using biomod2 package with continuous response variables

Question

I am looking to correlate crop area with climate variables and then predict for future if suitability of crop areas will change/remain same under different climate scenarios (in line with species distribution modelling where presence/absence of species is correlated with the environmental variables and hence gives an idea of what climatic conditions do species prefer). My response variable is area under crop (a particular crop) for 600 administrative units and I have some environmental variables as explanatory variables.

I was wondering if someone has experience using biomod2 package in R. biomod2 package allows user to run an ensemble of models (in contrast with running a single model) for species distribution modelling. It takes in species presence (coded as 1) and absence (coded as 0) and correlate it with the environmental variables giving an indication of what environmental variables species like based on where it is found. Since the biomod2 package needs species presence/absence data as response variable, in my case, what do you suggest I should do considering my response variable is a continuous data: Should I develop a threshold crop area value below which it will be considered 0 (i.e. absence of crop area) and above which it will be given the value of 1 (presence of crop area). Or is there any way I can use my crop areas as it is (I haven't tried it yet)

Answer 1

biomod2 is not handling continuous response variables.

You will have to switch to abundance/count based models (not supported in biomod2) or to convert your continuous data into binary variable if you want to work within biomod2 framework.

Both approaches are interesting and complementary. Comparing the outputs should give you a better idea of your species abundance and distribution across space.

Another interesting idea should be to weight presences by the abundances (Yweight argument) but this is still a under development part of biomod2 and should be used with cautious.