20080868 Ver 2_Appendix A -Entire_20190628

APPENDIX A
History and Methods by Parameter

APPENDIX A. HISTORY AND METHODS BY PARAMETER
Progress of the mine across a watershed is determined on an annual acreage basis. Some
monitoring parameters are collected once a year only and others are collected throughout the
year. Data collection for a parameter in a specific watershed may occur prior to the mine
progress through that watershed for that year or, in the case of parameters collected throughout
the year, mine progress through the watershed may have affected only the data collected in the
last portion of the calendar year. The NCDWQ agreed during the previous study that
"significant measurable results" were not likely to be determined "until 10 percent or more of the
basin was impacted". The 10 percent threshold was developed specifically for Alt E impacts in
Tooley Creek yet remains an important factor in the determination of the first post -Mod Alt L
year in a given creek.
To determine the percentage of basin impact and to increase accuracy of analysis, drainage
basin quantification is recalculated annually at the end of each mine advance year. Over the
course of the creeks study, basin acreage calculations have been refined as past activities were
uncovered and digital tools improved (e.g., LiDAR). In calculation of the percent impacts to
assign 2013 as pre- or post -Mod Alt L for a given creek (10 percent reduction threshold on
current basin), it was discovered that the drainage basins of the NCPC creeks had actually been
reduced prior to Alt E by the construction of a canal inside the Alt E boundary dug in the late
1970s or early 1980s. This canal cut off the NCPC creeks from portions of their historic
drainage basin and resulted in several corrections. It reduced the historic basin acreage, the Alt
E baseline acreage, and the basin remaining after Alt E for each NCPC creek shown in previous
annual reports. These reductions affected the pre -Alt E and pre -Mod Alt L basins and
subsequent percent reductions shown for some creeks in previous annual reports. Boundaries
of drainage basins have not been revised since 2013.
A. Low Flow Stream Monitorin
1. History
With the concurrence of regulatory agencies and subsequently, the Science
Panel (committee chosen to review this report), flow monitoring with weirs was discontinued at
the end of 2009 in Huddles Cut and the end of 2010 in Tooley Creek. While no monitoring had
occurred in Jacks Creek since 2005, the weir in Jacks Creek was removed in 2010 when the
weirs in Tooley and Huddles Cut were removed. However, PCS elected to install a new device
developed specifically to monitor low flow events in upper headwater stream mitigation sites in
the coastal plain in appropriate creeks monitored for this study. Two flow gauges were installed
in Porter Creek in April 2010 and four were installed in Duck Creek in March 2011. The most
downstream gauge in Porter Creek is within the Mod Alt L boundary and will not collect any
post -Mod Alt L data; and therefore is not part of the 2012 creeks report.
The upstream Porter Creek segment monitored by the other low flow gauge is
estimated to have a 117-acre drainage basin. One Duck Creek segment with two of the four
gauges is the headwaters of the northwest prong (-124-acre basin) and the other segment is
the south fork of the east prong (-151-acre basin).
Beta models of the low flow gauges were produced by Flowline Products and
installed in some PCS stream and wetland mitigation sites in 2010. Refinement of the beta
model and larger scale production by Remote Data Systems, Inc (RDS) has been stalled; future
use of these devices is uncertain. The gauges are used to document the duration and relative
magnitude of flow events occurring in shallow perennial, intermittent, and zero order stream

segments. The device is based on the principles of a variable area flow measurement device
and consists of a vertical baffle mounted on an axis inside a protective housing. Flow passing
through the housing causes the baffle to tilt on its axis, and the degree of tilt is recorded by the
internal electronics and logging device. The gauge is capable of reliably recording flows as low
as 0.5 gal/min. Once a month biologists download the gauge and visually determine the relative
degree of flow as well as water depth at each device location. Once downloaded from the flow
gauge, the voltage setting for each gauge is calibrated during analysis such that dry conditions
and times of no flow (from observer data) are graphically depicted as a "flatline" or zero. A flow
event is defined as flow >_0.5 gallons per minute (gpm) unless noted otherwise. Both rainfall in
the watershed and hydrology of adjacent wells are also used to double check the accuracy of
the flow gauge data.
By late 2013, the low flow gauges were becoming more prone to error and it was
increasingly evident that production of additional units was not likely. Therefore, PCS decided
to cease monitoring flow with these units as their inclusion in the creeks monitoring program
was completely voluntary.
2. Methods
Since March 2012 in all creeks at the intended locations of the low flow gauges,
biologists also make qualitative observations of flow (low, medium, high) and measure water
depth in the channel during the monthly download of adjacent wetland wells as indicated by the
red arrow shown on the monitoring locations figure for each creek. Evidence of recent flow is
noted and photographs or video are taken as appropriate. The observations are presented in
table form on an annual basis and photographs or video are retained at CZR and made
available upon request. With no further factory production anticipated and cessation of
monitoring of the low flow gauges, these observations will continue and will serve to document
flow events.
B. Salinity Monitoring
1. History
A CAMA permit (number 66-11) was received by PCS in May 2011 for
construction and/or rehabilitation of 23 piers located in the monitored creeks. Stainless steel
boxes were mounted on these piers to house the water level and salinity monitoring equipment.
Equipment was installed as the piers were completed. By July 2011 all salinity monitoring
equipment was in place per the plan and salinity monitoring began in 12 of the 15 creeks (six of
which were controls) identified in the plan for the first time in 2011. During 2013, salinity was
monitored at seven creeks that have had or will have their watershed disturbed by Mod Alt L
mine activities and at eight control creeks/locations. In 2018, a new CAMA General permits
covered new piers for the 15t" creek identified in the plan (Broomfield Swamp Creek) and its
nearby control creek (an unnamed tributary to South Creek, SCUT1). With the exception of the
control stations in South Creek, Pamlico River, PA2, Durham Creek, the two unnamed
tributaries to Durham Creek, the three stations on Huddles Cut (two upstream and one
downstream), and Broomfield Swamp Creek and SCUT1 which each have three salinity
stations, all other creeks monitored for salinity have an upstream (#1) and a downstream (#2)
salinity station. The control locations aid in the determination of whether changes that may
occur between pre- and post -Mod Alt L monitoring are also seen regionally.
During the same timeframe as the monitoring on creeks that were subject to, or
would be subject to, drainage basin reduction from the activities of the mine, control locations
were also monitored for salinity:
Uvtl

