Introduction

The formation of the Caribbean Sea coupled with that of the Panamanian Isthmus during the latest part of the Cenozoic Era resulted in one of the great natural experiments in evolution. The ‘Great American Biotic Interchange’ that took place in the continental communities has been largely addressed by studies of vertebrate animals, most specially mammals (Webb, 2006; Woodburne, 2010). Our understanding of the changes in marine environments is based on investigations of diverse invertebrate groups. Dynamic processes of the ocean currents, physical and chemical water mass conditions, and geographic configurations including Caribbean isolation along the geochronological sequence, produced direct influence over the Neogene Amphi-American marine communities (Corals: Johnson, Sánchez-Villagra, & Aguilera, 2009; Budd, Stemann, & Johnson, 1994, 1996; crustacean decapods: Schweitzer & Iturralde-Vinent, 2005; Collins, Portell, & Donovan, 2009; Aguilera, Rodrigues de Aguilera, Vega, & Sánchez-Villagra, 2010; echinoids: Mihaljević et al., 2010; mollusks: Jackson, Jung, Coates, & Collins, 1993; bryozoans: Cheetham & Jackson, 1996; Cheetham, Jackson, & Sanner, 2001; O’Dea, Herrera-Cubilla, Fortunato, & Jackson, 2004; O’Dea et al., 2007 O’Dea & Jackson, 2009). These faunal changes are characterized by extinction and turnover in Miocene–Pliocene times.

We address the paleobiogeography in this region and geological time using a rich and complex new database of fossil fish taxa, thus investigating for the first time the marine realm with the most diverse group of vertebrates. The data include almost 100 ‘families’, 236 genera, and 346 species, from 11 Neogene basins in Tropical America. We aim to produce the first qualitative and quantitative analysis of marine fish diversity of the past Caribbean biogeography.

We adopt the term “Gatunian Province” (Landau, Vermeij, & Marques da Silva, 2008) for the geographical area (proto-Caribbean) and the geochronological sequence (Miocene–Pliocene) of interest. This region presents a rich basis for investigation considering the geographic displacement, emersions, and submersions of sedimentary basins, transgression and regression of the sea level, changes in the oceanographic conditions, and species extinction and turnover. There is an ongoing debate over the geographic extension and geochronologic subprovinces involved (Woodring, 1974; Petuch, 1982, 1988, 2004; Landini, Carnevale, & Sorbini, 2002a; Landini et al., 2002b; Porta, 2003; Landau et al., 2008; Harnik, Jablonski, Krug, & Valentine, 2010; Távora, Rodrigues dos Santos, & Neto, 2010). Multiple tectonic and geographical variables were considered in the early model of Caribbean biogeography developed by Rosen (1975) based on living fauna and considering monophyletic clades. Rosen (1975) found for example the influence of the Pacific faunas on the Caribbean and mentioned a ‘track Pacific–Atlantic’. Although not comparable in extent or approach, our work on monophyletic clades of fish in geological time, represents an independent examination of a similar region to that examined by Rosen (1975).

The tectonic context

The geographic configuration of Tropical America during the Cenozoic is the result of the Caribbean plate tectonic interaction with the North American, South American, Nazca, and Cocos plates (e.g. Wadge & Bunke, 1983; Pindell & Barrett, 1990; Pindell et al., 2005; Bachmann, 2001; Iturralde-Vinent, 2004–2005). The tectonic dynamic involved the subduction of western Central America and the eastern Lesser Antilles, the displacement of the Greater and Lesser Antilles and the volcanic island arc coupled with the uplift of the Panama Isthmus (Wadge & Bunke, 1983; Coates & Obando, 1996; Hoernle et al., 2002). These events resulted in the faunal isolation of the Western Central Atlantic and the Eastern Central Pacific at about 3.5 Ma, and the displacement of the multiple sedimentary basins to different geographic locations through time. Prior to the formation of the Panamanian barrier, the trans-isthmus passage was an open seaway permitting a strong inflow of the equatorial Pacific current into the proto-Caribbean, characterized by rich nutrient concentrations to produce high primary productivity (Schneider & Schmittner, 2006; Newkirk & Martin, 2009). The superficial circumtropical current went in western direction through the Panamanian seaway, towards the Pacific Ocean (Iturralde-Vinent & MacPhee, 1999; Schweitzer, Iturralde-Vinent, Hetler, & Velez-Juarbe, 2006).

