Hello R-Sig-Eco Mailing List,
I have been familiarising myself with the MaxEnt GUI (v3.3.3k) and have been
successful in producing a range of models for species of interest to me. For
better reproducibility and a quicker workflow I have decided to use the Dismo
package and run MaxEnt through R. I hope this is the right mailing list for me
to contact.
I have noticed something strange when comparing the outputs of the two methods.
Using exactly the same environmental layers and occurrence records (n=81) I end
up with two wildly different predicted distributions when using anything but
the default application values (as per v3.3.3k).
For my first test I opened the MaxEnt GUI and, leaving all settings at default,
ran the model for my species. I then ran the exact same model with the
following code using Dismo and, when plotted, I got the same distribution (as
expected). x=expl is my rasterStack of environmental variables.
xx<-maxent(x=expl,p=spp.coords$garmani)
However, if I change settings to anything except default, the maps look totally
different. For example, if I set regularization=0.5 and prevalence=0.80 in the
GUI and run the model (keeping all other settings exactly as default) and then
do the same in R with the following code then the output is different.
xx<-maxent(x=expl,p=spp.coords$garmani, args=c(
'betamultiplier=0.5',
'defaultprevalence=0.80'
))
This is the code that I used to generate my maps within Dismo. As mentioned
previously, this code generates identical maps when using the default
applications settings but not when I specify my own arguments in Dismo.
map<- predict(object=xx,x=expl,na.rm=TRUE,format='ascii',filename=
"C:\\Users\\Simon\\Desktop\\nondefaultdismomap",overwrite=TRUE,progress='text',args=c('outputformat=logistic'),prj=TRUE,keepres=TRUE)
The AUCs and variable contributions for both models look nearly identical. I
would have assumed therefore that the maps wouldn't look so wildly different. I
have uploaded the maps to the following location so you can see for yourself
the discrepancy. Maps discrepancy<http://i.stack.imgur.com/zrSqJ.png>.
When analysing the two HTML outputs generated from the models I have only
noticed the one difference. When using the GUI with 81 presence records and
10,000 background points I am told that "10020 points used to determine the
Maxent distribution (background points and presence points)". However, when
using Dismo I am told "10081 points used to determine the Maxent distribution
(background points and presence points)". As I mentioned previously, all other
settings were kept the same (except the arguments as specified above), so
occurrence records haven't been used in cross-validating, for example.
A final point to mention that has me confused concerns adjusting the default
number of background points to sample. If I set an argument in Dismo to limit
background points (e.g. 'maximumbackground=2500') the HTML output in the model
gives me conflicting information. Under the 'Raw data outputs and control
parameters' section at the bottom of the page I am informed that "10081 points
used to determine the Maxent distribution (background points and presence
points)" but a little further down the page it says "maximumbackground:2500".
Does anyone know why two different numbers have been quoted here? Is there a
way for me to determine for certain whether 2500 points were used or 10,081? I
can confirm that my raster has may more non-NA cells than 10,000.
I was wondering if anyone else had come across these 'issues' before. Am I
doing something fundamentally wrong within R for the two distributions to be so
different, for example using the wrong syntax? I have received no errors when
running any models so I thought this possibility may be unlikely. I thought
that maybe I would have to specify all the arguments in Dismo (even if I wish
to keep them at default) but after some research I'm led to believe that Dismo
will just pull the GUI application defaults if you have not defined an argument
to say otherwise.
Finally, I can confirm that I am running the latest versions of R, Dismo, Java
and MaxEnt. My R and Java environments are x64.
Any help would be very much appreciated. Thanks for your time.
Simon
This message and any attachment are intended solely for the addressee
and may contain confidential information. If you have received this
message in error, please send it back to me, and immediately delete it.
Please do not use, copy or disclose the information contained in this
message or in any attachment. Any views or opinions expressed by the
author of this email do not necessarily reflect the views of the
University of Nottingham.
This message has been checked for viruses but the contents of an
attachment may still contain software viruses which could damage your
computer system, you are advised to perform your own checks. Email
communications with the University of Nottingham may be monitored as
permitted by UK legislation.
[[alternative HTML version deleted]]
_______________________________________________
R-sig-ecology mailing list
R-sig-ecology@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology
