3912000024 / December 26, 2002 / Page i
DEPARTMENT OF ENVIRONMENTAL PROTECTION
Bureau of Water Supply and Wastewater Management
DOCUMENT NUMBER:
3912000024
TITLE:
Field Data Collection and Evaluation Protocol for Deriving Daily and
Hourly Discharge Coefficients of Variation (CV) and Other Discharge
Characteristics
EFFECTIVE DATE:
October 13, 1998
Minor edits were made throughout (December 26, 2002)
AUTHORITY:
Federal Clean Water Act, PA Code Title 25, Chapter 92, 93, and 95 The
Clean Streams Law, 35 P.S. §§691.1 et seq.
POLICY:
It is the policy of the Department of Environmental Protection (DEP) to
use the best available data in the calculation or estimation of wasteload
allocations and NPDES effluent limitations. This often involves the use of
data that can only be collected in the field.
PURPOSE:
This document presents recommended data collection and analysis
procedures that can be used to determine two different discharge CVs
(Hourly and Daily) and the average monthly effluent concentration for
each parameter of concern. The discharge CVs (hourly and daily) and the
average monthly effluent concentrations derived from these data become
data inputs in DEP’s computer modeling program PENTOXSD.
APPLICABILITY:
DEP will use this guidance to provide technical details on how the
program carries out various calculations and applies the evaluated results
in preparing NPDES permits for dischargers.
DISCLAIMER:
The policies and procedures outlined in this guidance are intended to
supplement existing requirements. Nothing in the policies or procedures
shall affect regulatory requirements.
The policies and procedures herein are not an adjudication or a regulation.
There is no intent on the part of DEP to give the rules in these policies that
weight or deference. This document establishes the framework within
which DEP will exercise its administrative discretion in the future. DEP
reserves the discretion to deviate from this policy statement if
circumstances warrant.
PAGE LENGTH:
13 pages
LOCATION:
Volume 29, Tab 12
3912000024 / December 26, 2002 / Page 1
1.
INTRODUCTION
The NPDES permit limit specifies the allowable effluent quality and indicates if it is technology
or water quality based. If it is technology based, the limit is derived from the Long Term
Average (LTA) effluent quality of the treatment facility. If it is water quality based, the limit is
derived by: 1) computing Waste Load Allocations (WLAs) for each water quality criterion, 2)
translating them to LTAs, and then, 3) using the most stringent LTA to compute the Average
Monthly Limit (AML). In addition, a Maximum Daily Limit (MDL) is computed. Each of the
translations described here require effluent quality data for the treatment facilities.
For a given parameter, standard statistical techniques are used to describe the mean concentration
(long term average), variance, standard deviation and the coefficient of variation (CV). With a
limited amount of field data, it is possible to simulate an entire distribution (or set) of values and
then select a specified probability of occurrence. An average monthly effluent limit that must be
achieved a certain percentage of the time can also be computed. Generally, this is set at the
upper percentile (99
th
) so that the discharge will be in compliance nearly all the time. The most
significant piece of information used in all statistical translations is the discharge coefficient of
variation.
Based on the above discussions this document presents recommended data collection and
analysis procedures that can be used to determine two different discharge CVs (Hourly and
Daily) and the average monthly effluent concentration for each parameter of concern. The
discharge CVs (hourly and daily) and the average monthly effluent concentrations derived from
these data become data inputs in DEP’s computer modeling program (PENTOXSD). This
protocol will be used by both DEP and the applicant. This document describes:
• the sampling methods and procedures to follow for collecting the data,
• minimum quantity and frequency that samples should be collected,
• analytical methods that should be followed,
• reporting the analytical results, and,
• interpretation and evaluation of sample data.
(Note: The procedures used to compute water qualitybased effluent limits are described in the
“U.S. EPA Technical Support Document (TSD) for Water QualityBased Toxics Control, March
1991.” Specifically, they can be found in the “Chapter 5, Section 5.4, Permit Limit Derivation,
Pages 98105 and Section 5.5, Special Considerations in Use of Statistical Permit Limit
Derivation Techniques, Pages 105110.”)
2.
SAMPLING METHOD AND PROCEDURE
2.1
Sampling Techniques
Sample collection should be conducted or supervised by a person experienced in
performing wastewater sampling.
Specific requirements for sample containers, sample preservation, holding times, sizes,
etc. relevant to the applicable test methods must also be followed. When an EPA
approved test method is not available, DEP should be contacted for test procedures and
sample collection/analysis techniques.
