I have a number of rasters of environmental data (~10) which may be important predictors for modelling species presence and abundance at ~10 different locations.

I would like to know which of the rasters are important in explaining the variance in observed results. Is it appropriate to look for principal components of the rasters? Is this done across the entire extent, or only with respect to my observed sites? Any reference on this would be appreciated.

EDIT: It turns out that PCs are not useful in this case. Even though two rasters may have very large correlation across the landscape, if they are different at the narrow range of sites occupied by the modelled species, this is what really matters.

  • 2
    $\begingroup$ @winchester I have the feeling that analysis of variance would be more appropriate. $\endgroup$ – robin girard Mar 7 '11 at 21:47

Your idea about the "rasters" is not very clearly stated, but you might have a look at the paper by Borcard and Legendre (1994) and their later works on spatial eigenvector-based analyses to see if one of the approaches will fit to your problem.

Borcard, D., Legendre, P., (1994) Environmental control and spatial structure in ecological communities: an example using oribatid mites (Acari, Oribatei). Environmental and Ecological Statistics 1, 37–61.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.