NCS000533_OTHER_20140210

STOR-MWATER-DIVISION CODI-NG-SHEET
PERMIT NO.
DOC TYPE
❑ FINAL PERMIT
❑ MONITORING INFO
❑ APPLICATION
❑ COMPLIANCE
K'OTHER
DOC DATE
❑ `� 7� ba ) a
YYYYM M DD

C'OBC
We Make Chemistry Happen
February 10, 2014
Mr. John E. Skvarla, Ill, Secretary
NC Department of Environment and Natural Resources
217 West Jones St.
Mail Service Center 1601
Raleigh, NC 27699
RE: Oak -Bark Corporation
Request for Assistance— US EPA Issue
Dear Mr. Skvarla:
Bill Oakley and 1, Jim Barker, are residents of New Hanover County, and have a manufacturing plant in
Riegelwood, NC. There has been a manufacturing facility on or adjacent to our plant since 1883. The operation
was owned by the Wright family until 2004 when Bill and 1, employees of the company for over 30-years each,
purchased controlling interest in the operation that was known as Wright Chemicai/Wright Corporation. We
have been good stewards, employing hundreds of local citizens; and have complied with all NCDENR
reporting, filing, and regulatory requirements for the 30-years 1 have been associated with the facility. The
various branches of NCDENR with whom we interact will confirm this fact.
The purpose of this letter is to request your help. The Wright Chemical Company Site was listed on the
National Priorities List in 2011 due to the discovery of a former lead chamber acid plant that was used to
produce sulfuric acid from the 1880's to the early 1960's. This activity performed by prior owners created a
limited area of contamination characterized by arsenic, lead, sulfate and low pH values. What we are now
experiencing is creep of scope outside of the initial listing.
As outlined below, the USEPA is now conducting and/or proposing additional activities on acreage outside of
their initial focus of interest.
From February 27 to March 2, 2012 a USEPA contractor conducted field sampling activities at a number of
locations throughout the Wright Chemical property. These locations included a former 20-acre spray field that
operated and was closed out under a NCDENR permit, two lined ponds known as the Kelly Ponds which have a
current NCDENR permit, one lined soil/sludge monofill that has a current NCDENR permit, the main plant area
that is currently monitored semiannually under a NCDENR approved Corrective Action Plan, and some storm
drainage ways that are currently monitored and reported to NCDENR under storm water permits.
The results of the USEPA sampling were presented in a "Final Pre -Remedial Investigation Report" dated
October 2012. No contamination of any significance or actionable level was discovered. The report then
OAK -BARK CORPORATION'"* 1224 Old Hwy 87 Rd. •" RIEGELWOOD, NC 28456
Telephone: [phone]
ww-w.oak-bark.com

concludes with a "Remedial Investigation/Feasibility Study Scoping Recommendations". These
scoping recommendations include further investigation of operations at the property that
operate under current NCDENR permits, units that have been properly closed out under
NCDENR and the former Kaiser property that was specifically determined by the US EPA to not
be part of this Superfund investigation (letter dated October 21, 2009).
We respectfully ask for your support of our recommendation that the USEPA concentrate on
the original scope of correcting issues that existed with the lead chamber acid plant and leave
the remaining facility under the jurisdiction of NCDENR. The USEPA's commissioned study of
the site was done without consulting Oak -Bark, any of our long time outside environmental
advisors or NCDENR representatives.
Assuming you agree with our assessment, please contact USEPA and encourage them to
promote our recommendation. We are a small company, without deep pockets, and want to
use the resources we have in the most beneficial way to remedy a situation that occurred over
50-years ago.
Bill Oakley and I would welcome the opportunity to discuss this issue in more detail. I can be
reached by phone at 910-520-6518(Bill) and 910-617-2275(jim), and/or by email at
[e-mail] or [e-mail]. Thank you for your attention to this issue
and I look forward to working with your staff an a resolution.
Sincerely,
William E. Oakley
Chairman
James C. Barker
President
cc: john,[e-mail]
Rick Catlin, NC House of Representatives
John A. Andreasen

07
�0
0
Z
}'Fw�
czi y
4
a
t.f '
WI
' f I
I f/
I�
11
1
A
i r n
e ,
, I
i
O
I
v•
o t w Alp O
o
O�V �
g
0
6�
Ao
U0.
n
/ n
Pn
—�
J
3
WRIGHT CORPORATION �
STORM WATER MAP I STORM WATER MAP
.STEWAR 1-a200' —29-04 STORM WATER PLAN.dwg

o.A
S)A
rrf
HEXAMINE
PRODUCTION
... .. .. .........
..... . .... ... . ... .... 40 r.
HEXAMINE
0.0 UC T I El N
rrf
rrr
rrE
rrr
. ....... ....
...... ....
SPECIAL
PROJECTS DR41N
WHSE
HEXION
SW OUTFALL
008
kA A T N r-1 F c- T (-I E-
MAINTENANCE
SEPTIC DRAIN
FIELD

CREEK
L�NC'STON
U
SW O TTFALL .
003 .
(NC5000156)
67,500 SQ FT
013C
SW OUTFALL
007
(NC50DO532)
POND
55, S
Apq�, BK VH vHSE
Pipe
i
;
{!
%0
0
ti
Ob SILAR LLC
�SW OUTFALL
O 00NC5000533
LEGEND
MOMENTIVE
STORMWATER
OUT FALL # 3 , 4 , 6 , 9
SILAR, LLC
STORM WATER
OUT FALL # 1 , 2 , 5
OBC
STORMWATER
OUT FALL # 7
a VAPCRATIV v �! O )
MO LAGOON
_O 000r044 (LINED POND) °O �+
50 56) j�
� O O
47,400 SQ FT 42,300 SQ FT �•� ,
' LIQUID % 16 o
METH. STORAGE Q N[TRCGE
O TANKS (� ♦t b c�
O Jp �c, c`I' �;
I ! ' o BRIDGE RE SENTATIVE
r�1r U
t v� SANPLI POINT
ZRV10E) P E
001 I HLD FINE-'` ILAR,LLC
20,700 SQ FT O CHEMICALS SW ALL
O001 (NC5000533)
P [ IL LC
F— O
02
-M EN 7 173,400 SQ FT
�ot4,,i O c s iTCF
- 006'(Nc500015s0oat56)
�77,400 QFT
% � m
PAEKAGI STXAMI
4REA / ��; •11('
WAREHiIII
WAREHOUSE C7 DUCK POND
00 We
o
xaH[NE ® SPECIAL
Y'WAREPROJECTS
``� My PRq.JFC ]'S
Ronuct JV� aQ r
e;1 o HEXANINE �OO ,
PRODEXT113N SPECIA
WHSE StH 142,200 SQ FT v 0
pi HEXANINE
ODLs T1ON / O 63,OW SQ FT
Al
000
D /
,k"� OO O NAINT E�
O
ass.
a
a`
x
a`
4
N �
F[Tl MONAL LOCATl V-01-02-07
MOMF-NTWE `
SW O ALL
( 156)
SEPTIC DRAIN
r1ELD
UMKIM DM rD MMU eama* uaa v
ftxmaw eomvmM fD�T
4 ADDED SOUUM tOOTAM TO AWAS
ww= •ue rrea maom wm tp
u wm a ADDED Sim dews
MOMENTE-VAE ACME,N.C.
Specialty Chemicals Inc
STORM DRAIN RUNOFF
ml

J II I 1
wl I
I I
I !
IC I I
"N1Ila
4 !' ! 1sl (mil//A
110,
it
8 g 2 pan Cod
41
a
r
1 !
k � n
o g PIP 5 ®P o
0
vi
O
Cif o) ref
F � !
� �
1
i.
rr \
o � i
m $.
r,
17
+y,,
'
(30
� ❑ 1j f �
1
i
, � 1
j x
-- o
�z 1
z
Q
.�_.
a.LA
0 _
WRIGHT CORPORATION STORM WATER MAP STORM WATER_ MAP
E.L.STEWAR � 1--200' —29-04 STORM WATER PLAN.dwg