3912000024 / December 26, 2002 / Page 2
Each sampling event shall occur at the time representative of normal operation. This
means having all processes which contribute wastewater in normal operation, and with a
properly operating treatment facility, (where treatment is being provided) which is not
experiencing upset conditions.
Dischargers with two or more substantially identical outfalls, may request permission to
sample and analyze only one of those outfalls. Where such permission is granted, the
discharger may submit the results of the analysis as representative of the other
substantially identical outfalls. DEP will use the data to determine if effluent limits or
other requirements are necessary.
Samples should be collected from the center of the flow channel where turbulence is at a
maximum, or at a location specified in the NPDES permit, or at any discharge pipe or
location adequate for the collection of a representative sample.
2.2
Sample Types
Two different types of samples can be collected. Grab samples can be taken at the
appropriate time intervals specified below. A sampling event consists of one grab sample
and one analysis. Composite samples can be taken during the operation of the facility
and at the appropriate time intervals specified below using either manual or automatic
compositing. Manual compositing must be done by collecting at least six grab samples
during the actual hours of discharge over a 24hour period. These grab samples are
combined immediately and one analysis is performed on the “composite” sample.
Automatic compositing can be more frequent but not less than six samples per day.
2.3
Sample Definitions
A grab sample is an individual sample of at least 100 milliliters, collected over a period
not exceeding 15 minutes at a randomly selected representative time.
A composite sample is a combination of at least six sample aliquots of at least 100
milliliters, collected manually or automatically at periodic intervals during the operating
hours of a facility over a 24 hour period.
The composite must be flowproportional which can be achieved by one of following
three methods:
• Constant sample volume taken at time intervals, between samples, that are
proportional to the discharge flow.
• Constant time intervals between samples, with varying sample volume proportional to
total discharge flow at the time of sampling.
• Constant time intervals between samples, with varying sample volume proportional to
total discharge flow since the last sample was taken.
If the discharge rate is constant a constant sample volume can be taken at constant time
intervals.
3912000024 / December 26, 2002 / Page 3
2.4
Exceptions
For effluents from holding ponds or other impoundments with retention times of greater
than 24 hours, a minimum of one grab sample may be taken. This is because of the
expected uniformity in the flow and mass of pollutant. DEP may also waive automatic
composite sampling for any outfall for which it is demonstrated that use of an automatic
sampler is infeasible. In this case, a composite sample, derived from a minimum of six,
equally spaced, manual grab samples, will be considered representative.
3.
SAMPLING FREQUENCY
3.1
Determining Daily Coefficient of Variation.
A minimum of 12 separate composite samples must be taken during the normal operation
time of the facility to get a representative and relevant estimate of the dailyCV. These
composites will represent the daily average effluent concentration for each parameter.
Each sample should be taken during the lowest stream flow months of the year when the
design stream flow occurs. This is generally during the summer months July through
October, but for regulated streams, this could be at other times of the year.
Table 3.1  Determining the Daily Coefficient of Variation
Type of
Flow
Discharge Operations
Size (gpd)
# of Samples
Type of Sample(s)
POTW
Continuous
> 100,000
12 sampling events, one
per week taken over 12
week period
24Hour Composite.
(6 grabs, 4 hr
increments)
POTW
Continuous
<100,000
12 sampling events, one
per week taken over 12
week period
24Hour Composite.
(2 grabs, 12 hr
increments
IW
Continuous
Same as POTW
24Hour Composite
IW
Batch
Same as POTW except
taken during operation
and discharge of facility
Composite or Grab
Each sample should be a 24hour composite taken as specified in Section 1.2, above. For
discharges with flows greater than 100,000 gpd, a minimum of six grab samples must be
taken to make up the manual composite. These can be spaced apart in no less than 4hour
increments. The applicant/DEP can collect more than six grabs per day, however, they
must be spread over equal increments so that the sample is representative of the entire 24
hour period. If the compositing is done automatically, then the applicant/DEP can take
more but no less than six samples per day to make up this composite. For discharges
with flows less than or equal to 100,000 gpd, the daily minimum is reduced to 2 samples
spaced no less than 12 hours apart.
3912000024 / December 26, 2002 / Page 4
Sampling intervals should not be less than one per week. Therefore, the completion of
this study requires a minimum of 12 weeks.
Table 3.1, above, can be referred to for selecting type of operation, number of samples to
collect, and the type of samples for Daily Coefficient of Variation.
3.2
Determining Hourly Coefficient of Variation
A minimum of 4 consecutive days should be scheduled for sampling during the operation
time of the facility.
