Differing soil conditions such as texture, moisture, slope, and vegetation.

Prior surveys or any prior remedial treatments, or both.

Practical constraints, such as size.

Therefore, a site should be logically divided into units where each is sufficiently

uniform to suggest a single remediation approach and to justify characterization

units may vary widely in size but, for illustration in this document, we will

describe a unit that is 50 100 m (see Fig. 6, main text). Clearly, the division of a site

into units is a judgment and revisions may be needed as preliminary data become

available.

When at least four potentially contaminated sections (potential remediation units)

are present, four sample cores (046 cm [018 in.]) should be taken from each (for a

total of 16). If there are fewer than four units, the number of cores should be pro-

portionately increased. With more than six units, the number of cores can possibly

be decreased when there is great similarity among the units. However, this may be

impractical if units have widely different characteristics.

Cores are taken in pairs, 1 m apart, with each unit having two pairs. These pairs

are located at one-quarter and three-quarters of the distance along a diagonal

through the unit (BD in Fig. 6). This is a good arrangement if results show extreme

disparity between the means of the pairs. Then two additional sample pairs can be

taken later on the other diagonal (AC in Fig. 6). This forms a 2n factorial pattern.

Cores of 5.6-cm diameter, or other appropriate size, are taken to a depth of 46 cm

and divided into 0- to 15-, 15- to 31-, and 31- to 46-cm segments. Individual seg-

ments are placed in separate containers, such as Zip-LocTM bags, and carefully

homogenized by hand. Depending on the results from these initial samples, deeper

soil cores may or may not be necessary.

Before subsampling, a discrete sample is placed in an aluminum pie plate, or

other suitable container, where rocks, roots, and other debris are removed. After

further stirring and mixing, and coning and quartering (or equivalent), remove a

40-g portion. The four discrete samples for a given depth and unit are combined

and thoroughly homogenized to form one composite for each depth in each unit (a

minimum of 12 composites). Note that this assumes the action level of concentra-

tion being at least four times higher than the quantitation limit of the analysis

method. Any failure to meet this criterion will require using fewer discrete samples

in each composite if we wish to retain the capability for detecting "hot spots" for

surgical excavation.

On-site analytical methods require the extraction of 20-g subsamples. To obtain

these subsamples, transfer each bulk sample (discrete or composite) to a pie plate

and mix again before subsampling. A good way to do this is to cone and quarter

and then use a scoop to remove approximately 5 g of soil from each quarter.

27