When I try to make a CCA plot, only 4 out of 37 variables are displayed, this is the result when i input CCA.

Inertia Proportion Rank Total 1.0904 1.0000 Constrained 0.4789 0.4392 4 Unconstrained 0.6115 0.5608 15 Inertia is mean squared contingency coefficient Some constraints were aliased because they were collinear (redundant)

It's like the other columns of my data could be created as a linear combination of the 4 variables shown in the plot. may be those contained no extra information and hence are dropped from the analysis.

For this type of problem, should i use co-correspondence analysis (Co-CA) ?

you should check 37 environmental variable for collinearity before performing CCA.

Thank you for your help

The choice depends on (1) the nature of your data (hetero or homogeneous) and (2) the objective (linear or unimodal link). if your data are homogeneous a method with a linear link between (variables / species), IS appropriate (rda), whereas if they are heterogeneous you should apply an ordination with a unimodal link (cca).

the application of a Detrended Correspondence Analysis (DCA) a priori will allow to choosing the appropriate ordination where if the length of the first axis is

Dear Abdallah Aouadi , I agree with you that the choice is dictated by the type of the data. However, your conclusion at the end does not make sense to me.

If you expect a lot of variability/heterogenity in your data, this means that your data would be better with the unimodal type of the analysis, therefore CCA. Lower values of the length of the first axis suggest linear response of the depedent variables (eg., species) = RDA. Not other way around as you suggested.

I am using 2 as my threshold, instead of yours 5), therefore if the length of the first axis in my data is less then 2 = RDA, more then 2 = CCA. Not a foolproof, but quite robust as using unimodal method with linear data is not a big problem, but using linear method for unimodal data could be. :)

There is also a possibility to transform the data, so the RDA could be used. For a brief explanation of why, see response of Gavin Simpson here: https://stat.ethz.ch/pipermail/r-sig-ecology/2011-August/002280.html