• one in the Pamlico River established in 1998 (PS1),
• one in South Creek established in 1998 (SS1),
• two in Long Creek (LOCS1; LOCS2) established in 2011,
• two in Little Creek (LCS1; LCS2) established in 2011,
• one in PA2 (PA2S1) established in 2011,
• two in Duck Creek (DKCS1; DKCS2) established in 2011,
• one in DCUT19 (DC19S1) established in 2013,
• one in Durham Creek (DCS1) established in 2013, and
• three in a tributary to South Creek (SCUT1) established in 2019.
Muddy Creek is another control creek; however, only fish, sediment, and benthos
data are collected. Fish sampling has occurred weekly April through June every year since
1999 except for 2006, when no monitoring occurred for any parameter, and a handheld YS1 Pro
Plus Quatro is used to collect water quality data, including salinity, at each visit. Therefore,
salinity for those three months can be compared over the years to determine if there are any
patterns in that creek.
2. Methods
A YSI 600XLM multi -parameter water quality sonde (YSI) was utilized to record
salinity and water depth at each monitoring location during earlier monitoring (1999-2005) and
from January 2007 to August 2008. Salinity monitoring after August 2008 was performed with
In -Situ AquaTROLL 200 CTD loggers. The YSI sondes automatically calculated salinity
readings from conductivity and temperature. The salinity sensor range for each YSI was 0-70
parts per thousand (ppt), with an accuracy of +/- 1.0 percent of the reading, or 0.1 ppt,
whichever was greater. The resolution was 0.01 ppt. The depth sensor was a stainless steel
strain gauge pressure sensor with a range of 0-30 feet, an accuracy of +/- 0.7 inch, and a
resolution of 0.01 inch.
The Aqua TROLL 200 CTD logger is manufactured by In -Situ, Inc. Like the YSI
sondes, the Aqua TROLLs generate salinity readings from temperature and conductivity. The
salinity sensor range is 0-42 practical salinity units (psu), with an accuracy of +/- 0.5 percent of
the reading. The resolution is 0.001 psu. Practical salinity units (psu) are essentially equivalent
to parts per thousand (ppt); however, psu is considered a more appropriate descriptor since it
refers to the practical salinity scale that is used to calculate salinity (Reid 2006). The depth
sensor is a titanium/silicon strain gauge pressure sensor with a range of 0-35 feet, an accuracy
of +/- 0.003 inch, and a resolution of 0.001 inch.
The Aqua TROLLs are programmed to record a salinity and depth reading every
1.5 hours (16 readings per day). Each TROLL is downloaded every two weeks and both
conductivity and depth calibrated as necessary. The probes are also cleaned and batteries
checked and replaced as necessary. Sensors are located near the bottom of the stream to
ensure continuous data collection during most low water conditions. Depth readings are
compensated for the distance from the sensor to the creek bottom. Occasional gaps in the
continuous data in some years exist due to dead batteries, equipment malfunctions, and low
water levels not allowing sensors to be fully submerged. At the time of download, the handheld
YSI Pro Plus is used to determine the accuracy of the Aqua TROLL readings.
To aid in the interpretation of factors that may influence salinity fluctuations,
continuous salinity data from each Aqua TROLL are displayed on graphs along with the
continuous water level data from that monitor, daily rainfall from the nearest rain gauge (PCS
Aurora NOAA Station), and data from the Tar River U.S. Geological Survey flow gauge at
A-3

Greenville, NC (http://waterdata.usgs.gov/nwis/uv?02084000 (Appendix C). The Tar River
becomes the Pamlico River at the US Hwy 17 Bridge in Washington. These graphs were used
for a qualitative assessment of the relative effects of wind tides, local drainage basin input, and
Tar River input on salinity fluctuations in the monitored creeks. Wind data from the PCS NOAA
weather station were compared with the graphs for the 2012 monitoring year report to determine
if there was any effect from wind. Spearman Rank Order correlation tests were conducted to
test the correlation among rainfall, discharge, and salinity at the monitors. This wind analysis
may be conducted in subsequent years as applicable.
T-tests and ANOVAs were utilized to attempt to find differences between pre -
Mod Alt L salinity and post -Mod Alt L salinity but normality failed in the t-test using either
Shapiro -Wilk or Kolmogorov-Smirov, so a Mann -Whitney Rank Sum or Kruskal-Wallis One -Way
ANOVA on Ranks was used at the recommendation of the SigmaPlot® software.
C. Wetland Hydrology Monitoring
1. History
Monitoring of wetland hydrology occurred in Jacks Creek, Tooley Creek, and
Huddles Cut under the 1998 plan and was expanded to include seven new creeks under the
2011 plan, two of which (DCUT11 and DCUT19) were first monitored in 2013. In 2019,
Broomfield Swamp Creek (one of the three named creeks in the S33 Tract permitted to be
impacted by Mod Alt L) were included in the study along with SCUT1, a nearby unnamed
tributary to South Creek, as a control.
The location of wetland wells did not change from previous years for the three
creeks monitored under the 1998 creeks study, with the exception of the loss of three well
locations at Huddles Cut due to activities associated with the permitted mine advance, and the
addition of four more wells at Jacks Creek. The three wells lost at Huddles were in the most
upstream reaches of the west prong (HWW1, HWW10, and HWW11). Three of the new wells
added in Jacks Creek (JW5, JW7, JW9) were installed in close proximity to an original well
location so that a broader area of the floodplain could be monitored semi -continuously and a
fourth one (JW2B) was installed a short distance upstream of what was previously the most
upstream well.
2. Methods
Both manual wells and semi -continuous recorders are used to monitor the water
table. Each manual well consists of a 54-inch length of 1 1/4-inch diameter PVC well screen
(0.010-inch slots) and an 18-inch long riser made of solid -walled 1 1/4-inch diameter PVC pipe.
The well screen and riser are connected by a PVC coupler. The manual wells were installed to
a depth of 60 inches, with 12 inches of the riser extending above the ground. The top of the
riser is covered by a PVC cap, and a hole in the side of the riser provides air exchange during
water level fluctuations. Manual wells were installed with the original wells in Jacks, Tooley, and
Huddles to serve as controls to cross reference with the new electronic recording devices (WLs
— at the time). Manual measurements are now collected from the same well screen harboring
the electronic recorder for other monitoring sites.
WL/Ecotones (RDS products) were originally used to collect semi -continuous
water level data. In 2011, all new creek monitoring locations and all but two existing Ecotones
were replaced with Level TROLL 500 water level loggers manufactured by In -Situ. The Level
TROLL is housed inside a 3-inch diameter PVC well screen (0.010-inch slots) installed to a
depth of approximately 50 inches. The current configuration of cable and Level TROLL record
water levels over a range of about five feet, and installation varies according to site conditions,
/_QI

but in general most wells record water within 2 feet above and below the ground surface. The
TROLLs record the water level every 1.5 hours (16 times per day). To prevent damage by
bears, the aboveground portions of the well screens were surrounded by a fence enclosure
made of metal T-posts to encase the well with strands of barbed wire.
All monitoring wells were checked and downloaded once a month. Wetland
hydroperiods were calculated for each monitoring well during the growing season. A
hydroperiod is defined as consecutive days during the growing season that the water table is
within 12 inches of the surface or the surface is inundated, and is expressed as a percentage of
the growing season. For this project, the growing season is defined by the Beaufort County soil
survey (Kirby 1995) as 14 March through 24 November (256 days). Growing season dates
have recently been adjusted by the Regional Supplement to the Corps of Engineers Wetland
Delineation Manual: Atlantic and Gulf Coastal Plain Region (Version 2.0) (USACE ERDC 2010)
to match the Natural Resources Conservation Services' (NRCS) WETs tables. However, the
previously established soil survey growing season dates and Corps 1987 wetland definitions will
continue to be used for this report in order to maintain consistency with baseline years in
hydroperiod calculations. Hydroperiods were statistically compared to explore pre- and post -
disturbance differences and between creek variation. All statistical analyses were performed
with Sigma Plot 11.0 using a simple t-test, Kruskal-Wallis One-way Analysis of Variance
(ANOVA) on Ranks or Mann -Whitney Rank sum, depending on results of normality tests and
what was being compared.
Rainfall was recorded by tipping buckets at seven locations in the creek study
area, usually within a study creek basin itself, or at least in near proximity (e.g., Jacobs Creek,
PA2, and Drinkwater Creek use the same rain gauge).
D. Water Quality Monitoring. (Section III. D. was prepared by Dr. Enrique Reyes,
a faculty member of East Carolina University (ECU)).
1. History
Water quality monitoring sites on three creek systems were initially established at
the beginning of the creeks monitoring in 1998 as follows: (1) two locations in Jacks Creek, (2)
three locations on Tooley Creek, and (3) four locations on Huddles. These stations were
monitored in accordance with the 1998 plan and continued under the final 2011 plan. By
December 2011, two stations each in seven additional creeks (four control creeks and three
creeks to be impacted) had been added such that ten creeks designated for water quality
monitoring were part of the regular program (no water quality samples have ever been collected
in Muddy Creek). Once the salinity locations were established, collection/submission of water
quality samples was gradual in order for the ECU laboratory to ramp up their analysis and
throughput.
Water quality stations at two locations were added in the following creeks in
2011: Jacobs Creek- JCBWQ1 near the upstream salinity station and JCBWQ2 near the old
railroad trestle ; Project Area 2 (PA2) - PA2WQ1 at the upstream end of the main channel and
PA2WQ2 at the midstream salinity station; Drinkwater Creek- DWWQ1 at the upstream well
array and DWWQ2 near the upstream salinity station; Little Creek- LCWQ1 and LCWQ2, one at
each of the salinity stations; Long Creek- LOCWQ1 and LOCWQ2, one at each of the salinity
stations; Porter Creek- one downstream of the most upstream well array (PCWQ1) and PCWQ2
at the upstream salinity station; and Duck Creek- DKCWQ1 at the upstream salinity monitor
and DKCWQ2 at the downstream salinity monitor. With the addition of two stations in 2013 on
two small unnamed tributaries (UTs) to Durham Creek (DC11 WQ1 and DC19WQ1), all water
quality stations north of NC 33 designated in the study plan are in place. The headwaters of the