Methods

Data were obtained from sampling during field trips to Ecuador, Costa Rica, Panama, and Venezuela under the scope of the Panama Paleontology Project (PPP), with identifications and leadership in this portion of the work by the senior author and in co-operation with Werner Schwarzhans (Hamburg) in respect to joint publications in preparation. The local geology and stratigraphical reference follow Coates, Collins, Aubry, and Berggren (2004); (Coates, McNeill, Aubry, Berggren, and Collins, 2005), and Quiróz and Jaramillo (2010). The field trip to Brazil was under the scope of the Museu Emilio Goeldi Paleontology Project; the geological and stratigraphical references follow Rossetti and Góes (2004). Specimens from Trinidad and Dominican Republic were studied based on additional PPP samples and collections at the Museum of Natural History of Belgium, Museum of Natural History of Basel, and Smithsonian Natural History Museum. Data from Cuba, Puerto Rico, Grenadines, Tobago, and Jamaica as well as complementary information from the other countries were taken from published literature (Table 1). Our data were standardized at the generic level, as is common in this kind of study (Foote & Miller, 2007). This procedure served also to diminish distortions caused by preservation and sampling biases, endemic groups and nomenclatural uncertainties.

Table 1 References of fossil fish records by authors, formations, and age for each country-basin
Full size table

We grouped basins according to the countries studied, which are deemed good representatives of the Miocene–Pliocene sequence forming the proto-Caribbean Gatunian Province in terms of geographical, paleoenvironmental and geochronological sequence (Table 1). Biases are sometimes attributed to few samples of taxa in certain basins (e.g., Barbados Island) or the absence (lack of preservation or study) of teleosteans or elasmobranchs (e.g., Brazil and Dominican Republic, respectively).

Merging of different time periods for the analysis is justified as there is a well-defined and restricted geological time window represented by the data. Once more precise stratigraphical and temporal information on existing, as well as on new sites becomes available, a new analysis examining changing across the time window examined will be possible.

The binary data matrix includes 236 genera of fossil fishes from nine country-basins; these were analyzed using jointly ordination and classification techniques. Initially, the classification was accomplished using a non-hierarchical K-means algorithm that gathers similar variables in a pre-established number of groups (Legendre & Legendre, 1998). Such groups are geometrically as compact as possible around their respective centroids. The method is particularly efficient when classifying widely distributed variables with many null values without the need for establishing a relative scale of dissimilarity (Legendre, Ellingsen, Bjornbom, & Casgrain, 2002). The best number of groups present in a data set was defined by maximum value of the C–H pseudo-F-statistics (Calinski & Harabasz, 1974) for different partitions (number of groups), where:

$$ CH = \left[ {\frac{{R^{2} }}{{\left( {K - 1} \right)}}} \right] \cdot \left[ {\frac{{\left( {n - k} \right)}}{{\left( {1 - R^{2} } \right)}}} \right] $$

where,

$$ R^{2} = \frac{{\left( {{\text{SST}} - {\text{SSE}}} \right)}}{\text{SST}}, $$

K number of groups and n = 203, the number of genera.

SST is the total sum of squared distances to the overall centroid and SSE is the sum of squared distances of the objects to their group’s own centroids. The classification was accomplished using the free software K-MEANS2 (Legendre, 2001). In a simulation study involving many stopping rules for cluster analysis, Milligan and Cooper (1985) found that the Calinski–Harabasz criterion was the one that would best recover the correct number of groups.