4.8 X X 4A
`t
i`
/ X 6. 2 \ \
X 5.3
MOMENTP
SW OUTI~ALL
003
OAK -BARK
SW OUT1~ALL X
007 r.
WOM
X 75. 1
32.5
i'
32tRIDI
X
�
f
r
ti
34.3
X 32.4
y
X
t
X 33.3
X
11;
fl
ft
X 37.5
34t
uI
1
f
�
!If
fIr
1!I
I11
!1!
39, 2
X
1 rrr
X
-
36.5
40.4
X
_. v. X 36.7
-Y;
39 S X v
40.4 X
42.4 \
FUNCTIONAL LOCATION-0 1 -02-071
MOMENTIV E,
SW OUTFALL
009
SILAR,LLC
SW OUT);ALL
065
a€:k,.. ,
40.8
x X 42.9
X 45.9

Colurnbpl Cam
NC0087947 Currently receives wastewater from only one SIU (Momentive). Design flow rate is 0.125
MGD. Average flow currently is around 12% of that or 0.16 MGD. OutfaII 001 is to Livingston Creek.
,L.
NC0003395 - Only outfall 002 Is active. Non -contact cooling water, reverse osmosis reject, water
softener backwash, water softener regenerant and firewater tank blowdown. Discharge Is to Livingston
Creek. Outfall 001 is now connected to Columbus County WWTP.
WQ0034775 — Closed Loop Recycle for A/C chiller make-up from lined pond.
NCS000156 — NPDFS Stormwater
Slur
NCS000533 — NPDES Stormwater
Oak -Bark
WQ0003361— Closed Loop Recycle for Kelly Ponds Storage.
WQ0005699 — Land Application of Residual Solids
NCS000532 — NPDES Stormwater

�%VL �' L ~ Su►^
C�vt_ _GSS (t?GI�S
- l
s
_��.ov�� �ro��,:r,_w�S-•
Cry ---- -- - �=t_z#=c�- w���—�
4,,
t.� �es o�
_Dalt-�
d5_t I h�°t s soY�
+w
r
p VA 01�7
y�✓ -h -��s s —� Y.? � c a�
o f,6 i In Vol V e
�ayt -f'� o �n� � ►
Q —
1N
�i C % 00, � S �r
WAA-
IN
woo •ryt
V,
+0 'D Z>p
�? �s��.z --�
[7U1n�iV1 v
V-\.o--,o .
CJ 0 Op f tw
1
�l_[

•- �
,. �
` ,
_ �
+' �
� " �
� -
� �
_ � � � S � � � � � ��� r
,� � , � s�
� � � � _ _
� � � � _ ��
�
' -� � �
i s f
�� -
� _�
� 1
` r f � r
��� `
a. _ _
i � ,
� '. .. ��_ ,f ,
�`
n - -
r-
•*
�� ,
i
�\ �
., + -- .
�.
,- ��
1
'

0
--
w
rA
i
a V OVI
L�VLkzw����s
- �vt. t ��gA [ ✓z � vt S �"Y r � S --
A ok?X lnlO-V�- I d
?,Ind AZA2 S ova s crs L .
X R.D - E2/r) q M p s i
J-e?4-Q T6�1 7
vJ �a r al Le, iVLVo
�i+rirvl k
o S+AVl W4W BM�s ` 2� wr�� So Ve- P+�� .
a 0 cvviGr I? t S �r�►5 cGVLIL 14- U
C
�r I
\AA1 1/V el� elc to t' W
rP - OA b LA� Y1 0 4k
��VIVtic� ,

r !'
u
• � r
r.
I
3
I-
r _
t ,
4
I
i
't • �� } i.'� � � f 'I \ _ .{ It .•

WATER RESEARCH 41 (2007) 4186-4196
Available at www.sciencedirect.com
.:;�/ ScienceDirect
joufnal homcpagc: www.el5evier.com/locale/waLrc6
Design of stormwater monitoring programs
Haejin Leea, Xavier Swamikannub, Dan Radulescub, Seung jai Kim`,
Michael K. Stenstromd,*
allnited States Department of Agriculture, Natural Resources Conservation Service, El Centro Office, 177 N. Imperial Avenue, El Centro,
CA 92243, USA
eCalifornia Environmental Protection Agency, Los Angeles Regional Water Quality Control Board, 320 West 4th St. Suite 200, Los Angeles,
CA 90013-2343, USA
`Department of Environmental Engineering, Chonnam National University, 300 Yongbong-dong, Buk-gu, Gwangju 500-757, Korea
aDepartment of Civil and Environmental Engineering, University of California, 5714 Boelter Hall, Los Angeles, Los Angeles, CA 90095, USA
A R T I C L E I N F O
ABSTRACT
Article history:
Stormwater runoff is now the leading source of water pollution in the United States, and
Received 18 December 2005
stormwater monitoring programs have only recently been developed. This paper evaluates
Received in revised form
several stormwater monitoring programs to identify ways of increasing the likelihood of
4 January 2007
identification of high -risk dischargers and increasing data reliability for assisting in the
Accepted 10 May 2007
development of total maximum daily loads. No relationship was found between various
Available online 18 May 2007
types of industrial activity or landuse and water quality data. Stormwater data collected
Keywords:
with grab samples has much greater pollutant concentration variability than in potable
First flush
water or wastewater monitoring programs. Industrial land use is an important source of
Industrial general permit
metals. For grab samples, sampling time during the storm event will affect results. Data
Monitoring
from California, which has distinct dry periods, showed a seasonal first flush, whereas data
Municipal permit
from areas with more uniform rainfall throughout the year did not show a seasonal first
Stormwater
flush. Selecting a subset of sites from each monitored category using a flow -weighted
composite sampler is an alternative strategy, and may result in lower overall cost with
improved accuracy and variability in mass emissions, but may not be less successful in
identifying high -risk dischargers.
ID 2007 Elsevier Ltd. All rights reserved.
1. introduction
The completion of wastewater treatment plants mandated by
the clean Water Act has reduced pollution from point sources
to the waters of the United States. Parallel trends are
occurring in other developed countries and will occur in
newly developing countries over time. As a result, non -point
source pollution such as stormwater runoff is now the major
contributor to pollution of receiving waters. The problem of
stormwater pollution is growing worse because of continuing
development, which results in increased impervious surface
area. In response to this growth, regulatory agencies are requiring
'Corresponding author. Tel.: [phone]; fax: [phone].
E-mail address: stenstroOseas.ucla.edu (M.K. Stenstrom).
0043-1354/$ - see front matter z 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.w a tres.2007.05.016
stormwater monitoring programs implemented through the
National Pollutant Discharge Elimination System (NPDES) to
quantify and eventually reduce stormwater pollution.
The overall goals of stormwater monitoring programs
include the identification of high -risk dischargers, the devel-
opment of total maximum daily loads (TMDLs) and ulti-
mately the reduction of stormwater pollution. Most storm -
water monitoring programs are relatively new and evalua-
tions of their usefulness for satisfying these goals are only
now possible (Duke et al., 1998; Lee and Stenstrom, 20OS;
Pitt et al., 2003; Olding et al., 2004; Turner and Boner, 2004).
We evaluated 9 years of data generated by industrial facilities

PLEASE PR r CLEARLY OR TYPE
STATE OF NORTH CAROLINA IN THE OFFICE OF
Pickle, Ken ADMIIVISTRATiVE IMARINGS
( nL.un Gro Horti cul t '^ ' r g 1:47) PM
To: Matthews, Matt; Bennett, Bradley+; Georgoulias, Bethany
GIAI- Nif. Tan iipriata )
(your Fame) PEi-7110NMR, )
PETITION
FYI, no action now. V. ) FOR
N.C. Department of Environment and ) CONTESTED CASE HEARING
�? r co e� ation in Gn's r )
1 a ura I esour e. E,4WSion of. dater Oual.ity
�c rerererrr�3rr S rcr;��in's +baI report to me that Craig Coker had recommended the
(4s�sfa3�jtcitizi�y�r�z� of his repprt, but Zoo had him take it out to reflect their available
I hereby csk for a contested case hearing as provided for by North Carolina General Statute § 150B-23.because the Respondent has:
Ken (Briefly state facts showing how you believe you have been harmed by the State agency or board.) .
P '_ti.oner_her_eh ___seeks_rev_i ew of Respondent's Notice of Violation . NOV-2009-DV-0301
dfFUA:000,qR3, 200 copy a ace . e i toner en that it Fas vio a e en.
S M:
Subject: RE: NC Zoo update
(4) Amount in controversy S none yet (if applicable)
Gin, thanks for Copying me. (If more space is needed, attach additional pages.)
5 ecause of these facts, the State agency or board has: (check at least one from each column)
deprived me of property; X exceeded its authority or jurisdiction;
ordered me,togay a fine orcivil penalty;_or__ _ _ . _ X _acted_erroneousty;,
fj*p ptb#Nt9glsubstantially prejudiced my rights; AND _ j(__ failed to use proper procedure;
Sent: Friday, November 20, 2009 1:09 PM acted arbitrarily or capriciously; or
To: Pugh, Maryjoan- ]onesDavid
Mailed to act as required by law or rule.
(�gzFliekle, K&y-eAer 5,, 2009 (7) Your phone number( 919) 929,3905
Subject: FW: NC Zoo update
(8)Prirtyour full address: coo The Brough , Law firm, Ste. 800-A, 1829 E. Franklin St,, Chapel HillNC14
Thought you should know Ken Pickle called me today. Ken really wants to see us uqe the rain garden option. I spoke wiTh
I�t i 9dsm0Njaur budgeting constf tAs fob tPrairflArden option and how we discussed
r ig so reco ylUu 61.1allits. Tuld hlin about special zoo fund only can be usea ror r enhan ements; told hi y unlikely to see this as a funding opportunity; grant wise.., we just
OP{POIATw
f> sit JpMjt&M het y ��r @�3��1' o 1F' flf+ Ya t'wp6pnncy or
�BEhf�tmir)t�F$l�I�f'EE1�'psl3�'tcs to it so its not very attractive item for funding. In addition CWMTF don t
seem to have much money to dispense in an,yy event. This (eaves o eratin funds.... with the reductions and additional
business expenses we are taking and the cd4��i'ia�C,c�e would be able to skins out $40,000.
i E�Ers
� � � 'h 4 ?6P%*en efore budget folks will amend the budget is
budget peeple's mind th t this is a
rprost site i nie Est�p°�ace. old him DMJ and MJP said they need t• la a t`h I�i�an y a enext, Ken's
sYa e se of action is what we- dyle%Aat lhis p&TU;
) (upcod
s)
Gin
(14) This the day of , 20
fi&n • Pickle,
Sent: Friday, November 20, 2009 12:05 PM (your signature)
u
ffA,1vT6VV M9FMMkWE COPY to the Office of Administrative Hearings, 6714
1

WATER RESEARCH 41 (2007) 4186-4196
4187
in Los Angeles County covered under the statewide Industrial
StormWater General Permit (ISGP). The results of the study
suggest that parts of the current industrial stormwater
monitoring programs will not be helpful to identify high -risk
dischargers, nor will they be useful in estimating mass
emissions for developing TMDLs. The design and require-
ments of the monitoring programs do not produce data with
sufficient precision for decision making.
We also evaluated a number of stormwater monitoring
programs across the United States to determine their useful-
ness in achieving the dual goals, and we make recommenda-
tions for improvements in monitoring programs. Our analysis
included three industrial StormWater General Permit pro-
grams (ISGPs), Municipal StormWater Permit programs from
17 states and Los Angeles County, and UCLA's first flush
highway runoff characterization study sponsored by Califor-
nia Department of Transportation (Caltrans). Our results
should be helpful to planners and regulators to interpret
existing datasets and programs, as well as provide recom-
mendations for improving future programs.
2. Materials and methods
2.1. Data sources
Six stormwater monitoring programs were evaluated and are
summarized in Table 1. Three of the monitoring programs
were associated with industrial permits, two with monitoring
of municipal separate stone sewer systems, and the last was
associated with a research project to quantify discharges
from highways.
The ISGP had the most observations. ISGP programs require
facilities that discharge stormwater associated with indus-
trial activities to apply and obtain coverage under the ISGP.
Industries are categorized by Standard Industrial Classifica-
tion (SIC) codes. These monitoring programs serve as a model
for other programs in the United States and can be useful in
developing monitoring programs for other countries. In our
previous study, we analyzed data from the ISGP in Los
Angeles County from three wet seasons (1998-2001). In this
study, we extended the evaluations to nine wet seasons
(1992-2001), and we also analyzed eight wet seasons
(1993-2001) of data from a similar ISGP in Sacramento County.
Since industrial monitoring programs generally vary by state,
9 years (1995--2003) of stormwater data from the State of
Connecticut were also analyzed.
The StormWater Program for Municipal Separate Storm
Sewer Systems (MS4) is a nationwide program and is designed
to monitor and reduce sediment and pollution that enters
surface waters from separate storm sewer systems to the
maximum extent practicable. Under US EPA sponsorship, Pitt
et al. (2003) compiled and evaluated stormwater data from a
representative number of NPDES MS4 stormwater permittees.
Over 10 years of data from more than 200 municipalities
throughout the United States were assembled. The areas were
primarily located in the southern, Atlantic, central and
western parts of the United States. Only data from well -
described stormwater outfall locations were used in the
database. The Los Angeles County MS4 data, which were
not included by Pitt et al. (2003), were analyzed and added as a
separate program, as shown in Table 1.
The Department of Civil and Environmental Engineering at
UCLA has characterized stormwater runoff from three high-
way sites for Caltrans since 1999 (Kim et al., 2005; Khan et al.,
2006). The study was conducted to assess runoff water quality
and quantity from California freeways, with particular
emphasis on characterizing runoff during the early stage of
storm events or the first flush. The study used grab and flow -
weighted composite samples and measured a large range of
parameters.
Regulators and others need to understand that decision
making using stormwater monitoring results collected under
the current regulatory scheme is limited due to the data's
high variability. The variability is associated with natural
variability of stormwater, but is compounded by experimental
error, which is introduced through limited or poor sampling
and analysis techniques. To illustrate this difference, we
analyzed the coefficient of variation (CV) of other monitoring
activities and compared it to storm -monitoring variability. For
this analysis, two extra datasets that are not shown in Table 1
were used for comparison purpose. The two datasets are from
two typical water and wastewater treatment plants in
Southern California. The data from these plants were
collected over an entire year and used composite samples
and certified laboratories for analysis.
2.2. Methods
These datasets and others were all analyzed used Systat 10.2
(Systat Software, Inc., Point Richmond, CA), which is a well-
known program for performing a range of statistical analyses.
Means, standard deviations, coefficients of variations and
other simple statistics were performed using Systat proce-
dure Descriptive Statistics. Sysat's General Linear Models
(GLM) Routine was used to perform analysis of variance
(ANOVA), and data for Fig. 7 was generated using Systat's T-
test procedure.
Sampling methods, either grab or composite, are noted in
the description of the datasets and have large implications for
the variability and utility of the datasets. Grab samples are
discrete samples taken within a short period of time, usually
less than 15 min, and can be collected any time during the
storm event. The time of sampling, either early or late in the
runoff, can have important impacts and these impacts are
also discussed.
The datasets were collected directly form the agencies
responsible for the monitoring programs in machine-readable
formats. There was no opportunity to introduce coding or
transcription errors. The datasets were transmitted to the
regulatory agencies in the same way as we received them. The
only quality assurance procedures applied to the datasets
were those implemented by the agencies themselves.
Previous studies, even the original Nationwide Urban Run-
off Program (NURP) study (US EPA, 1983) and previous work
from our laboratory (Stenstrom et al., 1984; Fam et al., 1997;
Lau and Stenstrom, 2003), identified relationships between
water quality and land use in most cases. We expected to see
differences in water quality associated with different types of
landuse or industrial activity. In our previous work (Ha and

We will love only 'what we understand.
h1 ligiSerd� ANI- 't4RX4raM99 64 the consultant's report, and DWQ's preliminary review, and the Zoo's
penditures.
There are some persistent difficulties concerning the Zoo's access to funding. Gin reports that they will be able to fund
the very minimal measures recommended in the consultant's report. But they don't know how they will fund the two -
cell bioretention area that the consultant's report estimated at an additional $37,790. It is of course not budgeted for,
at the moment.
As I understand Gin's comments, funds for their regular yearly budget items are earmarked for specific uses, and the Zoo
has very limited freedom to spend outside of the earmarked purposes. As I understand Gin's comments, unless there is
some special and additional provision at the Departmental level, the Zoo cannot commit to installing anything beyond
what is recommended in the consultant's report. Her perspective is that even the possibility of a two-year delay before
bringing the bioretention area on line does not address their difficulty— ie, it doesn't really help their restrictions on
spending to have the extra time to get ready/raise funds/budget for it. To my limited understanding, it sounds like the
Zoo is essentially required to prioritize funds for improvements that visitors can see - - which is not the case with
improvements that we might require as far as stormwater discharges from the composting operation.
Gin and I agreed to the following course of action.
• I'll take a couple additional days to formalize my preliminary review with a little more detail and a little more
comprehensive coverage of the proposals in the consultant's report. Her Director will be back in town in
December, which is important for their progress on this issue.
• With my more complete review in hand, Gin suggests that the Zoo can take their funding difficulties up the chain
in the Department to see if additional funds can be identified, or if earmarked funds can be applied.
Although Gin and I didn't discuss it specifically, it probably makes sense for DWQ to internally revisit the conclusions
from my preliminary review to see if there may be some other acceptable alternative to the addition of the two -cell
bioretention area.
Gin -- did I get our conversation accurately?
Ken
E-mail correspondence to and from this address may be subject to the North Carolina Public Records tow, and may be disclosed to third parties.
I

i
:7
•
•
4188 WATER RESEARCH 41 (2007) 4186-4196
Table I - Summary of major monitoring datasets for this
study
Monitoring Monitoring Primary Period
Sampling type
Monitored parameters'
Number of
program area land use
observations
Industrial County of Industrial 1992-2001
Grab
Mandatory parameters
24,852b
General Permit Los Angeles,
are pH, SC, TSS, and O&G
CA
or TOC
Certain facilities must
analyze metals
Industrial State of industrial 1995-2003
Grab
Mandatory parameters
9589
General Permit Connecticut
are pH, COD, O&G, TSS,
Cu, Pb, Zn, TP, TKN, NO3,
and LCso
Industrial Countyof industrial 1993-2001
Grab
Mandatory parameters
810b
General Permit Sacramento,
are pH, SC, TSS, and O&G
CA
or TOG
Certain facilities must
analyze metals
Municipal 17 states in various 1991-2002
Flow -weighted
Nutrients, TSS, heavy
3765`
Permit the US land uses
composite
metals, and organic
Several pollutants such
toxicants, vary by state
as bacterial indicators
and O&G were collected
by grab sample
Municipal County of Various 1996-2001
Flow -weighted
Nutrients, heavy metals,
184
Permit Los Angeles, land uses
composite
and organic toxicants
CA
Several pollutants such
as bacterial indicators
such as pesticides and
herbicides
and O&G were collected
by grab sample
First Flush 405 and 101 Highway 2000-2002
Grab and flow -weighted
TSS, VSS, bacterial
351d
Highway Runoff freeway
composite
indicators, nutrients,
Characterization near UCLA,
O&G, COD, DOC, metals,
CA
hardness, and alkalinity
' Specific conductance (SC), total suspended solids (TSS), volatile suspended solids (VSS), oil and grease (O&G), total organic carbon (TOC),
dissolved organic carbon (DOC), chemical oxygen demand (COD),
copper (Cu), lead (Pb), zinc (Zn), aluminum (Al), iron (Fe), total phosphorus
(TP), total Kjeldahl nitrogen (TKN), and nitrate (NO3).
b The number of observations by parameter is shown in Table 2.
See Pitt (2004) for the number of observations for each parameter.
d Sample by grab and flow -weighted composite sample since 1999, However, we used only
grab samples and EMCs calculated from grab
f samples.in this paper.
Stenstrom, 2003), we successfully applied neural networks
(NNs) to identify relationships between water quality data
and various types of landuse (commercial, residential,
industrial, transportation, and vacant) using Los Angeles
County MS4 data. In this study, we used the same techniques
to analyze the ISGP data from the County of Los Angeles, the
County of Sacramento, and the State of Connecticut to
identify the various industrial activities as a function of
stormwater quality data. A multi -layer perception neural
model, the most common supervised NN, was used to
differentiate the various types of industrial activity using
Neural Connection 2.1 (SPSS, Inc. and Recognition, Inc.,
Chicago, IL).
Industrial and transportation landuses are generally known
as greater sources of heavy metals (generally six metals:
cadmium, copper, chromium, lead, nickel, and zinc) than
other land uses such as residential land use. We evaluated the
concentrations of metals in runoff from various landuse or
industrial categories to determine if landuse or source could
be associated with stormwater concentrations. We also
compared concentrations for each monitoring program, when
possible, to determine how many were outside the US EPA's
stormwater quality benchmarks.
The timing of a grab sample is important since the first
flush associated with many stormwaters creates a higher
concentration at the beginning of runoff. Therefore, the
sampling time during a storm event is important to avoid
the bias of the first flush and better characterize the mass
emission of the event. For example, Khan at al. (2006) showed
that oil and grease concentrations in samples from highways
in the first 15 min of runoff were several times higher than the
event mean concentrations (EMC). Samples early in a storm
event should be collected if the peak or maximum concen-
trations are desired.
The industrial sites are generally small, impervious water-
sheds and should experience a first flush. In so far as possible,

1V. Results
0 Table 12 provides a summary of the aforementioned model inputs.
•
•
Table 12. Summary of model inputs, two -segment marina, tidal amplitude method.
Input Parameter
Average Channel Depth
Average Marina Depth
Entrance Channel Surface Area
Marina Basin Surface Area
Tidal Amplitude
Return Flow Factor
Ambient Dissolved Oxygen
Saturation Dissolved Oxygen
Sediment Oxygen Demand
Decay Coefficient
Reaeration Coefficient
Channel Boat Activity
Marina Boat Activity
Value
8.0 ft
7.0 ft
97,800 ft2
950,289 ft2
1.0 ft
0 DIM
5.0 mg 1-1 to 8.0 mg 1-1
7.96 mg 1-1
1.5 g021m21day to 3.0 9021m21day
1 day-'
0.3 day-'
0 boat hours days
0 boat hours day-'
Table 13 summarizes the results from the DO model for the proposed marina entrance channel.
Table 14 summarizes the results from the DO model for the proposed marina basin. These
matrices display the break point along decreasing DO and increasing SOD concentrations in which
conditions result in DO concentrations below the State standard of 5.0 mg 1-1.
Table 13. Matrix of DO modeling results, Daniels Point proposed marina entrance channel. Average
ambient DO concentration of the canal system and the Neuse River was 7.2 mg 1-1.
DISSOLVED
OXYGEN
SEDIMENT OXYGEN DEMAND (902/m /day)
(mg 1-1)
1.5
2.0
2.5
3.0
8.0
-7.1 ;
6-8-
6_5 .
6.3-
7.5
6.8
6.5
6.2
6.0
7.2
6.6
6.3
6.1
5.8
7.0
6.5
6.2.
5.9
5 6 .
6.5
-6.2:
5.9
5.6
5.3
6.0
5.9
5.6
5.3
5.�
5.5
5.6
5:3
5.0rl
4.7
4.7
4.4
Shaded areas represent acceptable DO and SOD limits.
16

WATER RESEARCH 41 (2007) 4186-4196
4189
•
•
•
we have identified the sampling time and noted the implica-
tions of sampling time on monitoring results.
The decision of which storm events to sample is also
problematic. Climatic conditions, such as the existence of
long dry periods, may greatly impact pollutant emissions
from urban stormwater discharges. Most western parts of the
United States, including Los Angeles and Sacramento are dry
from May to September. This rainfall pattern creates a long
period for pollutant build-up and, therefore, the initial storm
of the wet season may have higher pollutant concentrations
than in later events (Lee et al., 2004). This phenomenon is
called a "seasonal first flush," and can be illustrated using
ISGP data from Los Angeles. These same climatic conditions
can be found in southern Europe, areas in South America,
Africa, and Australia.
The number of storms to be sampled during each monitor-
ing season is also an important parameter. Our analysis does
not address this decision, but we note that only one or two
storms each year is not likely to be representative. Leecaster
et al. (2002), in examining several monitoring programs in
California, recommended sampling seven storms per year
(approximately 50% of the storms in Southern California in a
typical year) to obtain small confidence intervals. Increasing
the number of storms or samples will create a burden to all
permittees, who will understandably question the benefits.
We suggest an alternative strategy to reduce the burden of
increased sampling requirements.
3. Results and discussion
Table 2 shows the basic statistics and US EPA benchmark
levels for the Los Angeles, Sacramento, and State of Connecti-
cut ISGP programs. The number of samples, mean, and CV are
shown. These results will be used in several of the following
sections.
3.1. Sampling method
Fig. 1 shows the CV (equal to the standard deviation divided
by the mean) for various routinely monitored water quality
parameters for water, wastewater and stormwater manage-
ment. The leftmost bars show the CVs for influent waste-
water quality parameters from a large west -coast wastewater
treatment plant. The CVs range from 0.09 to 0.2. The next set
of bars show the CVs for raw water quality parameters for a
large drinking water treatment plant. The CVs range from
0.07 to 0.3. The next group of bars shows the CVs from the
Municipal StormWater Permits (MS4), which range from 1.1 to
3.3 (Pitt, 2004). The final group of bars shows the CVs from the
ISGP data for Los Angeles County that are also tabulated in
Table 2. The ISGP data for Los Angeles County have the
highest CV among the various water data, and ranged from
2.8 to 14. The CVs from the ISGP data for the County of
Sacramento and the State of Connecticut are shown in Table 2
are also high when compared to the water and wastewater
monitoring results.
The variability in the stormwater data is several times the
mean value. With such high variability, it is virtually
impossible to make statistical inferences. Regulators and
users of stormwater monitoring data need to recalibrate their
expectations from stormwater monitoring programs. The
challenge in developing better monitoring programs is to
reduce the variability due to sampling errors and artifacts,
while retaining the intrinsic variability.
The sampling method is a major source of great variation.
For ISGP monitoring, grab samples are allowed, whereas
flow -weighted composite samples were collected in the
Table 2 -
Basic statistics
for three industrial
general
permit programs
Parameter
Unit
Benchmark level'
Los Angeles County
Sacramento County
Connecticut State
Sample no.
Mean
Cv
Sample no.
Mean
Cv
Sample no.
Mean
Cv
PH
pH unit
6.0-9.0
24,851
7.01
0.95
657
7.16
0.17
9617
6.32
0.20
TSS
mg/L
100
24,144
376
11.6
769
185
2.86
9617
124.
6.59
SC
pmos/cm
200
23,585
562
8.13
846
204
2.27
j O&G
mg/L
15
18637
16.6
14.3
286
11.3
1.61
TOC
mg/L
110
9714
50.1
5.23
399
31.4
2.12
SOD
mg/L
30
726
165.
10.7
COD
mg/L
120
1834
271.
2.77
50
154
1.58
9606
80.7
3.44
Al
mg/L
0.75
1618
10.1
12.2
46
3.44
1.66
Cu
mg/L
0.0636
3354
1.01
16.5
83
0.18
2,31
9596
0.13
7.56
Fe
mg/L
1
18"
25.5
6.39
82
7.49
2.03
Pb
mg/L
0.0816 -
3525
.2.96
14.1
78
4.48
3.82
9563
0.06
8.37
Ni
mg/L
1.42
1858
0,63
16.0
Zn
mg/L
0,117
5163
4.96
13.9
141
2.23
7.59
9614
0.51
7.79
P
mg/L
9606
0.45
4.30
TKN
mg/L
9608
2.50
3.11
NO3
mg/L
9613
1.19
2.73
24h LC,.
%
9628
0.82
0.38
48h LC50
%
9628
0.75
0.45
a Benchmark level suggested by US EPA.

Table 14. Matrix of DO modeling results, Daniels Point proposed marina basin. Average ambient
0 DO concentration of the canal system and the Neuse River was 7.2 mg I-'.
DISSOLVED
OXYGEN
SEDIMENT OXYGEN DEMAND (902lm /day)
(mg 1-1)
1.5
2.0
2.5
3.0
8.0
6.3
5.8
5.2 .°
4.7
7.5
6.2
,5'6
5:_1; y.
4.6
7.0
6.0
5.5
5:4
4.4
6.5
5:9
5.4
4.8
4.3
6.0
5.7
5.2"
4.7
4.1
5.5
5.6
5.1
4.5
4.0
5.0
5.3
5.0
4.4
3.8
Shaded areas represent acceptable DO and SOD limits.
V. Conclusions
As previously discussed, the present model assumes that there will be no additional runoff into the
system. Effective stormwater BMPs will be implemented in Phase II of the Dawson Creek
• Community to ensure that all runoff is graded away from the basin. No run-off will enter the basin
without prior treatment. These design features will aid in maintaining good water quality in the
basin and prevent the build up of pollutants,
Specific design parameters have been incorporated into the marina to increase flushing. Primarily,
water depths will be sloped from the head of basin towards the final depth contour in the river. The
use of marina best management practices will lessen the amount of pollutants entering the basin.
A "No Overboard Discharge" policy will be established, posted and enforced by an on -site dock
master. A stationary pump out station and properly trained staff will ensure disposal of waste.
Fuel service will also be supervised by trained marina staff. This marina will seek the NC Clean
Marina designation, and as such will be required to meet a wide variety of best management
practices for marina operations.
The DO model predicts DO concentrations will meet or exceed the State standard of 5.0 mg 1-'
within the proposed marina basin and entrance channel over a range of SOD concentrations
typical of estuarine mud (1.0 to 2,0 g021m2/day), While SOD concentrations higher than 2.0
902/M2lday would not be anticipated, it should be noted that during periods when sediment oxygen
demand is high and ambient DO concentrations are depressed, DO within the proposed basin has
the potential to drop below the State standard of 5.0 mg 1-1. In addition, ambient river monitoring
• data available through the USGS indicates that the Neuse River does experience episodic hypoxia,
especially during the summer months. As a result, the Daniels Point at Dawson Creek marina
17

4190
WATER RESEARCH 41 (2007) 4186-4196
•
0
•
16
12
vJ ❑ ❑ y U > m to it ❑ Z = c LO � U O
F Co 0 z 'E 0❑ 2 .> O F U N m ,5 O O N
U V
Y O 7 7
Q � '0 G
O O
U U
Fig. 4 - Coefficient of variation in various water sampling programs for a large wastewater treatment plant, a drinking water
treatment plant, and two stormwater monitoring programs, illustrating the different amounts of variability.
Municipal StormWater Permit program and water and waste-
water treatment plant monitoring programs. The use of grab
samples must increase the variability and CV It is universally
recognized that collecting flow -weighted composite samples
is better for stormwater monitoring. The result from a flow -
weighted composite sample is often called an EMC, which is
not only more representative than a grab sample, but can also
be used to estimate pollutant loading since the product of
EMC and total runoff volume is the pollutant load (Sansalone
and Ruchberger, 1997, Lee et al., 2002).
To illustrate the increased variability of grab samples over
composite samples, Fig. 2 is provided, which shows sampling
results from three storm events collected in a recently
completed highway runoff study (Stenstrom and Kayhanian,
2005). The results are reported for three different highway
runoff sites for three different storm events, and are
representative of approximately 70 storm events monitored
during the study. Values for hardness, chemical oxygen
demand (COD), dissolved organic carbon (DOC), oil and grease
(O&G), and ammonia (as N) are reported. Fig. 2 shows the
value of each grab sample, and compares it to the composite
sample, collected by a flow proportional automatic sampler,
which is denoted as "E" on the graphs (for EMC). No
composite is reported for O&G, which is usually not measured
with automated samplers. It is easy to observe that the
various grab samples can be 10 times greater or smaller than
the mean or EMC. Hence, the use of a single grab sample to
estimate mass emission rates or environmental impacts have
large error. Averaging 12 grab samples (horizontal line in Fig. 2)
is a much better estimate of the EMC, but still can be in error.
Unfortunately, collecting flow -weighted composite samples
is more difficult and expensive. The location and installation
of the sampler generally requires engineering and construc-
tion of sampling facilities, and training to operate the
samplers. There are approximately 3000 industrial permittees
in Los Angeles County alone, as compared to less than 50
monitoring programs for water and wastewater treatment
plants. Requiring composite sampling represents a large
financial burden. In addition, several water quality para-
meters, such as O&G, toxicity, and indicator bacteria are not
easily measured by automatic composite samplers.
3.2. Relationship between landuse activity and water
quality data
In a previous study, a multi -layer perception neural model
(Nerual Connection 2.1, SPSS Inc., Chicago, IL) was trained
with the basic water quality parameters for stormwater data
collected from specific landuses using composite samplers
(Lee and Stenstrom, 2005). The approach was applied to the
Los Angeles and Sacramento County ISGP data shown in
Table 2, and expanded to use three supervised, feed -forward

• basin will feature mechanic aerators to help maintain elevated DO concentrations in the basin
when ambient DO in the Neuse River is subpar. These devices will raise the level of DO in the
water and circulate floating debris out of the corners where it can be flushed naturally.
Significant impacts to water quality, specifically dissolved oxygen, are not anticipated as a result of
the proposed project.
VI. References
Chapra, S.C. 1997. Surface Water— Quality Modeling, McGraw-Hill Companies, Inc. New York,
New York. 844 p.
Environmental Protection Agency (EPA). 2001. National Management Measure Guidance to Control
Nonpoint Source Pollution from Marinas and Recreational Boating. EPA 841-B-01-005,
Mackenthun, A.A and H.G. Stefan. 1998, Effect of flow velocity on sediment oxygen demand:
Experiments. Journal of Environmental Engineering. 124:222-229.
NCDEHNR (North Carolina Department of Environment, Health, and Natural Resources). Division
of Water Quality. 1990. North Carolina Coastal Marinas: Water Quality Assessment Report No. 90-
01.
• Rizzo, W,M. and R.R. Christian, 1996, Significance of subtidal sediments to heterotrophically-
mediated oxygen and nutrient dynamics in a temperate estuary. Estuaries. 19(2B) 475-487.
Weiss, R. 1970. The solubility of nitrogen, oxygen and argon in water and seawater, Deep Sea
Research 17: 721-735.
0

0 0 0
Daniel Point Marina o Tide Study
108
8.15' 97.8"
NAVD88 11021 ---------------------- - - - - - -
Calibration
Point
96 -----------------------------------------------------------------------------------------------------
p au
•NOV
• T
z 9a--------- ---�. =------------ o o" o ::! -'--
s s _
D E II j I I
x •1 fie
i
•
j-- j i ----------------------------------------Zq -- :---
w +' I •--�---- ---------------
3 � � •
-78 �
• i
t `+
72 �------------------------- t---------------------------------------------------------------------------
66
- • n3�rscr -
JM Sanders
Note: Calibration point of well centered at 8.15 feet - 97.8 inches (NAVD88) Job # 02-07-593
Instrument Reads every Hour
Land Management Group, Inc.

4196
WATER RESEARCH 41 (2007) 41S6-4196
•
•
0
Sansalone, J.J., Buchberger, S.G., 1997. Partitioning and first flush
of metals in urban roadway stormwater. J. Environ. Eng. ASCE
123 (2), 134-143.
Stenstrom, M.K., Kayhanian, M., 2005. First flush phenomenon
Characterization. Report no. CTSW-RT-05-73-02.6, California
Department of Transportation, Division of Environmental
Analysis, CA, USA.
Stenstrom, M.K., Silverman, G.S., Bursztynsky, T.A., 1984. Oil and
grease in urban stormwaters. J. Environ. Eng. ASCE 110 (1),
58-72.
'Ibrner, B.G., Boner, M.C., 2004. Watershed monitoring and
modeling and USA regulatory compliance. Water Sci. Tech-
nol. 50 (11), 7-12,

0 0 0
Daniel Point Marina : Tide Study
Io8
8.15' (97.8"
) ---------------------------------------------
Calibration
Point
a;
96 1 -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
90AIN Y; --- - 0 ---------------
b OI
O
{
■ ��� b 1 I 0 �� � O II � � � I Q O i$ Y � I i O � 1 O
0,� - Q---I I I �i li 0 a
- - - - i i I I
84 �b- -�L I -I- �------�--f 0---- ^------------ -------- 4
u� i d e � �• @ o o `I �-o of a +_o
l p-��-�--
I 3d
I1 _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ - _ - _ - _ _ _ _ _ _ _ _ - _ - _ _ _ _ _ _ _ _ - _ - _ - _ _ - _ - _ - II _ _ -0_ -
d � �
72 1-____________________________________________________________________________________________________
66
l"
S.
e A3RF3CF
JM Sanders
Note: Calibration point of well centered at 8.15 feet w 97.8 inches (NAVD88) Job # 02-07-593
Instrument Reads every Hour
Land Management Group, Inc.

WATER RESEARCi? 41 (2007) 4186-4196
4195
•
•
•
4. Conclusions
Several stormwater monitoring programs, both national and
statewide, were evaluated to determine their usefulness in
identifying high dischargers and developing TMDLs. The
following conclusions and recommendation are made for
improving future programs:
1. Stormwater data, especially data in the Generalized
Industrial Monitoring Program that were collected using
grab samples, is highly variable, and typically much more
variable than data observed in other monitoring programs,
such as potable water or wastewater monitoring programs.
Reducing the variability from experimental error and
artifacts in data collection needs to be solved to improve
current monitoring programs, and provide for decision
making based upon collected data.
2. Statistical methods and a trained multi -layer perception
neural model were used to identify relationships
between water quality data and various types of industrial
activity based on SIC code and landuse but were
unable to detect the relationships. This was true for
landuses that have known differences in water quality,
and suggests that sampling methodology obscures the
relationships.
3. Data from better -designed monitoring programs show that
industrial landuse has the highest median concentrations
of cadmium, chromium, lead, and nickel, while highways
have the highest concentration of copper and zinc. In
addition, concentrations of metals exceeded the US EPA
stormwater benchmark level more frequently than the
basic water quality parameters. In anticipation of control-
ling metals for TMDLs and other purposes, future permits
should monitoring for metals for all industrial landuses,
unless it is known that metals will not be present.
Candidate metals include cadmium, chrome, copper, lead,
nickel, and zinc and were included in the reviewed
highway monitoring programs.
4. The timing of grab sample collection may affect results,
particularly for pollutants that experience buildup and
washoff, exhibiting a first flush. For these pollutants,
samples collected early in the storm will have higher
concentrations than the EMC and will be lower than the
EMC if collected late in the storm. If grab samples are
collected, they need to be collected at the appropriate
time. If peak concentrations are desired, samples collected
early in the storm are preferred; otherwise samples
collected in the middle of the storm are more representa-
tive.
5. Most places in California including Los Angeles and
Sacramento, which have distinct dry periods in the
summer, had higher pollutant concentrations in early
storm events than in later events, which documents the
existence of a seasonal first flush. Data from Connecticut,
with more uniform rainfall throughout the year, did not
show a seasonal first flush. Many places in the world have
distinct wet and dry periods and monitoring in these
locations should consider the impacts of seasonal first
flush on monitoring program goals. Samples collected
early in the season will better represent maximum
concentrations, but overestimate season mass loads.
Selecting a subset of permittees from each monitored
category using more advanced sampling techniques, such
as flow -weighted composite samplers, is a reasonable
approach for estimating representative loads from each
category and may result in lower overall cost with
improved accuracy and variability.
We expect that our results will be helpful to planners and
regulators to interpret existing datasets and programs, as
well as provide recommendations for improving future
programs. The results have particular value for California,
but are applicable worldwide and especially in areas with
Mediterranean climates.
R E F E R E N C E S
Davis, A.P., Shokouhian, M., Ni, S., 2001. Loading estimates of
lead, copper, cadmium, and zinc in urban runoff from specific
sources. Chemosphere 44, 997-1009.
Duke, L.D., Buffleben, M., Bauersachs, L.A„ 1998. Pollutants in
stormwater runoff from metal plating facilities, Los Angeles,
California. Waste Manage. 18, 25-38.
US EPA, 1983. Results of the Nationwide Urban Runoff Program,
vol. 1-Final Report. N71S access number PB84-18552, De-
cember 1983, Washington, DC.
Fam, S., Stenstrom, M.K., Silverman, G., 1987. Hydrocarbons in
urban runoff. J. Environ. Eng. Div. ASCE 113, 1032-1046.
Ha, H., Stenstrom, M.K., 2003. Identification of land use with
water quality data in stormwater using a neural network.
Water Res. 37, 4222-4230,
Han, Y, Lau, S.L., Kayhanian, M., Stenstrom, M.K., 2006. Correla-
tion analysis among highway stormwater pollutants and
characteristics. Water Sci. Technol. 53 (2), 235-243.
Khan, S., Lau, S.-L., Stenstrom, M.K., 2006. Oil and grease
measurement in highway runoff sampling time and event
mean concentrations. J. Environ. Eng. ASCE 132 (3), 415-422,
Kim, L.-H., Kayhanian, M., Lau, S.-L., Stenstrom, M.K., 2005. A new
modeling approach for estimating first flush metal mass
loading. Water Sci. Technol. 51 (3-4). 159-167.
Lau, S.-L., Stenstrom, M.K., 2003. Catch basin inserts to reduce
pollution from stormwater. Water Sci, Technol. 44 (7). 23-34,
Lee, H., Stenstrom, M.K., 2005. Utility of stormwater monitoring.
Water Environ. Res. 77 (3), 219-228.
Lee, J.H., Sang, K,W„ Ketchum, L.H., Choe, J.S., Yu, M.J., 2002. First
flush analysis of urban storm runoff. Sci. Total Environ. 293,
163-175.
Lee, H., Lau, S.-L., Kayhanian, M., Stenstrom, M.K., 2004. Seasonal
first flush phenomenon of urban stormwater discharges.
Water Res. 38, 4153-4163.
Leecaster, M.K., Schiff, K., hefenthaler, L.L., 2002. Assessment of
efficient sampling designs for urban stormwater monitoring.
Water Res. 36, 1556-1564.
Olding, D.D., Stelle, T.S., Nemeth, ].C., 2004. Operation monitoring
of urban stormwater management facilities and receiving
subwatersheds in Richmond Hill, Ontario. Water Qual. Res. J.
Can. 39 (4), 392-405.
Pitt, R., 2004. <http://unix.eng.ua.eduf-rpitt/Research/ms4iFi-
n al % 20Table % 20NSQD % 20v l_ 1 % 20020704.x is H ) .
Pitt, R., Maestre, A., Morquecho, R., 2003. Compilation and review
of nationwide MS4 stormwater quality data. In: Proceedings of
the 76th Water Environment Federation Technical Exposition
and Conference, Los Angeles, CA.

0 & 0
108
8.15' (97.8"
Daniel Point Marina : Tide Study
NAVD88iaz -------------------------------------- ----------------------------------------------------------
Calibration ,I
Point
96-------------------------------*--- I
s e is -: •I �-------------- ---- --� 46
o
U--- •
-------------
a I ! _!
a '�� • � o
o- - i 1 p------° i °,� °
M + f • Y
a I
3� -'----------�
---------78 -- &Vi-I--------------------------------------------- -•
-� ! ----------------'i-
1
- V
a n
II °
72 ---------------------------------------------------------
66 '— - - - - _ _ ._ - -- - - --- -- -
p�
�ay �a'� �a'� �a� �a� �ti��• �a'1 �o• �a'� �a"� �a'1 �a"1 �a� �a9 �,a'1 �a'i
• A3BF3CF
s
Note: Calibration point of well centered at 8.1.5 feet - 97.8 inches (NAVD88) Job # JM Sanders
93
Instrument Reads every Hour
Land Management Group, Inc.

4194
WATER RESEARCH 41 (2007) 4186- 4196
•
•
80
`c E
m E 40
80 0
200
60
L 150
U) E
(n u
~ 40 N 100
50
2a
30
Cn
20
U
0
f-
10
0.75
0.25
0 0.2 0.4 0.6 0.8 1
Normalized Cumulative Precipitation, 2000-2001
Fig. 6 - Concentrations versus normalized cumulative rainfall from Los Angeles general industrial stormwater permit
monitoring during 2000-2001 wet season, illustrating the high concentrations at the end of the dry season.
QOQ
100
10000
a
0
1000
CL
0
y 100
E
Z
10
1
56512
11 - x -- G.V. = 6.0
2511r C.V. = 4,0
0
\ 4128 — 0 C.V.= 2.0
A— C.V.= 1.0
630_ 0 00 `% AO —� C.V.= 0.8
�j\800'�3s4o �— C.V.=0.6
X` 2270 C,V,= 0.4
1580 \7580 a`--'1580 � 1580
A Ala x—_ i 180
�f 890
57d Epp �� S2A ~0--_570
394'�-_394
312 253
4:1 _1Q
q\\ _S130
\�a r\on _100
i7 64
z9 �+ �24 -fig- 26 17— 21 17
9� i2 13 ~G
4
�V
0 20 40 60 80 100
Percent difference in means
Fig. 7 - The required number of samples to detect a statistically significant difference between means as a function of relative
differences in means for an x value of 0.05 and a power of O.S.
lower than current costs. Even if only 10% of the permittees regulatory agency will have to develop appropriate methods
were sampled using composite samplers, the reliability of the for selecting sampling sites and distributing costs. Selecting a
data would be much improved and the CVs would be reduced. subset of each monitored category using more advanced
The remainder of the permittees could continue using grab sampling techniques is a reasonable approach and may result
samples or use some other program as mandated. A in lower overall cost with improved accuracy and variability.

•
108
8.15' (97.8")
NA VD88 102
Calibration
Point
3
96
90
84
78
72
66 L
o�
�P
Daniel Point Marina : Tide Study
---------------------------- - - -�-- --------------- - - - - -- ------------------------------------
I 0
0�`♦ b
tiI
--------t - i------------- ------ ------------------o--4e----------------------------------------
e
j
s: ,g
..rt
}r - -
------- - - - ---
------------ - - - - -------------- - - - - -- -
` ♦' A a
y i I
s 11
---------- - - - - - -- - - ------------------------------------------------------------- - - - - ----------
o�
llP IV 4-
Qt Q� Q� Q` Q` Qs 4t 4t Qt Q` 4`
�? �P $.P �.
A3BF3CF
JM Sanders
Note: Calibration point of well centered at 8.15 feet - 97.8 inches (NAVD88) Job # 02-07-593
Instrument Reads every Hour
Land Management Group, Inc.

•
r�
WATER RESEARCF! 41 (2007) 4136- 4196 4193
such as cadmium, copper, chromium, lead, nickel and zinc, rainfall, such as Connecticut, did not shown a seasonal first
unless they can demonstrate that these metals are not flush. This presence of a seasonal first flush is a dilemma for
present in their stormwater runoff. monitoring programs, which attempt to identify high dis-
chargers, as well as estimate central tendencies, such as
3.4. Sampling time in a storm event annual emissions.
In a the recently completed highway stormwater monitoring
project (Stenstrom and Kayhanian, 2005), grab samples were
collected every 15 min during the first hour of the storm and
additional grab samples were collected each hour for up to
8 h, Fig. 5 shows the impact of sampling time for TSS and total
Zn from highway sites. The ratio of observed concentration to
EMC is shown for more than 30 events for two wet seasons. If
the grab samples were the same as EMC, the points would all
appear along the horizontal line at 1.0. In general, the ratio of
the observed concentration to EMC is much higher than 1.0 at
the beginning of the storm and declines as the storm
progresses. It is obvious that collecting a sample in the early
part of the storm overestimates the EMC and the total load.
Khan et al. (2006) has examined this effect for sampling O&G
and concluded that collecting a grab sample 2-3h into a
typical storm more closely approximates the EMC than
sampling earlier or later in the storm. If grab samples are to
be used, the most appropriate time to sample should be
investigated.
3.5. Sampling time during wet seasons
Fig. 6 shows the median concentrations of specific conduc-
tance, TOC, TSS, and Zn as a function of the normalized
cumulative precipitation for ISGP data from Los Angeles
County. The concentrations of all parameters were highest in
the first few events, which is due to accumulation during the
long dry summer, and decreased as the season progressed. A
substantial concentration peak in the initial storm event
suggests the presence of a seasonal first flush. If only the first
storm is monitored, it will overestimate pollutant concentra-
tions in later storms. However, areas with more uniform
0 0 C C 0 C C C C C C
�2v O 0 N m O
N co M V V � CD
Sampling time (min.)
3.6. Sampling strategy
A routine response to inadequate data is to require more data
collection. It is temping to assume that additional sampling
will eventually overcome the variability of stormwater
monitoring programs. This conclusion is generally false,
because of the high variability, Fig, 7 shows the results of a
simulating a two -tailed t-test to determine the number of
observations required to detect differences in means between
two datasets. An n value (significance) of 0.05 and a power of
0.8 were assumed. The graph shows how many samples are
required to detect a statistically significant difference in
means. The graph is plotted as a number of required samples
as a function of relative difference in means, with the various
lines corresponding to different coefficients of variation. The
symbols indicate calculation points, with the numbers
corresponding to the ordinate values. Overcoming the varia-
bility by collecting additional samples will not be a good
approach. For Southern California, too few storms occur to
create the needed sampling opportunities.
An alternative approach might be used to select a subset of
representative industrial facilities from each major category.
The objective is to focus limited resources on fewer samples
to improve sampling methodology. In this way, the errors
introduced by a poor sampling technique might be avoided,
reducing the overall variability in the datasets.
This approach would be useful to estimate representative
loads from each major category, but would not be useful in
identifying high -risk dischargers. A large number of events
could be sampled using improved sampling technology, such
as flow -weighted composite samplers. The cost of such an
approach could be shared and the overall cost might even be
C 16
N
m
0
F 12
0
U
w B
c
0
U
0 4
Z
U7
a 0
0
O O O O O O O O O 0 0
+N^ N M M V V M
Sampling time (min.)
Fig. 5 - The ratio of observed concentration to EMC for TSS (left) and Total Zn (right) from UCLA First Flush Study during
2000-2001 and 2001-2002 wet seasons for highway runoff. Sampling time is the time since the beginning of measurable
runoff.