Durham Creek tributary DCUT11 will be impacted by the mine continuation and DCUT19 serves
as the control. After consultation with USACE and NCDWR in 2017, one of the S33 Tract creeks
permitted to be impacted was chosen to be monitored (Broomfield Swamp Creek); subsequently
a nearby creek was also included as control (SCUT1). In 2018, limited pre -Mod Alt L data
collection occurred at some locations in Broomfield Swamp Creek and SCUT1; however, as the
piers were not completed and only partial water quality samples were collected, 2019 will be
considered the first pre -Mod Alt L year with complete data for later comparison.
2. Methods
Water quality data were collected from study creeks throughout the year with 26
potential sampling periods; however, insufficient water depth resulted in some sites not being
sampled every period in various years. An annual summary list of samples collected from all
monitoring stations is made each year and included in an appendix in the annual report.
At the 2018 agency/Science Panel meeting to discuss the 2017 report, there was
discussion about consistency of water sample collection locations for analysis. Samples will
continue to be collected as shallow as 1 inch water depths, but with the same precautions
previously followed to avoid substrate collection along with the aqueous sample. In past dry
conditions at the normal collection location, samples were sometimes collected from the closest
pool or puddle in the stream channel and so noted on the data form. Going forward, when the
actual station location is dry, the sample can be collected from up to 20 feet upstream or
downstream, only if that sampled water in the channel is directly connected to receiving waters;
no puddles or pools will be sampled.
CZR biologists collected field measurements in conjunction with water sample
collections. Field measurements included water depth, Secchi disk depth, temperature, salinity,
conductivity, turbidity, dissolved oxygen, and pH. Water depth was measured to the nearest
0.25 inch in close proximity to the site where water samples were collected and all other
measurements taken. Temperature, salinity, conductivity, dissolved oxygen, and pH was
measured with an YSI 85 (older years) and a YSI Pro Plus Quatro (recent years) multi -
parameter water quality instrument. These measurements were made in the middle of the
water column when possible. Turbidity was measured with a Hazco DRT-15 Portable
Turbidimeter (early years) and a LaMotte 2020 turbidimeter (recent years). Turbidity water
samples were collected in the field and turbidity was measured at the time of collection. Care
was taken to exclude detrital particles from the substrate and surface in turbidity samples.
The creek water quality samples were collected in polyethylene bottles and
driven directly to the ECU lab in iced coolers at the end o the collection day. At the lab,
subsamples were taken for the various analyses. Pre-combusted Whatman 934-AH (glass
fiber) filters were used to separate particulate and dissolved fractions. The filtrate was stored
frozen in a polyethylene bottle for later analyses of total dissolved phosphorus (TDP), dissolved
orthophosphate (PO4-P), ammonium nitrogen (NH4-N), nitrate nitrogen (NO3-N), and dissolved
Kjeldahl nitrogen (DKN). The filter pads were also stored frozen for particulate nitrogen (PN),
particulate phosphorus (PP), and chlorophyll a determinations. Fluoride data, although not a
required monitoring parameter for the Creeks study, were included in reports and evaluations
through 2011. Parameters added to the study included dissolved organic carbon (DOC mg L-)
and total dissolved nitrogen (TDN mg L-'), with preliminary data in 2012 and with the first full
year of data beginning in 2013. Techniques used for these analyses are summarized in Table-
D1. These methods are identical to those used for the PCS Phosphate Pamlico River estuary
water -quality monitoring program (see Stanley 1997 for example).

Table-D1. Summary of techniques used for chemical and physical measurements.
Parameter
Technique
Reference
Total Dissolved Phosphorus
Persulfate Digestion
APHA (1998)
Particulate Phosphorus
Kjeldahl Digestion
APHA (1998)
PO4-P
Molybdate
EPA (1979)
NH4-N
Colorimetric
Solorzano (1969)
NO3-N
Cadmium Reduction
Strickland and Parsons (1972)
Dissolved Kjeldahl Nitrogen
Kjeldahl Digestion
APHA (1998)
Particulate Nitrogen
Kjeldahl Digestion
APHA (1998)
Chlorophyll a
Colorimetric
Strickland and Parsons (1972)
High temperature catalytic
Sugimura, Y., Suzuki, Y. (1998)
Dissolved organic carbon
oxidation with non -dispersive
Marine Chemistry, v24: 105-131
infrared detector
High temperature catalytic
Walsh, T.W., 1989; Marine
Total dissolved nitrogen
oxidation with
Chemistry, v26: 295-311
chemoluminsence detector
In order to reduce the amount of information presented in both graphical and
table format in this section, multivariate analysis approach was initiated in 2012 and has
continued since with some modifications. The addition of data from all years sampled - was first
analyzed for the 2013 report and divided into three distinct subsections: temporal variability for
all water quality stations analyzed separately across all years, spatial variability for all water
quality stations across all years, and temporal variability across pre -Mod Alt L and post -Mod Alt
L years (CZR 2014 annual report on data through 2013). In the 2013 report (on 2012 data),
only the calendar year data were analyzed and presented with multivariate techniques. In all
reports prior to 2013, data were summarized in up to five standard graphical formats and
depicted with box/whisker plots and standard deviations.
Temporal variability at water quality stations across years was analyzed by
employing a Principal Components Analysis (PCA) to recombine all water quality variables into
principal components that capture the intercorrelation between variables over time. PCA has
two primary uses: 1) to describe interrelationships between a matrix of intercorrelated variables
and 2) data reduction, i.e. to reduce a large matrix of intercorrelated variables into linear
recombinations of the original variables. PCA plots all original variables in multidimensional
space (one dimension for each variable), and then fits a regression line through the multiple
dimensions. The principle of least squares is used to fit the regression line and the resulting
variance explained is recombined into a principal component. This principal component (PC) is
a combination of the original variables and explains a fraction of the total variables. This
procedure is repeated, i.e. a new regression line is fitted to the remaining data, and a new
principal component is calculated and generated, until all variability is explained. The PCs

