Phylogenies can be used to predict future diversification but with limited success

Life on Earth is a story of constant change, where species split into new forms and others fade away. This process, known as diversification, happens at different speeds for different groups of animals. Some lineages, like birds, have exploded into thousands of species, while others have remained relatively static. Scientists have long tried to understand the rules behind this variation, hoping to identify which groups are likely to thrive and which might struggle in the future. To do this, they use family trees of life, called phylogenies, which map out how species are related. By looking at the tips of these trees—the living species at the very ends of the branches—researchers have developed mathematical tools to estimate how fast a lineage is currently splitting into new species. The ultimate goal is to use these estimates to predict which species will be the most successful in the coming millions of years, a question that matters deeply for conservation efforts and understanding the future of biodiversity.

A team of researchers set out to test whether these prediction tools actually work in the real world. While the methods had been tested on computer simulations, no one had ever checked if they could accurately forecast the future using actual data from nature. The team, led by Grace Ridder and Jan Smyčka, decided to put these methods to the ultimate test: time travel. Since they could not wait for the future to happen, they used a clever trick involving the past. They took large family trees of four-legged animals, including both living groups like birds and mammals, and extinct groups like dinosaurs. They sliced these trees at specific points in time, treating that moment as "today." They then used the standard prediction tools to estimate the future success of the lineages existing at that slice. Finally, they looked at what actually happened in the next few million years to see if the predictions matched reality.

The researchers found that these tools do have some ability to predict the future, but the success is far from perfect. In most of the groups they studied, the methods correctly identified that lineages with higher estimated rates of splitting would indeed produce more descendants later on. This means the tools are better than random guessing. However, the accuracy varied significantly depending on the animal group and the specific method used. For example, one method that relies on simple counting rules performed quite well across many different groups, consistently showing a positive link between its estimates and the actual number of new species that appeared. Another set of methods, which use more complex statistical models to account for how species go extinct, sometimes struggled. In some cases, particularly with certain groups of mammals like carnivores and hoofed grazers, these complex models failed to find any clear pattern, or even suggested the opposite of what actually happened.

The study also revealed that predicting the future of species is inherently difficult because evolution is a chaotic process. Even when the tools got the general direction right, they could not precisely predict the exact number of new species. The researchers calculated that the existing methods could explain only a small fraction of the variation in future diversity. Much of the outcome is driven by chance events that no model can foresee. Furthermore, the tools were much better at predicting the birth of new species than at predicting which species would go extinct. This is partly because the data available for living animals does not show us the lineages that have already died out, making it hard to learn from past failures. The researchers noted that the size of the family tree matters, too; the methods tended to work better on very large groups with thousands of species and less well on smaller groups with only a few hundred.

Ultimately, this work suggests that while we can make educated guesses about which lineages are poised for future growth, we must be cautious about the details. The tools are useful for spotting general trends, such as identifying which groups are generally more active in creating new species, but they cannot provide a precise forecast. The researchers found that different methods have different strengths: some are better at ranking species by their potential, while others are better at estimating the actual speed of evolution, though they are less stable. The study concludes that predicting the future of life is possible, but only to a moderate degree. The results depend heavily on the specific group of animals being studied and the quality of the data available, reminding us that the story of life is written with a mix of predictable patterns and unpredictable chance.