Far more species of organisms are found in the tropics than in temperate and polar regions, but the evolutionary and ecological causes of this pattern remain controversial1,2. Tropical marine fish communities are much more diverse than cold-water fish communities found at higher latitudes3,4, and several explanations for this latitudinal diversity gradient propose that warm reef environments serve as evolutionary ‘hotspots’ for species formation5,6,7,8. Here we test the relationship between latitude, species richness and speciation rate across marine fishes. We assembled a time-calibrated phylogeny of all ray-finned fishes (31,526 tips, of which 11,638 had genetic data) and used this framework to describe the spatial dynamics of speciation in the marine realm. We show that the fastest rates of speciation occur in species-poor regions outside the tropics, and that high-latitude fish lineages form new species at much faster rates than their tropical counterparts. High rates of speciation occur in geographical regions that are characterized by low surface temperatures and high endemism. Our results reject a broad class of mechanisms under which the tropics serve as an evolutionary cradle for marine fish diversity and raise new questions about why the coldest oceans on Earth are present-day hotspots of species formation.
Access optionsAccess options
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank M. Grundler for statistical and coding advice, and M. Venzon and A. Noonan for assistance with dataset assembly. We are grateful to the many institutions that curate the primary biodiversity data that underlie several of our analyses (see Supplementary Table 6). This research was carried out using computational resources and services provided by Advanced Research Computing at the University of Michigan, Ann Arbor. This work was supported in part by NSF grant DEB-1256330 (D.L.R.), an NSF DDIG grant to J.C. (DEB-1601830), an Encyclopedia of Life Rubenstein Fellowship to J.C. (EOL-33066-13) and by a Fellowship from the David and Lucile Packard Foundation (D.L.R.). P.F.C. was funded by a Gaylord Donnelley Postdoctoral Environment Fellowship (Yale) and through the ARC Centre of Excellence for Coral Reef Studies. We thank J. Johnson for creating the fish images in Fig. 3 and Extended Data Fig. 7.Reviewer information
Nature thanks O. Bininda-Emonds, O. Seehausen and the other anonymous reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
Major lineages are broken into subclades (top) to visualize fossil calibrations and are coloured by taxonomic order. Numbered nodes are described in the calibration report in the Dryad data repository. The same calibrations are red circles in the full phylogeny (bottom). A + E + L + P: Acipenseriformes, Elopiformes, Lepisosteiformes, Polypteriformes; A + E + S: Argentiniformes, Esociformes, Salmoniformes; B + H: Beryciformes, Holocentriformes; C + S + P: Centrarchiformes, Scombriformes, Perciformes; C + U: Chaetodontiformes, Uranoscopiformes; G + G: Gonorynchiformes, Gymnotiformes; G + O + S: Galaxiiformes, Osmeriformes, Stomiatiformes; P + Z: Percopsiformes, Zeiformes.
Extended Data Fig. 2 Relationships between mean speciation rates and latitude for 262 marine ecoregions using alternative methods for the computation of the cell rates.
a–c, λBAMM versus latitude. d–f, λDR versus latitude. Ecoregion rates are mean rates across all cells assigned to each biogeographical region. Arithmetic mean is the mean rate across all taxa inferred to occur in the cell; weighted arithmetic and weighted geometric means assign proportionately greater weight to species with small geographical ranges. Weighting schemes for speciation metrics are described in the Methods. g, Simultaneous autoregressive (SAR) spatial error models for the effects of absolute latitude on mean speciation rates for ecoregions. AIC1 gives the Akaike information criterion (AIC) for a linear model with a single slope and intercept term; AIC2 is the corresponding AIC for a breakpoint model that assumes no relationship (slope = 0) between absolute latitude and speciation rate for all values below some threshold, and a linear relationship for latitudes that exceed the threshold. SAR.I and SAR.IP are global Moran’s I estimates and associated P values for assessing the presence of residual spatial autocorrelation in the model residuals; OLS.I and OLS.IP are the corresponding values for ordinary least squares (OLS) regression that ignores spatial autocorrelation. All SAR models show highly significant effects of latitude on speciation rate, and breakpoint models provided a consistently better fit than models without a breakpoint. h, OLS and SAR models for the effects of absolute latitude on speciation rate for low-latitude grid ecoregions only. The slope ratio term gives the ratio of slopes for low-latitude ecoregions (below the corresponding breakpoint; g) to the slope for ecoregions with latitude above the breakpoint. Overall, there is a marginal effect of latitude on speciation rate for low-latitude ecoregions.