themselves are uncorrelated to each other and therefore may be used in further modeling
without violation of regression assumptions. The PCs can be related to the original variables by
examining the loadings on to each PC and these values represent the degree of correlation of
each original variable to the new PC. In order to examine how the PCs are related to one
another, a biplot is often generated that shows how the original variables are related to the first
two PCs. The final result of PCA is a set of PC values that represent new variables, made from
the original variables, but fewer in number and uncorrelated to each other. Plotting the PC
scores over the course of the years shows the temporal variability of multiple variables, without
the need to generate numerous plots. A PCA analysis was run for each water quality station and
the biplot, loadings, and PC time -series are presented.
Spatial variability among stations across all years was analyzed by comparing
the mean water quality parameters of each water quality monitoring station using a cluster
analysis approach. The approach groups water quality monitoring stations with similar
conditions together, demonstrating the relationship between each station. The benefit of this
over multiple years is to demonstrate how relationships between stations may change over time.
Water quality conditions for each group of stations were then summarized graphically. Briefly,
cluster analysis is a multivariate technique that analyzes similarity or dissimilarity computed
from a data matrix. For example, a single water quality monitoring station may be characterized
by multiple water quality measurements. Thus, one might ask: how similar are two water quality
stations from two different creeks based on all of the water quality measurements? While it is
straightforward to compare salinity values between the two creeks, the addition of more
parameters makes the comparison difficult. In order to accomplish this comparison the data
matrix of water quality values have been grouped by station and a dissimilarity matrix was
calculated. This is done by plotting the water quality values for each parameter in multivariate
space. Instead of fitting a regression line, as with PCA, the Euclidean distance is calculated
between the water quality values for each station. Values that are close to each other in space
have low dissimilarity and values that are for apart in space have high dissimilarity. Once the
dissimilarity matrix has been computed, the dissimilarity values for each station may be
clustered and displayed using a dendrogram (tree diagram). Height is typically used as the y-
axis for such a graph and the higher the height of a branch split, the more dissimilar the stations
that follow such a split.
Finally, the interannual variability in water quality variables for those creeks with
pre- and post -Mod Alt L data was compared. Pre- and post -mod Alt-L conditions were
compared using a one-way ANOVA and t scores and p-values are presented. Differences were
considered significant if p-values were < 0.05.
For all statistical analyses of laboratory results, when the limits of detection for a
specific method returned a "< the method detection limit", the "<" was removed and the limit
entered as a value in the spreadsheet. This value is the limit of 100 percent confidence in
detection; while the value may actually be less, if it were present it would not be higher than this
limit. This practice prevents many "gaps" in the graphical depictions and increases the value of
the statistical analysis.
E. Metals Sampling
1. History
From 1998 to 2010, the PCS sediment samples were analyzed by Dr. John
Trefry at the Florida Institute of Technology (FIT). In 2011 Dr. Trefry informed CZR and PCS
that his workload and research interests would prevent his involvement with future PCS
sediment analyses. To date, a search for an alternate university or commercial laboratory which
P

is both interested in the work and can duplicate the FIT laboratory report and perform total
metals digestion with either hydrofluoric or perchloric acid has been futile. (Commercial facilities
contacted do not digest samples with these acids due to special health, safety, and
instrumentation requirements; university facilities contacted either lacked the ability to commit to
the project or lacked the necessary equipment.) An alternate laboratory, SGS North America,
Inc. now SGS Analytical Perspectives (SGS) in Wilmington NC has performed the metals
analyses since 2011 for the water column and since 2012 for the sediments. In 2016, SGS
Analytical Perspectives acquired the assets of Accutest Laboratories USA and became SGS
Accutest-Orlando. In January 2018, it became SGS North America Inc. — Orlando. The PCS
metals sample analysis was transferred to the Orlando laboratory in 2017. Sediment samples
analyzed by SGS follow sample preparatory Method 3050B under the USEPA SW-846 series,
followed by inductively coupled plasma (ICP) or ICP/mass spectrometry analysis (ICP/MS) by
Method 6010C or Method 6020A. Without the use of the stronger acid(s) used by FIT, the
3050B method is not a total digestion technique but one that dissolves "elements that could
become environmentally available." The remaining residue after digestion by this method is
referred to as "environmentally inert material". However, without the stronger acids, the results
are not likely to be total metal concentrations, although the concentration measured is the total
amount recoverable by the method used. In previous creeks annual reports, the FIT lab
provided a summary report with tables of results and environmental interpretation but no raw
data. By their governing standards, commercial facilities provide a "report" of results with no
environmental interpretation or opinion. From 2013 - 2018, CZR contracted Dr. Jamie DeWitt of
ECU Pharmacology and Toxicology, to provide environmental interpretation of the metals data
and to serve as reviewer and advisor for this parameter; due to other commitments, a new
advisor(s) will be contracted for future review and interpretation of metals data.
Elevated metal concentrations in sediments can be evaluated within the context
of potentially harmful biological effects using the guidelines developed by Long et al. (1995). In
these guidelines, two assessment values are listed for several metals. An effects -range -low
value (ERL) and an effects -range -median value (ERM) are defined as the 10th and 50th
percentile, respectively, from an ordered list of concentration of substances in sediments that
are linked to a biological effect. Several authors have noted that sediment quality guidelines
should be used cautiously with an appropriate understanding of their limitations. For example,
Field et al. (2002) noted that the ERL is not a concentration threshold for a chemical in
sediment, above which toxicity is possible and below which toxicity is impossible. Instead,
according to O'Connor (2004), the ERL is a concentration "at the low end of a continuum
roughly relating bulk chemistry with toxicity." The ERL and ERM values are applied to the
sediments from this study with the caveats listed above.
2. Methods
Prior to the collection of sediment samples, the water column sample is collected.
Leaning over the bow of the boat while it is slowly underway, using a sample bottle provided by
the laboratory which contains HNO3 preservative, a gloved biologist fills the 1 L sample bottle
with creek water using a separate container which has been cleaned with deionized water and
alconox. The sample bottle is then sealed, labeled, and double bagged. As a back-up sample,
the biologist also collects a 0.5L bottle using the same process. Samples are kept in the dark in
iced coolers and delivered/shipped for receipt at the laboratory using the chain of custody form
provided by the laboratory. The samples are analyzed for concentrations of silver (Ag), arsenic
(As), cadmium (Cd), chromium (Cr), copper (Cu), iron (Fe), molybdenum (Mo), selenium (Se),
and zinc (Zn) (aluminum is not commonly analyzed in marine/estuarine water samples). In
2013, the sediments samples were also analyzed for total organic carbon (TOC) for the first
time. Standard laboratory protocols and procedures for these analyses are strictly followed.

For sediments, the ponar device is deployed and retrieved from the boat and the
collected sediment is dumped into a plastic tray from which -0.5 gallon of sediment is scooped
from the sample into a Ziploc bag using a plastic or stainless steel scoop. As the sediment
sample is scooped, sediment that may have touched the metal of the ponar is avoided. Each
bag is labeled with creek name and date and, to minimize the potential for leaks, each sample is
double bagged. The ponar itself, plastic tray, and scoop/spoon are thoroughly rinsed with
deionized water between each sample to avoid cross -contamination. A second sample is
collected from each station in case there is a problem with the shipment to the laboratory or a
problem encountered by the laboratory during analyses. The backup samples are kept at CZR
until results of the analyses are completed at which time the samples are discarded. The
sediment samples are delivered chilled to the laboratory. The samples are analyzed for
concentrations of aluminum (AI), silver (Ag), arsenic (As), cadmium (Cd), chromium (Cr), copper
(Cu), iron (Fe), molybdenum (Mo), selenium (Se), and zinc (Zn).
Since 2016, per 2015 EPA Publication SW-846, certified laboratories replaced
Method 6010C with Method 6010D and Method 6020A with Method 6020B for the ICP or
ICP/MS portion of the analysis. Bulk density of the sediments is measured by ASTM C 29/C
29M by a subcontractor to the SGS laboratory (CATLIN Engineers and Scientists, Wilmington
NC). Total organic carbon (TOC) measurement by Method 9060A was added to the SGS
sediment analysis in 2013.
For statistical analysis of data, a two-way ANOVA is not possible when only one
sample is collected per creek (no replicate samples) and when a year is not represented with
more than two creeks. While it is possible to treat individual creeks as pseudoreplicates, for
some years there are not a sufficient number of samples from the post -Mod Alt L creeks to
perform a two-way ANOVA with a repeated factor. In lieu of two-way ANOVA, pairwise
comparisons (i.e., Student's t-tests) were run for individual sediment metals and water column
metals. However, due to the lack of replicates per creek, creeks are treated as
pseudoreplicates, which decrease the quality of the statistical analysis.
For all statistical analyses of laboratory results, when the limits of detection for a
specific method returned a "< the method detection limit", the "<" was removed and the limit
entered as a value in the spreadsheet. This value is the limit of 100 percent confidence in
detection; while the value may actually be less, if it were present it would not be higher than this
limit. This practice prevents many "gaps" in the graphical depictions and increases the value of
the statistical analysis.
F. Vegetation Monitoring
1. History
Under the 2009 permit conditions, annual vegetation monitoring was no longer
necessary and the Corps suggested a wider interval (approximately every three to five years)
appropriate to each creek after baseline data were collected. In addition to the wider interval
that began in 2010, Hurricane Irene damage in 2011 severely limited access in Tooley Creek
and the main prong of Huddles Cut, both of which were on the calendar to be monitored that
year. The storm damage and high amount of standing water (in the main prong of Huddles Cut)
precluded any monitoring in 2011 in these areas of Tooley Creek and Huddles Cut main prong.
Additionally, the Science Panel agreed at the 2013 annual meeting that of all the
study parameters, vegetation is likely to have the longest "lag effect", the period of time between
impact and the response to that impact. Therefore, with the concurrence of the Science Panel,
it was determined that from 2013 forward, vegetation also does not need to be monitored during
A-10

the "lag effect", or transition years. A transition year is considered to be the year(s) when the
mine actually moves through a creek basin and also includes the first year after the mine has
completed all impacts to a basin.
The transition years vary depending on the size and shape of the creek basin
within the permitted mine area. While Drinkwater Creek is the first creek tracked with the
transition year approach as described above, because of proximity to the mine advance,
baseline data for Drinkwater Creek were only collected for one year (2011) before 2012 became
the first transition year. After the 2011 survey, it was agreed by the Science Panel that
monitoring in Drinkwater Creek, Tooley Creek and Huddles Cut would not occur in 2012.
Wetland vegetation data are collected from all non -transition creeks each year
that have had or will have their watershed disturbed by mining activities as well as from three
control creeks -Long Creek, DCUT19, and Duck Creek. The Science Panel and or agencies
will determine how many years of post -Mod Alt L data are sufficient in a specific creek or
whether additional gaps in monitoring can occur before all vegetation monitoring ceases in that
creek. At the August 2017 Science Panel meeting (2016 data report review) further
clarifications were discussed about the interval between post -Mod Alt L vegetation surveys. It
was agreed that as originally planned, five years of pre -Mod Alt L vegetation data would be
collected when possible, that four years of post -Mod Alt L data would be collected after the
transition year(s), and that after four years of post -Mod Alt L vegetation data collection the fifth
year would be skipped. In a certified letter sent to USACE and NCDWR on 12 September 2017,
PCS made further clarifications on the frequency and proposed that after the fifth year post -Mod
Alt L (the first skipped year), that vegetation be surveyed every other year until such time that
the agencies or Science Panel members deem otherwise. At the 2018 Science Panel meeting
(2017 data report review), the agencies agreed to the interval described in the PCS September
2017 letter. A copy of the August 2018 USACE letter of approval for the clarified vegetation
monitoring interval is included in Supplement 1 of the 2018 data report.
2. Methods
Drainage basin modifications are most likely to affect vegetation communities
found along the relatively narrow riparian wetlands upstream of the CAMA jurisdiction markers.
For this reason, vegetation assessments and monitoring sites were concentrated in these areas.
Data are collected in August or September each monitoring year. Most transects are in the
upper reaches of the creeks, although the distance from the mouths of the creeks differs among
the creeks. Vegetation transects at Jacks Creek, Tooley Creek, Long Creek, and Huddles Cut
are all between 3,000 and 5,000 feet (0.5 mile to 0.9 mile) from the mouth of the creek, Jacobs
and Drinkwater creeks' are between 5,000 and 6,000 feet (0.9 mile to 1.1 miles) from the mouth,
Duck and Porter creeks' are between 9,000 and 20,000 feet (1.8 miles to 3.7 miles), DCUT11
transects are between 700 to 2,800 feet (0.1 to 0.5 mile), and DCUT19 transects are between
1,900 and 3,800 feet (0.4 and 0.7 mile). Broomfield Swamp Creek transects are between 2,800
and 3,420 feet (0.53 and 0.65 mile) and transects in SCUT1 are between 3,986 and 4,550 feet
(0.75 and 0.86 mile).
In August 1998, vegetation monitoring sites were established in the vicinity of
each of the WL-80 continuous monitors (replaced by Ecotones in 2007 and Level TROLLs in
2011) located in the riparian wetlands of Jacks Creek, Tooley Creek and Huddles Cut. Five
vegetation transects were established in Jacks Creek, four in Tooley Creek, and 12 in Huddles
Cut. In 2007, vegetation monitoring transects were re-established in Huddles Cut, in 2010,
transects were re-established in Tooley Creek, and in 2011, transects were re-established in
Jacks Creek. Additional vegetation monitoring transects were established in 2011 in the vicinity
of Level TROLLs in five new creeks: one vegetation transect was established in Jacobs Creek,
US§

one in Drinkwater Creek, one in Long Creek (control), two in Porter Creek, and four in Duck
Creek (control). In 2013, new vegetation monitoring transects were established in DCUT11 and
DCUT19. In 2018, new vegetation monitoring transects were established in Broomfield Swamp
Creek and SCUT1; but were not monitored until 2019. These last two creeks represent the final
creeks to be monitored under the Mod Alt L permit.
Vegetation surveys focus on the shrub and herb layers. Compared to trees,
shrubs and herbs respond more quickly to changes in salinity and hydrology, and therefore
provide better indicators of changes in the vegetation over time. At each vegetation monitoring
transect, 10 permanent sample quadrats were established along a 40-meter transect that
proceeds on a random compass azimuth from the well, but may be adjusted to ensure the
transect stays in the riparian floodplain. The quadrats are arranged on alternating sides of the
transect axis, such that each quadrat shares a corner with the quadrat behind it and the quadrat
in front of it. However, no quadrat shares a boundary with any other quadrat. Each sample
quadrat consists of a 4-by-4 meter woody shrub vegetation plot with a 1-by-1 meter herb plot
nested in the proximal corner. These will be used throughout the duration of the study to
monitor density, coverage, and species composition of the herb and shrub strata layers.
Shrubs and woody vines, defined as woody plants greater than 3.2 feet in height
but less than 3 inches in diameter at breast height (DBH), were inventoried in each of the 10 4-
by-4 meter plots located in the vicinity of each well in the riparian wetlands. For each species,
the number of stems present was counted and percent cover estimated. Herbs, defined as all
herbaceous vascular plants regardless of height and woody plants less than 3.2 feet in height,
were inventoried in each of the 1-by-1 meter plots nested within the 4-by-4 meter plots. For
each species, the number of stems present was counted and percent cover estimated.
Qualitative descriptions of the overstory were made in the vicinity of each electronic well.
An importance value was calculated for each shrub and herb species present in
each transect. Relativized values of average percent cover, average stem count, and frequency
of occurrence in the 10 quadrats were used to calculate importance values. Dominant plants in
each transect were determined by applying the 50/20 rule to the importance values. The 50/20
rule was described in the 1989 wetland delineation manual (Federal Interagency Committee for
Wetland Delineation 1989) and is still used in delineating wetlands (Williams 1992, USACOE
2010). The 50/20 rule uses the quotient obtained from dividing each species' importance value
by the sum of all of the importance values for that transect (shrubs and herbs are treated
separately). This calculation expresses each species' importance value as a percentage of the
cumulative importance value of the entire transect. Beginning with the species having the
highest importance value and continuing in descending order, all species are listed until,
cumulatively, 50 percent of the overall importance value has been reached. These species,
along with any additional species that individually comprise at least 20 percent of the overall
importance value, are considered to be dominant.
To further assist in determining whether changes in the plant communities have
occurred, the tolerance of brackish conditions was assessed for each dominant species. The
determination of each species' tolerance was based on habitat descriptions provided in Radford
et al. (1968), Beal (1977), Godfrey and Wooten (1979, 1981), Odum et al. (1984), Eleuterius
(1990), eFloras (2008), and Weakley (2015). A species was considered tolerant of brackish
conditions if any of the habitats listed were brackish, even if most of the habitats were fresh.
The percentage of dominant species intolerant of brackish conditions (viz. solely freshwater
species) was calculated for each transect. A linear regression analysis was performed for each
transect using Sigma Plot 11.2. Pre- and post -Mod Alt L years were analyzed independently of
each other so that a comparison could be made. Any significant changes in the salinity of the
USVA

