Table 10. Relative percent differences estimated by the off-site labo-
ratories and on-site by the technology developer for the sample
duplicates.
Sample
CRREL
Ref. lab
HM 2000
no./ID
(% RPD)
(% RPD)
(% RPD)
GRO
SG-8, 18
5.1
65
46
SG-10, 19
0
120
67
WG-11, 14
0
21
140
DRO
NA*
SDM-12, 21
35
31
SDM-15, 22
NA
20
16
WDM-4, 18
NA
52
24
RRO
M7, M8
NA
5.1
0
*Not analyzed.
limited number of photographs that represent different
known (U.S. Navy 1999, Hanby 1998). Here, two
TPH concentrations. Therefore, one would expect that
independent laboratories established concentrations for
an increase in accuracy would accompany this more
PE water and soil samples contaminated with GRO
sophisticated measurement technology. In comparison
(Tables 5a and b) that were within 12% of the certified
to the visual method of analysis, however, the HM 2000
or expected concentration. In addition, two certified PE
yielded some values that were false negatives (Table
samples of DRO compounds in soil were distributed
5d, WDM-1 and -2) and one that was greater than the
for analysis. The average value reported by the tech-
expected value by more than 10 (Table 5b, WG-7).
nology developer for these same PE samples was, in
Therefore, about 10% of the values (3 out of 29) esti-
one case, 59% higher, while in the other seven cases it
mated for QA samples by the HM 2000 failed to meet
was more than 250% higher than the concentrations
the criterion that is currently applied to the visual
verified by the reference laboratory. Looking at the
method of analysis.
values estimated for the matrix spike samples shows
The samples that the HM 2000 had the most diffi-
that, in only two cases, was the visual method, and in
culty with were background and HPLC water samples
one case, was the HM 2000 method, able to yield an
average value within 25% of the expected concentra-
spiked to between 0.48 and 1 mg TPH/L (Tables 5c
and d). The reported detection limit for both the visual
tions (Table 9). In one instance, for the visual method,
and HM 2000 methods is stated to be 0.1 mg TPH/L
this was clearly fortuitous, since the two values were
(Hanby 1998). The inability of the HM 2000 to esti-
separated by a factor of 2.9 (Table 5a, SG-20 and
mate values that were at least within an order of mag-
SG-21).
nitude for waters spiked at 0.48 or 0.98 mg TPH/L
One of the other reasons for developing a spectro-
shows that this detection limit cannot always be
photometric method of analysis with a digital-readout
achieved. Furthermore, even when comparing the val-
was that this approach would allow for an assessment
ues yielded for a PE sample (Table 5c, SDM-1 through
of precision. Looking at the relative standard devia-
4) with a TPH concentration close to the mid-point of
tions established for the PE samples shows that
the HM 2000 was incapable of achieving a high-
failed to distinguished itself as being superior to the
degree of precision. That is, this method cannot
achieve the levels of accuracy (i.e., 25%) and preci-
visual method of analysis.
It has been stated that these on-site methods of
sion (i.e., 15% RSD) that are associated with the more
estimating TPH in environmental matrices are capable
rigorous statistical analyses that are applied to field and
of producing concentrations within 25% or better of
laboratory analytical methods for the analysis of PE
the concentration established by accepted methods of
samples.
analysis, when the specific contaminant of concern is
The analysis problems experienced with the HM
15