Mar Biol DOI 10.1007/s00227-009-1224-z

ORIGINAL PAPER

Latitudinal gradients in species richness for South American Mytilidae and Ostreidae: can alternative hypotheses be evaluated by a correlative approach? Alvar Carranza · Omar Defeo · Juan Carlos Castilla · Thiago Fernando L. V. B. Rangel

Received: 20 January 2009 / Accepted: 8 May 2009 © Springer-Verlag 2009

Abstract We tested to what extent mean sea surface temperature, geometric constraints in range size frequency distributions (the mid-domain eVect) and geographical coastline distance to the equator are related to species richness of coastal Mytilidae and Ostreidae in the PaciWc and Atlantic coasts of South America (excluding islands). The location and magnitude of the peaks in species richness, as well as the shape of the pattern, varied between oceans. Results were not biased by spatial autocorrelation, although strong multicollinearity among predictor variables was detected. However, these regional-extent regression models suggest diVerences in the causal factors that explain richness gradients of studied bivalves in South American coasts, most likely related to historical events such as the Southeastern PaciWc Pleistocene mass extinction of bivalves. Our results reinforced the conclusion that there is no single best explanatory cause for the latitudinal gradient in species richness and showed that the correlative Communicated by F. Bulleri. A. Carranza (&) · O. Defeo UNDECIMAR, Facultad de Ciencias, Iguá 4225, CP 11400 Montevideo, Uruguay e-mail: [email protected] A. Carranza · O. Defeo Dirección Nacional de Recursos Acuáticos, Constituyente 1497, CP 11200 Montevideo, Uruguay J. C. Castilla Facultad de Ciencias Biológicas, Center for Advanced Studies in Ecology and Biodiversity (CASEB), PontiWcia Universidad Católica de Chile, Casilla 114-D, Santiago, Chile T. F. L. V. B. Rangel Department of Ecology and Evolutionary Biology, University of Connecticut, Storrs, CT 06269, USA

approach is not useful when predictor variables are strongly correlated.

Introduction The latitudinal diversity gradient is probably the best documented of all observed biogeographical patterns, and is remarkably consistent across habitats and taxonomic groups (Hillebrand 2004). The pattern consists in a poleward decline in the number of species, peaking at the tropics in terrestrial as well as in coastal marine and pelagic biotas. However, and despite the huge amount of work on the subject, a consensus on the explanation for the phenomenon is far for being reached (Harrison and Cornell 2007). The disparity of opinions is illustrated by the number of hypotheses proposed to explain the latitudinal gradient of species diversity. In this vein, Mittelbach et al. (2007) distinguished three main explanations for the gradient: (1) ecological hypotheses that focus on mechanisms of species coexistence and the maintenance of species diversity; (2) evolutionary hypotheses that focus on rates of diversiWcation; (3) historical hypotheses that focus on the duration and extent of tropical environments along Earth’s history. Empirical support for ecological hypotheses includes those presented by Roy et al. (1998), who showed that diVerences in species richness along comparable transects for the North-western Atlantic and North-eastern PaciWc are correlated with mean sea surface temperatures (SST). This led to the suggestion that enhanced diversity may be related to some measure of productivity (e.g. species-energy hypothesis, see Currie 1991; Roy et al. 2000). Among the historical explanations, empirical support for the time and area hypothesis is given by observations that show that temperate taxa are often younger than, and nested within, tropical

123

Mar Biol

taxa, and that diversity is positively correlated with the age and area of geographical regions (Mittelbach et al. 2007). The diversiWcation rate evolutionary hypothesis is mainly supported by phylogenetic evidence for higher rates of diversiWcation in tropical clades and paleontological data (Mittelbach et al. 2007). The diversiWcation rate and time and area hypotheses predict a decrease in species richness with geographical distance to diversiWcation centers. Thus, the coastline distance from the equator can be related to species richness since the dispersal of species or spread of clades must occur along the shelf. In addition to these “deterministic” hypotheses, recent studies provided empirical support for the null geometric constraint models (GCMs) and the mid-domain eVect (MDE) these models produce, in the absence of abiotic and biotic gradients aVecting the distribution of species along geographic (e.g. Colwell et al. 2005 and reference therein) and other environmental domains (Carranza et al. 2008). This pattern takes the form of a unimodal diversity curve with a middomain maximum. However, rejection of any of these particular hypotheses is diYcult: while the mechanisms proposed by the hypotheses diVer, their primary predictions are not mutually exclusive (Algar et al. 2007). Nonetheless, it is still possible to analyze when predictions or variables associated with each theory increase or not the statistical Wt of the models. Bioengineering shallow marine bivalves are conspicuous members of nearshore communities and of utmost ecological importance due to its ability to create, modify and/or increase habitat heterogeneity (Gutiérrez et al. 2003; Commito et al. 2005, 2006; Prado and Castilla 2006). These ecosystem engineers (see Jones et al. 1994) provide several ecosystem services and control the availability of resources to other species, by modifying several environmental features (Gutiérrez et al. 2003; Prado and Castilla 2006; Borthagaray and Carranza 2007). These species occur in estuaries, lagoons and rocky shore coastal ecosystems worldwide. Oyster and mussel beds often increase local species richness due to generation of tri-dimensional habitats (Commito et al. 2006; Prado and Castilla 2006; Borthagaray and Carranza 2007). As such, they can have strong ecosystem-level impacts, including direct eVects on the diversity patterns of associated species along large spatial scales. Besides their ecological or socioeconomic importance, South American Mytilidae and Ostreidae are diverse enough in both the Caribbean and South-western Atlantic coasts (hereafter Atlantic) and central and South-eastern PaciWc coast (hereafter PaciWc), thus providing a unique opportunity to downscale predictions from both ecological and historical-evolutionary hypotheses. However, most of the studies on latitudinal gradients have been carried out in the Northern hemisphere, with much less eVort devoted to