Extended Data Fig. 3 Relationships between speciation rate and latitude for alternative speciation rate metrics and for endemic taxa only.
a, b, Global maps of λDR and λBAMM-TC, as in Fig. 1. c, Global map of endemic species richness, by grid cell. ‘Endemic’ taxa are those that are restricted to a single MEOW realm; an endemic taxon can occur in multiple grid cells provided all grid cells are contained within a single realm. d, Relationship between speciation rates (λDR) and latitude for ecoregions (n = 232), computed using realm endemics only. e, Relationship between speciation rates (λBAMM-TC) and latitude for ecoregions, computed using realm endemics only. f, SAR spatial error models for the relationship between ecoregion speciation rates and absolute latitude, for which ecoregion means are computed from single-realm endemics only. Weighting schemes for assemblages are described in the Methods. SAR modelling results are presented as in Extended Data Fig. 2g and show a strong correlation between latitude and speciation rate.
a, Negative relationship between species richness and mean speciation rate (λDR) for individual grid cells. b, Negative relationship between mean annual sea surface temperature and mean speciation rate. c, d, Same as a, b, but for BAMM with time-constant rate regimes (λBAMM-TC). Grid cells as in Fig. 1 (n = 16,150). See Fig. 2 for comparison. e, Correlation between mean speciation rate for MEOW biogeographical provinces and two measurements of regional endemism. ‘Occupancy (provinces)’ measures endemism as the inverse of the mean number of provinces occupied by each taxon that occurs in a particular province. ‘Range size’ is the inverse mean range size across all taxa occurring in a given province. High values of endemism indicate that a given region consists of species that are found in fewer additional provinces, or of species with smaller geographical ranges. The bottom two rows show the correlations between the endemism parameters and latitude.
a, λDR for all marine species with genetic data (n = 5,229) as a function of the latitudinal (centroid) midpoint of their geographical range. Non-phylogenetic OLS regression with quadratic term is overlaid on points to denote trend in mean rates. b, λBAMM for the same taxon set. c, Sliding window analysis of λDR distributional quantiles in speciation rates by individual taxa with respect to latitudinal midpoint. Contours denote quantiles from 0.10 to 0.90, in 0.10 increments, with a sliding window size of 6°. Dark red line is the median rate. d, Distributional quantiles of λBAMM for all species with respect to latitudinal midpoint; dark red line is median rate.
a, Mean speciation rates for taxa from low latitudes (<30°), intermediate latitudes (30–60°) and high latitudes (>60°) computed using the interval method. Per-taxon interval-based rates were computed for time intervals between 0.25 and 50 million years before present. Time-averaged speciation rates for high-latitude fishes are much higher than those inferred for low-latitude fishes, even across timescales that exceed 20 million years. b, Rate differential between high-latitude and low-latitude taxa as a function of interval duration. c, Speciation-over-time curves reconstructed using the time-varying rates model in BAMM for 14 clades of high-latitude (blue) and low-latitude (red) fishes. Inset numbers for each panel give the numbers of low-, intermediate- and high-latitude (from left to right) taxa from each clade for which geographical range data are available. Low-latitude clades were selected to represent high-diversity and iconic reef-associated clades that contribute substantially to the tropical diversity peak in marine fishes. With the possible exception of gobies, there is no signal of early, rapid speciation in low-latitude or tropical shallow-water clades.
Extended Data Figure 7 Speciation rates in deep-sea fishes and the phylogenetic structure of high-latitude fish diversity.
