I don't know CANOCO that well, but a RDA sounds appropriate.

You can check whether your data are proper for this method or which other statistical methods you could use on this page: http://mb3is.megx.net/gustame.

if you have too many environmental variables it could also be a good idea to not include all of them in the model.

You can use RDA or PCA. For more data as mentioned by Tobias, you can use multiple regression  analysis to find out some important varibales to use in your model. Regards

Have a look at the Primer-E site for ideas, it's a good stat program suitable for your work.

http://www.primer-e.com/index.htm

Tom

CANOCO is mainly  for  multivariate gradient/ordination analyses. For  simple correlation analysis you can use the past software, which is free and easy to run.

http://nhm2.uio.no/norlex/past/download.html

Seid