creek should be reflected by a shift in this percentage. Comparison of variability in the
composition of the vegetative community pre- and post -Mod Alt L was determined using
similarity percentages (SIMPER). SIMPER identifies the percent contribution of each individual
species contributing to dissimilarity (or similarity) between groups (in this case pre and post),
hence, it detects the species that are most important in causing the known dissimilarity (or
similarity).
Another analysis was performed using the wetland indicator status (Lichvar et al.
2016) of the dominant plants. The percentage of dominant species with a wetland indicator
status of FACU or drier was calculated for each transect but linear regression analysis was not
performed for this parameter. Any major change toward drier conditions should be reflected by
a change in this percentage.
G. Fish Monitoring
1. History
Monitoring of fish assemblages occurred in Jacks Creek, Tooley Creek, Muddy
Creek, and Huddles Cut under the 1998 plan. Under the 2011 plan, fish assemblage monitoring
was expanded to include seven new creeks in 2011 and two additional creeks (DCUT11 and
DCUT19) in 2013. In 2018/2019, the final creek under the 2011 plan, Broomfield Swamp Creek
(and its control, SCUT1), were added. The geomorphic characteristics of the seven new study
creeks were described in the 2011 PCS Creeks Report (CZR et al. 2012), the geomorphic
characteristics of the two DCUTs are described in Section I-B of the 2013 report, and
Broomfield Swamp Creek and SCUT1 geomorphology were described in Supplement 2 of the
2018 report.
2. Field Methods
If a monitored stream is too shallow and narrow to sample using a trawl, fyke
nets are used to sample fish (Huddles Cut, DCUT11 and DCUT19). Each fyke net sample
event is conducted using two fyke nets (one net fished upstream and one fished downstream)
anchored for a set time of approximately 16 hours (late afternoon until the following morning).
Each fyke net is deployed across the entire width of the sampled stream and consists of 0.25-
inch mesh net with four 21-inch hoops, a 6-inch throat, and a 22-foot wingspan.
For monitored streams large enough to trawl, each fish trawl sample is
conducted using a two -seam otter trawl. The trawl was constructed with a 10.5-foot head rope,
0.25-inch bar mesh wings and body, and 0.12-inch bar mesh cod end. The trawl is towed for
approximately one minute, covering approximately 75 yards, from a beginning point marked in
the field with flagging (fixed GPS coordinates) near the mouth of each monitored creek. Tow
direction is always toward the creek mouth.
All fish captured by either method are identified and counted. Those that are
easily identified in the field are released; others are preserved for later identification. Total
length is measured to the nearest millimeter for the first 30 specimens of each species.
Representative photographs of sample stations are taken during the first sample event and on
other occasions and are on file with CZR Incorporated.
Water quality data are collected with a YSI Pro Plus Quatro multi -parameter
instrument prior to deployment and retrieval of fyke nets and/or before each trawl at each creek.
Parameters measured include temperature, dissolved oxygen (DO), salinity, conductivity, pH,
and both Secchi depth and water depth. In the deeper creeks, excluding Secchi depth,
measurements for all other water quality parameters are taken at both surface and near bottom
A-13

levels. Estimates of the percentage of the water surface covered by submerged aquatic
vegetation (SAV) are also made. Water quality data are examined with regard to how well each
site provided habitat appropriate for the preservation of fish communities. Particular attention is
given to dissolved oxygen, as low DO levels are commonly implicated in fish kills.
3. Statistical Methods
Multivariate cluster analysis among creeks/years analyzed spatial variability of
fish species composition and abundance. Fish data are transformed to reduce the multiplicity of
variance among creeks/years and assembled into a Bray -Curtis dissimilarity matrix. Bray -Curtis
has the advantage of incorporating not just the presence of similar species but also the relative
abundance of species found between sample locations and is therefore a comprehensive
descriptor of community similarity. Bray -Curtis (Bray and Curtis 1957; Clarke et al. 2006) is
defined as:
Cbc= 100*1- �; LyLizy I, where
1i(Yi1+Yi2)
y, and y2 are the total number of the itb species at site 1 and 2.
The resulting dendrogram created from the Bray -Curtis dissimilarity matrix is
tested for significant cluster/group differences among fish assemblages using a similarity profile
test (SIMPROF) at the 5 percent level (P = 0.05) for individual creeks and the 1 percent level (P
= 0.01) when all creeks are included in the matrix. Comparison of variability between fish
assemblage clusters/groups is determined by means of similarity percentages (SIMPER).
Similarity percentages (SIMPER) identifies the percent contribution of each individual fish
species contributing to dissimilarity (or similarity) between clusters/groups, hence, it detects the
fish species that are most important in causing the known dissimilarity (or similarity).
Analysis of similarity (ANOSIM) is a nonparametric test that is used to compare
interannual variation in fish species abundance and composition to detect spatial differences
between pre- and post -Mod Alt L years for creeks with drainage basin reduction and related
control creeks. Significance was set at 0.05.
A two -sample t statistic (T-Test) is used to compare interannual variation in
individual fish species catch -per -unit -effort (CPUE) and abundance to detect spatial differences
between pre- and post -Mod Alt L years for creeks with drainage basin reduction and related
control creeks. Significance is set at 0.05. When normality failed and the data do not meet
assumptions for a parametric test, a nonparametric test is used (Mann -Whitney Rank Sum
Test).
Biota and/or environmental matching (BEST) using the BIOENV method was
used to analyze the relationship between the multivariate fish data and both in -situ collected
hydrographic data and ECU water quality data. As with fish data, water quality data are also
transformed to reduce the multiplicity of variance among creeks/years. Following transformation
of the water quality data, Euclidean distance is calculated between the water quality variables
and a dissimilarity matrix is formed. BIOENV then calculates all possible combinations of
variables between the two dissimilarity matrices; biota and/or environmental matching (BEST) at
this moment displays the environmental variables that best correlate to the multivariate fish
assemblage clusters/groups.
A-14