123

the analysis of patterns in the southern hemisphere. Previous work on South American benthic invertebrates suggested diVerences in the shape of diversity patterns between the South-Eastern PaciWc and South-Western Atlantic coasts (Astorga et al. 2003; Valdovinos et al. 2003). In this paper, we quantiWed latitudinal gradients in species richness of South American Mytilidae and Ostreidae, most of which can be considered as ecosystem engineers, along the Atlantic and PaciWc coasts of South America, in a latitudinal range spanning from 13° to 55°S. Three major alternative, non-exclusive hypotheses are evaluated in a context of multiple causality: the time and area/diversiWcation rate hypothesis (merging evolutionary and/or historical explanations), the species–energy hypothesis (linked to a contemporary ecological process) and the eVects of geometric constraints on range size frequency distributions (RSFD) (hereafter MDE, a “neutral” explanation). To this end, we evaluate the relationships between species richness and mean SST (a proxy for solar energy input), coastline distance to the equator (as a proxy of the predictions of historical and evolutionary hypotheses), and the expected richness pattern emerging from MDE null models. Further, we provide an explicit analysis of multicollinearity between predictor variables and discuss the potential implications of these issues on the interpretation of the causal factors linked to the observed patterns.

Materials and methods Empirical data Species lists for South American Mytilidae and Ostreidae were compiled for the Atlantic and PaciWc coasts of South America based on information provided by the Western Atlantic Molluscs Database (Malacolog 4.0.), Rios (1994), Coan et al. (2000), Scarabino (2003), Zelaya (2005), Keen (1971), Paredes et al. (1991), Linse (1999) and Valdovinos (1999). Once the species list for each coast was determined, we checked all the available literature to determine the southernmost and northernmost records for each one, thus obtaining their latitudinal ranges. The inclusion in the database took place after verifying the validity of the name and geographic ranges provided elsewhere with a reasonable certainty, after checking for potential synonymies. Insular taxa (i.e. those endemic to eastern Island, Diego Ramírez and Galapagos Islands in the PaciWc and San Pedro and San Pablo and Trindade Islands in the Atlantic) were not included. We restricted the database to those species whose bathymetric range was partially or totally contained within the innermost portion of the continental shelf (i.e. <50 m depth). Introduced species (i.e. Perna viridis, Crassostrea gigas) were not considered, thus restricting our analyses to

Mar Biol

endemic species of the geographic domain studied here. We recognize, however, that it is extremely diYcult to separate the eVects of human activity, especially in prehistoric times, from “natural” processes aVecting the geographical range of the native species. Species richness was estimated at several virtual stations placed each 5° latitude along the gradients, as the number of species whose ranges intersected that point, assuming a continuous distribution within range endpoints. This assumption is usually not a source of signiWcant bias for most taxa (Colwell et al. 2004). The resulting richness values for each virtual station are referred here as interpolated richness. When necessary following Cardelús et al. (2006), we adjusted for range underestimation by adding 1° to species ranges represented by single literature records. Without this, single-site species would otherwise have been ‘lost’ between virtual stations during midpoint randomization (Brehm et al. 2007). Sea surface temperature The eVect of solar-energy input on species richness (species-energy hypothesis), was evaluated by means of a correlative approach between this variable and interpolated species richness. To this end, we calculated the annual mean SST in each virtual station by averaging values from the Wve nearest pixels (Fig. 1). This was done with data extracted from an annual SST satellite image (Aqua-

MODIS, resolution 9 km) processed with SeaDas software. Although SST may not be a good proxy for productivity, the latter is diYcult to measure in the sea. Thus, temperature has frequently been used as a proxy measure, thereby confounding temperature and productivity eVects on diversity (Clarke and Gaston 2006). For this reason, and because we were interested also in the statistical problems related with the evaluation of hypotheses, we use a single annual SST image, early and widely used by ecologists in the evaluation of this hypothesis (e.g. Valdovinos et al. 2003). Although this may not represent an exact description of interannual mean values, it represents properly the macroscale SST gradient along both coasts. Geographic distance To evaluate the eVects of geographic coastline distance from each virtual station to latitude 0° station (the point where coastline intersects the equator), we measured this distance on a 1:3,000,000 digital map of South America (geographic coordinate system) for the Atlantic and PaciWc coasts. This was done using the Measure tool’s snapping option in ARCGIS 9.2, which uses a standard 8 pixels of the display as the snapping tolerance. Minimum step-length used was 50 km, while near-straight portions of the coastline were measured in toto. The coastline distance obtained for each virtual station was used as a proxy to evaluate the predictions of both evolutionary and historical hypotheses.

Fig. 1 Map of South America showing the locations of the virtual stations and the distribution of Annual Sea Surface Temperature

123

Mar Biol

Null models The MDE is the increasing overlap of species ranges toward the center of a shared, bounded domain due to geometric boundary constraints in relation to the distribution of species’ range sizes, producing a peak or plateau of species richness toward the center of the domain (Colwell 2006). To estimate predicted richness under the assumptions of GCMs, namely a random placement of interpolated richness, we drew ranges at random from the empirical RSFD and placed them randomly within the latitudinal domains using model 4 of the computer application RangeModel 5 (Colwell 2006). To this end, the location of each species latitudinal ranges within the domain (in our case given by the northernmost and southernmost range endpoints) was randomly reassigned (sampling without replacement) and then the predicted richness at each of the virtual stations was recorded after 1,000 randomizations, reporting the mean richness and its 95% conWdence interval (CI) for each station. This was done using the entire latitudinal domain (60°N–60°S), in order to avoid the artiWcial truncation of species ranges. Resampling empirical RSFDs balances the risk of underestimating and overestimating the role of MDE (Colwell et al. 2004; Algar et al. 2007). The predicted mean richness along the stations places in each coast were then used in multiple regression analysis, in order to incorporate MDE as an explanatory variable on an equal statistical footing with other candidate explanatory variables.

