Even though ERF is an estuarine salt marsh,
concentrations of white phosphorus.
large areas undergo prolonged seasonal drying.
The two compositing methods tested were
We have found that white phosphorus particles
found to meet the specific objectives for which they
are not persistent in these areas, and natural
were designed. The stirred compositing method
attenuation is one remedial process under inves-
permitted us to rank areas by average concentra-
tigation. Secondly, some ponds within the marsh
tions. Three mixing protocols were tested, all of
have been temporarily drained to promote dry-
which yielded data that approached a normal
ing of contaminated sediments. Remediation will
distribution (unlike the data from discrete sam-
be considered successful if the amount of white
ples). Of the three protocols, mixing a volume of
phosphorus in the sediments declines to a level
sediment with an equal volume of water en-
that is no longer hazardous to waterfowl. The
hanced mixing and consequently reduced vari-
need to evaluate remedial success led to this
ability between subsamples. In addition, when
study.
white phosphorus concentrations were based on
For this study we chose to sample an area that
dry sediment mass, there were no significant dif-
was known to contain white phosphorus that
ferences between the mean concentrations found
was available to feeding waterfowl: the DWRC
by manually stirring sediment and those found
pen 5 that was used for chemical bird repellent
by sampling the suspended sediment in a 1:1
studies. This pen is located near the edge of a per-
sediment:water mixture.
manently flooded pond, and, since 1990, when
The sieved compositing method enabled us to
we began our studies, has always been covered
determine the presence of solid white phospho-
with water. In the DWRC study, within this 7-
rus particles.
20-m area, a total of 54 mallards, in groups of six,
For future applications of these compositing
were penned for various lengths of time. Of these
methods, what size grid would we recommend?
54 mallards, 29 died of apparent white phospho-
The size of the grid depends on the acceptable
risk of missing a single hot spot (β), if one exists.
rus poisoning. In the studies where the mallards
were confined continuously for one week or
In all likelihood, more than one hot spot exists in
more, all the birds died (Table 2). This pattern of
highly contaminated areas of ERF or other impact
areas, as we found in the 7- 20-m area sampled
mortality says that, if the mallards took enough
"samples" from the sediments within the pen,
for this study. If more than one hot spot exists,
they eventually would hit a hot spot. Modeling
and we are still interested in finding only a single
our sampling approach after the behavior of the
hot spot with a probability of 90%, how might the
ducks, we also took numerous samples; however,
grid size be changed? If we know the number of
we chose the sample points systematically so that
hot spots (H), we may use a binomial distribution
we would have some idea of our risk of missing a
to compute the grid spacing needed to find a cer-
hot spot.
tain number (h) of these spots at an acceptable
The results from the discrete samples taken on
risk (Gilbert 1982):
a 1.82-m-square grid indicated that the white
phosphorus particles within this 7- 20-m area
Probability of hitting exactly h of H hot spots
were indeed localized in "hot spots" (Fig. 3). Spa-
H!
(1 - β) β(H - h).
h
=
tial heterogeneity was enormous, with concentra-
h!(H - h)!
tions varying over six orders of magnitude.
Unfortunately, we have no idea of how many
Recall the 25-m sampling interval originally used
hot spots exist in ERF. Gilbert (1982) presents two
at ERF; only one discrete sample would have
options when this knowledge is unavailable. The
been taken from an area this size. Given the range
first is to guess a likely number of hot spots,
of concentrations detected in this study, discrete
based on historical information, and use the bino-
samples taken at 25-m intervals will help to
mial distribution as shown above. The second is
roughly define the bounds of the contamination
more complicated, and involves choosing an
but cannot be used to draw conclusions about the
a priori Poisson distribution and using a com-
level of spot contamination unless blank samples
pound binomial distribution. We lack sufficient
predominate.
information for either approach. However, we
The assumption that the radii of the hot spots
can determine the extremes of grid spacing by
are around 1 m appears to be valid. For example,
systematically increasing the number of hot spots
two hot spots were located in row 7 (Fig. 3), and
in our calculation and seeing how grid size may
the samples adjacent to each had relatively low
16
Previous Page
