Accounting for biotic interactions through alpha‐diversity constraints in stacked species distribution models. Issue 9 (11th February 2017)

• Record Type: Journal Article
• Authors: Gavish, Yoni; Marsh, Charles J.; Kuemmerlen, Mathias; Stoll, Stefan; Haase, Peter; Kunin, William E.
• Editors: Freckleton, Robert
• Is Part Of: Methods in ecology and evolution. Volume 8:Issue 9(2017)
• Page Start: 1092
• Page End: 1102
• Publication Date: 2017-02-11
• Subjects: alpha‐capacity -- beta diversity -- coexistence -- competition -- freshwater environment -- gamma diversity -- macroecological models -- random‐forest -- spatial ecology -- species distribution models; Ecology -- Periodicals; Evolution -- Periodicals; 577
• DOI: 10.1111/2041-210X.12731
• Languages: English
• ISSNs: 2041-210X

Abstract

Summary: Species distribution models (SDM) are widely used to predict occupancy patterns at fine resolution over wide extents. However, SDMs generally ignore the effect of biotic interactions and tend to overpredict the number of species that can coexist at a given location and time (hereafter, the alpha‐capacity). We developed an extension of SDMs that integrates species‐level and community‐level modelling to account for the above drivers. The alpha‐adjusted SDM takes the Probabilities of Occurrence (PoO) for all species of a community and the site's alpha‐capacity and adjusts the PoO, such that: (i) their sum will equal the alpha‐capacity as predicted by probability theory; and (ii) the adjusted PoO are dependent upon the relative suitability of each species for that site. The new method was tested using community data comprising 87 freshwater invertebrate species in an LTER watershed in Germany. We explored the ability of the method to predict alpha and beta‐diversity patterns. We further focused on the effect on model performance at the species‐level of the error associated with modelling alpha‐capacity, of differences in gamma diversity (the size of the community) and of the type of community (random or guild‐based). The models that predicted alpha‐capacity contained considerable error, and thus adjusting the PoO according to the modelled alpha‐capacity resulted with decreased performance at the species level. However, when using the observed alpha‐capacity to mimic a good alpha‐capacity model, the alpha‐adjusted SDMs usually resulted in increased performance. We further found that the alpha‐adjusted SDM was better than the original SDM at predicting beta‐diversity patterns, especially when using similarity indices that are sensitive to double absences. Using the alpha‐adjusted SDM approach may increase the predictive performance at the species and community levels if alpha‐capacity can be assessed or modelled with sufficient accuracy, especially in relatively small communities of closely interacting species. With better models to predict alpha‐capacity being developed, alpha‐adjusted SDM has considerable potential to provide more realistic predictions of species‐distribution patterns.