5

How do you generate a species distribution probability map from GAM models? Or in other words, how do you set up a GAM equation in GIS to produce probability of species distribution?

It is simple to apply generalized linear models (GLM) or equation in GIS due to the independent variables have a liner relationship with dependent variable but for GAM, I am not able to do it because the parameters in GAM are non-linear predictors and also their estimates (β) are not be calculated.

1
  • Please be weary of overfit. I like GAM's well enough but there is a well known issue here. – Jeffrey Evans Jan 14 '15 at 2:50
1

I do not believe that there is a package for GAM's in Python. I would recommend the gam package in R. For spatial predictions you can utilize the raster package to wrap the generic predict function in a spatial prediction. There is ample information on implementing these types of models. If you would like some introduction on working with spatial objects in R I have some tutorials on my website (evansmurphy.wix.com/evansspatial) and there is plenty of additional information to be found through an internet search.

0
0

Probably the most popular tool for fitting and projecting Species Distribution Models (SDMs) is R http://www.r-project.org/. There are several very good packages (e.g. 'dismo', 'raster') which offer a lot more functionality for SDM than the common GIS softwares like Arc or QGIS.

If you have not had much experience with R there will inevitably be a learning curve, but if you will be doing a lot of modelling then it might be worth learning.

1
  • Please be specific. There is no implementation for gam's in dismo. The raster package provides capacity for spatial predictions but not for actual implementation of Generalized Additive Models nor SDM's. – Jeffrey Evans Jan 14 '15 at 2:49
0

I haven't used this toolset myself yet but some researchers at Duke University built the Marine Geospatial Ecology Tools. It looks like you can run R script and GAM models integrated directly into ArcGIS.

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.