The principal components analysis (PCA) uses the calculated covariance and variance between localities, for this reason this is a Q-PCA sensu Fasham (1977). The Q-PCA analysis was run using the free software PAST (Hammer, Harper, & Ryan, 2001), and the K-mean for the groups was expressed in the PCAs planes using the concentration ellipse level of 80%. Following Jackson, Somers, and Harvey (1989) the size effect of the linear relationship between the PCA axis 1 and the relative frequency of species occurrence was established with the largest occurrence.

The confidence ellipses of genera groups were formed by computing the three following procedures using PAST software (Hammer et al., 2001) as proposed by Abdi, Dunlop, & Willians (2009): (1) the coordinates of the center of the ellipse of each group in the factorial plane (Axis II and Axis III), (2) the angle between the major axis of the ellipse and the first dimension of the plane, and (3) the relative size of the minor axis compared to the major axis of the ellipse. To perform these procedures the center of the ellipse must be set at the center of mass of the points. Thereafter, the ratio between the minor axis and the major axis is calculated as done in the ratio of the second to the first eigenvalue. Finally, the angle of rotation is given by the first eigenvector. For a 70% confidence interval, the length of the axes of the ellipse is set to ensure that the ellipse comprehends 70% of the points. It is also important to note that the value of 70% was chosen for visual clarification as to avoid overlapping the ellipses.

Results and discussion

The Miocene–Pliocene fish faunal assemblage established in the Tropical American seaway before the uplift of the Panamanian isthmus is called here the proto-Caribbean fauna, because the Caribbean Sea proper was only formed after the Pacific and Atlantic isolation was complete.

In our assesment the proto-Caribbean fauna is arranged in four different paleobiogeographic patterns of countries/basins and genera assemblages following specific regional paleoenvironments. In terms of faunal composition, age and geographic distribution these four paleobiogeographic patterns can be classified as subprovinces of the Neogene Gatunian bioprovince.

The isolated vector of Venezuela (V) exhibits a pattern of co-occurrence and singularity of fossil fish genera indicated in Fig. 1a (Group 1). Fig. 1b (Group 4) shows Venezuela characterized by a larger frequency of genera of the proto-Caribbean.

Fig. 1
figure 1

Principal components analysis arranged by genera and country-basin (B Brazil, C Cuba, CR Costa Rica, D Dominican Republic, E Ecuador, J Jamaica, P Panamá, T Trinidad, V Venezuela), and ovoid area of fossil fish assemblage (group species 1, 2, 3 and 4). a Axis II versus axis III. b Axis I versus axis II

Full size image

Since the Neogene until today, the marine sediment basins of Venezuela represent a faunal subprovince with unique features, which distinguish it from the rest of the Caribbean in diversity of ecosystems represented (estuaries, coastal lagoons, mangrove, rocky bottom, muddy bottom, corals and sea-grasses area, sandy beaches, shallow waters and depth waters, upwelling coastal process, bays and oceanic islands). A similar conclusion was reached by Landau et al. (2008) in their revision of the paleobiogeography of the mollusks of the South Caribbean. Based on their study Landau et al. (2008) recognized a Colombia–Venezuela–Trinidad subprovince. Our analysis though shows large qualitative and quantitative differences between Venezuela and Trinidad (T). An example of the unique composition in the Venezuelan subprovince is the Lower Miocene assemblage of the Cantaure Formation (Nolf & Aguilera, 1998) and that of the Upper Miocene to Lower Pliocene Cubagua Formation (Aguilera & Rodrigues de Aguilera, 2001). The first represents a shallow water environment, and the Cubagua Formation an environment of intense coastal upwelling. The comparisons with Trinidad could be expanded in the future with the study of the faunas from the Springvale Formation (Landau et al., 2009). As for Colombia, the available information on Neogene fossil fish is still very preliminary and does not allow conclusions in respect to the suggestion by Landau et al. (2008).

