Climate, Critters, and Cetaceans: Cenozoic Drivers of the Evolution of Modern Whales

See allHide authors and affiliations

Science  19 Feb 2010:
Vol. 327, Issue 5968, pp. 993-996
DOI: 10.1126/science.1185581


Modern cetaceans, a poster child of evolution, play an important role in the ocean ecosystem as apex predators and nutrient distributors, as well as evolutionary “stepping stones” for the deep sea biota. Recent discussions on the impact of climate change and marine exploitation on current cetacean populations may benefit from insights into what factors have influenced cetacean diversity in the past. Previous studies suggested that the rise of diatoms as dominant marine primary producers and global temperature change were key factors in the evolution of modern whales. Based on a comprehensive diversity data set, we show that much of observed cetacean paleodiversity can indeed be explained by diatom diversity in conjunction with variations in climate as indicated by oxygen stable isotope records (δ18O).

Modern cetaceans (Neoceti), the mysticetes and odontocetes, show a number of mass-feeding adaptations beyond the immediate demands of an aquatic existence (1). Whereas mysticetes have become edentulous and rely on baleen to filter food from the water, odontocetes have evolved the ability to search for prey by means of echolocation. What unites these two different adaptive strategies is their effectiveness in terms of mass feeding: Whereas mysticetes obtain enormous amounts of small prey by filtering vast quantities of water, odontocetes may be able to use their biosonar to locate the vertically migrating layers of plankton with their associated grazers and predators known as deep scattering layers (1). To support such large and abundant apex predators, the ecosystems exploited by cetaceans must be extremely productive, and the energy captured by primary producers must be transmitted very efficiently through the food web (2). Compared with photosynthetic bacteria and nannoplankton, diatoms, the dominant marine producers of today, are relatively large organisms (3) and are thus likely to be at the base of a food web with relatively fewer intermediate consumers (1, 2). This shortening of the food web reduces the amount of trophic fractionation between the original photosynthetic event and the final consumption by apex predators, such as cetaceans, thus allowing the latter to forage more efficiently and grow larger, more abundant, and more diverse as a result (1).

We investigated whether the rise of diatoms to dominance may have triggered the radiation of neocetes (1) by fitting a set of a priori models (Table 1) to a comprehensive genus-level cetacean diversity data set (n = 204) downloaded from the Paleobiology Database (4), as well as by assessing the explanatory power of the different models using the second-order Akaike’s Information Criterion (AICc) and Akaike weights (wi) (5, 6). Apart from diatom and nannoplankton species diversity, which we downloaded from the Neptune database (7) (comparable dinoflagellate data were unavailable), our set of potential predictors also comprised oxygen isotope records (expressed as δ18O values), a proxy reflecting both temperature and global ice volume (8). These records allowed us to test for the effect of primary production (1) and climate change (9), respectively. In addition, we included two measures of global marine rock abundance to account for the potentially biasing effect of the variable amount of preserved sedimentary rock on our cetacean diversity data (10). Our first estimate consisted of the total number of fossiliferous marine formations; our second estimate was a subset of the former and included only those formations that have produced any vertebrate fossils, to account for potential preservational biases. Both were downloaded from the Paleobiology Database (4).

Table 1

Comparison of a number of a priori models attempting to explain cetacean paleodiversity based on the PaleoDB diversity data as sampled per bin. δ18O, oxygen isotope records used as proxy for climate change (8); diatom, diatom species diversity (Neptune database) (7); nanno, nannoplankton species diversity (Neptune database) (7); rock, total number of fossiliferous marine formations as downloaded from the PaleoDB (4). All models also included geologic stage duration as a nonoptional predictor to account for the potentially biasing effect of unequal Cenozoic stage durations. The R2 shown is the generalized R2 proposed by (26) and (27). The preferred models are shown in bold.

View this table:

In addition to the sampled-in-bin cetacean diversity data, we also included a biologically adjusted diversity estimate in our analysis to investigate whether this would result in a strengthening of any observed relationships between the diversity and environmental data. For this adjusted estimate, taxa were ranged through time bins in which they had not actually been sampled, provided that the taxon in question had already been recorded in at least one earlier and one later bin. We accounted for variations in stage length by including stage duration as a nonoptional predictor in all our models. Furthermore, we accounted for non-normality and nonconstant variances, which are often associated with count data like ours, by square-root-transforming cetacean diversity in all analyses, and for temporal autocorrelation by fitting autoregressive models to our original data and using generalized least squares where appropriate (4). We ran two sets of models based on our two estimates of rock abundance. However, because the results of both sets were similar, we decided to focus on the total number of marine formations [see (4) for further details].

Out of our models, the combination of diatom species diversity and δ18O values was most strongly associated with cetacean diversity (Fig. 1 and Table 1). This was true for the sampled-in-bin data both for the whole of Neoceti and for mysticetes and odontocetes viewed separately, with mysticete diversity being particularly well predicted by this model (Tables 1 and 2). By contrast, we could not clearly distinguish between the model including diatom diversity and δ18O only and the model including diatom diversity, δ18O, and rock abundance as far as the ranged-through data were concerned, the latter model having an Akaike weight of nearly 0.1. Because the diatom and δ18O model is nested within the latter, we were able to test whether the more parameter-rich model explained our data significantly better by using a likelihood ratio test. This test showed a significant improvement of the model also including rock abundance over the model including just diatom diversity and δ18O for the neocete ranged-through data (Χ2 = 6.50, P = 0.011) but not for the sampled-in-bin data (Χ2 = 0.02, P = 0.879). Interestingly, in all of the favored models the link between phytoplankton and cetacean diversity was restricted to diatoms only, as nannoplankton failed to explain much of the variance in the cetacean data both on its own and in combination with diatoms.