data may cause. In a spatial regression, spatial dependency is entered in the regression model as relationships, e.g. between the independent and the dependent variables. Alternatively, if regression residuals were not spatially autocorrelated, further analysis by spatial regression was not considered necessary (Diniz-Filho et al. 2002). Multicollinearity To assess the impact of collinearity among the variables in a regression model we used the condition number (CN) and the variance inXation factor (VIF). The CN was calculated as the square root of the ratio between the Wrst and last eigenvalue of the correlation matrix of regression predictors (Graham 2003). The CN measures the stability of sensitivity of the matrix of predictors (the smaller the CN, the more reliable the regression model). On the other hand, VIF was calculated for each predictor variable, measuring the amount of variance in the regression coeYcient of the predictor due to collinearity. VIF ranges from 1 (absence of multicollinearity) to positive inWnity. There is no formal VIF value for determining the presence of multicollinearity, but VIF values exceeding 10 are commonly used as indicators of signiWcant multicolienarity (Graham 2003). We used SAM v3.0 (Rangel et al. 2006) for spatial analysis, model selection and multicollinearity analysis. SAM is freely available at http://www.ecoevol.ufg.br/sam.

Regression analysis

Results

We evaluated the predictions of each one of the three hypotheses using multiple linear regressions to explore multivariate explanations for the latitudinal patterns in interpolated species richness, incorporating distance, SST, and MDE as predictor variables. Since strong multicollinearity is expected, we were not aiming to disentangle between unique and shared contributions of the variables, but, instead, to select a Wnal model that explain the largest variability in the response (Graham 2003), and to examine which variables are entered in the model. To this end, we selected the best model using the Akaike information criterion (AIC; Burnham and Anderson 1998) after examining all possible models. Regression residuals were examined for spatial autocorrelation based on Moran’s I for a connectivity matrix constructed under the distance criterion (2,000 km for Atlantic data and 1,500 km for PaciWc data). The lag distances were chosen to minimize the number of connections between points but ensuring at least one connection per point. Connections were weighed according to the distance between points. If regression residuals were spatially autocorrelated, an explicit spatial regression is needed in order to control for bias that spatially structured

Species richness was very similar between coasts: 43 species in the Atlantic and 41 in the PaciWc (Table 1). Shared species Mytella guyanensis, Lithophaga patagonica, Choromytilus chorus, Brachidontes blakeanus, Crenella decussta, Aulacomya atra, Modiolus patagonicus, Perumytilus purpuratus and Semimytilus algosus represented nearly 10% of the total taxa. Latitudinal patterns, however, markedly diVered between oceans: in the Atlantic, species richness declined monotonically with latitude after peaking at latitude 10°N, whereas species richness in the PaciWc decreased to a local minimum at 15°S, slightly increasing southwards until 55°S (Fig. 2). Multicollinearity analysis for the Atlantic dataset showed that the explanatory variables used in the regression are not independent. The CN for the full model was high (CN = 25.564). When the multicollinearity in each predictor was evaluated using VIF, Distance (VIF = 32.199), MDE (VIF = 136.809), and SST (VIF = 89.276) were similar, supporting the overall multicollinearity in the model estimated by CN. A principal component analysis, used to synthesize the information provided by the three predictors, showed that the Wrst component alone explains

123

Mar Biol Table 1 Species list and distributions for South American Mytlidae and Ostreidae

Ostreidae Crassostrea corteziensis (Hertlein, 1851)

PaciWc

Crassostrea gasar (Dautzenberg, 1891)

Atlantic

Crassostrea rhizophorae (Guilding, 1815)

Atlantic

Ostrea chilensis (Philippi, 1845)

PaciWc

Ostrea columbiensis (Hanley, 1846)

PaciWc

Ostrea conchapila (Carpenter, 1857)

PaciWc

Ostrea cristata (Born, 1778)

Atlantic

Ostrea equestris (Say, 1834)

Atlantic

Ostrea iridescens (Hanley, 1854)

PaciWc

Ostrea libella (Weisbord, 1964)

Atlantic

Ostrea lixula (Weisbord, 1964)

Atlantic

Ostrea palmula (Carpenter, 1857)

PaciWc

Ostrea puelchana (d’Orbigny, 1842)

Atlantic

Undulostrea megodon (Hanley, 1846)

PaciWc

Loha angelica (DeRochebrune, 1895)

PaciWc

Lopha gibsonsmithi (Macsotay and Campos, 2001)

Atlantic

Mytilidae Adula soleniformis (d’Orbigny, 1842)

PaciWc

Amygdalum americanum (Soot–Ryen 1955)

PaciWc

Amygdalum dendriticum (Muhlfeld, 1811)

Atlantic

Amygdalum politum (Verril and S. Smith, In Verril, 1880)

PaciWc

Amygdalum sagittatum (Rehder, 1935)

Atlantic

Aulacomya atra (Molina, 1782)

PaciWc–Atlantic

Botula cylista (Berry 1959)

PaciWc

Botula fusca (Gmelin, 1791)

Atlantic

Brachidontes adamsius (Dunker 1857)

PaciWc

Brachidontes blakeanus (Melvill and Standen, 1914)

PaciWc–Atlantic

Brachidontes darwinianus (d’Orbigny, 1842)

Atlantic

Brachidontes domingensis (Lamarck, 1819)

Atlantic

Brachidontes exustus (Linnaeus,1758)

Atlantic

Brachidontes granulata (Hanley 1843)

PaciWc

Brachidontes modiolus (Linnaeus, 1767)

Atlantic

Brachidontes playasensis (Pylsbryii and olsson, 1935)

PaciWc

Brachidontes puntarenensis (Pylsbry and Lowe, 1932)

PaciWc

Brachidontes rodriguezii (d’Orbigny, 1842)

Atlantic

Brachidontes semilaevis (Menke, 1849)

PaciWc

Brachidontes solisianus (d’Orbigny, 1842)

Atlantic

Choromytilus chorus (Molina, 1782)

