Multiple response variables that show a linear response to an ecological gradient may often be summarised by principal components analysis (PCA). Examining a PCA solution can help identify what combination of response variables best distinguishes between sampling units (objects).
If you're more interested in detecting and characterising underlying variables (i.e. 'factors') which have either not been measured or cannot be measured, consider factor analysis (FA). This method is very similar to PCA, however, its objective is specifically to identify latent variables which underlie the measured response variables. This may be useful in approaching an ecosystem with unknown drivers in a data-centric manner. If the knowledge of the system is scant, it may be premature to 'bundle' variables into speculative factors with no evaluation of those factors. FA allows this to be approached in a more formal manner in certain circumstances.
Alternatively, if you're interested in whether and how your response variables support the separation of your objects into pre-defined groups, you may consider discriminant analyses. These analyses pre-suppose a confident grouping of objects which is at least likely to be supported by the measured response data.