A minimum of 24 grab samples should be collected during this time period for discharges
with flows in excess of 100,000 gpd. For discharges that are less than or equal to
l00,000 gpd a minimum of 4 grab samples should be collected. Each individual grab
sample will represent one discrete hourly event in the given day. These grabs will
approximate the hourly effluent concentration for each parameter and will be used to
determine the hourlyCV.
Each sample should be taken during the lowest stream flow months of the year when the
design stream flow occurs. This is generally during the summer months July through
October, but for regulated streams, this could be at other times of the year.
For discharges with flows in excess of 100,000 gpd, all samples should be grabs spaced
apart in equal 4hour increments. A total of six samples should be collected during any
one day. The applicant/DEP can collect more than six grabs in each day provided that
they are spread over equal increments. The applicant/DEP must not compress the
schedule by collecting more samples per day for fever days. For discharges with flows
less than or equal to 100,000 gpd, the applicant can collect 1 grab sample per day over 4
consecutive days.
Table 3.2, below can be referred to for selecting type of operation, number of samples to
collect, and the type of samples for Hourly Coefficient of Variation.
Table 3.2  Determining the Hourly Coefficient of Variation
Type of
Discharge Operations
Flow Size
(gpd)
# of Samples
.
Type of
Sample(s)
POTW
Continuous
>100,000
6 sampling events per day over
4 consecutive days. Samples
taken minimum 4 hours apart
Grab
POTW
Continuous
<100,000
1 Sample per day over
4 consecutive days
Grab
IW
Continuous
Same as POTW
Grab
IW
Batch
Same as POTW except taken
during operation and discharge
of facility
Grab
3912000024 / December 26, 2002 / Page 5
3.3
Discharge Flow Information
It is highly recommended that the applicant collect discharge flow information concurrent
to the sample collection. The flows may be estimated for facilities that are not equipped
for continuous flow monitoring.
4.
ANALYTICAL METHODS
Analytical methods promulgated in 40 CFR Part 136 or other DEP approved test methods must
be used where applicable. If no method has been promulgated by EPA for a particular pollutant,
use a suitable method for measuring the level of the pollutant provided that a description of the
method or a reference to a published method is attached to the results. The description shall
include the sample holding time, preservation technique, and quality control measures. DEP
reserves the right to review the methodology used and require further analysis, if necessary.
In the determination of Free Cyanide (Group B, 16M), since there is no EPA approved method,
DEP’s Bureau of Laboratories has developed a method to be used. Contact the Bureau of
Laboratories, P.O. Box 1467, Harrisburg, PA 17120 (7177874669) for a copy of this
procedure. The applicant must specify the analytical detection level achieved for the pollutant
and associated wastewater matrix effects. If the applicant cannot achieve the minimum
acceptable detection levels, then an explanation must accompany the application. In addition, a
list of the acceptable analytical methods is contained in Pa. Code Title 25, Chapter 16, Water
Quality Toxics Management StrategyStatement of Policy, Table 2.
5.
INTERPRETATION AND EVALUATION OF SAMPLE DATA
First, the selection of the appropriate statistical approach requires an assumption about the
distribution of these data which is that they follow a lognormal distribution. Once this
assumption is made, the procedures from the EPA TSD, Appendix E (March 1991) can be used.
There are two possible sets of computations that could be performed. This will depend on
whether or not there are any “nondetect’ values in the sample set. If all of the values are
“detectable” then the lognormal statistics are used. If some of the values are “notdetectable”
then the deltalognormal method is used. Each is described below.
The same procedures described below can be used to evaluate and compute the statistics for the
samples collected regardless of whether they are daily averages or hourly (grab) samples. The
purpose of these computations is to determine the coefficient of variation for each parameter
sampled. In addition, the average monthly effluent concentration can be computed. This value
could be used in any scientific modeling that DEP performs.
3912000024 / December 26, 2002 / Page 6
5.1
LogNormal Statistics
Compute each effluent concentration in a lognormal distribution:
where,
X
i
= Observed Value
Y
i
= Natural Log of Each Measurement
Compute the mean and variance of the data. The equations are:
µ
y
=
Σ$%
yk
i
/
(4.1a)
σ
2
y
= Σ
[ (
y
i

μ
y
)
2
]
/
()
k
 1
(4.1b)
where
µ
y
= Mean of the Log Normal Values
k
= Sample Size of the Data Set
2
σ
y
= Variance of the Log Normal Values
Compute the Long Term Average and Variance for the lognormal distribution. The
equations are:
E(x) exp
(
+
2
/ 2)
μ
y
σ
y
(4.1c)
V(x) exp
(2
+
2
)×[exp(
2
)1 ]
μ
y
σ
y
σ
y
(4.1d)
where
E(x) = Long Term Average for the observed data
V(x) = Variance for the observed data
Finally, compute the coefficient of variation. The equation is:
CV(x) =
[exp(
σ
2
y
)1]
0.5
(4.1e)
Compute the average monthly effluent concentration
(for daily average values only)
AMEC(x) = E(x) × [( In[CV(x)
2
/n+1] )
.5
 0.5 × In(CV(x)
2
/n+1)].