PaciWc–Atlantic

Crenella abbotti (Altena, 1968)

Atlantic

Crenella caudiva (Olsson, 1961)

PaciWc

Crenella decussata (Montagu, 1808)

PaciWc–Atlantic

Crenella gemma (Olsson and McGinty, 1958)

Atlantic

Crenella magellanica (Linse, 2002)

Atlantic

Gregariella coarctata (Carpenter 1857)

PaciWc

Gregariella coralliophaga (Gmelin, 1791)

Atlantic

Lioberus castaneus (Say, 1822)

Atlantic

Lithophaga (Diberus) plumula (Hanley, 1843)

PaciWc

Lithophaga antillarum (d’Orbigny, 1853)

Atlantic

123

Mar Biol Table 1 continued

Ostreidae Lithophaga aristata (Dillwyn, 1817)

PaciWc

Lithophaga atenuata (Deshayes, 1836)

PaciWc

Lithophaga nigra (d’Orbigny, 1853)

Atlantic

Lithophaga patagonica (Orbigny, 1846)

PaciWc–Atlantic

Litophaga aristata (Dillwyn, 1817)

Atlantic

Litophaga bisulcata (d’Orbigny, 1842)

Atlantic

Litophaga peruviana (d’Orbigny, 1853)

PaciWc

Modiolus americanus (Leach, 1815)

Atlantic

Modiolus capax (Conrad, 1837)

PaciWc

Modiolus carvalhoi (Klappenbach, 1966) Modiolus Wcoides (Macsotay and Campos, 2001)

Atlantic

Modiolus patagonicus (d’Orbigny, 1842)

Atlantic–PaciWc

Modiolus pseudotulipus (Olsson, 1961)

PaciWc

Modiolus rectus (Conrad, 1837)

PaciWc

Modiolus squamosus (Beauperthuy,1967)

Atlantic

Musculus lateralis (Say, 1822)

Atlantic

Musculus viator (d’Orbigny, 1842)

Atlantic

Mytella charruana (d’Orbigny, 1842)

Atlantic

Mytella guyanensis (Lamarck, 1819)

PaciWc–Atlantic

Mytella tumbeziensis (Pylsbry and Olsson, 1935)

PaciWc

Mytilus edulis chilensis (Hupe 1854)

PaciWc

Mytilus edulis platensis (Linnaeus, 1758)

Atlantic

Perna perna (Linnaeus, 1758)

Atlantic

Perumytilus purpuratus (Lamarck, 1819)

PaciWc–Atlantic

Semimytilus algosus (Gould 1850)

PaciWc–Atlantic

Septifer zeteki (Hertlein and Strong, 1946)

PaciWc

Fig. 2 Latitudinal variations in South American bioengineering bivalve species in the PaciWc (Wlled circle) and Atlantic (open square) coasts of South America

99% of the variance in all predictors. The regression model estimated for the PaciWc dataset also suVered from multicollinearity, although it was not as severe as the Atlantic dataset. The CN for the full PaciWc model was 8.593, whereas the VIF for Distance was 16.767, MDE was

123

Atlantic

14.637 and SST was 15.86. The Wrst Principal Component captured 97% of the variation in the PaciWc predictors, strongly suggesting that they are indistinguishable. Regression analyses also showed diVerent patterns between ocean coasts: R2 values for the Atlantic were much higher than their PaciWc counterparts (Table 2). The R2 obtained for the PaciWc (Model: MDE + Distance + SST) was 0.821, whereas in the Atlantic (same model) R2 was 0.968. The best single predictor of the latitudinal gradient in the Atlantic was MDE (R2 = 0.960), whereas in the PaciWc the single best predictor was SST (R2 = 0.353). No signiWcant spatial autocorrelation was detected in regression residuals (all Moran’s I > 0.05). Univariate relationships between interpolated species richness and explanatory variables are shown in Fig. 3. A model selection criterion based on AIC suggests that in the Atlantic dataset a regression model containing only MDE as a predictor variable has 0.279 probability of being the best model among models, as measured by its AIC weight. On the other hand, a full model is certainly the worst model, since it adds too much multicollinearity without adding explanation: there is only a probability of 0.016 that the full model is the best one. For the PaciWc dataset,

Mar Biol Table 2 Model selection for interpolated species richness for Atlantic–Caribbean and PaciWc coasts separately

Models

AIC

R2

Variable contribution


P value

PaciWc ocean MDE + sea surface temperature + Distance

49.736

0.821

MDE

¡4.502

Sea surface temperature

<0.001

1.951

0.026

¡0.005

0.019

MDE

¡1.971

0.011

Distance

¡2.479

0.003

¡1.849

0.015

2.359

0.004

Distance

0.003

NS

Sea surface temperature

1.187

NS

¡0.536

NS

0.454

NS

Distance MDE + distance

51.657

MDE + sea surface temperature

52.543

0.681

0.658

MDE Sea surface temperature Sea surface temperature

56.519

0.353

Distance

56.524

0.353

MDE

59.891

0.161

Distance + sea surface temperature

60.710

0.360

Atlantic ocean MDE

19.667

0.96

Distance

21.211

0.958

Distance + sea surface temperature*

20.382

0.968

Distance Sea surface temperature MDE + distance*

21.032

0.967

MDE In multivariate models, standardized regression coeYcients (
) are shown together with the associated P value AIC Akaike information criterion * SigniWcant correlations between independent variables (R2 > 0.90; P < 0.05). The best models are highlighted in bold and italics

0.519

NS

¡0.468

NS

MDE

0.735

NS

Sea surface temperature

0.246

NS

MDE

0.101

NS

Sea surface temperature

0.377

NS

¡0.504

NS

Distance Sea surface temperature

21.472

0.954

MDE + sea surface temperature*

23.467

0.961

MDE + sea surface temperature + distance

Distance

the full model was the best among all possible models (AIC weight = 0.586).