Fig. 1

Comparison of neocete (A), mysticete (B), and odontocete (C) paleodiversity with diatom paleodiversity (D) and global δ18O values (E). Cetacean diversity is shown as sampled-in-bin data as downloaded from the Paleobiology Database (gray) and as a ranged-through estimate (black). Error for the δ18O curve is shown as mean standard error (SE) multiplied by 100. The picture of a diatom is after Haeckel (28).

Table 2

Estimated best-fit model parameters for the neocete, mysticete, and odontocete data sets. δ18O, oxygen isotope records used as proxy for climate change (8); diatom, diatom species diversity (Neptune database) (7); rock, total number of fossiliferous marine formations as downloaded from the Paleobiology Database (4); st. dur., geologic stage duration; the latter was included as a nonoptional predictor in all models to account for the potentially biasing effects of unequal Cenozoic stage durations.

View this table:

The link between diatom diversity and observed cetacean diversity supports the hypothesis that diatom-based primary production has been an important driver of neocete evolution (1). Similarly, the observation that climate change also has a role to play is not surprising in light of recent research that has demonstrated substantial temperature-dependent variations in the diversity of extant cetaceans (9). Finally, the observation that the model including diatom diversity and δ18O seems to explain mysticete diversity relatively better than odontocete diversity is reasonable, considering the relatively lower trophic levels (copepods, krill, and small fish) that many mysticetes feed on (11). Because these results depend on our choice of environmental and biological variables used as predictors here, they could substantially change if other or additional variables were used. Nonetheless, our study offers an entirely biological explanation of cetacean diversity, which, if correct, would seem to imply that the abundance of fossiliferous rock, previously proposed to affect or even overwhelm any biological signal in paleodiversity data sets (10, 1214), does not exert a major bias on cetacean paleodiversity.

One peculiar observation is the curious increase in explanatory power of rock abundance in combination with diatom diversity and δ18O when the ranged-through data were used. Aside from the possibility of this being a coincidence, such an increase in fit in the face of a crude (fig. S1) but inherently biologically motivated correction of paleodiversity seems counterintuitive. One potential explanation might be found in the way the effects of rock abundance are commonly interpreted. Although often treated as a simple bias (10, 12), it has also been proposed that both rock abundance and diversity may be influenced by a common third factor, such as sea-level change (15), thus potentially making rock abundance a covariate, rather than a determinant of observed cetacean diversity. Assuming that enough rock has been preserved per stage to overcome an initial and inevitable small-scale link between rock abundance and the number of fossil cetacean taxa preserved, it might thus be expected that in the presence of presumably genuine drivers of diversity, such as plankton abundance or climate change, rock abundance should explain none or hardly any of the diversity patterns observed. However, the amount of preserved sedimentary rock may be influenced by a complex interplay of factors, such as changes in sea-level or climate, influencing both deposition and erosion, all of which may also have an effect on cetacean diversity to various degrees. Assuming the ranged-through estimate to be a better reflection of actual cetacean paleodiversity than the data as sampled per bin, it might thus be possible that the increase in the explanatory power of rock abundance might reflect the effects of one or more genuine common-cause drivers of both cetacean evolution and rock abundance not represented in our models. This view may be supported by the negative coefficient for rock abundance in our ranged-through model (Table 2), which implies that an increase in rock abundance is linked with a decrease in neocete diversity—an observation clearly inconsistent with the interpretation of the latter as a simple bias, at least in the case of cetacean paleodiversity.

One prominent hypothesis regarding the evolution of neocetes is that the onset of the Antarctic Circumpolar Current (ACC) may have triggered the radiation of modern whales by greatly increasing the availability of nutrients in the upper layers of the sea through deep mixing in the Southern Ocean (1, 1619). Indeed, the effects of the ACC provide the Southern Ocean with one of the highest surface concentrations of silica, the major component of diatom frustules, anywhere in the world (20, 21). Although most of this silica is used up by local diatom growth, the water leaving the area in the form of Subantarctic Mode Water still supplies high concentrations of other nutrients, such as nitrate, to the world’s oceans, supporting as much as 75% of global export production north of 30°S in the process (20). Together with local sources of silica, particularly in the North Pacific (20), the ACC thus seems to offer a credible mechanism supporting the high rates of biological production needed to sustain large apex predators such as cetaceans. Although paleontological evidence suggests that neocetes appeared and possibly started to radiate in the latest Eocene close to the Eocene-Oligocene boundary (22, 23), the time of the actual establishment of the ACC is still a matter of debate (24, 25), and thus no firm conclusion can be drawn. However, our results imply that, if the onset of the ACC indeed triggered the evolution and diversification of neocetes, it likely must have done so through a great increase in diatom-based productivity, possibly by increasing the bioavailability of silica and other nutrients in the Southern Ocean and coastal upwelling zones around the world through deep-mixing occurring around Antarctica (1).

Supporting Online Material

Materials and Methods

Fig. S1

Tables S1 to S3


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank three referees for their comments, as well as G. Hunt for advice and help with R programming and M. Katz for the provision of data. L. Kavalieris, I. Jolliffe, M. Carrano, D. Thomas, M. Benton, and the participants of the August 2009 Encyclopedia of Life synthesis meeting on marine tetrapod evolution at the Field Museum, Chicago, provided advice and helpful discussions about earlier versions of this work and related topics. F.M. was supported by a University of Otago Postgraduate Scholarship. This is Paleobiology Database publication 108.
View Abstract

Navigate This Article