•
[7
4192
WATER RESEARCH 41 (2007) 4186-4196
60
15
3
+
34
21
270
A` /I 1
Cd
a�V. ',
n - n- -n Cr
z✓ 7
x— x—x Cu
\
7--7- -v Pb
40
10
2
C /
'\
A---A--A Ni
20
14
180
\ ` 1
x
20
5
1
14
7
90
J
fl
D
0
0
0
0
`G`a\ z
CP OQ
Fig. 3 - Median concentrations of metals in stormwater runoff from industrial, freeway, commercial,
residential, and open
land uses.
100
-,.c 80
E
t
60
a
w
C5
� 40
0
C
d 20
0
i1Connecticut1�
Sacramento
Los AngeleslMassachusetts�
`l
r
6• r •A
pH (6-9) COD TSS Copper Lead Zinc
(120) (100) (0.0636) (0.0816) (0.117)
Parameter
Fig. 4 - Percent of observations greater than US EPA Benchmark concentrations from four industrial permits (numbers below
parameter value indicate US EPA benchmark value in mg/L, except for pH). Values from Massachusetts not available for all
parameters.
Connecticut for all industrial categories, whereas only certain source of metals from industrial landuse is not only from
industrial facilities are required to analyze for metals in industrial activity, but also may be from building materials,
• California. There may be logic for including or excluding such as roof material and siding material of the industrial
metals in specific industrial permits, but the current mon- facilities (Davis et al., 2001). In anticipation of controlling
itoring results showed no remarkable differences in mon- metals for TMDLs and other purposes, any future permit
itored pollutant parameters among industrial categories. The should require industrial stormwater monitoring for metals

�'• ` i' + tir. i A ? �'�� '� ' y` -:�'k ,� S �' ���ir�k141 ;.7 _ � ��'�•'� >r
'+ _ Y � M fi 'f'1 .l,�,�• '� .. 4 w' ..� ,`�r +�,L�a�F :F � i . ,Y
Ja �� r -�. . i5� � � -� r '�c '�* �# ������ : 9i �N r x� 4� r�; � ,•.,r,
Cd
.L5! wli
Mil
Eh
�a��'� �i n�a ���t��rr1 Tt,la�, �'xn�f�'� � � '•',i 1#`i. r�'�• � ,� `''i.
+,'�, �,S: aid r 5:.. a+� "`it���}. J Jp r. ,'{� 4 f.•� z ` i. !. i„-
late .r � �+� � ♦ s'p���� ,n i''7� � � 7� 'Ls T r �'� �' i�• r i' .D-O- � ;t'. `Sy it M16" ,m-,'aJ'J,f if E'i 6'Py M $' •"
' f!iYr^L J s hM1 Rlr ' }� 5 �,�Jr ,....•?.,/ ! ti
qff
' .ray rr �':�` � � � ' �,�., r . ti d �.. •r�� ;
P ' k, r �-
�s ,.y rr 'y «4 �ri6•�'ia�Nk.Yt _✓ r� .. .ta..� ,r' � M . .
D4
rjr5 c+� ,dz +d ^cM.{ 1 p '�}.,+ r���! S{l �,'1' �q:s 1 _ jL�. --' • �w•Nt+ ��
i r ?
3,voif
i;t 4e'
1
` 1 f✓ n` 3 r.i xs tt .1 _j
J.. w J '4+i 4 r,. •r .lro i q wrJr I
• L • ~ r ! { II
.j" ^�
# - ?�' `� ° iCt,4^r y ri;9✓:.tx r ((^^{{]]`�� 'aj• I� f r.
q -•
DO r.
rr r
Neuse River

WATER RESEARcx 41 (2007) 4186-4196
4191
•
0.1
Hardness COD DOC O&G NH3-N Hardness COD DOC O&G NH3-N
Fig. 2 - Scatter diagram of the values of grab samples for hardness (mg/1 CaCO3), chemical oxygen demand (COD), dissolved
organic carbon (DOC), oil and grease (O&G), and ammonia nitrogen (NH3-N), compared to the composite samples (E) colIected
• with automatic flow -weighted samplers and the mean of the grab samples (horizontal line). Composite samples not collected
for oil and grease. Upper left panel: Site 3, February 11, 2005; upper right panel: Site 2, October 26, 2004; lower right panel: Site
2, October 16, 2004; lower left panel: Site 1, October 26, 2004.
r1
u
NNs (multi -layer perceprton, radial basis function, and
Bayesian network), Because of the small number of observa-
tions, it was not possible to train using metals data. For
Connecticut's data, both basic water quality parameters and
metals data were sufficient to be used as input parameters.
For output parameters, eight major industrial categories
(based on SIC code) were selected, based on their prevalence
in case number —food and kindred products (20); chemical
and allied products (28); primary metal industries (33);
fabricated metal products (34); transportation equipment (37);
motor freight transportation and warehousing (42); electric,
gas and sanitary services (49); and whole trade -durable goods
(50). Outliers in the dataset were defined as greater than 1.5
times the interquartile range plus the 75% data or less than
25% data minus 1.5 times the interquartile range, and were
removed from the analysis. This procedure is consistent with
techniques used and documented in the Systat manuals. The
neural models were extensively trained using various archi-
tectures; however, the performance of all models was very
poor. Only 20% of data case are correctly classified which is not
significant. These results confirmed that either no relationship
exists between the industrial categories based on SIC code and
the stormwater runoff quality, or that sampling errors are so
large that they obscure any relationship.
The overall conclusion of this exercise is that the current
monitoring programs are unable to discern a relationship
between industrial activity and stormwater quality. The
causes are probably a combination of imprecise nature of
SIC codes as well as the errors introduced by poor monitoring
technique, such as the use of grab samples and non-standard
timing for sample collection.
3.3. Sampling parameters
Fig. 3 shows metal concentrations as a function of landuse
(highway data from Han et al., 2006; other landuse calculated
from Pitt's (2004) dataset. Industrial landuse has the highest
median concentrations for cadmium, chromium, lead, and
nickel, but copper and zinc are the highest from highways.
Similar results were found from Los Angeles County's
Municipal StormWater Permit program, which were not
included in Pitt's dataset. industrial landuse had the highest
concentrations of aluminum, nickel, and zinc, but copper was
highest in transportation landuse (not shown in this paper).
The results confirm that industrial landuses are greater
sources of heavy metals than other landuses. In addition,
the concentration of metals more frequently exceeds the US
EPA's stormwater benchmark than for basic water quality
parameters (Fig. 4). For example, zinc concentrations ex-
ceeded the benchmark concentration approximately 90% of
samples from the ISGP in Los Angeles County.
Including metals in industrial stormwater monitoring
programs should be a high priority, but their presence varies
among states and industrial categories. It is mandatory in