Do you work with species distribution modelling from the perspective of climate change? In that case, this paper is FOR YOU. Working with species distribution, but don’t really care about climate change? This paper is still FOR YOU.
In this article Buisson and colleagues tackle the fundamental problem of uncertainty in projecting species distribution shifts under climate change. The work is pretty breath taking: 35 species, 100 partitions of data, 7 different statistical modelling methods, 3 climate models, 4 emission scenarios and 3 different time steps (2020, 2050 and 2080), resulting in a staggering 8 400 different projection per species per time step. Not to mention the numerous PCAs and one phylogenetic tree done within this study.
This paper shows, statistically quite convincingly, that the greatest contributors to the variation in model results are in fact the statistical methods we use for projecting the likelihood of a species to occur in given climate conditions. Thus, the selection of modelling method, or even the decision of using just one instead of several, should be done with care. Unfortunately this paper does not go on by making any suggestions about which methods to choose, but instead underlines the need of using several modelling methods and presenting not just the results across them, but also the uncertainty related to them.
Buisson and colleagues have really done a comprehensive and good work: The paper is full with technical details, tests and analyses, but the text has been written so that the reader does not get lost in the jungle of residuals, autocorrelations and standard deviations. The only weakness we could find was the fact that the individual model performance and how this relates to the high variation in model results was not reported. Reading between the lines it could be that the authors have decided to save this the sequel. And we can’t really blame them – this paper definitely had information worth of several papers. Excellent job!
Final conclusions: You have to read this :)!
Buisson, L., Thuiller, W., Casajus, N., Lek, S. and Grenouillet, G. (2010) Uncertainty in ensemble forecasting of species distribution. Global Change Biology, 16, 1147-1157.