Discussion We showed that macroscale species richness patterns for South American Mytilidae and Ostreidae diVered between Atlantic and PaciWc coasts and that this heterogeneity may be related to regional diVerences in the eVect of diVerent environmental features (e.g. SST, coastline conWguration) and stochastic (e.g. MDE on RSFD) processes. The shape

25.399

0.968

of the latitudinal gradients showed marked asymmetries between ocean coasts: while species richness in the Atlantic linearly decreased with latitude, the PaciWc pattern depicted a more complex trend. However, we were not able to statistically distinguish the relative eVect of predictor variables due to the strong multicollinearity detected, especially in the Atlantic coast. Since we are unable to disentangle the causality of the patterns based on statistical analysis, our discussion will focus on assembling pieces of evidence. In this study, the models that best explained the observed patterns diVered between oceans: while MDE in species distribution ranges

123

Mar Biol Fig. 3 Univariate relationships between interpolated species richness and the predictions of the best models for the PaciWc (a) and Atlantic (b) coasts

alone best explained the Atlantic pattern, a three-variable model best explained the patterns for the PaciWc coast, where MDE showed the highest contribution among individual predictors. However, the regression coeYcient for MDE in the best model for the PaciWc was negative, meaning that although this variable improved the statistical performance of the model, it is not really a candidate explanation for the PaciWc pattern. On the other hand, the eVect of MDE in the Atlantic pattern can be explained at least in part, since maximum and mean range size for Atlantic taxa was higher than for their PaciWc counterparts. Large-ranged species (Amygdalum sagittatum, Gregariella coralliophaga, Brachidontes exustus, Ostreola equestris and Crenella decussata in the Atlantic and C. decussata, Lithophaga

123

(Diberus) plumula, L. aristata, P. purpuratus and Modiolus rectus) in the PaciWc, disproportionately drive geographic patterns of species richness due to repeated counts for each one over wide portions of the domain, whereas species with small ranges are counted over a limited portion (Jetz and Rahbek 2002). Our best model for the PaciWc pattern is in agreement with Valdovinos et al. (2003), who did not Wnd a relationship between mollusk richness and SST, even after removing the potentially confounding eVect of latitude. However, our results contrast with Roy et al. (2000), who showed a strong latitudinal diversity gradient closely related to mean SST for some 930 North-Eastern PaciWc marine shelf infaunal and epifaunal bivalves. The authors showed that most

Mar Biol

bivalve clades within broad functional groups conformed to the general latitudinal trend, the exception being the deposit-feeding protobranchs. On the other hand, in a study conducted between 29°S and 36°S (Central Chile), Broitman et al. (2001) found that a signiWcant proportion of the regional variation in abundance for some of the most abundant invertebrate functional groups within each tidal level was associated with the latitudinal gradient in annual mean SST, as documented for mussels in the mid-intertidal zone. This points out to a potential role of the PaciWc SST gradient in driving large scale ecological patterns. However, these studies did not perform explicit analyses of multicollinearity. Direct comparisons of the shape of the PaciWc and Atlantic invertebrate patterns in South American coasts have seldom been reported. Astorga et al. (2003) showed that diversity decreased with increasing latitude for Brachyuran and Anomuran crustaceans in both oceans, and that spatial variations in SST explained diversity patterns of both groups at large, but not small (<5°) scales. These authors suggested that SST might diVerentially aVect taxa with contrasting modes of development, thus inXuencing diversity patterns. South American Mytilidae and Ostreidae are primarily pelagic developers, though all species of the genus Ostrea incubate their eggs after fertilization and release larvae. In particular, Tiostrea chilensis has no planktonic larval phase, incubating larvae up to the pediveliger, settlement stage (Millar and Hollis 1963). However, CranWeld and Michael (1989) documented the ability to release some planktonic larvae in the New Zealand population of T. chilensis, suggesting that this may play a role in maintaining gene Xow. Besides, free living, planktonic larvae, has been reported for the mussel genera Brachidontes (MonteiroRibas et al. 2006), Perna (Lasiak 1986), Adula (Lough and Gonor 1971), Aulacomya (Kennedy 1977), Perumytilus, Semimytilus (Navarrete et al. 2002), Choromytilus (Bellolio et al. 1996), Mytilus and Modiolus (De Schweinitz and Lutz 1976). Thus, we are conWdent that we have avoided the potential confounding eVect of contrasting developmental modes. Fortes and Absalão (2004)showed that richness of benthic invertebrates in both coasts decreased with increasing latitude. These authors found a much higher number of species in the Atlantic, as well as a sharper decrease of PaciWc species south of the Equator. However, the shape of the PaciWc pattern depicted by Valdovinos et al. (2003; Fig. 1C) is strikingly similar to our results. Physical diVerences between Atlantic and PaciWc coasts could account for dissimilar distribution patterns reported here. For instance, diVerences in latitudinal ranges of some coastal ecosystems (e.g. mangroves) and in some topographic features of the coastline (e.g. the fractal nature of the southernmost PaciWc coast) could aVect the distribution patterns