Fish Guild Analysis
Each species caught in trawl or fyke nets was assigned to one of six trophic
guilds: zoobenthivore, zooplanktivore, piscivore, herbivore, omnivore, or detritivore (see Table 1
for each species' designation). For all species the predominant prey item was chosen in order to
assign species to guilds and, for most species, guilds were assigned by information available on
FishBase (Froese and Pauly 2016). Definitions for guilds are as follows (from Elliot et al. 2007):
• zoobenthivore — feed on invertebrates associated with the substratum;
• zooplanktivore — feed on zooplankton, including crustaceans, fish eggs,
and fish larvae;
• piscivore — feed on finfish and large invertebrates (e.g., squid);
• herbivore — feed on plants, including algae and phytoplankton;
• omnivore — feed on both plants and animals;
• detritivore — feed on detritus.
Information for six species (chain pipefish, mud sunfish, naked goby, redear
sunfish, sheepshead minnow, and swamp darter) was not available on FishBase. Guild
designations for these fish were assigned based on previously published scientific literature.
These six fish species represented about 1.5 percent of the total catch over all years.
For each sample event, the number of fish in each guild was determined. Then,
relative abundance of each guild was calculated by dividing the number of fish in each guild by
the total number of fish caught in that particular sample. Relative abundance was used as
opposed to total abundance in order to focus on compositional changes in community structure
through time rather than changes in total numbers.
Statistical analysis of the guilds proceeded in a similar manner as the fish and
benthos sections of previous reports. For each creek, a Similarity Profile (SIMPROF) was used
to determine clusters among the different years. Similarity Percentage Analysis (SIMPER) was
then used to determine how clusters differed. For the impacted creeks that have pre- and post -
Mod Alt L data, an Analysis of Similarity (ANOSIM) was used to determine if trophic guild
designation differed based on Mod Alt L status. Through the 2016 data collection year, or
creeks with more than five years of data, a Biota and/or Environmental Matching (a.k.a. BEST
routine) was conducted to determine the set of environmental variables most important in fish
trophic guild structure. In 2017, the BEST routine was used again but if a strong correlation
(Spearman rank correlation > 0.60) with one or more environmental variables was displayed, a
multivariate analysis of variance (MANOVA) was used to determine the significance of the
variables (p < 0.05).
H. Benthos Monitoring
1. History
Monitoring of benthic macroinvertebrate communities occurred in Jacks Creek,
Tooley Creek, Muddy Creek, and Huddles Cut under the 1998 plan and was expanded to
include seven of the new creeks identified under the 2011 plan; two more creeks (DCUT11 and
DCUT19) were first monitored for benthos in 2013. In 2018, Broomfield Swamp Creek in the
S33 Tract portion of the Mod Alt L permit and its control creek, SCUT1, were added into the
study for a total of 14 creeks; full monitoring began in these two creeks in 2019.
A-15

2. Field Methods
a. Macrobenthos Sweeps
Timed sweep methodology is used to sample benthic macroinvertebrates
along the shoreline at each upstream and downstream station in the monitored creeks following
NCDENR, NCDWR standard operating procedure (SOP) for timed sweep methods (NCDENR
2006). The timed sweep samples consists of 10-minute collections with a D-frame net in
representative shoreline and near -shore habitats. Water quality data are collected with a YSI
Pro Plus Quatro multi -parameter instrument during the sweep sampling. Parameters measured
include temperature, dissolved oxygen (DO), salinity, conductivity, pH, and both Secchi depth
and water depth. In the deeper creeks, excluding Secchi depth, measurements for all other
water quality parameters are taken at both surface and near bottom levels. Water quality data
are examined with regard to how well each site provided habitat appropriate for the preservation
of benthic macroinvertebrate communities.
Within each sampling station, three replicate samples are collected from
the same three locations as in previous years. Sweep contents are sieved in the field through a
0.5 mm mesh screen, preserved and returned to the laboratory for benthic sorting, enumeration,
and identification to the lowest practical taxonomic level (usually species). Each replicate is
enumerated, and the mean number of individuals per taxa is calculated for all taxa collected at
each station. An Estuarine Biotic Index (EBI) is calculated for each upstream and downstream
station in each creek (NCDENR 1997):
EBI = jLSKj*n;j, where
N
SV; is the ith taxa's sensitivity value,
n; is the ith taxa's abundance value (1, 3, 10, 30,or 100),
N is the sum of all abundance values.
An EBI is a mean of the water quality sensitivity values for each taxon in the sample, weighted
by abundance values of the taxa. Only those taxa for which sensitivity values are available can
be used to calculate an EBI. An EBI can be used at all salinities to make comparisons and
assess differences in site water quality (NCDENR 1997). Theoretically, an EBI can range from
a low of one to a high of five. A high EBI indicates many intolerant taxa and good water quality
at a location, while a low EBI is indicative of stressed conditions (NCDENR 1997). All sensitivity
values for taxa encountered between 2000 through 2005 and 2009 through 2016 were verified
and assigned by Larry Eaton, Surface Water Protection, NCDWQ in an effort to standardize and
evaluate benthic data from past study years. However, Larry Eaton retired in 2017 and no
tolerance values have been assigned since his retirement. To date, there is no one within NC
Division of Water Resources (NCDWR, formerly Division of Water Quality or DWQ) with
equivalent estuarine expertise to continue to make assignments.
b. Macrobenthos Ponars
Five ponar grabs are taken from a boat near mid -stream at each
upstream and downstream station in all 15 creeks in this study. Little Creek, Duck Creek,
DCUT11, DCUT19, and SCUT1 upstream stations and both upstream and downstream stations
at Huddles Cut and are not accessible by boat; therefore a manual ponar grab method was
A-16

used to obtain the five samples. Just as with benthic sweeps above, basic hydrographic data
are collected at each ponar sample station.
Collected sediments are placed in 1 gallon plastic bags, with a full bag
constituting a sample. Samples are then sieved in the field through a 0.5 mm mesh screen. All
material retained on the screen are preserved and returned to the laboratory for benthic sorting,
enumeration, and identification to the lowest practical taxonomic level (usually species). For
each taxon, the mean number of individuals per grab is calculated. An EBI is calculated for
each sampling station based on total individuals of each taxon collected from all five replicate
grabs.
3. Statistical Methods
Multivariate cluster analysis
Multivariate cluster analysis among creeks/years/stations analyzed spatial
variability of taxa richness and abundance for both sweep and ponar collections. Benthic data
are transformed to reduce the multiplicity of variance among creeks/years/stations and
assembled into a Bray -Curtis dissimilarity matrix. Bray -Curtis has the advantage of
incorporating not just the presence of similar taxa but also the relative abundance of taxa found
between sample locations and is therefore a comprehensive descriptor of community similarity.
Bray -Curtis (Bray and Curtis 1957; Clarke et al. 2006) is defined as:
Cbc= 100*1- 1�; LyLl:y I where
1i(Yi1+Yi2)
y, and Y2 are the total number of the ith species at site 1 and 2.
The resulting dendrogram created from the Bray -Curtis dissimilarity matrix is
tested for significant cluster differences among benthic macroinvertebrate communities using a
similarity profile test (SIMPROF) at the five percent level (P = 0.05). Comparison of variability
between benthic macroinvertebrate community clusters/groups is determined by means of
similarity percentages (SIMPER). Similarity percentages (SIMPER) identifies the percent
contribution of each individual taxon contributing to dissimilarity (or similarity) between
clusters/groups, hence, it detects the taxa that are most important in causing the known
dissimilarity (or similarity); however, such comparisons of statistical significance must be made
with care since very abundant taxa were not fully enumerated (counts stop at 100 for each
sweep and ponar grab samples).
Analysis of similarity (ANOSIM) is a nonparametric test that is used to compare
interannual variation in taxa abundance and composition to detect spatial differences between
pre- and post -Mod Alt L years for creeks with drainage basin reduction and related control
creeks. Significance was set at 0.05.
Biota and/or environmental matching (BEST) using the BIOENV method was
used to analyze the relationship between the multivariate benthos data (both sweep and ponar)
and both in -situ collected hydrographic data and ECU water quality data. As with benthos data,
water quality data are also transformed to reduce the multiplicity of variance among
creeks/years. Following transformation of the water quality data, Euclidean distance is
calculated between the water quality variables and a dissimilarity matrix is formed. BIOENV
then calculates all possible combinations of variables between the two dissimilarity matrices;
biota and/or environmental matching (BEST) at this moment displays the environmental
A-17

