[R] Ecological Niche Modelling on R

[Clément Calenge]

>   It's very kind of Stephen to plug my book, but it's notwhat you're looking for. 
>   You need to read more about this general topic, and aboutthe particular packages: try
> http://www.unine.ch/CSCF/grasp/grasp-r/index.htmlhttp://www.unine.ch/CSCF/grasp/
>   Based on downloading grasp , it doesn't look as thoughit will handle presence-only data, though -- you may needto look further.
>  It doesn't look like adehabitat is what you want.From Calenge, Clement. 2006. The package "adehabitat" for the R software: A toolfor the analysis of space and habitat use by animals. Ecological Modelling 197,no. 3-4 (August 25): 516-519. doi:10.1016/j.ecolmodel.2006.03.017.
> ' ... the “adehabitat” package for the R software, which offers basic GIS(Geographic Information System) functions, methods to analyze radio-trackingdata and habitat selection by wildlife, and interfaces with other R packages.'
>   General advice about "I want to do X in R" -- (expandingon Stephen's advice above):
> 1. read about X in general (perhaps you have already done this);2. search for R packages and functions that do what you want  (you've already done this, although you misidentified "adehabitat"3. install those packages and see what they do.  Look at thedocumentation included with the packages, including any citationsreferenced.  Try the examples.4. If you don't know enough R to understand the examples or howto get your data into R, back up and read the introductory Rdocumentation.
>   

Actually, the confusion could be explained by the fact that many 
analyses methods (and especially factor analyses) originally developed 
in community ecology and biogeography to study the niche are also used 
in habitat selection studies (e.g., OMI analysis, ENFA, etc.). As the 
statistical issues (predict the species/animal presence on an area, 
given the value of environmental variables) and type of data 
(presence-only data to be compared with a sample/census of "available" 
units, etc.) involved in studies of the niche and habitat selection are 
often similar, the methods used are often similar too... However, most 
of the functions in adehabitat implement /exploratory/ methods of the 
ecological niche, and methods suitable for prediction are rare in the 
package (except one or two functions which have already been used for 
that, such as mahasuhab or domain, but they are probably not the best 
choice given your aim)... The package grasp may indeed be a better 
choice if your aim is prediction...

But I concur with Ben and Stephen on the fact that you should first read 
the (large) literature on niche modelling before choosing the method 
that seems appropriate to your data/issue, and then search R 
archives/package for a solution. a good start:

@ARTICLE{Elith2006,
  author = {Elith, J. and Graham, C.H. and Anderson, R.P. and Dudik, M. 
and Ferrier,
    S. and Guisan, A. and Hijmans, R.J. and Huettmann, F. and Leathwick,
    J.R. and Lehmann, A. and Li, J. and Lohmann, L.G. and Loiselle, B.A.
    and Manion, G. and Moritz, C. and Nakamura, M. and Nakazawa, Y. and
    McC. Overton, J. and Peterson, A.T. and Phillips, S.J. and Richardson,
    K. and Scachetti-Pereira, R. and Schapire, R.E. and Soberon, J. and
    Williams, S. and Wisz, M.S. and Zimmermann, N.E.},
  title = {Novel methods improve prediction of species distributions 
from occurrence data},
  journal = {Ecography},
  year = {2006},
  volume = {29},
  pages = {129-151}
}

and references therein.
Cheers,



Clément Calenge.

-- 
Clément CALENGE
Office national de la chasse et de la faune sauvage
Saint Benoist - 78610 Auffargis
tel. (33) 01.30.46.54.14