of coastal taxa, limiting distribution ranges especially for organisms closely related to speciWc habitats (e.g. mangrove oysters and rocky shore mussels). In addition, Fortes and Absalão (2004) evaluated the role of the size of biogeographic provinces in both oceans in driving richness patterns, concluding that this factor strongly aVects the form of the pattern. For example, a species-area eVect may be related with the location of the Atlantic peak in species richness (10°N), since the continental shelf is wider in the Atlantic coast than in the PaciWc. We are unaware of similar studies for bivalves in the Atlantic, but Floeter and Soares-Gomes (1999) reported a similar decreasing trend in gastropod species richness, though they worked with data binned in three biogeographic regions (10–20°S, 20– 30°S and 30–40°S). Based on results for regional scale studies of terrestrial ecosystems, Harrison and Grace (2007) suggested that explanations to exceptions to the general positive productivity-richness relationship may be related to the evolutionary history of the particular species pool in question. Without invoking these explanations, the weaker correlation between SST and richness found at the PaciWc may result from weak correlations between mean annual SST and productivity itself, or to the eVects of strong temporal variability in environmental conditions (e.g. severe El Niño Southern Oscillations). In this vein, physical-oceanographic processes are responsible for diVerences in the shape of the SST and productivity gradients along both coasts: Acha et al. (2004) recognized four frontal zones in the Atlantic: Atlantic upwelling zone; the temperate estuarine zone; the Patagonian tidal zone and the Argentine shelf-break zone, which occupies most of the Atlantic side. In contrast, the PaciWc coast is dominated by two large fronts, appearing simplest than the Atlantic coast in terms of frontal richness (Acha et al. 2004). However, The Humboldt Current System in the PaciWc includes one of the most productive upwelling areas of the world, extending from southern Chile (»42°S) up to Ecuador and the Galapagos Islands near the equator (Thiel et al. 2007). This area includes our 15°S station where the species richness minimum for the PaciWc was observed, thus, suggesting weak links between productivity and richness. In addition, while suggesting that the relationship between SST and diversity is consistent with a species-energy hypothesis, Harrison and Grace (2007) recognized that the linkages between SST and diversity remain unclear. In the same line, and as pointed out by Ricklefs (2007), present-day environmental variables such as temperature and productivity have an inXuence on large-scale patterns of species diversity (e.g. Currie 1991; Roy et al. 1998; Currie et al. 2004). However, the strong correlations between these variables and diversity gradients could simply reXect how environmental conditions inXuence spatial patterns of species distributions or

123

Mar Biol

how they constrain interactions between species that determine local coexistence. Present day conditions may thus not be always a good proxy to evaluate causalities of present-day patterns: there is a need to include geological timescale processes and evolutionary events shaping them (evolutionary macroecology). For instance, Rivadeneira (2005) made an exhaustive analysis of bivalve fossil record in Quaternary and Pliocene sites along the southeastern PaciWc, concluding that the comparative low bivalve species richness in the southeastern PaciWc, when contrasted with the northeastern PaciWc, seems to be related to southeastern PaciWc bivalve mass extinctions that occurred during the late Pleistocene, which devastated more than 75% of the species. Further, for the Magellan region (South of ca. 41°S), the bivalve faunal extinction would be part of an even older geological process associated with global cooling and glacial advances. Valdovinos et al. (2003) hypothesized that the southeastern PaciWc trend in mollusk diversity (increasing richness toward the poles) has been the result of higher diversiWcation of mollusks at higher latitudes, south of 42°S, linked to the use of discrete refugia due to the presence of large archipelagoes, fjords, and convoluted interconnected channels. Rivadeneira (2005) presented evidences that mass bivalve extinctions in the Peruvian zone would be associated with the establishment of a shallow oxygen minimum zone, when the Humboldt upwelling system reached its present state (see also Rivadeneira and Marquet 2007). None of these events have been reported for the Atlantic realm of South America. However, it has been shown that coral reef development enhanced molluscan diversity in the late Neogene and Quaternary of the southwestern Caribbean (Johnson et al. 2007), and that a previous regional faunistic enrichment took place owing to molluscan invasions from the Caloosahatchian province (North Carolina to Florida and the Yucatan peninsula) during Miocene to early Pleistocene (Vermeij 2005). Both phenomena may account for the Atlantic peak in species richness observed in this work. Therefore, it is our view that in analyzing present day diversity gradients, macroecological analysis must also take a closer look at geological and evolutionary processes, since the present-day bio-physical and ecological scenery may not be enough to give a comprehensive answer. This appears to be the case in our study case. In fact, the disparity of opinions, illustrated by several explanatory hypotheses, indeed responds to the results of focusing in diVerent ecological and evolutionary processes.

Conclusions We demonstrated that there is no single best explanatory cause for the latitudinal diversity gradients in South American

123

Mytilidae and Ostreidae, and that only a well carried simulation experiment would be able to distinguish the role of each variable. The location and magnitude of the peaks and the shape of the pattern varied between oceans, even when considering a similar number of closely related species. This reinforces the concept that the shape of latitudinal diversity gradient could be aVected by physical regional features, as the size of the province and shelf area, as well as for pass evolutionary process. The latter issue is the key to understanding the causality of present-day patterns. Acknowledgments Financial support to A. C. and O. D. by The Nature Conservancy, The Kabcenell Family Foundation and the project UTF/URU/025/URU (Uruguay) is acknowledged. T. F. L. V. B. R. is supported by a CAPES/Fulbright fellowship. J. C. C. acknowledges support from CASEB, FONDAP, Project 15001-0001. T. Blackburn and D. J. Currie are acknowledged for the valuable comments made on a previous version of this manuscript. Special thanks to L. Ortega (Dirección Nacional de Recursos Acuáticos) F. Scarabino (Museo Nacional de Historia Natural y Antropología, Uruguay), M. Lee and L. Prado (PontiWcia Universidad Católica, Chile) that provided invaluable bibliography. A. C. acknowledges Marina and Estela for encouragement and support.

References Acha EM, Mianzan HW, Guerrero RA, Favero M, Bava J (2004) Marine fronts at the continental shelves of austral South America: physical and ecological processes. J Mar Syst 44:83–105. doi:10.1016/j.jmarsys.2003.09.005 Algar AC, Kerr JT, Currie DJ (2007) A test of metabolic theory as the mechanism underlying broad-scale species-richness gradients. Glob Ecol Biogeogr 16:170–178. doi:10.1111/j.1466-8238.2006. 00275.x Astorga A, Fernández M, Boschi EE, Lagos N (2003) Two oceans, two taxa and one mode of development: latitudinal diversity patterns of South American crabs and test for possible causal processes. Ecol Lett 6:420–427. doi:10.1046/j.1461-0248.2003.00445.x Bellolio GC, Toledo A, Dupré E (1996) Larval development of Choromytilus chorus (Molina, 1782) reared in laboratory. Sci Mar 60:336–353 Borthagaray AI, Carranza A (2007) Mussels as ecosystem engineers: their contribution to species richness in a rocky littoral community. Acta Oecol 31:243–250. doi:10.1016/j.actao.2006.10.008 Brehm G, Colwell RK, Kluge J (2007) The role of environment and mid-domain eVect on moth species richness along a tropical elevational gradient. Glob Ecol Biogeogr 16:205–219 Broitman BR, Navarrete SA, Smith F, Gaines SD (2001) Geographic variation of southeastern PaciWc intertidal communities. Mar Ecol Prog Ser 224:21–34. doi:10.3354/meps224021 Burnham KP, Anderson DR (1998) Model selection and inference: a practical information-theoretic approach. Springer, New York, 353 pp Cardelús C, Colwell RK, Watkins JEJ (2006) Vascular epiphyte distribution patterns: explaining the mid-elevation richness peak. J Ecol 94:144–156. doi:10.1111/j.1365-2745.2005.01052.x Carranza A, Colwell RK, Rangel TFLVB (2008) Distribution of megabenthic gastropods along environmental gradients: the middomain eVect and beyond. Mar Ecol Prog Ser 367:193–202. doi:10.3354/meps07596 Clarke A, Gaston KJ (2006) Climate, energy and diversity. Proc R Soc Lond B Biol Sci 273:2257–2266. doi:10.1098/rspb.2006.3545

Mar Biol Coan EV, Scott PV, Bernard FR (2000) Bivalve seashells of Western North America: marine bivalve mollusks from Arctic Alaska to Baja California. Santa Barbara Museum of Natural History, Santa Barbara, p 764 Colwell RK (2006) RangeModel a Monte Carlo simulation tool for assessing geometric constraints on species richness. Version 5. User’s Guide and application published at: http://viceroy.eeb. uconn.edu/rangemodel Colwell RK, Rahbek C, Gotelli N (2004) The mid-domain eVect and species richness patterns: what have we learned so far? Am Nat 163:E1–E23. doi:10.1086/382056 Colwell RK, Rahbek C, Gotelli NJ (2005) The mid-domain eVect: there’s a baby in the bathwater. Am Nat 166:E149–E154. doi:10.1086/491689 Commito JA, Celano EA, Celico HJ, Como S, Johnson CP (2005) Mussels matter: postlarval dispersal dynamics altered by a spatially complex ecosystem engineer. J Exp Mar Biol Ecol 316:133–147. doi:10.1016/j.jembe.2004.10.010 Commito JA, Dow WE, Grupe BM (2006) Hierarchical spatial structure in soft-bottom mussel beds. J Exp Mar Biol Ecol 330:27–37. doi:10.1016/j.jembe.2005.12.015 CranWeld HJ, Michael KP (1989) Larvae of the incubatory oyster Tiostrea chilensis (Bivalvia: Ostreidae) in the plankton of central and southern New Zealand. N Z J Mar Freshw Res 23:51–60 Currie DJ (1991) Energy and large-scale patterns of animal and plantspecies richness. Am Nat 137:27–49. doi:10.1086/285144 Currie DJ, Mittelbach GG, Cornell HV, Field R, Guégan J-F, Hawkins BA, Kaufman DM (2004) Predictions and tests of climate-based hypotheses of broad-scale variation in taxonomic richness. Ecol Lett 7:1121–1134. doi:10.1111/j.1461-0248.2004.00671.x De Schweinitz EH, Lutz RA (1976) Larval development of the northern horse mussel, Modiolus modiolus (L.), including a comparison with the larvae of Mytilus edulis L. as an aid in planktonic identiWcation. Biol Bull 150:348–360. doi:10.2307/1540677 Diniz-Filho JAF, de Santána CER, de Souza MC, Rangel TF (2002) Null models and spatial patterns of species richness in South American birds of prey. Ecol Lett 5:47–55. doi:10.1046/j.14610248.2002.00289.x Floeter SR, Soares-Gomes A (1999) Biogeographic and species richness patterns of gastropoda on the Southwestern Atlantic. Rev Bras Biol 59:567–575. doi:10.1590/S0034-71081999000400006 Fortes RR, Absalão RS (2004) The applicability of Rapoport’s rule to the marine molluscs of the Americas. J Biogeogr 31:1909–1916. doi:10.1111/j.1365-2699.2004.01117.x Graham MH (2003) Confronting multicollinearity in ecological multiple regression. Ecology 84:2809–2815. doi:10.1890/02-3114 Gutiérrez JL, Jones CG, Strayer DL, Iribarne OO (2003) Mollusks as ecosystem engineers: the role of shell production in aquatic habitats. Oikos 101:79–90. doi:10.1034/j.1600-0706.2003.12322.x Harrison S, Cornell HV (2007) Introduction: merging evolutionary and ecological approaches to understanding geographic gradients in species richness. Am Nat 170:S1–S4. doi:10.1086/519011 Harrison S, Grace JB (2007) Biogeographic aYnity helps explain productivity-richness relationships at regional and local scales. Am Nat 170:S5–S15. doi:10.1086/519010 Hillebrand H (2004) On the generality of the latitudinal diversity gradient. Am Nat 163:192–211. doi:10.1086/381004 Jetz W, Rahbek C (2002) Geographic range size and determinants of avian species richness. Science 297:1548–1551. doi:10.1126/ science.1072779 Johnson KG, Todd JA, Jackson JBC (2007) Coral reef development drives molluscan diversity increase at local and regional scales in the late Neogene and Quaternary of the southwestern Caribbean. Paleobiology 33:24–52. doi:10.1666/06022.1 Jones CG, Lawton JH, Shachak M (1994) Organisms as ecosystem engineers. Oikos 69:373–386. doi:10.2307/3545850

