Datasets:
patrickbdevaney
/
mia9

Dataset card Data Studio Files
xet
 Community
1
text
stringlengths
0
6.44k
-
- NO2
-
and dissolved inorganic nitrogen (DIN) was calculated as
NOX
-
+ NH4
+
. Details of sampling methodology and laboratory analysis have been
described elsewhere (Boyer and Fourqurean 1999; Boyer and Briceño, 2007).
Preservation procedure for DERM samples was by acidification of unfiltered
water samples, so adsorbed species and labile compounds, if present, would be
incorporated into solution. FIU samples were not acidified and either unfiltered for total
nutrients or filtered for dissolved nutrients; hence, direct comparison of some analytical
results between DERM and FIU is not recommended. A detailed inter-laboratory
comparison analysis was completed by Hunt and Todt (2006) using stations from FIU
and DERM programs. Samples were either co-located (four stations) or within a few
hundred meters of each other, and most reported measurements were from surface
samples. Only salinity and dissolved oxygen were reported for depths in excess of 1
meter. Salinity, DO, NOx and CHLa values reported from both programs were similar
and acceptable correlation were observed for salinity, DO and NOx between samples
collected within a 24-hour window. Total phosphorous values were slightly different
between programs perhaps due to differences in sampling schedule.
Major differences between DERM ammonia and FIU ammonium values were
impossible to reconcile stoichiometrically, perhaps due to differences in sample
preservation and analytical methods, let alone time and spatial differences in sampling
events. Acidification of unfiltered water samples by DERM would incorporate nitrogen
species, either contained in labile compounds or loosely bound on suspended particles,
to the liquid phase. This combination of methodological differences would render higher
total ammonia values for DERM than corresponding FIU ammonium data. In summary,
FIU’s dissolved ammonium measurements (Boyer and Briceño 2006) are not
comparable to DERM’s total ammonia measurements (Zhang et al. 1997). Each
parameter seems to express a totally different chemical character of BB waters.
9
Figure 1.1: Location of FIU sampling sites in Biscayne Bay.
Figure 1.2: DERM sampling sites in Biscayne Bay and its watershed
10
National Park Service data includes daily and monthly averages based on 15 min
samplings of temperature, conductivity, salinity and depth, from 47 sampling sites (Fig
1.3). Monitoring began in 2004 and it is still active.
Figure 1.3: Location of BNP sampling stations in Biscayne Bay.
The USGS data set contains groundwater, rain and well data. NOAA data sets
include rain and salinity measured along about-monthly cruises of BB. NADP data set
contain chemical Ca, Mg, K, Ca, NH4, NO3, Cl, SO4, pH, conductivity and precipitation.
Besides nutrients and physical chemical properties of surface waters in the bay
and its immediate watershed, we collected data for metals, pesticides, herbicides, and
organic and inorganic pollutants in waters and sediments to identify areas most
impacted by these contaminants. Among the studies, usually short-term and lasting
less than 4 years and/or covering restricted areas, we could mention those of Brand
(1986), Cantillo and Lauenstein (2004), Corcoran (1984, 1984b), McKinley (1995) and
Carnahan (2005, 2008).
11
QUALITY ASSESSMENT / QUALITY CONTROL
Water quality data was subjected to detailed QA/QC procedures. The initial steps
are summarized in Figure 1.4. In general it involved corrections for negative values,
flagging values below detection limit (BDL), detection of data gaps and potential
outliers, and correction of miscellaneous typographical errors and formatting. Detection
limits (DL) posed a special complication because of their variability along the period of
record (POR) and/or lack of reporting levels. When DLs were available, values below
detection limit (BDL) were flagged for further statistical analysis with applicable methods
(Helsel 2005). Special care was taken on suspected statistical outliers, because
ecological data may contain extremely high (i.e. CHLa data) or extremely low (i.e.
salinity) values which are in fact valid system responses, so their deletion or substitution
would impair a correct analysis. Outlier identification in data sets was part of the data
screening process done routinely before further statistical analysis. The simplest case
was the identification of univariate outliers in symmetric distributions where extreme
data values to the right or left are the obvious outlier candidates. Generally in our case,
most potential outliers to the left are bundled together in the non-detects and are
incorporated as such in the statistical treatment. In skewed distribution, as in most
ecological data sets, the suspect outliers are likely to be the extremes of the longer tail
(extreme high).
Figure 1.4: Summary of QA/QC procedures and summary statistics calculation for WQ data
12
Criteria for assessing if a value is an outlier are complicated (Burke 2009;
Montgomery 2001) because ecological data do not necessarily follow a specific type of
statistical distribution. We applied the golden rule of ―no value should be removed from
a data set on statistical grounds alone‖. First, we used box-plots as non-parametric tool
to display data without making any assumptions of the underlying statistical distribution
to isolated potential outliers. The box and- whisker plot is a powerful statistic as it shows
the median, range, the data distribution as well as serving as a graphical, nonparametric
ANOVA. The center horizontal line of the box is the median of the data, the top and
bottom of the box are the 25th and 75th percentiles (quartiles), and the ends of the
whiskers are the 5th and 95th percentiles. The notch in the box is the 95% confidence
interval of the median. When notches between boxes do not overlap, the medians are
considered significantly different.
Each suspected outlier was analyzed individually to assess its validity when the
overall time-series was displayed together with its corresponding z-scored running-sum
(Zcusum) plots. If the value belonged to a dataset where extreme values are expected,
like in CHLa or salinity, we verified if the occurrence(s) responded to seasonality or to
extreme events of canal flows or storm events. If so, the value was not considered an
outlier favoring knowledge-based criteria over purely statistical considerations. If the
value was a unique (way above/below any other value) and isolated occurrence, not
appearing periodically and/or was not accompanied by high values before and/or after
its occurrence, it was flagged as outlier and was discarded before calculation of
summary statistics. This approach, although time-consuming and perhaps tedious,
renders the best results. Major complications arise when analyzing metal and pollutant
concentrations. These data are usually the result of eventual sampling responding to
specific needs, hence, data is scattered in space and time, values show extreme
spread, and a direct connection with other data sets is rarely possible. Finally, after
QA/QC procedures were completed, data were consolidated in a Master Database as