Environmental Engineering Reference
In-Depth Information
Clustering used a log-likelihood distance measure, with number of clusters set at
ten to divide sub-catchments into matched pairs. Sub-catchment pairs were then
randomly assigned to control and treatment groups. To identify homogeneous
groups of sub-catchments, a second two-step cluster analysis was performed with
number of clusters determined by Schwarz's Bayesian Information Criterion.
Differences in mean pollutant concentrations between treatment and control sub-
catchments were evaluated by paired
t
-tests, with pairs matched by cluster and
collection date. Pollutant correlations with catchment characteristics, rainfall
amount and frequency were evaluated by Spearman's rank test. All statistical tests
were performed in SPSS v15
[7]
.
Results
Of 69 samples analyzed after 3 weeks monitoring, 100% of As, 66% of Cd and
88% of Pb concentrations were below analytical detection limits, therefore these
metals were excluded from subsequent data analysis. Principal component analysis
extracted three factors representing 88% of variation in the remaining six parame-
ters: The first factor correlated with Fe, Ni, Cr, and Cu, the second factor correlated
with Zn, and the third factor correlated strongly with TPH concentrations.
Therefore, parameters other than Cu, Zn and TPH were redundant and could be
dropped from analysis without losing substantial information.
Prior to the commencement of the E&E program, 150 GAC samples were collected
from 20 sub-catchments over approximately 8 weeks. Bag recovery was 93.8%.
High water caused by heavy rain prevented collection from four sub-catchments
on Nov 8, and six sub catchments on Nov 15. Heatherton drain outfalls were
submerged prior to collection on Nov 8, although accumulated pollutant concentra-
tions at these four sub-catchments were similar to previous weeks.
After the E&E program finished, 114 GAC samples were collected from 20 sub-
catchments over approximately 6 weeks. Bag recovery was 95%. Sample bags were
missing from sub-catchment H1 on Oct 24, and H4 on Nov 11 and Dec 6. Heavy
rain in the week preceding Dec 6 also caused the loss of bags from sub-catchments
D11, D4, and S1.
Cluster analysis, using the data collected prior to the commencement of the E&E
program, separated the 20 sub-catchments into ten control-treatment pairs, sum-
marized in Table
1
. Four homogeneous sub-groups were separated, corresponding
roughly to the spatial location of sub-catchments. Settlement and Heatherton groups
corresponded with sub-catchments draining into the Settlement and Heatherton
drains. Sub-catchments serviced by the Dunlops drain were separated into two
groups, Dunlops A and Dunlops B. Sub-catchments in Dunlops A were generally
larger than Dunlops B, with higher levels of hydrocarbon and lower levels of Zn
pollution (Table
1
).
As both Dunlops A and Settlement groups had an odd number of members, sub-
catchment D6 was included in the Settlement group for between group comparisons.
