Choosing bioclimatic variables for maxent distribution

[Pankaj Koparde]

Dear all,

I am interested in finding potential distribution of a bird species. I did background search and found that one needs to look out for ecology of the species and extract variables that can comment on species presence. Based on this I short-listed few variables (BIO 2,4,11,15,19 and altitude). I ran maxent with cross-validate option keeping 25% points for testing. In the maxent output I got to know that 2 of my variables are not contributing to the model (permutation importance). Also in a correlation analysis, I found 2 of the variables highly correlated. I removed all these variables and now I have a set of 4 variables. Running maxent on these variables yields a distribution in which all the variables are contributing to the model. Is this the right way to remove variables? Only limited data is available on the species that I am studying, so I am not sure if the variables I have selected are the best representatives of the species presence! 

Thanks. 

Pankaj Koparde

Research fellow,

Department of Biodiversity, Garware College, Pune, India

[Christopher Ilori]

Hi Pankaj:

In a similar study of mine, I ran all variables and removed those that have zero contribution (based on advice from an expert in SDM field). I then used the remaining variables in my model....

Using bioclimatic variables only, I obtained a different set of contribution of variables compared to using both bioclimatic, land use, topography etc together.

My advice is (though may not be absolutely right), use different combinations of variables as for example, if vegetation contributes more to bird distributions in a location, we should not forget that vegetation cannot act in isolation; rainfall parameters are as well important.

Having said that, it is good to point out in your study that your results are based on the variables you considered. 

Not sure if you can pick anything from this?

--
You received this message because you are subscribed to the Google Groups "Maxent" group.
To unsubscribe from this group and stop receiving emails from it, send an email to maxent+un...@googlegroups.com.
To post to this group, send email to max...@googlegroups.com.
Visit this group at http://groups.google.com/group/maxent.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Pankaj Koparde]

Dear CO,

Thank you for the reply. 

If I use lots of variables and remove them selectively based on their permutation importance until I get >0 pemutation importance, wouldn't I be over-fitting the model? I mean each variable is going to contribute sooner or later as the model develops the distribution. 

Where can i get variables for vegetation cover for India? I do not have Arc GIS, so I am not sure how to convert .grd files to .ascii file which maxent accepts. For bioclimatic variables I extracted data from DIVA GIS.

Thanks. 

[Christopher Ilori]

Hello

I only know sources for NDVI, LAI and other proxies for vegetation indices. The vegetation I used during my study is White's, which is for Africa.

--

[Christopher Ilori]

Hello 

I only know sources for NDVI, LAI and other proxies for vegetation density. The vegetation I used during my study is White's, which is for Africa