Importance of nurse plants structuring plant communities is well‐appreciated at local scales, yet the effect of a single nurse on large scales has been neglected in analyses. So far, studies only use environmental gradients within one type of ecosystem and tend to generalize the nurse effects.
To assess how the effect of a single nurse species is modulated by different environmental settings, interactions between the shrub Vachellia caven and the surrounding plant communities were evaluated at 481 paired plots (outside vs. underneath the plant crown), in 39 sites across two distribution ranges, the Mediterranean west and the mostly subtropical east of the Andes Mountains (covering ca. 2 × 106 km2).
Cover, abundance and richness of perennial plants underneath and outside V. caven were used as response variables to estimate an index indicative of plant interactions (relative interaction index [RII]) and tested how this was affected by the rainfall gradient and distribution range.
Overall, RII responses to rainfall gradients had low conditional R2 (~0.25) in this large scale of analysis, but were significantly different between ranges: the RII followed a quadratic trend across the rainfall gradient in the western range, while this relationship was positive and close to linear at the eastern range.
Then, by projecting the RII models (i.e. for abundance, cover and richness) spatially through a consensus map, we show that most positive effects of V. caven are geographically found in dissimilar areas: the central part of Chile (western range) and across the Paraná River (eastern range).
When local fine‐scale predictors (i.e. annual herbs' cover and height, and herbivores' faeces cover) were used to model each response variable at the plot level (underneath or outside V. caven), we observed similar trends as when we considered only the large‐scale predictors.
Synthesis. Here, we show that the effect of the same nurse species on neighbouring plant communities can be very different depending on ranges of distribution, stressing that its ecological function cannot be generalized and not only depends on local factors but also is large‐scale context‐dependent.