The second subprovince identified by analysis of the fossil fish faunas is the arc of western orientated basins composed of Jamaica (J), Dominican Republic (D) and Trinidad (T) islands. Jamaica and Hispaniola (Haiti and Dominican Republic) moved eastwards along the Eastern Fault in the northern limit of the Caribbean plate during Cenozoic time (Wadge & Bunke, 1983, Pindell et al., 2005). Structural elements affecting Trinidad include a complex mix of contraction, extension, and strike-slip. In the early Cenozoic, deformation and localized uplift of some elements of the passive margin began in the Trinidad region. Through the Oligocene, the leading edge of the Caribbean plate advanced eastward imposing a transpressional deformation onto the passive margin of eastern Venezuela and Trinidad. In the late Miocene, Caribbean-South American relative motion became fundamentally translational (Lingrey, 2007). This subprovince is defined here as the Antillean subprovince and it corresponds in the analysis to the co-occurrence and the exclusivity of the group expressed in Fig. 1a (Group 3); the mean of the species frequency is expressed in Fig. 1b (Group 2).

Given that the angle resulting from the PCA is larger than 45° between Trinidad-Dominican Republic-Jamaica in relation to Panamá-Ecuador, they should best be allocated into two separate subprovinces.

The third subprovince is thus composed of Panama and Ecuador (Fig. 1a; Group 2 and overlap with Group 3). Ecuador has a singular faunal group restricted to the less genera frequency of occurrence pattern (Fig. 1b; Group 3). This pattern reflects the Pacific faunal influence into the proto-Caribbean, because Ecuador, and to some extent also Panama, show species diversity essentially typical for the East Pacific, particularly as to the benthopelagic fauna.

Previous research on Ecuador Pliocene benthonic foraminifera, mollusks, and teleostean fishes (Landini et al., 2002a, b), suggested association within a Panamanian subprovince for this region. Our results for the Miocene–Pliocene are in agreement with the proposition of Landini et al. (2002a, b), extending thus the geochronological range of the subprovince into earlier times equivalent to the proto-Caribbean.

Based on the functional structure of bryozoans and mollusks, O’Dea et al. (2006) hypothesized geochronological abiotic and biotic factors affecting the invertebrate community as consequence of the uplift of the Panamanian isthmus. The change in the water current patterns and the collapse of primary productivity did not lead to synchronous evolutionary responses with immediate extinction and turnover, but rather exhibited a time lag of 2–1 Ma. In contrast to this Pleistocene effect, Johnson et al. (2009) demonstrated that during the Oligocene–Miocene transition a major change occurred in the coral community of the proto-Caribbean with almost 50% of diversity disappearance in the lower Miocene.

The fourth subprovince is here named Costarican and is composed only of Costa Rica (CR) showing the co-occurrence pattern and the singular faunal group expressed in Fig. 1a (Group 4) characterized in the Fig. 1b by high frequency of genera occurrence in this region (Group 2). The absence of fossil fish records in adjacent Nicaragua and Honduras produces regional gaps, to be considered in future studies.

The short vectors of Brazil (B) and Cuba (C) shown by our analysis and the isolated data from Puerto Rico, Tobago, and Grenadine (Table 1) represent preliminary data. However, the associated vertebrate fauna from the lower-middle Miocene Cojimar Formation from Cuba (Iturralde-Vinent, Hubbell, & Rojas, 1996) and the lower Miocene Pirabas Formation from Brazil (Ramos, Santos, Costa, & Toledo, 2009), include similar remains of crocodiles, dugongs (Sirenia), turtles and land mammals (e.g., a rodent). Rosen (1975) had already suggested a close biogeographic link between Cuba and Brazil based on the occurrence of the swamp eel Ophisternon.

The fossil sharks and rays’ records from Brazil, Cuba, Puerto Rico, Tobago, and Grenadine are few demersal elasmobranchs widely distributed during the Neogene.

