Comparison Criteria for
Environmental Chemical Analyses of
Split Samples Sent to Different Laboratories
Corps of Engineers Archived Data
CLARENCE L. GRANT, THOMAS F. JENKINS, AND ANAND R. MUDAMBI
alistic numerical comparison criteria. The major
INTRODUCTION
shortcoming of the archived data is the absence
In conjunction with hazardous waste remedial
of any basis for assuming that the QA results are
activities, the U.S. Army Corps of Engineers
more accurate than the QC data being evaluated.
(USACE) requires that quality control (QC) and
In fact, a contractor laboratory generating QC re-
quality assurance (QA) samples be collected and
sults on one project may serve as the QA labora-
analyzed by a contract laboratory and a USACE
tory on another project. Lacking evidence to the
laboratory, respectively. Procedures for these ac-
contrary, we must assume similar uncertainties
tivities are prescribed in Chemical Data Quality
associated with both data sets.
Management for Hazardous Waste Remedial Activi-
The archived data received were as follows:
ties (USACE 1990). The QC and QA samples rep-
a) Forty-nine data reports for metals in soils
resent about 10% of the field samples and are
from the New England Division (NED) labora-
splits or co-located samples. Splits are used for
tory, labeled M1-1 to M1-49, and 75 reports from
analyses such as metals in soils where bulk sam-
the North Pacific Division (NPD) laboratory, la-
ples can be homogenized and subdivided in the
beled M3-1 to M3-75. These results were from 37
project locations in eight states and involved 16
samples are required for volatile organic com-
laboratories.
pounds (VOCs) in soil and water samples to re-
b) Sixty-six data reports of volatile organic com-
duce losses in handling, and are often preferred
pounds (VOCs) in soils supplied by the Missouri
River Division (MRD), labeled VS2-01 to VS2-66
water samples. Since co-located samples are not
(VS2-60 to VS2-66 used GC methods instead of
homogenized, the contribution of natural sample
GC/MS), and three reports from the NPD, labeled
variability can increase differences compared to
VS3-1 to VS3-3. These results were from 26 project
splits. Field personnel also collect rinsate blanks,
locations in 15 states and involved 21 laborato-
trip blanks for VOCs, and background soil and
ries.
c) One hundred and two data reports for total
ples is kept blind to analysts and laboratory per-
petroleum hydrocarbons (TPH) in soils supplied
sonnel until data are in deliverable form.
by the NPD, labeled TS3-01 to TS3-102, and 61
Acceptability of contractor laboratory data is
reports from the MRD, labeled TS2-01 to TS2-61.
determined by a USACE QA laboratory. Guide-
Analysis was by Method 8015M, GC-FID or equi-
lines exist for this assessment except for the num-
valent. These results were from 17 project locations
erical comparison of QC and QA analytical re-
in three states and involved eight laboratories.
sults from split or replicate samples. Currently,
limits for these comparisons are not based on tech-
soils supplied by MRD, labeled ES2-01 to ES2-71,
nical criteria and vary from one QA laboratory to
and eight reports from NPD, labeled ES3-01 to
another.
ES3-08. Analyses were by method 8330. These re-
It was decided that an analysis of archived QC/
sults were from eight project locations in eight
QA data might be useful in establishing some re-
states and involved eight laboratories.