One approach is to take a random sample of sites, apply treatment to one site, and then see if the treated site becomes an outlier. Oksanen points out that one problem with this approach is that treatment may 'merely' be affecting the variability - not the mean. Our comment would be that affecting the variability would still be a treatment effect, albeit not necessarily affecting the mean.

Another approach is the multiple BACI design as described in the More Information page on multiple group/period designs.