7942

• UCREFRP Catalog Number: 7942
• Author: Schmitt, C. J., et al.
• Title: National Pesticide Monitoring Program
• USFW Year: 1983.
• USFW - Doc Type: Organochlorine Residues in Freshwater Fish, 1976-79.
• Copyright Material: NO

5
pounds were eluted with 30 mL of 4% ethyl acetate in
benzene. These eluates were concentrated and made ready
for GLC analysis along with Florisil eluate B. Aliquots of
1 µL were injected into a Packard 805 gas chromatograph
equipped with electron capture detectors. The glass column
(1.8 m x 4 mm i.d.) was packed with 1.5% OV-17 +
2.0% OV-210 on 100/120 Gas Chrom-Q. Temperatures
were as follows: column, 200° C; injector, 225° C; and de-
tector, 300° C.
Fractions giving chlordane response were analyzed on
a column (1.8 m x 4 mm i.d.) containing 3% SE-30 on
100/120 Gas Chrom-Q at 180° C to resolve cis- and trans-
chlordane from trans- non achlor.
The 1978-79 cross-check analyses were done at the Great
Lakes Fishery Laboratory, FWS, Ann Arbor, Michigan.
The analytical procedures used there were described by
Snyder and Reinert (1971) and Hesselberg and Nicholson
(1981).
Data Handling and Statistical Analyses
Improved analytical methods freed the residue values
determined for the collection years 1976 through 1979 from
the intercompound interferences reported for earlier NPMP
organochlorine analyses (Schmitt et al. 1981). The present
data were therefore amenable to more rigorous statistical
testing than were those for previous years. We analyzed
transformed values of percent lipid, wet-weight residues,
and lipid-weight residues by analysis of variance (ANOVA),
using the mixed nested-crossed model
Y4ki = µ + Li + Pi + LPij + Sk(LP) + Ei,kl
where Yi;kl = transformed value from location i, collection
period j, species k, and sample l;µ = the grand mean;
Li = random location effect; P, = fixed collection-period
effect; LPij = location-collection period interaction; Sk
(LP) = random effect of species within each location and
collection period; and Eijkl = error (among samples of the
same species). The rationale for using this model to test for
collection-period differences and the details of the overall
testing procedures were described by Schmitt (1981).
We used the Statistical Analysis System (SAS Institute,
Inc. 1979), available through the University of Mis-
souri-Columbia computer system (Amdahl 470 V7), for all
data handling and statistical analyses. For trend analyses,
the following descriptive statistics were computed for each
compound, by collection period: mean (after appropriate
transformation), minimum and maximum concentration,
and percentage of stations at which each compound was
detected. Least-squares means (SAS Institute, Inc. 1979),
which are adjusted for the number of observations per cell,
were examined throughout. Percent lipid values were nor-
malized by using the angular transformation (Schmitt
1981). Lipid-weight residue concentrations were computed,
and both lipid-weight and wet-weight concentrations were
normalized by applying the (log,„ [residue concentration +
1.0]) transformation before computation of means and
further statistical analyses. Schmitt et al. (1981) discussed
the applicability of these procedures to residue-monitoring
data. In addition to descriptive statistics, we computed
annual matrices of product-moment correlation coefficients
to illustrate trends in the co-occurrence of compounds.
To examine more thoroughly for temporal trends, we re-
analyzed some NPMP results from 1974 (Schmitt et al.
1981). After recomputing the ANOVA for data from 1974
through 1979, we analyzed for differences among collec-
tion periods, using the sum-of-squares simultaneous test pro-
cedure to compare means (Sokal and Rohlf 1969). We then
used analysis of covariance (Snedecor and Cochran 1967)
to examine the significance and direction of the observed
changes in lipid-weight residues from 1974 through 1979.
The level of significance used in all statistical tests was
P s 0.05.
We used several statistical techniques to compare our re-
sults with corresponding values reported by the cross-check
laboratories. Summary statistics were computed for each
laboratory's results after the data were transformed (as
described), and percent occurrence was compared. For each
compound, the product-moment correlation coefficient (r)
between corresponding values was computed to give a
general indication of agreement. We also used linear regres-
sion (Snedecor and Cochran 1967) to test for overall agree-
ment, computing slope (b) and intercept (a) values for the
relationship y = a + bx, where y = cross-check value and
x = original CNFRL value. We then tested the coefficients
to determine whether the slopes (b) differed significantly
from 1.0 or the intercepts (a) differed from 0.0 (perfect
agreement between laboratories would yield relations in
which a = 0.0, b = 1.0, andv = 1.0). Overall differences
between laboratories were tested by using two-way ANOVA
for paired comparisons (Sokal and Rohlf 1969).
Results and Discussion
Characteristics and Limitations of the Data
Among the 112 active NPMP stations (Fig. 1), collections
were made at 106 (94%) in 1976-77 and at 108 (96%) in
1978-79; 102 stations (91 %) yielded data in both 1976-77
and 1978-79 (Table 2). This 102-station subset of data from
both periods contained 591 samples (95% of the 1976-79
data). There were 78 stations (out of a possible 113) with
complete data from 1974 through 1979 that could be
analyzed for temporal trends; this subset contained 753 ob-
servations (Table 2).
As reported previously by Schmitt et al. (1981), the
species collected at the NPMP stations have varied from year
to year. A total of 58 taxa were sent to CNFRL for residue
determinations in 1976-79 and 71 taxa in 1970-74. Overall,
the five most frequently collected species changed little from