Hidden Taxonomic Bias Could Flip the Map of Evolutionary Diversity

One of the oldest patterns in biology is also one of its most seductive: species pile up toward the tropics, thinning steadily toward the poles. For decades, ecologists and evolutionary biologists have debated why. Are tropical lineages simply producing new species faster? Do temperate regions suffer from younger clades, harsher climates, or less time to accumulate diversity? A growing body of research has tried to settle the question by estimating diversification rates—the balance of speciation against extinction—directly from evolutionary trees built from DNA. But a new study published in Frontiers of Biogeography sounds a sharp warning: the very trees scientists rely on may be quietly distorted by an overlooked bias in how well species are catalogued around the world, and even a modest distortion could be enough to reverse the conclusions.

The study, led by researchers examining the interaction between two famous gaps in biological knowledge, focuses on what biologists call the Linnean shortfall—the fact that most of Earth’s species have never been formally described—and the Darwinian shortfall, our ignorance of how species are related to one another on the tree of life. These shortfalls are not distributed evenly. Temperate species, particularly in Europe and North America, have been studied, named, and split into distinct entities far more thoroughly than their tropical counterparts. The result is a geographically structured pattern that Freeman and Pennell dubbed the Latitudinal Taxonomic Gradient, or LTG: tropical species tend to be poorly delimited, meaning that many currently recognized tropical species may actually be bundles of several distinct species waiting to be separated once better data arrive.

On the surface, this might seem like a technicality. Surely discovering a few more tropical species would only strengthen the well-known latitudinal diversity gradient, since the tropics are already richer? The new study shows why that intuition fails. Estimates of diversification rates from phylogenies are exquisitely sensitive to the lengths of the youngest branches on the tree. When a poorly delimited tropical species is split into two—say, a subspecies is elevated to full species status—a brand-new, extremely short branch is added near the tips of the phylogeny. Short terminal branches are exactly what rate-estimating methods interpret as recent, rapid speciation. Subdivide enough tropical species, and the southern region appears to be an evolutionary hotbed, even if the underlying diversification dynamics are identical everywhere.

To quantify how much taxonomic incompleteness would be needed to distort the picture, the researchers built a simulation framework. They generated thousands of random phylogenies containing between 50 and 500 species, assigning each species to a northern or southern region based on the first split at the root of the tree. The southern clade was deliberately made richer, holding roughly 75 percent of the species, to mimic the real latitudinal diversity gradient. Crucially, the simulations began from a scenario in which diversification rates in the two regions were equal—any apparent difference emerging later would therefore be purely an artifact of taxonomic splitting. The team then ran splitting cycles, sequentially subdividing randomly chosen southern species and attaching small new branches close to the tips, with each new daughter species assumed to be phylogenetically close to its previously described relative, reflecting how real species descriptions tend to cluster among close relatives.

Two independent methods were used to detect when the artificial splitting had changed the apparent pattern. The first, known as the DR statistic, estimates diversification rates from the inverse of each species’ evolutionary distinctiveness, a measure popularized in large-scale analyses of birds and other vertebrates. Mean DR values for the two regions were compared using a Phylogenetic Generalized Least-Squares model, which accounts for the non-independence of related species. The splitting cycle stopped as soon as this comparison became statistically significant, identifying the minimum number of new species needed to manufacture a latitudinal gradient in diversification where none existed before. The second method, GeoSSE, extends the classical State Speciation and Extinction models to geographic settings, estimating speciation, extinction, and dispersal rates separately for two regions while accommodating species that occur in both. Here the researchers conservatively kept only simulations in which the fitted model did not initially beat a null model of equal rates across regions, then tracked the Akaike Information Criterion as splits accumulated, stopping when the model with regional differences outperformed the null by more than three AIC units.

The results were strikingly consistent. For the DR method, about 65 percent of simulations started with no significant difference between regions, and on average it took the addition of only about 14 percent new daughter species in the southern region for a significant gradient to appear. Once that threshold was crossed, mean DR values in the south had risen by a median of 26 percent, and the correlation between the proportion of species added and the magnitude of the rate shift was an almost perfect 0.97. The GeoSSE analyses told the same story with slightly different numbers: in the roughly 18 percent of simulations that initially showed no regional differences, adding about 15 percent new southern species was enough to push the model comparison past the threshold, a figure that dropped to 11.3 percent when newly described species distributed across both regions were also counted. In most cases, these modest additions less than doubled the estimated southern speciation rate.