a, Formal test of the relationship between speciation rate and depth classification for tropical fishes. ‘Classification’ is the criterion used to define fishes as deep sea versus shallow water; mean depth (200 m) thus classifies all fishes with mean depth greater than 200 m as deep sea. Among tropical fishes, there is no effect of depth state on speciation rate. b, Phylogenetic composition of high-latitude fish diversity by taxonomic order, across all marine fishes (top) and for the subset of species with genetic data (bottom). High latitude is defined as having a centroid midpoint greater than 45° north or south. Only the three most species-rich high-latitude orders are labelled. Most high-latitude marine fishes are Perciformes. c, Phylogenetic and geographical structure of the diversity of Perciformes. The latitudinal range of each perciform species in the phylogenetic dataset is shown, along with the corresponding speciation rate (λBAMM). Latitudinal ranges from species with speciation rates that are faster and slower than the median rate are shown in red and blue, respectively. High-latitude and rapidly speciating clades are nested within slowly speciating tropical lineages, and speciation rates for high-latitude taxa of Perciformes are higher than those observed in tropical lineages. Mean speciation rates for high-latitude species (>45°, n = 376) are faster than those observed for tropical (<25°, n = 287) species (tropical: λDR = 0.16, λBAMM = 0.15; high latitude: λDR = 0.30, λBAMM = 0.23). For polar species (>60°, n = 105), these rate differentials are even more extreme, with mean λDR = 0.38 and λBAMM = 0.31.
Extended Data Figure 8 Latitudinal gradient in speciation rate for cell assemblages inferred from occurrence data.
Cell assemblages (n = 843) and species latitudinal midpoints were inferred from a non-redundant merge of four primary occurrence-based biodiversity databases (GBIF, OBIS, Fishnet2 and VertNet). a, λBAMM for cell assemblages as a function of latitude. b, λDR as a function of latitude. c, SAR spatial error models for the effects of absolute latitude on mean speciation rates for grid cells. AIC1 is a linear model with a single slope and intercept term; AIC2 is the corresponding AIC for a breakpoint model that assumes no relationship (slope = 0) between absolute latitude and speciation rate for all values below some threshold, and a linear relationship for latitudes that exceed the threshold. All other column headings as in Extended Data Fig. 2g. Results indicate a strong effect of latitude on speciation rate and are nearly identical to results obtained using the dataset of the primary map. d, Effects of absolute latitudinal midpoint for individual taxa on corresponding tip speciation rates, as assessed using FiSSE. Each row gives the results of FiSSE using a different threshold for classifying lineages as tropical and temperate. λ0 and λ1 denote estimated speciation rates (similar to λDR) for tropical and temperate lineages, respectively. All column headings are identical to those shown in Extended Data Table 1. Results are nearly identical to those obtained using explicit range reconstructions and reveal a pervasive effect of latitude on lineage-level speciation rates, regardless of the threshold used to classify species.
a, Relationship between terminal branch lengths and absolute latitudinal midpoint; means are shown for all species falling into a given bin (±2.5° from the focal value, n = 15). Mean branch lengths decrease with increasing latitude, reflecting faster speciation at high latitudes. b, Relationship between the estimated speciation rate for each taxon (λDR, n = 5,155) and the sampling fraction for the corresponding family-level clade to which the taxon belongs; the sampling fraction is simply the percentage of known taxa from the family that were represented in the phylogenetic dataset with genetic data. There is no clear relationship between the sampling fraction and the estimated speciation rates. c, Multiple regression analysis (OLS) of the relationship between taxon-specific speciation rate (λBAMM or λDR) and two predictors (latitude and family-level sampling fraction) in a multiple regression framework (n = 5,155). If the relationship between speciation rate and latitude is driven by progressively greater (or lower) genetic taxon sampling as a function of latitude, the sampling fraction term should explain a large fraction of the overall sums of squares. Even when sampling fraction is included as a covariate, the overwhelming fraction of variance is explained by latitude. For both λDR and λBAMM, more than 98% of the total sums of squares is explained by latitude and not sampling. d–f, Test for the effects of molecular evolutionary rate variation and latitudinal bias in speciation rate. d, Relationship between root-to-tip branch length sum for uncalibrated (non-ultrametric) RAxML phylogeny and midpoint latitude for each marine taxon (n = 5,149). e, f, Relationship between root-to-tip distance and λDR. There is effectively no relationship between the total path length for individual tips and their absolute latitudinal midpoint (Pearson r = 0.020). Plots in e and f emphasize tropical (midpoint latitude <25°; n = 3,481; red) and temperate–polar (midpoint latitude >45°; n = 567; blue) taxa, respectively, all other taxa are shown in grey. Overall relationship between (log)λDR and the rate of molecular evolution (root-to-tip sum) is weak but positive (Pearson r = 0.130) and inconsistent with the hypothesis that slow rates of molecular evolution at high latitudes results in fast but spurious estimates of speciation rate.