this is the most common method used to infer a disturbance has occurred. it compares the impacted area with a control or reference site that is far enough away from the potential disturbance to be unaffected, yet close enough to experience similar environmental conditions. because no two sites can ever be exactly alike, the control site should reflect at least the potential to support all ecosystem function the impacted site did prior to disturbance. if one treats the impact as an experiment, reference sites (natural areas without impacts) can be thought of as control sites (areas not receiving an experimental treatment). sampling is conducted similarly at both impact and control sites, with replicate samples taken. sample locations within sites are usually selected randomly, with covariables or stratification to account for strong environmental gradients. disturbance is detected if the statistical variability between sites is greater than variability of the replicates within the sites. the c-i design reflects a limited sampling opportunity in that long before any sampling can be conducted, a site has already been deemed as impacted, for example where a known disturbance has occurred