University of Wollongong
Browse

A simple post-hoc method to add spatial context to predictive species distribution models

Download (727.33 kB)
journal contribution
posted on 2024-11-14, 15:36 authored by Michael Ashcroft, Kristine FrenchKristine French, Laurie ChisholmLaurie Chisholm
Methods to incorporate spatial context into species distribution models (SDMs) are underutilised, with predictions usually based only on environmental space and ignoring geographic space. The goals of this study were to demonstrate a relatively simple post-hoc method to include spatial context in SDMs and to quantify the improvement over purely niche-based models. The method involved producing a standard niche-based model using established techniques, such as Maxent, and then calculating the neighbourhood average of the model output in geographic space. In effect, we tested whether the spatially averaged model output was better at predicting species distributions than the raw model output. We demonstrated the method using 32 tree species on the Illawarra Escarpment and found the area under the receiver operating characteristic curve (AUC) increased by a mean of 0.021 using this method. The improvements were largest for eucalypts, which have poor dispersal ability and clustered distributions. Improvements were smaller for moist rainforest species, which were restricted to small areas with sufficient shelter from hot, dry northwesterly winds. We conclude that it is relatively easy to add spatial context into species distribution models using this post-hoc method, and the resulting models are better for predicting species’ distributions.

History

Citation

Ashcroft, M. B., French, K. O. & Chisholm, L. A. (2012). A simple post-hoc method to add spatial context to predictive species distribution models. Ecological Modelling, 228 17-26.

Journal title

Ecological Modelling

Volume

228

Pagination

17-26

Language

English

RIS ID

50513

Usage metrics

    Categories

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC