Alien plants are more often mycorrhizal than natives, except in grasslands and drylands

An average that reverses itself

A paper published online on 17 August 2026 in Nature Ecology & Evolution, and in print in the September issue at volume 10, pages 1644–1657, puts a fungal question to a field that usually answers with seeds, growth rate and grazing. Sara Giulia Cazzaniga and colleagues at ETH Zurich matched 440,788 vegetation plots against database records of the mycorrhizal type each plant species forms. Globally, the result runs against the popular assumption: non-native plants are more likely to be mycorrhizal than the native plants growing beside them.

The average conceals two opposed patterns. It holds, in the authors' words, across most forested temperate regions, and in boreal forests and tundra as well — even though those biomes are generally dominated by ectomycorrhizal plants, the incomers there are more often mycorrhizal than their native neighbours, and most often arbuscular-mycorrhizal. Across most grasslands and drylands it turns over, and non-natives are more often the ones doing without a fungus. Tropical grasslands and shrublands show it most sharply: the models predict that 17.8 per cent of non-native species there are non-mycorrhizal, nearly double the 9.0 per cent predicted among natives. Those are model-estimated mean proportions carrying standard errors, not a tally of counted species, and they describe that one biome rather than the world.

What the data are, and are not

The vegetation came from sPlot version 4.0, a compilation of 2.5 million georeferenced plot surveys made between 1888 and 2022, with plot sizes from 1 to 1,000 square metres. Native or non-native status was assigned where two databases agreed: GloNAF, covering more than 10,000 plant species across 1,029 geopolitical regions, and Plants of the World Online. Mycorrhizal type came from FungalRoot. Only plots already holding at least one non-native species were kept, which left 440,788 plots with 39,731 plant species from 14 biomes, surveyed between 1899 and 2022. The paper carries 102 authors, sPlot being pooled from the vegetation archives of many national teams.

It is worth stating plainly what this study did not involve. No fungus was collected, no soil sequenced, no root examined under a microscope. Mycorrhizal type here is an attribute attached to a plant species in a database, and where species-level information was missing it was inferred from the genus. Nor is this a comparison of invaded vegetation with uninvaded vegetation: only plots where non-native species already occur were retained, which the authors list among the study's limitations.

Two kinds of disturbance, pulling apart

The two disturbance measures point in opposite directions, and collapsing them into a single statement would reverse half the paper. The structural disturbance index captures the frequency and intensity of natural and human-caused events such as fire, flooding, logging and land conversion; it was built from twenty years of Landsat 5, 7 and 8 imagery covering 1999 to 2019 and tested on the 249,947 plots surveyed in that same window. High values raise the probability that a non-native plant is non-mycorrhizal, globally and in most biomes. Mediterranean forests are the exception: there, higher generalised disturbance went with more mycorrhizal non-natives, not fewer.

The global human modification index version 3 describes stable pressure instead — built-up areas, croplands, pasture, population density, night-time lights, roads, railways and navigable waterways — and was tested on the 295,056 plots surveyed between 1991 and 2020. Under that measure the relationship broadly inverts, high human pressure generally going with a greater prevalence of mycorrhizal species among non-natives, and the biomes swap sides. Grasslands, tropical moist broadleaf forests, boreal forests and tundra, pushed towards non-mycorrhizal strategies by generalised disturbance, are pushed towards mycorrhizal ones by human pressure, while Mediterranean and temperate forests do the reverse. A sentence saying that disturbance favours plants without fungi, written without naming which index, gets half of this wrong.

The authors offer an interpretation rather than a test: repeated disturbance probably depletes fungal inoculum where arbuscular-mycorrhizal fungi dominate and their dispersal is limited, whereas settled human pressure appears to preserve or even facilitate the partnership — not least because people themselves move mycorrhizal crops and ornamentals around.

Flexibility that earns nothing

Textbooks have long favoured the facultatively mycorrhizal plant as the ideal invader — one that can take a fungal partner but does not depend on it. The plots say otherwise. Facultatively mycorrhizal species are consistently under-represented among non-natives relative to natives, and disturbance does not flip that: they remain under-represented at low and at high disturbance alike. It would be wrong, though, to say disturbance leaves them untouched. Their relative proportion among non-natives rose under high human pressure in every biome.

The reversal also needs its units named. The earlier work that credited symbiotic flexibility measured success as the number of regions a species naturalised in and the geographic range it covered; this paper measures how often a species turns up inside a vegetation plot. Those are two different currencies of success, and a finding in one does not automatically cancel a finding in the other.

Where the evidence stops

The authors write that their study is correlational and that experiments will be needed to test the mechanisms. They also write that FungalRoot remains incomplete, covering only a small fraction of the global flora with marked geographic and taxonomic biases, and that some regions of the world are considerably undersampled compared with others — the cover figure shows it in the hatching over ground more than 600 kilometres from any plot, and in the broad grey of no significant difference. A rarefied re-analysis reproduced the main trends, which is reassuring without conjuring data where none exists. The open-access full text carries the complete list of caveats.

For a reader outside the modelling, the checkable point is not that alien plants can manage without fungi. It is that the answer depends on what opened the ground for them. A verge cleared by logging and a lawn in the middle of a city are both disturbed, and on these models they suit plants with different fungal strategies. The pre-processed data and the analysis code are public, on Zenodo and GitHub, with one limit the authors state outright: plot coordinates were deliberately shifted so that the surveyed locations cannot be pinpointed. The sums can be redone; the map cannot be walked.