomic distinctness, whereas at high latitudes communities showed a lower mean and large variation in distinctness. Our finding that Δ+ of annelids and crustaceans showed a stronger positive correlation to depth than latitude (Figs 3b,c and 4b,c) suggest that depth may be an important factor influencing the relatedness of species. Thus, average taxonomic distinctness was influenced by habitat type. This is in contrast to that obtained by Warwick & Clarke (1998) for nematodes from the coasts of Britain and Chile, but in this case all samples were from shallow coastal waters. However, the finding that Δ+ for molluscs was not related to depth (Fig. 4d) shows that the impact of depth is not consistent between phyla.

The finding that for all taxa combined Δ+ was negatively related to latitude and depth, whereas for 71% of the fauna (annelids and crustaceans) Δ+ was positively related, and not significantly related for 20·1% of the fauna (molluscs) (Figs 3 and 4), is surprising and requires further examination. Changes in the proportions of species in the three dominant phyla between areas (Fig. 2), although quite small, might be one reason explaining the decrease in Δ+ from south to north for the combined phyla. Molluscs, for example, have a high overall Δ+ relative to annelids and crustaceans (Table 5), and molluscs have the lowest proportion of species in the northernmost area (Fig. 2). The number of species in each phylum varied between one and 344 species (Table 2). A possible explanation for the inconsistency in Δ+ might be that the taxonomic classification system also varies between different phyla. It might be argued, as for example Harper & Hawksworth (1994) have done in relation to Simpson's index, that taxonomically related biodiversity measures might be more meaningfully applied to a single phylum than to all taxa combined.

To summarize, the dominant phyla displayed different patterns of Δ+ and therefore one taxonomic group could not be taken to represent the others in terms of taxonomic distinctness. This is in agreement with other measures of biodiversity used to study benthic patterns on the Norwegian shelf (Ellingsen 2001, 2002; Ellingsen & Gray 2002). Thus, using a particular taxon to predict patterns in overall biodiversity is not straight forward.

relations with other measures of biodiversity

We found no clear relationship between the two univariate biodiversity measures, Δ+ and S, and therefore assemblages with lower species richness do not necessarily have a smaller average taxonomic range than those with many species. Warwick & Clarke (1995) reported a decrease in taxonomic distinctness (Δ*) of macrobenthos along a gradient of increasing environmental contamination around an oilfield in the North Sea where species diversity remained constant. They concluded that taxonomic distinctness might be a more sensitive univariate index of community perturbation than species diversity. However, these findings were not supported by data from other oilfields in the North Sea (Somerfield, Olsgard & Carr 1997). Using a data set on nematodes from coastal sites in the UK Warwick & Clarke (1998) demonstrated a decrease in Δ+ at polluted sites compared with ‘clean’ sites. In a study of marine fish communities of the northern North Sea Hall & Greenstreet (1998) found that Δ, Δ*, and two more conventional measures of diversity behave in a broadly similar manner, and there was no clear relationship to increased fishing effort. Likewise, Rogers et al. (1999) found that a reduction in average taxonomic range of fish communities between the western waters of the UK and the southern North Sea mirrored a similar pattern in the number of taxa represented. In a study of corals at Ko Phuket, Thailand, Brown et al. (2002) found that both univariate measures of species diversity and taxonomic distinctness (Δ*) decreased with the level of physical stresses. Thus, some studies show that assemblages that are more species rich contain species that have a wider average taxonomic range than assemblages that are less species rich, i.e. that species richness and taxonomic distinctness behave in the same way with respect to environmental stress. However, no generality can be attached to these findings. Patterns of taxonomic distinctness may be a result of both natural processes and anthropogenic disturbance at a local scale, and these two factors are not always easily separated.

All measures of species diversity rely on good taxonomy. Although, the definition of a species may vary to some degree in different taxonomic groups (Harper & Hawksworth 1994), there is fairly widespread acceptance of the species concept, and it is clear that traditional species richness and equitability measures can usefully be applied to combinations of phyla. However, measures of taxonomic distinctness across all phyla must be considered with more care, as taxonomic ranks (e.g. families, orders) may differ between different phyla. The Linnaean hierarchical taxonomy and classification is constantly being modified based on cladistics and phylogeny. An updated classification of the crustaceans (Martin & Davis 2001) based on phylogenetic information and traditional Linnaean ranks resulted in an increase of 197 families compared with a previous classification (Bowman & Abele 1982). Such revisions may well occur in other phyla in the future. There is also debate over the best way to name and classify organisms in light of current understanding of evolution and biodiversity (e.g. Pennisi 2001). In the traditional system organisms are grouped into ranks, such as genus, family and order, whereas in one new naming system, known as ‘PhyloCode’ (http://www.ohio.edu/phylocode), organisms are assembled into ‘clades’, defined as any set of organisms with a common ancestor (Cantino & de Queiroz 2004).

Clarke & Warwick (1999) pointed out that Δ+ are not constrained only to hierarchies with fixed points of genus, family, order, etc., but carry over naturally and forcefully to a continuous phylogeny in which the branch lengths are fully determined, for example by genetic distances (e.g. Nei 1996). Linnaean classifications are a simple, discrete approximation to a fully fledged phylogeny, but still better than no classification (Humphries, Williams & Vane-Wright 1995). Furthermore, Clarke & Warwick (1999) and Rogers et al. (1999) showed strong insensitivity of Δ+ to major variations in the assumed branch step lengths between taxa ranks. Branch step lengths are currently determined by the fixed points of Linnaean classifications and, naturally, the further these are away from a true phylogenetic ranking, the less meaningful the final interpretation is likely to be (but see Rogers et al. 1999).

Whether it is appropriate to calculate taxonomic distinctness across different phyla, or instead restrict calculations of this measure to within different phyla (as advocated by, e.g. Clarke & Warwick 1999), is arguable. There is a trade-off between the taxonomic breadth of an assemblage under consideration and the likelihood that taxonomic ranks within the assemblage are consistent. Within a typical marine assemblage containing a mixture of phyla, one must take into account the fact that differences between samples may result from differences in the application of taxonomic ranks within phyla. If the analysis is restricted to a single phylum the problem in no way disappears because no, for example, polychaete or crustacean groups is likely to have received the same amount of attention, and at the same time the hierarchy is truncated by one rank. The closer one approaches groups in which there is a greater probability of consistency in use of taxonomic nomenclature, the less useful information on relatedness between species is likely to remain. The investigator must therefore carefully decide which aspect of this trade-off is more pertinent to the hypothesis being tested.

Acknowledgements

OLF (The Norwegian Oil Industry Association) permitted us to use the data, and Det Norske Veritas and Akvaplan-Niva prepared them. KEE acknowledges the support of the Research Council of Norway. This work was also supported, in part, by the United Kingdom Department for Environment, Food and Rural Affairs (Defra) through projects AE1137 and CDEP 84/5/295, and is a contribution to the biodiversity element of PML's core strategic research project. We thank Prof J. S. Gray, Dr F. Olsgard and Dr M. J. Anderson for valuable discussions.

Ancillary