We have a new paper out at MEE. It is a guide on how to use Joint Species Distribution Models (JSDMs) originally known (to us) as Multivariate Probit models (MVP), a name which thankfully didn’t stick.

Without delving into the details, this is a way to use species co-occurrence data when modelling species distributions. Why is this important?

I really like the point that Clark et al. (in press) make. Think of your favorite species. If you were asked to predict if that species occurs at a site, which information would you rather have–the average temperature in January or a list of the other species that occur there?  I would certainly choose the species list.

Species do not occur independently of others, so it’s important to consider that when modelling their distributions.

One output of the model is how much correlation between species we can attribute to shared environmental variables and how much is leftover (residual) correlation. Species that co-occur more frequently than we expect given the environment (positive residuals) might indicate some sort of dependence (e.g. mutualism) or may simply mean we haven’t captured all relevant environmental variables. Negative residuals may indicate a number of processes ranging from competitive exclusion to allopatric speciation. Note: these models will not tell you whether species are actually ‘interacting’. For that you would need experiments and lots of time or at the very least models of population dynamics.
Mick McCarthy has provided a much more detailed description of our paper here:

Or, if you want even more details, here is the paper:

References

Clark, J.S., Gelfand, A.E., Woodall, C.W. & Zhu, K. (in press) More than the sum of the parts: Forest climate response from Joint Species Distribution Models. Ecological Applications, http://dx.doi.org/10.1890/13-1015.1.

Golding, N. (2013) Mapping and understanding the distributions of potential vector mosquitoes in the UK: New methods and applications. Doctor of Philosophy, University of Oxford.

Ovaskainen, O., Hottola, J. & Siitonen, J. (2010) Modeling species co-occurrence by multivariate logistic regression generates new hypotheses on fungal interactions.Ecology, 91, 2514-2521.

Pollock, L.J., Tingley, R., Morris, W.K., Golding, N., O’Hara, R.B., Parris, K.M., Vesk, P.A., and McCarthy, M.A. (in press) Understanding co-occurrence by modelling species simultaneously with a Joint Species Distribution Model (JSDM). Methods in Ecology and Evolution. http://dx.doi.org/10.1111/2041-210X.12180