I haven't done habitat use studies, analyzing which local and landscape-level factors best explain bird species occurrence. for several decades, and am wondering if CCA (canonical correspondence analysis) is still considered valid (and used). Looks like most of the references are from the 1980s and 1990s, so I'd be interested to hear if some other method is used instead.
CCA is still widely used in ecology. For example, in my database of article data I have over 40 papers in marine ecology using the ACC of 90 papers for the year 2013 ....
In R, the library Vegan and especially ADE4 make available these functions.
Note: extensive documentation in French and English related to library ADE4 :
http://pbil.univ-lyon1.fr/ADE-4/home.php?lang=eng
Especially the biplot graphs of CANOCO are very useful to give a quick overview where certain species can be found with respect to which environmental variables.
Envrionmental variables with a strong impact on species distribution will have high (absolute) correlation coefficients with axis 1 and/or 2. CANOCO also gives you a p-values for the overall analysis (but you cannot say, which environmental variable contibutes how much to the variance explained).
There is also an option, where you can check the distribution of individual species along the first two axes. Sometimes a species has more than one maximum, which makes interpretations more difficult.
Please keep in mind, that CCA assumes linear relationships for the environmental variables, whereas species distributions are expected to be unimodal.
We have used CCA in the attached paper. Have a look and see if it can help you.
Best,
G.Lemperiere
You can also use RDA (canonical redundancy analysis) with a Hellinger transformation (for species data), more appropriate for linking community structure and environmental variables in a multivariate context.
CARO
Thanks to all for the responses. I will certainly look further at CCA and also consider RDA (as Caroline suggested).
Hi Timothy, CCA comes in play especially when you relating many responses to many independent variables, of course there are some assumptions that needs to be met. There is another method called partial least square (PLS) regression also can be used where it is relaxed in its assumption, i.e. is not as restrictive as CCA. I am attaching a publication and a link to PLS paper and an EPA reports, hope that will be of help for you.
http://www.hindawi.com/journals/isrn/2011/571749/ref/
http://www.epa.gov/esd/land-sci/pubs-posters.htm
- Badges
- Science method
- Similar topics
- Biological Science
- Ecology