(4.1f)
where,
n = number of samples per month used to derive the average (n=4).
3912000024 / December 26, 2002 / Page 7
5.2
Delta LogNormal Statistics
Compute each effluent concentration in a lognormal distribution and compute the ratio
of nondetected values to the total sample size.
Υ
i
=
In
(
x
i
)
(4.2a)
where,
x
i
= Observed Value
Y
i
= Natural Log of Each Measurement
Next, compute the mean and variance. The equations are:
µ
Υ
= Σ
$%
y
1
/ (
k
−
r
)
(4.2b)
= Σ[(  )
2
]/(  1)
1
σ
2
y
y μ
y
k r
(4.2c)
where,
μ
y
= Mean of the Log Normal Values
2
σ
y
= Variance of the Log Normal Values
r
= Number of nondetected values in the data set
k = Sample Size of the data set
Complete the Long Term Average and Variance for the lognormal distribution. The
equations are:
E
(
x
) =
δD
(1
δ
)exp(
μ
y
+ 0.5
σ
2
y
)
(4.2d)
V(x) =
()1
δ
exp(2
μ
y
+
σ
2
y
)[exp(
σ
2
y
)()1
δ
]+
δ
()1
δ D
[
D
2exp(
μ
y
+0.5
σ
2
y
)]
(4.2e)
where,
D
= Analytical Detection Limit
E(x) = Long Term Average for the observed data
V(x) = Variance for the observed data
δ
=
r
/
k
Finally, compute the coefficient of variation. The equation to use is:
CV(x) = (V(x))
0.5
/ E(x)
(4.2f)
3912000024 / December 26, 2002 / Page 8
5.3
Department’s Computer Program
DEP has developed a computer program
(TOXCONC)
that can be used to assist the
applicant in computing the necessary statistics. In the case of daily average values, the
program can also compute the average monthly effluent concentrations. This is a
spreadsheet application developed to run in Lotus 123 and is available for both the MS
DOS and Windows operating systems. The user enters the analytical data for each
measurement taken. These will either be the individual grab samples or the composites
(which represent daily averages.) The program then computes the statistics described
above, internally, and simply displays the resulting coefficient of variation and average
monthly effluent concentration. TOXCONC can accept up to 60 samples for each
parameter being evaluated. Daily and hourly data, however, should be entered in
separate columns. Each input data file has a capacity to accept data up to 10 different
parameters. The program produces a report that includes a listing of the data inputted and
the resulting statistics. These values are then used in the PENTOXSD computer
program. For more general information on this program, including the user’s guide and
systems disk, please contact the Bureau of Water Supply and Wastewater Management.
For case specific questions, please contact the appropriate field office.
5.4
Computing the Coefficient of Variation using LogNormal Statistics
When all of the values are reported as detectable, the following procedure is used.
Table 5.1  Example of Reporting Results for Daily
CV Studies
Discharge Statistics
Detection Level
(1)
= _____
Date
Sample
Conc.
mg/l(x
i
)
Ln of
Sample
y
i
= In(x
i
)
Variance
(y
1
?
y
)
2
7/18/94
0.04
3.2189
0.5071
7/25/94
0.05
2.9957
0.2391
8/1/94
0.06
2.8134
0.0940
8/15/94
0.11
2.2073
0.0897
8/22/94
0.11
2.2073
0.0897
8/29/94
0.09
2.4079
0.0098
9/5/94
0.075
2.5903
0.0070
9/12/94
0.08
2.5257
0.0004
9/19/94
0.06
2.8134
0.0940
9/26/94
0.1
2.3026
0.0417
10/3/94
0.225
1.4917
1.0304
µ
y
=
2.5067
σ
y
2
=
0.2203
Summary Statistics
LTA
= 0.0910
CVd
= 0.4964
V(x)
= 0.002
AMEC = 0.0027
(1)
Use only if some of the samples are less than detection level, these samples
are reported as zero along with the detection level to compute the delta log
normal statistics.