Keen AM (1971) Sea shells of tropical west America: marine mollusks from Baja California to Peru, 2nd edn. Stanford University Press, Stanford, p 1064 Kennedy VS (1977) Reproduction in Mytilus edulis aoteanus and Aulacomya maoriana (Mollusca; Bivalvia) from Taylors Mistake, New Zealand. N Z J Mar Freshw Res 11:255–267 Lasiak T (1986) The Reproductive cycles of the intertidal bivalves Crassostrea cucullata (Born, 1778) and Perna Perna (Linnaeus, 1758) from Transkei coast, Southern Africa. Veliger 29:226–230 Linse K (1999) Mollusca of the Magellan region. A checklist of the species and their distribution. Sci Mar 63:399–407 Lough RG, Gonor JJ (1971) Early embryonic stages of Adula californiensis (Pelecypoda: Mytilidae) and the eVect of temperature and salinity on developmental rate. Mar Biol (Berl) 8:118–125. doi:10.1007/BF00350927 Millar RH, Hollis PJ (1963) Abbreviated pelagic life of Chilean and New Zealand oysters. Nature 197:512–513. doi:10.1038/ 197512b0 Mittelbach GG, Schemske DW, Cornell HV, Allen AP, Brown JM, Bush MB, Harrison SP, Hurlbert AH, Knowlton N, Lessios HA, McCain CM, McCune AR, McDade LA, McPeek MA, Near TJ, Price TD, Ricklefs RE, Roy K, Sax DF, Schluter D, Sobel JM, Turelli M (2007) Evolution and the latitudinal diversity gradient: speciation, extinction and biogeography. Ecol Lett 10:315–331. doi:10.1111/j.1461-0248.2007.01020.x Monteiro-Ribas W, Rocha-Miranda F, Romano RC, Quintanilha J (2006) Larval development of Brachidontes solisianus (Bivalvia, Mytilidae), with notes on diVerences between its hinge system and that of the mollusk Perna perna. Braz J Biol 66:109–116. doi:10.1590/S1519-69842006000100014 Navarrete SA, Broitman B, Wieters EA, Finke GR, Venegas RM, Sotomayor A (2002) Recruitment of intertidal invertebrates in the Southeast PaciWc: interannual variability and the 1997–1998 El Niño. Limnol Oceanogr 47:791–802 Paredes C, Huamán P, Cardoso F, Vivar R, Vera V (1991) Estado actual del conocimiento de los moluscos acuáticos en el Perú. Rev Peruana Biol 6:5–47 Prado L, Castilla JC (2006) The bioengineer Perumytilus purpuratus (Mollusca: Bivalvia) in central Chile: biodiversity, habitat structural complexity and environmental heterogeneity. J Mar Biol Assoc UK 86:417–421. doi:10.1017/S0025315406013282 Rangel TFLVB, Diniz-Filho JAF, Bini LM (2006) Towards an integrated computational tool for spatial analysis in macroecology and biogeography. Glob Ecol Biogeogr 15:321–327. doi:10.1111/ j.1466-822X.2006.00237.x Ricklefs RE (2007) History and diversity: explorations at the intersection of ecology and evolution. Am Nat 107:S56–S70. doi:10.1086/ 519402 Rios EC (1994) Seashells of Brazil, 2nd edn. Fundação Universidade de Rio Grande, Museu OceanográWco, Rio Grande do Sul, p 328 Rivadeneira M (2005) Macroecología evolutiva de los bivalvos marinos de la costa PaciWca de Sudamerica. PhD Thesis, Facultad de Ciencias Biológicas, Universidad Católica de Chile, Santiago Rivadeneira MM, Marquet PA (2007) Selective extinction of late Neogene bivalves on the temperate PaciWc coast of South America. Paleobiology 33:455–468. doi:10.1666/06042.1 Roy K, Jablonski D, Valentine JW, Rosenberg G (1998) Marine latitudinal diversity gradients: Tests of causal hypotheses. Proc Natl Acad Sci USA 95:3699–3721. doi:10.1073/pnas.95.7.3699 Roy K, Jablonski D, Valentine JW (2000) Dissecting latitudinal diversity gradients: functional groups and clades of marine bivalves. Proc R Soc Lond B 267:293–299. doi:10.1098/rspb.2000.0999 Scarabino F (2003) Lista sistemática de los Bivalvia marinos y estuarinos del Uruguay. Commun Soc Malacol Urug 8:229–259 Thiel ME, Macaya E, Acuña W, Arntz H, Bastias K, Brokordt P, Camus JC, Castilla LR, Castro M, Cortés CP, Dumont R,

123

Mar Biol Escribano M, Fernández DA, Lancellotti JA, Gajardo CF, Gaymer I, Gómez AE, González HE, González PA, Haye JE, Illanes JL, Iriarte G, Luna-Jorquera C, Luxoro PH, Manríquez V, Marín P, Muñoz SA, Navarrete E, Pérez E, Poulin J, Sellanes A, Sepúlveda W, Stotz F, Tala A, Thomas CA, Vargas JA, Váquez A, Vega A (2007) The Humboldt current system of Northern-central Chile. Oceanographic processes, ecological interactions and socioeconomic feedback. Oceanogr Mar Biol Annu Rev 45:195–344 Valdovinos C (1999) Biodiversidad de moluscos chilenos: base de datos taxonomica y distribucional. Gayana (Zool) 63:59–112

123

Valdovinos C, Navarrete SA, Marquet PA (2003) Mollusk species diversity in the Southeastern PaciWc: why are there more species towards the pole? Ecography 26:134–139. doi:10.1034/j.16000587.2003.03349.x Vermeij GJ (2005) One-way traYc in the western Atlantic: causes and consequences of Miocene to early Pleistocene molluscan invasions in Florida and the Caribbean. Paleobiology 31:624–642 Zelaya DG (2005) The bivalves from the Scotia Arc islands: species richness and faunistic aYnities. Sci Mar 69:113–122. doi:10.3989/ scimar.2005.69s2113