variables that best correlate to the multivariate benthic macroinvertebrate community
clusters/groups.
In addition to the dendrogram created from the Bray -Curtis dissimilarity matrix for
ponar collections, Shannon -Wiener diversity index (H') (Shannon 1948), is also used to detect
spatial differences in species diversity in ponar samples collected at all 13 creeks. The index is
defined as:
H'=- 1p; Inp;, where
n;= number of individuals of the it" species,
N= total number of individuals of all species, and
pi= n;/N.
Fuzzy Correspondence Analysis
For each species collected in the ponar grabs, two different characteristics were
analyzed: trophic level and functional feeding guild. Trophic level describes the position of a
species in a food chain and consisted of four levels: detritivore, herbivore, carnivore, and
parasite. Functional feeding guild describes the mechanism by which each species acquires
food resources and/or nutrients and consisted of six categories: gatherer/collector,
filterer/collector, scraper, grazer, shredder, and predator.
Each species ever collected in a ponar grab was assigned a number between 0
and 3 for each of the four trophic level categories and each of the six functional feeding guild
categories. A '0' indicated that the species had no affinity to that particular category whereas a
`3' indicated that the species had high affinity to that particular category. An advantage to this
approach is that a species could be assigned nonzero values in multiple categories, which
addresses species that are omnivorous (e.g., herbivore and carnivore), or species that display
multiple or facultative feeding modes (e.g., predator and gatherer/collector). Omnivory and
multiple feeding modes are extremely common among benthic macroinvertebrates.
Prior to species designations into categories, several sources were consulted
including the Environmental Protection Agency (EPA), the United States Geological Survey
(USGS), and the Georgia Environmental Protection Division (GAEPD). Peer -reviewed scientific
publications were also consulted for species not covered in the aforementioned three sources.
Final species designations were reviewed by Dr. Larry Eaton of the North Carolina Department
of Environmental Quality Division of Water Resources (NCDEQ DWR) and any necessary
adjustments were made. A list of all the species and their respective designations is included in
Appendix I (only on CD).
The list of species designations was analyzed using a fuzzy correspondence
analysis (FCA). An FCA is similar to an ordinary correspondence analysis except that species
can have membership in multiple categories. The outputs of an FCA are standardized, normally
distributed scores for a certain number of axes for each species. Each axis represents some
type of differentiation of the categories. For example, a particular axis of the FCA may
differentiate parasite from all the other trophic levels by assigning species that are parasites a
relatively large negative (or positive) number. For the initial FCA analyses, a total of six axes
were produced which differentiated the other trophic level categories and functional feeding
guild categories; however, a different number of axes could be selected. Thus, each species
was described by a separate score for the different axes, each score describing a different
combination of trophic level and/or functional feeding guild.
MM

To characterize the trophic level and functional feeding guild of the benthic
macroinvertebrate community of each creek, each species' FCA score on each of the axes was
multiplied by its relative abundance for each ponar grab conducted in the creek. Relative
abundance is computed as the abundance of a particular species divided by the total
abundance of the sample. Species with 0 abundance are given a relative abundance of 0. A
sum of the FCA scores weighted by the relative abundance of each species was obtained. This
process was repeated for each of the axes from the FCA. Thus, the benthic macroinvertebrate
community of every ponar grab across all creeks, years, locations (upstream vs. downstream),
and sample (1-5) is described by this summed value for each of the FCA axes. For example, a
ponar grab sample that has a relatively large negative or positive number for a particular axis
relative to the others indicates that the benthic macroinvertebrate community was dominated by
that particular category (either a trophic level or a guild depending on the particular axis). These
summed weighted FCA scores were used in subsequent statistical analyses.
Spatial and Temporal Variability of Benthic Guilds
Multivariate cluster analysis among creeks/years/locations analyzed spatial
variability of benthic trophic level/functional feeding guild scores of all six FCA axes. The scores
were assembled into a Euclidean distance matrix.
The resulting dendrogram created from the Euclidean distance matrix is tested
for significant cluster differences among benthic trophic level/functional feeding guild scores by
a similarity profile test (SIMPROF) at the five percent level (P = 0.05; in some years, a more
conservative threshold may be set). Comparison of variability between benthic trophic
level/functional feeding guild clusters/groups is determined by similarity percentages (SIMPER).
Similarity percentages (SIMPER) identifies the percent each axis contributes to the distance
between clusters/groups, hence, it detects the FCA axes that drive the known distance;
however, such comparisons of statistical significance must be made with care since very
abundant taxa were not fully enumerated (counts stop at 100 for each ponar grab sample).
Analysis of similarity (ANOSIM) is a nonparametric test used to compare
interannual variation in benthic trophic level/functional feeding guild scores to detect spatial
differences between pre- and post -Mod Alt L years for creeks with drainage basin reduction and
related control creeks. Significance was set at 0.05 (in some years, a more conservative
threshold may be set). As done for the fish guilds, in 2017, the BEST routine was used for the
benthic guilds to determine environmental variables of importance to guild structure; if a strong
correlation (Spearman rank correlation > 0.60) with one or more environmental variables was
displayed, a multivariate analysis of variance (MANOVA) was used to determine the
significance of the variables (p < 0.05).
Mixed -model analysis of Pre- and Post -Mod Alt L Creeks
A fully factorial mixed -model analysis of variance (ANOVA) was used to
determine how macroinvertebrate guilds in impacted creeks have changed in post -Mod Alt L
years, and how any changes in guild structure compared to control creeks over identical time -
frames. Separate models were run for each of the six axes from the FCA described above and
for each impact -control creek comparison. For each model, the FCA axis was the dependent
variable, Creek and Mod Alt L status were fixed independent variables, and Ponar subsample
was a random independent variable. Pre- and post -Mod Alt L status for the control creek
sample years were assigned to match those of the impacted creek. Five ponar subsamples
were taken in each year for all creeks, thus these subsamples were treated as a random effect
A-19

in order to avoid pseudoreplication. Each year of sampling was treated as a true replicate.
Models were fit with the Ime() function in R, and the anova() function was used to conduct null
hypothesis testing.
Alpha values for determining the significance of main and interaction effects were
set at 0.05 (in some years, a more conservative threshold may be set). For each model, the
significance of the interaction between Creek and Mod Alt L Status was determined. A
significant interaction indicated that the change in benthic macroinvertebrate guild composition
(i.e., depending on the particular guild Axis examined) from pre- to post -Mod Alt L differed
between the impact and control creeks. This interaction would indicate that the
macroinvertebrate community in impacted creeks responded differently than a control creek to a
change in Mod Alt L status. Although less relevant, significance of the main effects of Creek
and Mod Alt L Status was also determined. A significant Creek effect indicated that the two
creeks differ in the particular guild Axis analyzed, and a significant Mod Alt L Status indicated
that pre -Mod Alt L Status guild structure was different than post -Mod Alt L Status guild structure
among both creeks combined.
A-20