Below is the problem.

A multi-site agricultural experiment is being planned in which several
varieties of a given crop are going to be allocated to plots within each
site using a randomized complete-block layout.   The sites have been chosen
to differ dramatically over the range of the environments in which the crop
is grown.  Varietal differences are not expected to homogeneous over the
sites, and the best performing varieties at some locations are not expected
to be the best performing varieties at others.


Question:
how can I statistically approach the grouping of the sites based on
differential varietal responses over sites.  How to apply cluster analysis on this problem.

Thanks.