1.1 MB pdf - Bolsa Chica Lowlands Restoration Project

SECTION 3: ANALYSIS
concentrations, potential toxicity that may result following inundation of sediments that are
currently dry may be estimated, and remedial actions to mitigate this toxicity may be
planned.
Simple linear regression analyses of COPEC concentrations in pore water on those in
sediment were performed using SAS (1994) PROC REG. All models were considered
significant if p#0.05. The results of these analyses (scatterplots and associated regression
analyses) for all chemicals and data transformations are presented in Appendix H. Of
70 chemicals considered, significant regressions based on untransformed data were
obtained for only seven chemicals (aldrin, arsenic, beryllium, alpha chlordane, endosulfan
sulfate, mercury, and phenanthrene). The highest r 2 for these chemicals was 0.28. Significant
regression models were obtained for 10 of 70 chemicals (acenaphthene, aldrin, arsenic,
beryllium, alpha chlordane, copper, endosulfan I, endrin aldehyde, total DDT, and thallium)
based on natural-log-transformed data. The highest r 2 from these data was 0.24.
Because significant relationships between concentrations of COPECs in pore water to those
in sediment were observed for few chemicals, and these relationships generally accounted
for less that 25 percent of variation, the pore water bioassay results cannot be used to predict
potential toxicity from pore water associated with the sediments. The pore water bioassays
may, however, be used to evaluate potential toxicity at locations where pore water already
exists.
Correlations Between COPECs in Sediment
Many COPECs have been detected in sediment from throughout the Bolsa Chica Lowlands
(using the entire ERA Sampling and Analyses dataset). To aid in streamlining future data
collection, correlation analyses were performed among analytes detected in sediment to
determine which chemicals were consistently detected in association with each other.
Analyses for chemicals whose occurrence is highly correlated may be reduced, such that
only those chemicals that are the best indicators are analyzed for.
Correlation analyses were performed using SAS (1994) PROC CORR. The resulting
correlation matrix is presented in Table 3-22.
Principal Components Analyses for COPECs in Sediment
Principal components analyses (PCA), a multivariate statistical technique, is another
approach to reduce the dimensionality (i.e., number of variables) associated with COPECs in
sediment. PCA is a statistical technique that linearly transforms the original numerical
variables to a substantially smaller set of uncorrelated variables that represent most of the
information in the original dataset (Dunteman, 1989). These uncorrelated variables, known as
principal components, are a linear combination of the original variables and are sorted in
decreasing order of the amount of variability in the original dataset they explain. Backcorrelation
of the principal component scores with the original data provides an indication of
which parameters each principal component represents. PCA is generally a data evaluation
method used to explore underlying relationships among variables in a large dataset.
PCA was performed on the sediment data previously used for correlation analyses using
SAS (1994) PROC PRINCOMP. (Note: the complete output from the principal components
analyses is included in Appendix H.) A total of 47 principal components was generated, of
which the first 9 accounted for 79.5 percent of the variance in the sediment data (Table 3-23).
ERA REPORT 3-36 SAC/143368(003.DOC)
7/31/02