Rare Species May Be Safer Than They Look: Dispersal Diversity Shields Forests From Extinction

One of the most stubborn puzzles in ecology is deciding which species are truly on the brink. Conservation biologists have long leaned on abundance as the default alarm bell: if a species is rare, it must be at risk. But a new study published in Nature Ecology & Evolution argues that this intuition is only half the story. By importing mathematical tools from the physics of disordered systems, a team led by Davide Bernardi of the University of Padova, together with collaborators including Simon Levin of Princeton University and Andrea Rinaldo of EPFL, has built a framework that measures how close each species in a community sits to competitive exclusion—and found that the diversity of dispersal strategies, not sheer numbers, is what buffers many species against local extinction.

The core of the work is a single, measurable quantity the authors call the competitive balance. In classic metapopulation theory, a species persists when its local extinction rate is lower than the so-called metapopulation capacity, typically captured by the largest eigenvalue of a dispersal kernel matrix. That logic works for a lone species, but in a community of many competitors, meeting the threshold is necessary yet not sufficient. The new framework shows that long-term persistence in an interacting metacommunity is instead governed by how a species’ competitive balance compares with those of all its neighbors—a collective property that emerges nonlinearly from the whole community rather than summing individual contributions.

What makes the approach practical is its refusal to demand data that ecologists rarely have. Mechanistic models of interacting communities usually require estimates of demographic rates, interaction coefficients and dispersal parameters, which are notoriously difficult to measure in species-rich ecosystems. Statistical tools such as species distribution models or population viability analyses are more data-friendly but often treat species in isolation. The Padova-led team sidestepped the problem by modeling the interaction–dispersal kernel entries as effectively random, drawn from a species-specific gamma distribution whose parameters can be inferred directly from spatial abundance patterns—the kind of coarse-grained census data that forest monitoring plots already collect.

The bridge between theory and data is a quantity called the vacancy-adjusted abundance, or VAA. For each species in a subplot, the VAA is the fraction of space it occupies divided by the locally unoccupied space. This normalization upweights patches that are nearly saturated, so the same relative space-use contributes differently depending on how crowded a site is. The result has a clear ecological reading: it rescales local occupancy to reflect a species’ effective colonization effort. Fitting the theoretical VAA distributions to empirical histograms yields two interpretable parameters per species—dispersal heterogeneity and effective dispersal strength—without any direct measurement of vital rates or interaction coefficients.

To validate the framework, the researchers turned to three spatially complete forest censuses: the Barro Colorado Island tropical plot in Panama, the Pasoh Forest Reserve plot in Malaysia and the Michigan Big Woods temperate plot in the United States. Each plot was partitioned into 50-meter square subplots, and the space used by each species was estimated using an established allometric relation between trunk diameter and crown area. The theory captured the diversity of distributional shapes across species with striking accuracy. In out-of-sample tests, parameters fitted on one half of a plot predicted the mean space-use of species in the other half with coefficients of determination of 0.97, 0.93 and 0.95 for the three datasets respectively.

The framework also outperformed ecological null models. Against a neutral negative-binomial model with community-level parameters, and against a species-specific variant matched in parameter count, the new theory achieved absolute goodness-of-fit consistent with a correctly specified model across all three forests, and beat the alternatives in relative comparisons. Perhaps most tellingly, the fitted parameters carried real biological signal: in the Barro Colorado Island data, gap-exploiting and intermediate tree species showed significantly higher dispersal heterogeneity than shade-tolerant species, consistent with the episodic local dominance that light-demanding trees achieve when canopy gaps open. Effective dispersal strength, by contrast, did not differ significantly between guilds.

The deeper payoff comes when the competitive balances of all species are computed from the fitted parameters. Although dispersal-related parameters varied widely among species in all three forests, the competitive balances themselves were tightly clustered around their mean. The theory predicts exactly this: a group of competing species can coexist as long as their balances are not too dissimilar, with full coexistence possible only within a limited region of balance space surrounding the equal-balance condition. Equal balance does not mean identical traits—simulations with species sharing the same balance but different dispersal strategies produced distinct abundances—so the narrow clustering reflects a community-wide trade-off rather than trait uniformity.

From this clustering, the team derived a vulnerability index: the distance between a community’s position in competitive-balance space and the critical boundary below which a given species cannot persist. A negative value signals persistence; a positive value means eventual competitive exclusion. Applied to the three forests, every analysed species fell safely below its threshold. More provocative was what the index revealed about rarity. Abundance and vulnerability were strongly correlated only near the extinction threshold; far from it, species with low average abundance were not necessarily in danger of competitive suppression. In simulations, species with positive vulnerabilities saw their stationary abundance collapse toward zero, but away from the threshold, abundances spread across a wide range, confirming that being rare does not automatically mean being doomed.

The explanation lies in spatial strategy. Holding mean abundance constant, the researchers found a significant negative partial correlation between the spatial heterogeneity of a species’ space-use distribution and its vulnerability across all three datasets. Two Barro Colorado species made the point vividly: Prioria copaifera and Pterocarpus rohrii differ in total stem count by only about two percent, yet the former’s much more heterogeneous spatial distribution gives it a lower vulnerability. Intuitively, a heterogeneous dispersal strategy lets a species carve out spatial niches where it is hard to displace, trading some overall abundance for local strongholds. This trade-off between abundance, its spatial variability and vulnerability may be a key mechanism sustaining the coexistence of many species in the same forest.

The authors are careful about limits. The vulnerability index predicts final outcomes, not the timescales over which species approach them, and transient dynamics varied substantially in simulations. At very low abundances, demographic randomness will matter, and the framework cannot disentangle local competition from facilitation, both of which are folded into the inferred kernel parameters. Still, the applications are tantalizing: the same machinery could probe how an invasive species with a particular competitive balance might reconfigure a community, or whether suppressing a single species would trigger cascading exclusions like dominoes. An open-source package, vulntool, implements the core calculations. For a field racing to anticipate a potential mass extinction, the message is quietly radical: to know which species are truly at risk, look not just at how many individuals remain, but at how variedly they spread themselves across the landscape.

Subject of Research: A statistical-physics framework quantifying species vulnerability to competitive exclusion in metacommunities using spatial abundance data

Article Title: Dispersal diversity buffers species vulnerability to local extinction

Article References: Bernardi, D., Nicoletti, G., Padmanabha, P., Suweis, S., Azaele, S., Levin, S. A., Rinaldo, A., & Maritan, A. (2026). Dispersal diversity buffers species vulnerability to local extinction. Nature Ecology & Evolution. https://doi.org/10.1038/s41559-026-03195-y

Image Credits: AI Generated

DOI: 10.1038/s41559-026-03195-y

Keywords: community ecology, metacommunity, dispersal, competitive exclusion, species vulnerability, coexistence, statistical physics, forest plots, biodiversity, spatial ecology, dynamical mean-field theory, conservation