Temporal validation plots: quantifying how well correlative species distribution models predict species' range changes over time

Authors

  • Giovanni Rapacciuolo,

    Corresponding author
    1. Berkeley Initiative in Global Change Biology, University of California Berkeley, Berkeley, CA, USA
    2. Department of Life Sciences, Imperial College London, Ascot, Berkshire, UK
    3. Centre for Ecology & Hydrology, Wallingford, Oxfordshire, UK
    4. Grantham Institute for Climate Change, Imperial College London, London, UK
    Search for more papers by this author
  • David B. Roy,

    1. Centre for Ecology & Hydrology, Wallingford, Oxfordshire, UK
    Search for more papers by this author
  • Simon Gillings,

    1. British Trust for Ornithology, The Nunnery, Thetford, Norfolk, UK
    Search for more papers by this author
  • Andy Purvis

    1. Department of Life Sciences, Imperial College London, Ascot, Berkshire, UK
    2. Grantham Institute for Climate Change, Imperial College London, London, UK
    3. Department of Life Sciences, Natural History Museum, London, UK
    Search for more papers by this author

Summary

  1. The use of data documenting how species' distributions have changed over time is crucial for testing how well correlative species distribution models (SDMs) predict species' range changes. So far, however, little attention has been given to developing a reliable methodological framework for using such data.

  2. We develop a new tool – the temporal validation (TV) plot – specifically aimed at making use of species' distribution records at two times for a comprehensive assessment of the prediction accuracy of SDMs over time.

  3. We extend existing presence–absence calibration plots to make use of distribution records from two time periods. TV plots visualize the agreement between change in modelled probabilities of presence and the probability of observing sites gained or lost between time periods. We then present three measures of prediction accuracy that can be easily calculated from TV plots.

  4. We present our methodological framework using a virtual species in a simplified landscape and then provide a real-world case study using distribution records for two species of breeding birds from two time periods of intensive recording effort across Great Britain.

  5. Together with existing approaches, TV plots and their associated measures offer a simple tool for testing how well SDMs model species' observed range changes – perhaps the best way available to assess their ability to predict likely future changes.

Ancillary