In other words, a taxonomic bias of less than 20 percent—well within the plausible range of undiscovered or misdelimited species in many tropical groups—could be sufficient to create, invert, or erase a latitudinal gradient in diversification rates. The implications reach into some of the most cited results in macroecology. The researchers point out that analyses of New World birds reporting inverted gradients, with higher speciation in temperate zones but only small rate differences, could plausibly reflect problems in tropical species delimitation. Similarly, the diversification gradients reported in landmark phylogenetic studies of birds are small enough that adding fewer than 20 percent new tropical species would flip them. And while the Linnean shortfall in birds is comparatively modest, the situation is far worse in amphibians, reptiles, plants, and insects, where the proportion of undescribed and poorly delimited species is likely much higher, making diversification estimates for those groups even more fragile.

The study also reframes a long-standing debate about how to handle incomplete knowledge of biodiversity. Some researchers have argued that building better phylogenies—mitigating the Darwinian shortfall—could effectively bypass the urgent need to describe and delimit species, since phylogenetic diversity metrics are often more robust than simple species counts for conservation and macroecological purposes. Freeman and Pennell had already challenged that optimism, and the new simulations reinforce their skepticism: when the goal is estimating diversification rates, the two shortfalls interact rather than substitute for each other. Adding new species to the tree, which is precisely what resolving the Linnean shortfall entails, simultaneously reshapes branch lengths and therefore the Darwinian picture. There is no way to fix one without confronting the other, and in some scenarios the newly described species may also reveal gaps in distributional knowledge—the Wallacean shortfall—when a split produces a species restricted to only one region.

Where does the field go from here? The authors outline two major research avenues. First, biologists need realistic estimates of how large the Latitudinal Taxonomic Gradient actually is across different groups of organisms, since biases vary enormously from well-surveyed birds to barely explored tropical trees and insects. Predictive models of the Linnean shortfall, built from the socioeconomic and historical geography of taxonomic research—which regions and lineages attract collectors, funding, and expertise—could identify which species are most likely to be poorly delimited. Second, those predictions can be folded directly into empirical analyses: rather than treating current phylogenies as finished products, researchers can simulate splits of poorly delimited species into hypothetical close relatives and test whether estimated diversification gradients survive the perturbation. Only gradients that remain robust under plausible levels of taxonomic uncertainty should be treated as strong evidence for the evolutionary mechanisms driving diversity patterns.

The broader lesson is a humbling one for a discipline that increasingly treats phylogenies as ground truth. The latitudinal diversity gradient is real and enormous, and no reasonable number of new species descriptions will erase it. But the finer question of whether the tropics are also engines of elevated speciation—a question that shapes theories from tropical niche conservatism to density-dependent diversification—may currently rest on estimates too fragile to trust. Until taxonomic effort becomes more geographically balanced, or until analyses explicitly account for the branches that undiscovered species will eventually add, the evolutionary story behind life’s grandest pattern remains, in a very literal sense, unfinished. The missing species are not just absent names in a catalogue; they are absent branches, and their absence bends the tree of life in ways that science is only now learning to measure.

Subject of Research: How the Latitudinal Taxonomy Gradient biases estimates of diversification rates underlying latitudinal diversity gradients

Article Title: Effects of the interaction between Linnean and Darwinian shortfalls on diversification gradients

Article References: Tavares, G. R. G., Frateles, L. E. F., & Diniz-Filho, J. A. F. (2024). Effects of the interaction between Linnean and Darwinian shortfalls on diversification gradients. Frontiers of Biogeography, 17, Article e131169. https://doi.org/10.21425/fob.17.131169

Image Credits: AI Generated

DOI: 10.21425/fob.17.131169

Keywords: biogeography, latitudinal diversity gradient, Linnean shortfall, Darwinian shortfall, speciation, phylogenetics, GeoSSE, species delimitation, macroecology, taxonomy, Effects, interaction