Table 2. Orthogonal array of experimental trials
trials (not including equilibrium trial B) formed
and trial sequence.
an L4 array. Statistical analysis of these results
confirmed that stripping performance for this
Temperature
Flow rate
Random
source water was controlled by TPH removal and
(Factor A)
(Factor B)
sequence
that TPH removal was controlled by liquid tem-
(C)
Trial
(L/min)
of trials
perature, not flow rate. Liquid temperature was
1
8.89
18.9
5
therefore varied at a third level in the remaining
2
trials.
(Equilibrium A)
8.89
75.7
1
3
8.89
75.7
4
4
15.56
18.9
7
Sampling methods
5
15.56
75.7
8
Two types of aqueous samples, VOC and TPH,
(Equilibrium B)
22.20
18.9
2
were collected and analyzed during each trial. A
6
22.20
18.9
3
nested design of eight trials, six samples per trial,
7
22.20
18.9
9
and three analyses per sample was performed.
8
22.20
75.7
6
The samples collected and analyzed for a typical
trial of the 8 6 3 design is shown in Table 3.
air stripping performance conditions (U.S. EPA
More than 300 analyses were performed to pro-
1991).
duce statistically reliable data from the eight main
A full factorial design of six experimental tri-
trials in this experiment. To determine sampling
als results from varying temperature at three lev-
variability and analytical precision, triplicate in-
els and flow rate at two levels. This design pro-
fluent and triplicate effluent samples were col-
vided an experimental resolution level of 4,
lected, each analyzed in triplicate. Sampling peri-
meaning all main effects and factor interactions
ods were spaced 2.8 minutes apart at the 75.7-L/
can be statistically estimated by analysis of vari-
min flow rate, and 11.2 minutes apart at the 18.9
ance (ANOVA). The six-trial design was made
L/min flow rate. These periods are equal to the
more robust by conducting two additional trials,
detention times under plug-flow conditions, and
duplicates of trials that were excessively variable
were used so that approximately the same slug of
or critical to predicting performance. The re-
influent water was sampled at the effluent.
searcher chose to repeat the trials conducted at
the factor level extremes (trial 1 at 8.89C with
75.7 L /min [48F/20 gpm], and trial 3 at 22.2C
Analytical procedures
with 18.9 L /min [72F/5 gpm]). The two pairs of
VOC samples were collected in EPA-cleaned
40-mL vials with Teflon-lined septa (Eagle-Picher
replicated trials were treated as simple duplicates
Environmental Science and Technology, Miami,
and used to determine experimental variability
Okla.). Samples were stored at 4C and analyzed
between trials. The array of all trials, equilibrium
within six days. Analysis was by headspace gas
and main, and the actual random sequence in
chromatography (HS/GC) techniques and a pho-
which they were conducted is shown in Table 2.
toionization detector as presented by Hewitt et al.