Figure 1a shows the spatial distribution of the genera complexes in four ovoid areas that explain the 70% of the probability of occurrence. The first group is formed by 79 genera and represents the more diverse assemblage in terms of taxa and paleoenvironments (Table 2). This group is characterized mostly by the presence of shallow water demersal fishes, associated with bathypelagic and benthopelagic taxa that reflect productive and environmental stability in this zone. The second group formed by 39 genera is characterized by the presence of shallow water fishes and pelagic planktivores from the neritic zone. The third group formed by 61 genera is characterized by the presence of a large assemblage of fishes that inhabit shallow waters over mud bottom, some of them being associated with brackish environments. This group of fish is indicative for inner platform slope environments and some nocturnal migratory planktivore bathypelagic and carnivore benthopelagic taxa were present. The fourth group is composed of 20 genera, most of them shallow water sharks and rays. Large extinct lamnid sharks, medium sized hemigaleid and carcharinid sharks are present in this group, all of which were widely distributed circumtropically.

Table 2 Functional groups (1–4) organized by genera, trophic preference and habitat
Full size table

The qualitative analysis of the faunal composition based on a comprehensive species list, though preliminary in many aspects (Table 3), shows extinct genera (e.g. Megaselachus megalodon), extinct species (e.g. Hemipristis serra), and an extremely high turnover rate of species. The list of fossil fishes presented here is the most comprehensive overview and compilation of the Neogene Tropical faunal community, with shallow water (e.g. Mugil), brackish (e.g. Aspistor), demersal (e.g. Lepophidium), bathypelagic (e.g. Diaphus), benthopelagic (e.g. Heptranchias), and pelagic (e.g. Mobula) faunal elements. Benthic fish representatives exhibit preference for mud bottom (e.g. Ariosoma) or sandy bottom (e.g. Stellifer). Paleodepth indicators show a range from near shore (e.g. Equetus: 0–50 m) to continental slope (Maurolichus: 300–500 m depth), and some are known for their diurnal vertical migration (e.g. Lampadena).

Table 3 Fossil marine fish species occurrence by country-basin
Full size table

Summary and conclusions

The fish fossil record in the proto-Caribbean presents identifiable patterns about the pre-isthmus paleobiogeography, denoting the influence of the Pacific current into the Amphi-American fish distribution in the Gatunian Province. We identified the Venezuelan, Antillean, Panamanian, and Costarican subprovinces based on a multivariate analysis of genera. These subprovinces are related to the regional geology and its tectonic dynamics coupled with the oceanographic conditions, paleodepth, and paleoenvironment.

The Caribbean geographic isolation formed as consequence of the Panamanian Isthmus and led to the Plio-Pleistocene extinction and high turnover of the species composition best documented until now in the invertebrate benthic fauna (Aguilera et al., 2010; Budd et al., 1994, 1996; Cheetham & Jackson, 1996; Cheetham et al., 2001; Collins et al., 2009; Jackson et al., 1993, 2009; Mihaljević et al., 2010; O’Dea et al., 2004, 2007; O’Dea & Jackson, 2009; Schweitzer & Iturralde-Vinent, 2005). However, the nektonic fish fauna reflects the impact of the proto-Caribbean fish fauna at the generic level in local paleoenvironment community changes in neritic, estuarine and deep water assemblages. The overall diversity remains largely the same, except for a few lamnids, such as the giant-toothed white shark and the wide-toothed mako shark.

The concept of a Gatunian Province is a valid concept in terms of the geographical and geochronological relationship with the proto-Caribbean paleofauna of invertebrate groups studied so far. In addition, we propose the Venezuelan, Antillean, Panamanian, and Costarican subprovinces as clearly recognizable assemblages of marine fishes (Fig. 2).

Fig. 2
figure 2

The proto-Caribbean subprovinces recognized in this work (map modified from Google map). Question marks denote unknown fossil fish record and star symbols denote sedimentary deposits with potential for fossil fish occurrences

Full size image