3912000024 / December 26, 2002 / Page 9
Table 5.2  Example of Reporting Results for Hourly CV Studies
Discharge Statistics
Detection Level
(1)
= _____
Date
Sample
Time
Sample
Conc.
mg/l(x
i
)
Ln of
Sample
y
i
= In(x
i
)
Variance
(y
1
 µ
y
)
2
9/1/94
12:00 am
0.03
3.5066
1.1459
4:00 am
0.062
2.7806
0.1187
8:00 am
0.097
2.3330
0.0106
12:00 pm
0.092
2.3860
0.0025
4:00 pm
0.1225
2.0996
0.1132
8:00 pm
0.099
2.3125
0.0152
9/2/94
12:00 am
0.054
2.9188
0.2330
4:00 am
0.02
3.9120
2.1784
8:00 am
0.09
2.4079
0.0008
12:00 pm
0.175
1.7430
0.4804
4:00 pm
0.1
2.3026
0.0178
8:00 pm
0.225
1.4917
0.8920
9/3/94
12:00 am
0.125
2.0794
0.1272
4:00 am
0.09
2.4079
0.0008
8:00 am
0.117
2.1456
0.0844
12:00 pm
0.1
2.3026
0.0178
4:00 pm
0.09
2.4079
0.0008
8:00 pm
0.07
2.6593
0.0498
9/4/94
12:00 am
0.04
3.2189
0.6127
4:00 am
0.04
3.2189
0.6127
8:00 am
0.176
1.7373
0.4884
12:00 pm
0.345
1.0642
1.8821
4:00 pm
0.109
2.2164
0.0483
8:00 pm
0.06
2.8134
0.1424
µ
y
2.4361
σ
y
2
0.3865
Summary Statistics
LTA
= 0.107
CVd
= 0.7048
V(x)
= 0.0056
(1)
Use only if some of the samples are less than detection level, these samples
are reported as zero along with the detection level to compute the delta log
normal statistics.
3912000024 / December 26, 2002 / Page 10
5.5
Computing the Coefficient of Variation using Delta LogNormal Statistics
Whenever there are some “notdetected” values in the data set, the method for computing
the coefficient of variation is slightly different. This takes into account the effects of
having some values less than the detection limit, knowing that these values are not equal
to zero.
6.
REPORTING RESULTS OF ANALYSES
When sampling process, noncontact cooling, miscellaneous, or sanitary wastewater, all results
must be reported in concentration (mg/l or ug/1) and/or as mass units (lbs/day), except where
other units of measurement apply (temperature, pH, color, radioactivity, fecal coliform).
All reporting of values for metals must be in terms of “total” metal unless the approved
analytical methods for the metal inherently measure only its dissolved form (e.g. hexavalent
chromium).
As indicated in Section 4 above, the applicant is expected to provide information on the level of
detectability used (achieved).
Two summary tables have been prepared as examples on how to submit the results from these
studies. Table 5.1 provides an example of how to report the results when doing a study to
determine the daily coefficient of variation and compute the average monthly effluent
concentration. Table 5.2 provides an example of how to report the results when doing a study to
determine the hourly coefficient of variation. DEP will accept these reports in hardcopy. DEP
will also accept the results in electronic format, either in ASCII, Microsoft Excel, or word
processing format, or, data files created while using the TOXCONC program.
DEP prefers the following method for reporting the results for each parameter evaluated (in order
of preference):
• Submission of the TOXCONC reports (hard copy) along with the electronic data files,
• External computations of statistics and submission of data in tabular format, above, and,
• Submission of the actual laboratory analysis results.
Sections 5.1 and 5.2, above, can be used as a guide to doing the computations by hand. Use the
equations in Sections 5.1 and 5.2 to compute the long term average, variance, coefficient of
variation and the average monthly effluent concentration. The equations for the two columns in
each table are provided in the appropriate table. One table should be completed for each
parameter of concern. If the applicant uses TOXCONC, submit the hard copy reports along with
the data files for each analysis performed.
3912000024 / December 26, 2002 / Page 11
REFERENCES
U.S. EPA Technical Support Document (TSD) for Water QualityBased Toxics Control, March 1991.
Federal Register, 40 CFR Part l36
Chapter 16  Pennsylvania Code  Title 25 Environmental Protection www.pacode.com
3912000024 / December 26, 2002 / Page 12
Sample Data Sheet
Discharge Statistics
Detection Level
(1)
= _____
Date
Sample Time
Sample Conc.
mg/l(x
i
)
Ln of Sample
y
i
= In(x
i
)
Variance
(y
1
 µ
y
)
2
µ
y
=
σ
y
2
=
Summary Statistics
LTA
=
CVd
=
V(x)
=
(1)
Use only if some of the samples are less than detection level, these samples are
reported as zero along with the detection level to compute the delta lognormal statistics.
Back to top