Method selection for species distribution modelling: are temporally or spatially independent evaluations necessary?

Method selection for species distribution modelling: are temporally or spatially independent evaluations necessary?