# 7.2.7 Decision Mechanism 7: Evaluating Oversized DUs

An oversized DU is one that is larger than can be justified based upon site objectives but is evaluated nevertheless because of practical considerations. While oversizing DUs is strongly discouraged, there are some situations where it is unavoidable. Examples include DUs that are larger than the home range of some of the species of interest in an ecological risk assessment or are larger than the exposure area for some of the receptors/risks of interest in a human health risk assessment, such as when acute exposure to soil is a concern. In this situation, the DU evaluated actually consists of a few to perhaps thousands of smaller, ideally but impractically sized DUs. The problem faced is determining what information the sampled DU can provide concerning concentrations in the smaller sub-DUs.

There are no optimal answers to solve this dilemma, unfortunately, because typical ISM sampling designs are devoid of spatial information on contaminant distribution within the DU. This is not a new problem, as it has been documented in the literature for composite sampling and there are a number of approaches for estimating high-end concentrations within a sampled area. The simplest of these is to multiply the mean value from the composite (or ISM sample) by the number of increments. Note: A computationally equivalent approach is to use the average concentration but divide the soil *action level* by the number of increments. This approach represents the situation in which all of the contaminant is present in one of the increments. Given the number of increments in a standard ISM design, this approach is *extraordinarily conservative* and can yield quite high values. Given the conservative nature of this method, it is useful only to support “no further action” decisions or decisions to characterize the area further. Other approaches that are less conservative include multiplying the average concentration by the square root of the number of increments or more complicated formulas (for an example, see Barnett and Bown 2002).