A comparison of simultaneous autoregressive and generalized least squares models for dealing with spatial autocorrelation

Authors

  • S. Beguería,

    Corresponding author
    1. Aula Dei Experimental Station, CSIC, Campus de Aula Dei, PO Box 202, 50080 Zaragoza, Spain
    Search for more papers by this author
  • Y. Pueyo

    1. Department of Environmental Sciences, Copernicus Institute, Utrecht University, PO Box 80.115, 3508 TC Utrecht, The Netherlands
    Search for more papers by this author
    • Present address: Geography Department, University of Zaragoza, Campus de San Francisco, 50009 Zaragoza, Spain.


*Correspondence: S. Beguería, Aula Dei Experimental Station, CSIC, Campus de Aula Dei, PO Box 202, 50080 Zaragoza, Spain.
E-mail: sbegueria@eead.csic.es

ABSTRACT

Aim  In their recent paper, Kissling & Carl (2008) recommended the spatial error simultaneous autoregressive model (SARerr) over ordinary least squares (OLS) for modelling species distribution. We compared these models with the generalized least squares model (GLS) and a variant of SAR (SARvario). GLS and SARvario are superior to standard implementations of SAR because the spatial covariance structure is described by a semivariogram model.

Innovation  We used the complete datasets employed by Kissling & Carl (2008), with strong spatial autocorrelation, and two datasets in which the spatial structure was degraded by sample reduction and grid coarsening. GLS performed consistently better than OLS, SARerr and SARvario in all datasets, especially in terms of goodness of fit. SARvario was marginally better than SARerr in the degraded datasets.

Main conclusions  GLS was more reliable than SAR-based models, so its use is recommended when dealing with spatially autocorrelated data.

Ancillary