United States
Environmental Protection
Agency
Office of Water
(4303)
EPA821-B-01-011
December 2001
FINAL
&EPA Coal Remining Statistical Support
Document
-------
-------
EPA-821-B-01-011
COAL REMINING
STATISTICAL SUPPORT DOCUMENT
DECEMBER 2001
Office of Water
Office of Science and Technology
Engineering and Analysis Division
U.S. Environmental Protection Agency
Washington, DC 20460 .
-------
-------
Coal Remining Statistical Support Document
Acknowledgments
This document was developed under the direction of William A. Telliard and John F. Fox of the
Engineering and Analysis Division (BAD) within the U.S. Environmental Protection Agency's
(EPA's) Office of Science and Technology (OST). EPA gratefully acknowledges Roger
Hornberger, Michael W. Smith, Daniel J. Koury, and J. Corey Cram of Pennsylvania's
Department of Environmental Protection (PADEP) for their contributions to this document. EPA
also wishes to thank DynCorp Information and Enterprise Technology for its invaluable support
in developing this document.
Disclaimer
The statements in this document are intended solely as guidance. This document is not intended,
nor can it be relied upon, to create any rights enforceable by any party in litigation with the
United States. EPA may decide to follow the guidance provided in this document, or to act at
variance with the guidance, based on its analysis of the specific facts presented/ This guidance is
being issued in connection with amendments to the Coal Mining Point Source Category.
The primary contact regarding questions or comments on this document is:
John F. Fox
Engineering and Analysis Division (4303)
U.S. Environmental Protection Agency
Ariel Rios Building, 1200 Pennsylvania Avenue, N.W.
Washington, DC 20460
Phone: 202/260-9889
email: fox.john@epa.gov
Acknowledgments
-------
Coal Reminine Statistical Suwort Document
Acknowledgments
-------
Coal Remining Statistical Support Document
TABLE OF CONTENTS
Page
LIST OF TABLES . ... . Hi
LIST OF FIGURES .. v
Section 1.0 Introduction 1-1
1.1 Remining Program History 1-2
1.2 Pennsylvania DEP Remining Permitting Procedures 1-4
1.2.1 Pre-existing Discharges . 1-4
1.2.2 Baseline Pollution Load-Determination and Compliance Monitoring ..1-5
1.3 IMCC Evaluation of State Remining Programs 1-13
References 1-15
Section 2.0 Characteristics of Coal Mine Drainage Discharges 2-1
2.1 Impact of Stream Flow Variation on Water Quality Parameters 2-9
2.2 AMD Discharge Types and Behaviors . 2-13
2.3 Distributional Properties of AMD Discharges . 2-18
References '. 2-23
Section 3.0 Statistical Methodology for Establishing Baseline Conditions and Setting
Discharge Limits at Remining Sites 3-1
3.1 Objectives, Statistical Principles, and Statistical Issues 3-1
3.2 Statistical Procedures for Calculating Limits from Baseline Data 3-3
3.2.1 Method 1 3-4
3.2.2 Method2 3-5
3.2.3 Accelerated Monitoring 3-5
References 3-8
Section 4.0 Baseline Sampling Duration and Frequency 4-1
4.1 Power and Sample Size 4-1
4.2 Sampling Plan 4-5
References • • • • 4-7
Section 5.0 Long-term Monitoring Case Studies 5-1
5.1 A Comparison of Seven Long-term Water Quality Datasets 5-2
5.1.1 Sampling Interval 5-8
Table of Contents I
-------
Coal Remining Statistical Support Document
5.1.2 Duration of Baseline Sampling 5-24
5.1.3 Effects of Discharge Behavior on Baseline Sampling 5-25
5.1.4 Year-to-Year Variability . 5-26
5.2 The Effects of Natural Seasonal Variations and Mining Induced Changes
in Long-term Monitoring Data 5-26
5.2.1 Markson Discharge 5-29
5.2.2 Tracy Discharge 5-35
5.2.3 Swatara Creek Monitoring Station 5-40
5.2.4 Jeddo Tunnel Discharge 5-42
5.3 Case Studies :.. 5-45
5.3.1 Fisher Discharge 5-45
5.3.2 McWreath Discharge ... 5-51
5.3.3 Trees Mills Site 5-55
5.4 Conclusions 5-70
References 5-72
Appendix A: Example Calculation of Statistical Methods Performed in
BMP Analysis A-l
Table of Contents
-------
Coal Remining Statistical Support Document
List of Tables
Section 1.0 Introduction
Table 1.2a: Baseline Pollution Load Summary .• •. 1-8
Section 2.0 Characteristics of Coal Mine Drainage Discharges
Table 2.0a: Examples of High Alkalinity in Pennsylvania Mines
2-4
Section 3.0 Statistical Methodology for Establishing Baseline Conditions and
Setting Discharge Limits at Remining Sites
Section 4.0 Baseline Sampling Duration and Frequency
Table 4.1a: Statistical Triggers as Modified for Final Regulation: Percentage of
Discharges Declared to Exceed Baseline Level.
4-4
Section 5.0 Long-term Monitoring Case Studies
Table S.la: Long Term Acid Mine Drainage Datasets 5-2
Table S.lb: Comparison of Median Acidity and Iron Loads by Sample Period and Interval.. 5-5
Table 5.1c: Comparison of Median Acidity and Iron Loads by Baseline Sampling Year . . . 5-7
List of Tables
ill
-------
Remining Statistical Support Document
IV
List of Tables
-------
Coal Remining Statistical Support Document
List of Figures
Section 1.0 Introduction
Figure 1.2a: Algorithm for Analysis of Mine Drainage Discharge Data 1-7
Figure 1.2b: The Quick Trigger Process 1-10
Figure 1.2c: Dunkard Creek pH . .. • • • - 1-12
Figure 1.2d: Dunkard Creek Manganese 1-13
Section 2.0 Characteristics of Coal Mine Drainage Discharges
Figure 2.0a: Distribution of pH in Bituminous Mine Drainage . . 2-7
Figure 2.0b: Distribution of pH in Anthracite Mine Drainage 2-8
Figure 2.1a: Annual Variability in Streamflow at Dunkard Creek 2-10
Figure 2.1b: Sulfate Concentration vs. Streamflow at Dunkard Creek 2-12
Figure 2.2a: Acidity vs. Streamflow in Arnot Mine Discharge 2-14
Figure 2.2b: Inverse Loglinear Relationship between Acidity and Streamflow 2-15
Figure 2.2c: Streamflow and Acidity in SchuyMU County 2-16
Figure 2.2d: Streamflow and Acidity in Coal Refuse Pile 2-16
Figure 2.3a: Frequency Distribution of Sulfate at Dunkard Creek , - - 2-18
Figure 2.3b: Stem-and-leaf Diagram of pH (Arnot 003) , • • • 2-20
Figure 2.3c: Stem-and-leaf Diagram of Discharge • • 2-21
Figure 2.3d: Stem-and-leaf Diagram of Log Discharge - - 2-21
Figure 2.3e: Stem-and-leaf Diagram of Acidity 2-22
Figure 2.3f: Stem-and-leaf Diagram of Log Acidity - - • • 2-22
Section 3.0 Statistical Methodology for Establishing Baseline Conditions and
Setting Discharge Limits at Remining Sites
Figure 3.2a: Method 1 . • • 3'6
Figure3.2b: Method2 • • 3'7
Section 4.0 Baseline Sampling Duration and Frequency
Section 5.0 Long-term Monitoring Case Studies
List of Figures
-------
Coal Remitting Statistical Support Document
Figure S.la:
Figure S.lb:
Figure S.lc:
Figure S.ld:
Figure S.le:
Figure 5.1f:
Figure S.lg:
Figure S.lh:
Figure S.li:
Figure 5.1 j:
Figure 5.1k:
Figure 5.11:
Figure 5.1m:
Figure S.ln:
Figure 5.1o:
Figure S.lp:
Figure S.lq:
Figure S.lr:
Figure 5.1s:
Figure S.lt:
Figure 5.1u:
Figure S.lv:
Figure 5.1w:
Figure S.lx:
Figure S.ly:
Figure S.lz:
Figure S.laa:
Figure S.lab:
Figure 5.2a:
Figure 5.2b:
Figure 5.2c:
Figure 5.2d:
Figure 5.2e:
Figure 5.2f:
Figure 5.2g:
Figure 5.2h:
Figure 5.2i:
Figure 5.2j:
Figure 5.2k:
Figure 5.21:
Figure 5.2m:
Figure 5.2n:
Figure 5.2o:
Figure 5.2p:
Arnot-3 Acidity Loading (1980-1981)
Arnot-3 Flow Data Comparison
Arnot-3 Iron Load Data Comparison
Arnot-3 Acidity Load Data Comparison
Arnot-3 Monthly Flow Comparison
Arnot-3 Monthly Acidity Load Comparison
Arnot-4 Acidity Load Data Comparison
Arnot-4 Acidity Load (1980-1983)
Clarion Acidity Load Data Comparison
Clarion Iron Load Data Comparison •
Clarion Acidity Load (1982-1986)
Clarion Iron Load (1980-1983)
Ernest Acidity Load Data Comparison
Ernest Iron Load Data Comparison
Ernest Acidity Load Data Comparison
Ernest Acidity Load (1981-1985)
Fisher Monthly Acidity Load
Fisher Iron Load Data Comparison
Fisher Acidity Load Data (1982-1987) •
FisherIronLoadData(1982-1987) • • • >
Hamilton-8 Acidity Load Data Comparison • •
Hamilton-8 Iron Load Data Comparison . .
Hamilton-8 Acidity Load (1981-1985) .. .:
Hamilton-8 Iron Load (1981-1985)
Markson Acidity Load (1984-1986) - •
Markson Iron Load (1984-1986)
Markson Acidity Load Data Comparison
Markson Iron Load Data Comparison
Mine Discharge Map • •
Markson Time Plot (Flow, pH, Acidity, Iron, Manganese, Sulfate)
Markson Time Plot (Flow & Sulfate) ,
Markson Time Plot (Flow & Acidity)
Markson Time Plot (Flow & Iron) -
Markson Time Plot (How & Manganese)
Markson Time Plot (Flow 1994-1997)
Tracy Airway Time Plot (Flow, pH, Iron, Manganese, Sulfate) .. .
Tracy Airway (Flow 1994-1997)
Tracy Airway (Flow & Sulfate) •
Tracy Airway (Flow & Iron)
Tracy Airway (Flow, pH, Manganese)
Swatara Creek Flow and Sulfate Data •
Swatara Creek Flow and Suspended Solids Data
Swatara Creek Flow and Iron Data
Jeddo Tunnel Discharge and Wapwallopen Creek Flow Data
5-10
5-10
5-11
5-11
5-12
5-12
5-13
5-13
5-14
5-14
5-15
5-15
5-16
5-16
5-17
5-17
5-18
5-18
5-19
5-19
5-20
5-20
. 5-21
. 5-21
. 5-22
. 5-22
. 5-23
. 5-23
. 5-28
. 5-30
. 5-30
. 5-31
. 5-32
. 5-32
. 5-34
. 5-35
. 5-37
. 5-37
. 5-38
. 5-39
. 5-41
. 5-41
. 5-42
. 5-43
VI
List of Figures
-------
Coal Remining Statistical Support Document
Section 5.0 Long-term Monitoring Case Studies (cont.)
Figure5.2q: JeddoTunnelHowData • 5-44
Figure 5.2r: Precipitation Data From Hazleton, PA 5-44
Figure 5.3a: Fisher Mining MP1 (Flow, Iron, Manganese, Sulfate) 5-47
Figure 5.3b: Fisher Mining MP1 (Net Acidity) 5-48
Figure5.3c: Fisher Mining MP1 (AcidLoad) . • • 5-48
Figure 5.3d: Fisher Mining MP1 (IronLoad) 5-49
Figure 5.3e: Fisher Iron Load Box Plot 5'50
Figure 5.3f: Fisher Net Alkalinity BoxPlot 5-50
Figure 5.3g: McWreathDl (Flow, Net Acidity, Iron, Manganese, Sulfate) . 5-52
Figure 5.3h: McWreathDS (Flow & Net Acidity) -. : 5-53
Figure 5.3i: McWreathDS (Flow &Iron) 5-53
Figure 5.3J: McWreathD4 (How & Net Acidity) • • - 5-54
Figure 5.3k: Trees Mills Site Map '••••' 5-56
Figure 5.31: Trees Mills Drill Hole Data 5-57
Figure 5.3m: Trees Mills MP1 (How, Manganese, Aluminum, Net Acidity, Iron, Sulfate) . 5-59
Figure 5.3n: Trees Mills MP1 (Acid, Iron, Manganese Load) - 5-60
Figure 5.3o: Trees Mills MP2 (How, Iron, Manganese, Aluminum, Net Acidity, Sulfate) . 5-61
Figure 5.3p: Trees Mills MP2 (Acid, Iron, Manganese Load) . 5-64
Figure 5.3q: Trees Mills MP3 (How, Iron, Manganese, Aluminum, Net Acidity, Sulfate) . 5-65
Figure 5.3r: Trees Mills MP3 (Acid, Iron, Manganese Load) • • • 5-67
Figure 5.3s: Trees Mills MP6 (Acid, Iron, Manganese Load) 5-68
Figure 5.3t: Porter Run (Alkalinity: Upstream & Downstream) 5-69
Figure 5.3u: Beaver Run (Alkalinity: Upstream & Downstream) '. 5-70
List of Figures
vn
-------
Coal Remitting Statistical Support Document
viii
List of
-------
Coal Reminins Statistical Supvort Document
Section 1.0
Introduction
Acid mine drainage has been produced by coal mining operations in the Appalachian Coal
Region of the eastern United States and elsewhere for many years, resulting in extensive
surface-water and ground-water pollution. The Federal Clean Water Act (CWA), the Federal
Surface Mining Control and Reclamation Act (SMCRA), and associated state laws require coal
mine operators to take steps to prevent or control the production of acid mine drainage, and to
treat acid mine drainage from active and reclaimed surface mining operations so that
point-source discharges meet the applicable effluent limitations found at 40 CFR part 434.
Much of the acid mine drainage occurring in the Appalachian Coal Region is emanating from
abandoned surface and underground mines that were mined and abandoned prior to the
enactment of SMCRA and the CWA. According to the Appalachian Regional Commission
(1969), 78 percent of the acid mine drainage produced in northern Appalachia is associated with
inactive or abandoned mines. More recent U.S. Geological Survey reports (Wetzel and Hoffman,
1983, 1989) provide summaries of surface-water quality data and patterns of acid mine drainage
problems throughout the Appalachian Coal Basin. A set of companion reports (Hoffman and
Wetzel, 1983, 1989) contain similar information for the Interior Coal Province of the Eastern
Coal Region of the United States. Current EPA data document that the number one water quality
problem in Appalachia is drainage from abandoned coal mines, resulting in over 9,700 miles of
acid mine drainage polluted streams. A 1995 EPA Region III survey found that 5,100 miles of
streams in four Appalachian states are impacted by acid mine drainage, predominantly from
abandoned coal mines. Pennsylvania alone accounts for approximately 2,600 acid mine drainage
impacted stream miles.
The remaining coal reserves in these abandoned mine land areas frequently make them attractive
for the active mining industry; but traditionally, potential liability for the treatment of the
abandoned mine drainage established a disincentive to the permitting and remining of these
Introduction
1-1
-------
ilRe
rt Document
areas. If a pre-existing pollutional discharge of acid mine drainage was occurring within the area
or on an area hydrologicaUy connected to the permit area, mine operators often faced liability to
treat the discharge to best available technology economically achievable (BAT) effluent
standards (40 CFR part 434).
1.1 Remining Program History
In the 1980s, changes to the CWA (see 1987 Amendment to Section 301; the Rahall
Amendment) and state mining laws (e.g., 1984 Amendment to the Pennsylvania Surface Mining
Conservation and Reclamation Act (PA SMCRA)) provided incentives to mine operators to
remine areas with pre-existing pollutional discharges. Pursuant to these changes in state and
federal laws, the flow and water quality characteristics or "baseline pollution load" of these
pre-existing discharges must be documented prior to the commencement of the remining
operation. Under this program, the mine operator submits a surface mining permit application
including: (1) sufficient baseline pollution load data, and (2) a pollution abatement plan which
demonstrates how the remining operation proposes to eliminate or reduce the pre-existing
pollution. The regulatory authority completes a "best professional judgement (BPJ) analysis"
pursuant to Section 402(a) of the CWA as part of the permit review process. A BPJ-based
remining permit may be issued that requires the mine operator to treat the pre-existing discharges
only if the baseline pollution load has been exceeded, and then only treat the discharges to
baseline pollution load levels rather than to conventional BAT effluent standards. The
procedures to determine the level of treatment required to meet baseline is not standard, and is
dependent on various site-specific elements of the BPJ.
BPJ is defined as: "The highest quality technical opinion forming the basis for the terms and
conditions of the treatment level required after consideration of all reasonably available and
pertinent data. The treatment levels shall be established in accordance with Sections 301 and 402
of the Federal Water Pollution Control Act (33 USC §§1311 and 1342)." BPJ-determined
effluent limits must be based upon BAT or any more stringent limitation necessary to ensure the
1-2
Introduction
-------
Coal Remining Statistical Support Document
discharge does not violate state in-stream water quality standards. Theoretically, BPJ-determined
treatment levels can range from the pre-existing baseline level to the conventional BAT limits.
The BPJ analysis is a BAT analysis in miniature, specific to an individual mine site, rather than
an entire class of industrial wastewater discharges (i.e., surface coal mining). For a remining
permit, the analysis should consider the cost of treatment to conventional surface mining BAT
levels, as well as the cost of achieving pollution load reduction through the implementation of a
pollution abatement plan. The permit writer also should consider any unique factors pertaining
•to the proposed remining operations and any potential adverse or beneficial non-water quality
environmental impacts.
The Rahall Amendment to the Federal Clean Water Act in 1987 provided a foundation for the
development of effective remining programs in many coal mining states. Between 1984 and
1988, EPA and the Commonwealth of Pennsylvania Department of Environmental Protection
(PADEP) cooperated in a remining project. The purpose of that project was to develop an
effective remining program pursuant to the 1984 Amendments to PA SMCRA, that would not be
in conflict with the provisions of the Federal Clean Water Act and the associated 40 CFR part
434 regulations. Pennsylvania promulgated remining regulations on June 29, 1985, that were
approved by the U.S. Office of Surface Mining Reclamation and Enforcement (OSMRE) on
February 19, 1986. ' ,
The work products of the PADEP/EPA cooperative study included preliminary treatment costing
and remining costing studies in 1986 (prepared by Kohlmann Ruggiero Engineers (KRE), and
Phelps and Thomas of the Pennsylvania State University), the development of the REMINE
computer software package and Users Manual in 1987, and associated technical reports in 1987-
and 1988. Included in these technical reports was the final treatment costing study (Kohlmann
Ruggiero Engineers, P.C., 1988) and a series of eight water quality statistical reports prepared by
Dr. I C. Griffiths of the Pennsylvania State University. Since these unpublished statistical
analyses of mine drainage datasets are relevant to baseline pollution load statistics, they are
Introduction
1-3
-------
Coal Re
Statistical Suvvort Document
presented in an abridged form in a companion volume to this report, prepared by US EPA (2001;
EPA-821-B-01-014).
1.2 Pennsylvania DEP Renaming Permitting Procedures
Since 1985, PADEP has issued approximately 300 remining permits, with a 98 percent success
rate. A successful remining site is one that has been mined without incurring treatment liability
as the result of exceeding the baseline pollution load of the pre-existing discharges. Data from
112 of these sites that have been completely reclaimed have been used by EPA in evaluating Best
Management Practice (BMP) performance (see EPA Coal Remining BMP Guidance Manual).
The elements of establishing baseline pollution loads and measuring compliance that are
provided in the Pennsylvania program guidance on the BPJ process are summarized below.
1.2.1 Pre-existing Discharges
Various relationships exist between the permit boundaries of surface coal mine sites and the
location of pre-existing pollutional discharges. The simplest relationship exists where a single
abandoned mine drainage discharge point is located within, or closely adjacent to, the proposed
surface mine permit boundaries, and the proposed mine is the only active mining operation. A
more complex relationship occurs where numerous pre-existing discharges from the same coal
seam, or frommultiple coal seams, are located within the proposed surface mine permit area or
are hydrologically connected to that permit area. In addition, there are situations where more
than one active mining operation is hydrologically connected to the same pre-existing discharge.
The PADEP program includes considerations for: (a) monitoring and baseline data collection of
single and multiple discharges, (b) establishing baseline pollution load of single or multiple
discharges through statistical methods, and (c) determining compliance with BPJ determined
effluent limits for the pre-existing discharges through statistical methods and permit conditions
(for individual operators and multiple operators).
1-4
Introduction
-------
Coal Remining Statistical Support Document
La many cases, pre-existing pollutional discharges may occur in the form of numerous discharge
points, an of which emanate from a hydrologically discrete ground-water flow system Ground-
water flow paths may change during and following remining such that new discharge points
appear, former discharge points disappear, and/or the distribution of flow rates between
discharges changes. Where this situation is likely to occur, it is usually advantageous to
designate hydrologic units. Each unit must be a hydrologically discrete area such that ground
water from one hydrologic unit does not flow to a different hydrologic unit. Hydrologic unit
boundaries must be determined for situations where two or more discharges are to be aggregated
for load calculations.
Discharges may be combined either naturally or by man-made controls to a single monitoring
point, provided that the combination of discharges does not affect the pollution load
measurement and that discharges from different hydrologic units are not combined. It is usually
desirable (in terms of cost to the operator, permit writing, and compliance monitoring) for the
permit applicant to minimize the number of monitoring points needed.
The permit applicant must perform a baseline pollution load statistical determination for each
monitoring point. Where multi-discharge hydrologic units are defined, the baseline statistics
should be calculated for the aggregate pollution load from the hydrologic unit. That is, loads are
summed for all the discharges in the hydrologic unit on a given date. Baseline pollution load
determination of a hydrologic unit requires sampling and analysis of each discharge on the same
date using an equal number of samples from each discharge. The baseline pollution load is then
reported as the combined pollution load from the hydrologic unit.
1.2.2 Baseline Pollution Load - Determination and Compliance Monitoring
The process of establishing a realistic baseline pollution load for a mine site requires knowledge
of hydrology and statistics. An adequate number of samples must be collected at sufficient time
intervals to represent seasonal variations throughout the water year (October through September).
The statistical components of establishing baseline pollution load include characterizing the
Introduction
-------
Coal Reminine Statistical Support Document
patterns of variation and measuring central tendency, so that any mining-induced changes in
pollution load can be distinguished from seasonal and random variations. During active mining
and post-mining, individual sampling events and the statistical summary of data collected over
successive water years are compared to the pre-mining baseline statistics.
An algorithm for the statistical analysis of mine drainage discharge data was developed by
Griffiths (1987) for use in the Pennsylvania program and elsewhere (Figure 1.2a). The algorithm
included a simple quality control approach using the exploratory data analysis methods
developed by Tukey (1977), and used bivariate statistical methods and time series analyses,
where appropriate (e.g., research purposes documented in eight statistical reports by Griffiths,
1987 and 1988). In practice, almost all of the remining permits issued under the Pennsylvania
program have used the Baseline Pollution Load Statistical Results Summary presented in Table
1.2a The five statistical calculations (range, median, quartiles, 95 percent confidence interval
about the median, and 95 percent tolerance interval) are based upon Tukey's exploratory data
analysis methods and order statistics. Alternative statistical calculations may be used in place of
the calculations identified on Table 1.2a, provided that the permit applicant demonstrates that the
alternative calculations are statistically valid and applicable. For example, the mean and variance
may be used if the data are normally distributed. The REMINE computer software package
developed by EPA, PADEP and the Pennsylvania State University was integrated with the
MIMTAB statistical software package1, which includes statistical and graphical methods to
perform all of the steps in the algorithm presented in Figure 1.2a.
1MINITAB is a commercial software package fromMinitab, Inc., © 1986, 3081 Enterprise Drive,
State College, PA 16801
; • " . Introduction
1-6
-------
Coal Reminine Statistical Supvort Document
Figure 1.2a: Algorithm for Analysis of Mine Drainage Discharge Data
Raw
Data
1. Initial examination.
2. Adjust missing values to *.
3. Examine for unusual values.
4. Graph discharge or Log discharge vs days.
5. Check for unequal intervals, missing data, and extremes
MINITAB
6. Univariate statistics.
DESCRIBE - Summary Statistics.
HISTOGRAM: symmetry?
8. Examine and Edit Outliers
9. Bivariate analysis.
Var. (x) vs. pH
Var (x) vs. Disch. or Log Disch.
Association r2
Cross-correlation.
Regression if required.
10. Time series plots (TSPLOT) for each variable.
1. Search for missing values.
2. Periodicity?
3. Outliers
4. Quality Control Graphs. .
11. Box - Jenkins Time Series Analysis
1. Identification: Acf, Pacf, Acf, Pacf
2. Estimation of Model parameters.
3. Residuals to check for outliers.
4. Forecasts (wher required).
12. Sampling bv Simulation
1. Choose samples (18 for example) according to recommended procedure.
2 Test by quality control graphs.
3. TSPLOTs using mean, median and various multiples of standard deviation.
13. Adopt procedure for routine analysis in the field.
Introduction
-------
Coal Remining Statistical Support Document
While the permit applicant is responsible for submitting the baseline pollution load data and
statistical summary, the permit reviewer must check the calculations to ensure that the results are
correct. In addition, the reviewer must examine the distribution of the data to determine whether
a logarithmic transformation is appropriate. If logarithmic transformation results in a more
normal distribution curve, log-transformed data should be used in determining the baseline.
Each discharge point or hydrologic unit will have a baseline pollution load summary.
Compliance of discharge points is determined by comparing monthly sample analysis results to
the determined baseline pollution load. A discharge point is considered to be in compliance as
long as the sample analysis indicates that the pollution load does not exceed either the 95 percent
tolerance limit (item 4, Table 1.2a) or the 95 percent confidence interval about the median (item
5, Table 1.2a). The confidence intervals around the median are calculated using the equation
noted in Table 1.2a, and taken fromMcGill, Tukey, and Larsen (1978).
Table 1.2a: Baseline Pollution Load Summary
Mine ID: Mine Name:.
Hydrologic Unit ID:.
f of Satnptest * s'
Statistical Results ' ;
1. Range Low:
High:
2. Median
3. Quartiles Low:
High:
4. Approximate 95% Low:
tolerance limits High:
5. 95% Confidence Int Low:
at> out median* High:
Flow
<££«*)' '
Loading In Pounds Per Bay
Acidity
Iron
Manganese
Aluminum
Sulfates
*Notc: Confidence intervals about median = M+/- 1.58[1.25R/(1.35xSQR(N)] where:
M = median, R = range between quartiles, and SQR(N) = the square root of the number of samples (McGill
etal., 1978).
1-8
Introduction
-------
Coal Reminins Statistical Support Document
An excursion (i.e., apparent violation) of the baseline pollution load occurs when the result of a
sample analysis exceeds the 95 percent tolerance limit, or when a median of sample results
obtained over a new water year is outside the bounds of the 95 percent confidence interval about
the original baseline median. An excursion of the baseline pollution load above the 95 percent
tolerance limit is known as a "quick trigger" violation. A more subtle and long-term trend of the
pollution load above the 95 percent confidence interval about the median is known as a "subtle
trigger" violation.
The 95 percent tolerance limit is determined by ranking the data in order of magnitude and
dividing the data into 32 increments. The 95 percent tolerance limits correspond to the lowest
and highest of the 32 increments. For datasets containing 16 or fewer samples, the approximate
95 percent tolerance limits correspond to the smallest and largest sample values. The upper 95
percent tolerance limit is the "quick trigger" or "critical" value mechanism for monthly
monitoring data. If two consecutive monthly samples exceed the upper 95 percent tolerance
limit, weekly monitoring is initiated. The quick trigger values are provided in the Surface Mine
Permit. Permit conditions specify quick trigger monitoring and compliance steps shown in
Figure 1.2b.
Determination of long-term compliance with the subtle trigger typically involves comparison of
the pollution loading data for successive water years to the 95 percent confidence limit about the
median for the baseline. The 95 percent confidence limits are also based on baseline data (Table
1.2a) and given in the Surface Mine Permit. The 95 percent confidence interval is defined as the
range of values around the median in which the true population median occurs with a 95 percent
probability. This value is used to determine if statistically significant changes in median
pollution loads have occurred between the baseline monitoring period and water years during
mining and postmining. Permit conditions specify the process that will be used to determine
compliance with the subtle trigger.
Introduction
-------
Coal Re
Statistical Support Document
Figure 1.2b: The Quick-Trigger Process
Monthly Sample
Is 95% tolerance limit violated?
Yes
I
Is this the second
consecutive month
that the 95%
tolerance limit
exceeded?
Yes
Commence weekly
sampling.
Do 4 weekly samples
exceed 95% tolerance
limit?
Yes
Notify Dept. and commence
treatment unless determined
that increase in pollution load
is not attributed to the mining
operation.
Are 4 consecutive weekly
samples in compliance?
No
Continue Treatment.
-No-
No
Continue monthly
sampling.
-No-
Yes
1-10
Introduction
-------
Coal Reminins Statistical Support Document
Box plots can be used to easily compare the baseline pollution load (or concentration) for
different analytes to successive water-year datasets. Box plots can be constructed to show the
median value, the 95 percent confidence limits around the median and the upper and lower
quartiles and range of data. The length of the box corresponds to the interquartile range (IQR)
equal to the 75th percentile minus the 25th percentile. Therefore, 50 percent of the data will fall
within the range given by the length of the box. The upper whisker (the t-shaped line above the
upper end of the boxes) extends to the largest value less than or equal to the 75th percentile plus
1.5 times the IQR. Likewise, the lower whisker extends to the smallest value greater than or
equal to the 25th percentUe minus 1.5 times the IQR. Any value that is beyond the whiskers is
known as an extreme value. Extreme values less than 1.5 IQRs away from the nearer whisker (or
equivalently, less than 3 IQRs away from the edge of the box) are represented by an open circle.
Extreme values beyond 1.5 IQRs away from the nearer whisker are represented by an "x".
Figures 1.2c and 1.2d are examples of box plots of water quality data from the Dunkard Creek in
Greene County, Pennsylvania (this acid mine drainage impacted stream segment is featured in
several other figures in Section 2.0). Figure 1.2c shows variations in the range, median and
quartiles of pH distributions for three time periods corresponding to significant changes in
mining regulation and acid mine drainage (AMD) control and abatement in Pennsylvania. The
box plot of pH data from 1950 to 1965 represents all available data (N=54) at this monitoring
point prior to the Pennsylvania Clean Streams Law requiring that active mines treat acid mine
drainage. This law went into effect in 1966. The box plot of pH data from 1983 to 1997
(N=175) represents the time period following the approval of Pennsylvania for primacy to
regulate the Federal SMCRA of 1977. Primacy provided for significant increases in staff and
resources for permitting, inspection and enforcement of active mine sites, and funds to reclaim
abandoned mine sites with AMD problems. The median pH of 6.9 for the data (N=l 12) for the
intermediate time period (1966-1982) is significantly different than the median pH of 3.95 for the
time period prior to the 1966 AMD treatment requirement. Figure 1 .2d shows box plots of
manganese concentrations for the same monitoring point and same time periods as Figure 1.2c.
The median and interquartile range for the 1966-1982 data (N=107) is significantly less than that
of the 1950-1965 data (N = 14); and the range of the manganese concentrations in the 1983-1997
- ; - :
Introduction
-------
Coal Remining Statistical Support Document
data (N = 173) is less than half of the range for the two previous time periods. Additional
examples of box plots and explanations of their origin and development are contained in Tukey
(1977), McGffl, Tukey, and Larsen (1978), Veleman and Hoaglin (1981) and Helsel (1989).
Figure 1.2c: Dunkard Creek pH
9.CH
8.1 E
7.2E
6.3=
5.4=
4.5E
3.6E
2.7E
1-8E
0.9-
0.0
s V X^ """*'
b^]
x
X
1950-1965
1966-1982
1983-1997
1-12
Introduction
-------
Coal Remining Statistical Support Document
Figure 1.2d: Dunkard Creek Manganese (mg/L)
10000-On
9000.0E
8000.0=.
7000.0=
6000.0E
5000.0=
4000.0=
3000.0=
2000.0E
1000.0=
0.0=
X
I*. :
I '- ',
M> .v-
' " "/
X
o X
X'
s e
~T~ ~T~
fj / j^j I ^^J^—-!^
1950-1965 1966-1982 1983^1997
Monitoring and compliance inspections are conducted periodically (i.e., quarterly). Reviews of
the monthly monitoring data for the purpose of comparing current data to the baseline data,
checking for subtle and quick trigger violations, and noting any data trends, are conducted
annually.
1.3 IMCC Evaluation of State Remining Programs
The Interstate Mining Compact Commission (IMCC) is a multi-state governmental organization
representing the natural resource and environmental protection interests of its member states,
including extensive interaction with EPA, OSMRE, and other federal agencies. The IMCC
organized a national remining task force in 1996, with representation from EPA, OSMRE and
member states, in order to develop and promote various remining incentives that would
accomplish significant abandoned mine land reclamation and associated water quality benefits.
A product of the EVICC Remining Task Force is a discussion paper on water quality issues
Introduction
1-13
-------
Coal Re
Statistical Support Document
related to coal renaming, for which EPA, OSMRE, and IMCC jointly solicited comments in
February 1998 from a wide range of environmental, industry, and government agency
commentors/respondents.
In a related effort that is part of a cooperative project with EPA and OSMRE, the IMCC solicited
and compiled responses from 20 states on their renaming program experiences. This compilation
of responses provides extensive information on the number of Rahall-type remining permits
issued in the states, the contents of these permits (including the availability of baseline pollution
load analyses and data), and the types and effectiveness of BMPs employed during remining
operations. A summary of these responses is included as Appendix C of EPA's Coal Remining
BMP Guidance Manual. IMCC also submitted 61 data packages to EPA from 6 member states.
These data packages include pre-, during, and post-mining water quality data, BMP
implementation plans, raining operation plans, geology and overburden analysis data,
abandoned mine land conditions, and topographic maps.
Based upon review of the IMCC solicitation responses, 61 data packages, and discussions with
state agency representatives, it is evident that baseline pollution load data requirements vary
widely from state to state. Pennsylvania, Virginia and some other states generally require a
rninimumof 12 monthly samples of pre-existing pollutional discharges to calculate the baseline
pollution load and characterize seasonal variations throughout the water year. One state water
quality agency has advocated the use of 52 weekly samples to characterize baseline pollution
load, which may have been a disincentive to remining. That state has only been able to issue a
handful of Rahall remining permits. At the other end of the scale, another state has developed a
draft sampling protocol to establish baseline pollution load with only 6 monthly samples, similar
to the background sampling requirement for determining "probable hydrologic consequences" in
most state surface mining permits. The draft protocol divides the water year into high-flow,
low-flow, and intermediate-flow periods, and contains requirements for sampling each of these
periods and considering transition periods and other hydrologic factors.
1-14
Introduction
-------
Coal Remining Statistical Support Document
The establishment of the baseline pollution load is largely a statistical exercise because many
pre-existing discharges are known to be highly variable in flow and/or water quality. The use of
statistics is necessary to quantify these variations and to summarize the behavior of the
discharges, which may be related to seasonal variations or other hydrologic factors. Statistical
analysis of the baseline pollution load data also enables a distinction to be made, after remining
has commenced, between normal seasonal variations and mining-induced changes in pollution
load that will require initiation of treatment.
The baseline must consist of an adequate number of samples of sufficient intervals and duration,
in order to provide adequate protection for both the industry and the regulatory authority against
false triggers. The greater the number of samples and range of hydrologic conditions represented
by the baseline pollution load determination, the greater the likelihood that the baseline pollution
load determination is statistically and hydrologically sound. In attempting to establish the
baseline pollution load with a relatively small number of samples, there is an inherent risk of
under representation. In establishing national standards or guidelines for baseline pollution load,
careful consideration must be given to determining the optimum number of samples, the
associated time intervals, and sampling duration in order to achieve statistical and hydrologic
credibility, without being overly burdensome, costly, or impractical.
References
Appalachian Regional Commission, 1969. Acid Mine Drainage in Appalachia. Appalachian
Regional Commission: Washington, DC.
Griffiths, J.C., (No date). Report No. 1: Reconnaissance into the use of MINITAB using
Hamilton 01, 08 Data U.S. Environmental Protection Agency, Office of Water:
Washington, DC.
Griffiths, J.C., 1987. Report No. 2: Time Series Analysis of Data fromHamilton 08. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987. Report No. 3: Analysis of Data from Arnot 001, 003, 004. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Introduction
1-15
-------
ilSu
ent
Griffiths, J.C., 1987. Report No. 4: Analysis of Data from the Clarion Site. U.S. Environmental
Protection Agency, Office of Water: Washington, DC.
Griffiths J.C., 1987. Report No. 5: Analysis of Data fromErnest Refuse Pile, Indiana County,
Pennsylvania. U.S. Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987. Report No. 6: Analysis of Data from the Fisher Deep Mine. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1988. Report No. 7: Analysis of Data from the Markson Site. U.S. Environmental
Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1988. Report No. 8: Synopsis of Reports 1 - 7. U.S. Environmental Protection
Agency, Office of Water: Washington, DC.
Helsel, D., 1989. Boxplots: A Graphical Method for Data Analysis. US Geological Survey,
' Bureau of Systems Analysis Technical Memorandum No. 89.01: Reston, VA.
Hoffman S A and K.L. Wetzel, 1983. Summary of Surface Water Quality Data, Interior Coal
Province, Eastern Region, October 1978 to September 1982. US Geological Survey
Open File Report 83-941: Harrisburg, PA.
Hoffman, S.A. and K.L. Wetzel, 1989. Distribution of Water Quality Indicators of Acid Mine
Drainage in Streams of the Interior Coal Province, Eastern Coal Region of the United
States. US Geological Survey Water Resources Investigations Report 89-4043:
Harrisburg, PA.
KohlmannRuggiero Engineers, P.C., 1988. Development of a Wastewater Treatment Cost
Analysis Module for the Coal Remining BPJ Computer Software Package (Preliminary
Draft). U.S. Environmental Protection Agency: Washington, DC.
McGill, R., J-W. Tukey, and W. A. Larsen, 1978. Variations of Box Plots. The American
Statistician, Vol. 32, No. 1, pp. 12 - 16.
Phelps, L.B. andM.A. Thomas, 1986. Assessment of Technological and Economical Feasibility
' of Surface Mine Reclamation, Stage One Research. Pennsylvania Department of
Environmental Resources Contract No. 898621: Harrisburg, PA.
Tukey, J.W., 1977. Exploratory Data Analysis. Addison Wesley Publishing Company: Reading,
'MA.'
U.S. EPA, 1986. Preliminary Engineering Cost Manual for Development of BPJ Analyses for
Coal Remining, Four Case Histories in Pennsylvania. Prepared for U.S. Environmental
1-16
-------
Coal Reminins Statistical Support Document
Protection Agency, Office of Water, Industrial Technology Division Energy and Mining
Branch, Washington D.C. by Kohlmann Ruggiero Engineers, P.C.
U.S. EPA, 2001. Statistical Analysis of Abandoned Mine Drainage in the Establishment of the
Baseline Pollution Load for Coal Remining Permits. Prepared for US Environmental
Protection Agency, Office of Water by Pennsylvania Department of Environmental
Protection and DynCorp I&ET. EPA-821-B-01-014.
VeUeman, P.P. and D.C. Hoaglin, 1981. Applications, Basics, and Computing of Exploratory
Data Analysis. Duxbury Press: Boston, MA.
Wetzel K.L. and S.A. Hoffman, 1983. Summary of Surface Water Quality Data, Eastern Coal
Province, October 1978 to September 1982. US Geological Survey Open File Report
83-940: Harrisburg, PA.
Wetzel, K.L. and S.A. Hoffman, 1989. Distribution of Water Quality Characteristics that May
'indicate the Presence of Acid Mine Drainage in the Eastern Coal Province of the United
States. US Geological Survey Hydrologic Investigations Atlas HA-705: Harrisburg, PA.
Introduction
-------
Coal Remining Statistical Support Document
1-18
Introduction
-------
Coal Remining Statistical Support Document
Section 2.0 Characteristics of Coal Mine Drainage Discharges
Acid mine drainage is generated when sulfide minerals, principally pyrite (FeS^, are exposed to
increased amounts of air and water in the oxidizing and non-alkaline environment of a surface or
underground mine. The sulfide minerals typically occur in coal beds as well as in strata
overlying and underlying the coal. Weathering and aqueous dissolution of the sulfide mineral
oxidation products, including dissociated sulfuric acid and metals (e.g., Fe, Mn, Al), produces
surface and groundwater degradation. Explanations of the chemical reactions by which acid
mine drainage is produced from pyrite and other iron sulfide minerals are found in Singer and
Stumm (1970), Kleinmann et al. (1981), Lovell (1983), Evangelou (1995), and Rose and
Cravotta (1998). Additional references presenting data and discussion of factors related to pyrite
oxidation rates include Emrich (1996), McKibben and Barnes (1986), Moses and Herman
(1991), Watzlaf (1992), and Rimstidt and Newcomb (1993). These reactions also are presented
and discussed in. Section 2.0 of EPA's Coal Remining Best Management Practices Guidance
Manual.
While pyrite is the most commonly reported producer of AMD, other mineral species including
the sulfide mineral marcasite (FeSj), and sulfate minerals jarosite (KFe3(SO4)2(OH)6) and alunite
(KA13(SO4)2(OH)6), are capable of producing acidic drainage at surface and underground mine
sites. Sulfate minerals are generally secondary weathering products of pyrite oxidation.
Nordstrom (1982) shows the sequence by which these minerals can form from pyrite. Many
secondary sulfate minerals have been identified that are typically very soluble and transient in the
humid eastern United States. These minerals form during dry periods and are flushed into the
ground-water system during precipitation events. The sulfate minerals that contain iron,
aluminum, or manganese are essentially stored acidity and will produce acid when dissolved in
water. Sulfate minerals such as melanterite, pickeringite, and halotrichite occur commonly in
Appalachian Basin coal-bearing rocks. Additional information about these sulfate minerals is
found in Cravotta (1994), Lovell (1983), Rose and Cravotta (1998), and Brady et al. (1998).
Characteristics of Coal Mine Drainage Discharges 2-1
-------
Coal Remitting Statistical Support Document
Acid mine drainage is the most frequently described and most environmentally damaging type of
coal mine drainage. However, other damaging types can occur due, principally, to geologic
factors and influences from mining and reclamation practices. According to Rose and Cravotta
(1998):
"Coal Mine drainage ranges widely in composition, from acidic to alkaline, typically with
elevated concentrations of sulfate (SO4), iron (Fe), manganese (Mn), and aluminum (Al)
as well as common elements such as calcium, sodium, potassium, and magnesium. The
pH is most commonly either in the ranges 3 to 4.5 or 6 to 7, with fewer intermediate or
extreme values... Acidic mine drainage (AMD) is formed by the oxidation of pyrite to
release dissolved Fe2+, SO42' and H+, foUowed by the further oxidation of the Fe2+ to Fe3+
and the precipitation of the iron as a hydroxide ("yellow boy") or similar substance,
producing more H+.. .In contrast, neutral or alkaline mine drainage (NAMD) has
alkalinity that equals or exceeds acidity but can still have elevated concentrations of SO4,
Fe, Mn and other solutes. NAMD can originate as AMD that has been neutralized by
reaction with carbonate minerals, such as calcite and dolomite, or can form from rock that
contains little pyrite. Dissolution of carbonate minerals produces alkalinity, which
promotes the removal of Fe, Al and other metal ions from solution, and neutralizes
acidity. However, neutralization of AMD does not usually affect concentrations of SO4."
The rate of AMD production and the concentrations of acidity, sulfate, iron, and other water
quality parameters in mine drainage are dependent upon numerous physical, chemical, and
biological factors. According to Rose and Cravotta (1998):
"Many factors control the rate and extent of AMD formation in surface coal mines. More
abundant pyrite in the overburden tends to increase the acidity of drainage, as does
decreasing grain size of the pyrite. Iron-oxidizing bacteria and low pH values speed up
the acid-forming reaction. Rates of acid formation tend to be slower if limestone or other
neutralizers are present. Access of air containing the oxygen needed for pyrite oxidation
is commonly the limiting factor in rate of acid generation. Both access of air and
exposure of pyrite surfaces are promoted by breaking the pyrite-bearing rock. The
oxygen can gain access either by molecular diffusion through the air-filled pore space in
the spoil, or by flow of air which is driven through the pore space by temperature or
pressure gradients..."
Numerous studies have evaluated the distribution of total sulfur contents and pyritic sulfur
contents within coals and overburden strata. In some of these studies, investigations have
examined the significance of pyrite morphology, especially the framboidal form with high
surface area.
2-2 Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
AMD discharges in Pennsylvania range in flow from seeps of less than 1 gallon per minute
(gpm) to abandoned underground mine outfalls such as the Jeddo Tunnel near Hazleton, PA
where a flow greater than 150,000 gpm (40,000 gpm average flow) has been measured. Table
2.0a presents typical and extreme examples of acidity, alkalinity, and related water quality
parameters in coal mine drainage (from surface mines, underground mines, and coal refuse piles)
and nearby well and spring samples. These water samples were compiled from data in
Hornberger and Brady (1998) and Brady et al. (1998) to illustrate mine drainage quality
variations in Pennsylvania. Similar variations in mine drainage quality exist in West Virginia,
Ohio, and other states in the Appalachian Basin. Acidity and alkalinity concentrations greater
than 100 mg/L are shown in bold in Table 2.0a.
Some of the most extreme concentrations of acidity, iron, and sulfate in Pennsylvania coal mine
drainage, have been found at the Leechburg Mine refuse site in Armstrong County, and at surface
mine sites in Centre, Clinton, Clarion, and Fayette Counties (Table 2.0a). Acidity concentrations
of seeps from Lower Kenning Coal refuse at the Leechburg site exceed 16,000 mg/L, while the
sulfate concentration of one sample exceeds 18,000 mg/L. Schueck et al. (1996) reported on
AMD abatement studies conducted at a backfilled surface mine site in Clinton County. A
monitoring well that penetrated a pod of buried coal refuse produced a maximum acidity
concentration of 23,900 mg/L prior to the implementation of the abatement measures.
Characteristics of Coal Mine Drainage Discharges
2-3
-------
Coal Remitting Statistical Support Document
Table 2.0a: High Alkalinity Examples in Pennsylvania Mine Discharges
Site Name
Willow Tree
Susan Ann
Bertovich
Smith
Brown
Trees MiUs
State Line
Cover Hin
Fruitful!
LaurelHffl
#1
Morrison
Stuart
Clinger
Leechburg
*Eran
Swiscambria
AIbert#l
Snyder#l
Lawrence
Graff Mine
** Old 40
Orcutt
Stratigraphic
Waynesburg
Waynesburg
Redstone
. Upper &
Lower
Bakerstown
Lower
Bakerstown
Brush Creek
Upper &
Lower
U. Ereept to
U. Kittng.
Upper
Upper
Middle
Lower
Lower
Kittanning
Lower
Lower
Lower
Lower
L. Kittng. &
Clarion
Clarion
pH
7.8
3.3
31
7.7
74
75
8.1
3.6
68
7.8
8.1
7.0
2.8
6.8
2.4
2.2
4.2
3.1
6.9
2.2
7.8
27
2.2
3.9
Alkalinity
379.0
0.0
0.0
246.0
626.0
0.0
210.0
0.0
189.4
238.0
484.0
308.0
0.0
190.0
0.0
0.0
5.0
0.0
114.0
0.0
274.0
0.0
0.0
0.0
Acidity
0.0
1500.0
378.0
0.0
0.0
3616.0
0.0
168.1
no data
0.0
0.0
0.0
1290.0
0.0
16718.0
23900.0
88.0
1335.0
0.0
5938.0
no data
9732.0
10000.0
5179.6
Fe
0.12
324.40
74.80
M.47
1.65
190.40
<0.30
0.83
0.21
0.01
0.97
0.63
56.70
<0.30
> 300.0
5690.00
0.09
186.00
1.10
2060.00
0.0
1959.81
3200.00
2848.0
Mn
0.04
89.70
9.14
0.27
1.05
13.50
1.37
14.60
0.40
0.01
1.98
3.49
49.20
1.28
19.30
79.00
24.20
111.00
3.14
73.00
1.13
205.30
260.0
349.0
SO,
me/L
165.0
2616.0
1098.0
122.0
1440.0
1497.8
416.0
787.0
68.2
458.0
590.0
327.0
1467.0
184.0
18328.0
25110.0
1070.0
3288.0
264.0
3600.0
1645.0
4698.0
14000.
11120.
Flow
£pm
1.0
<1.0
2.0
0.0
no data.
13.0
no data
1.8
4.0
60.0
no data
<1.0
no data
0.0
2.0
0.0
no data
55.0
0.0
0.0
10.0
35.0
0.0
0.0
Comments
ire-mining
eep near sealed deep
mine entry
Deep mine discharge
it water at lowwall
ump
pring near cropBne
Deep mine discharge
'ost-mining seep from
>ackfiUed spoil
Discharge from
bandoned pit below
ite
Logan spring
Deep mine discharge
?oe of spoil seep
Seep near collection
ditch
Seep, sandstone
overburden
Pit water
Seep from coal refuse
disposal area
Monitoring well in
backfilled spoil
Seep, freshwater
paleoenvironment
Spoil discharge,
brackish paleoenviron.
Pit water, marine
paleoenvironment
Pit water, sandstone
overburden
Seep above road
Spoil discharge
Monitoring well in
backfilled spoil
Spoil water from
piezometer
2-4
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Site Name
Cousins
Zacherl
Horseshoe
Wadesville
Stratigraphic
Interval
Clarion
Clarion
Mercer
Llewellyn
pH
7.6
2.3
2.3
6.7
Alkalinity
me/L
130.0
0.0
0.0
414.0
Acidity
me/L
0.0
9870.0
1835.0
0.0
Fe
me/L
7.15
2860.00
194.00
3.61
Mn
me/L
0.30
136.60
27.00
3.37
SO4
1D2/L
71.0
7600.0
2510.0
1038.0
Flow
apm
0.0
no data
700.0
no data
Comments
Pit water, glacial till
influence
Toe of spoil discharge
Abandoned deep none
discharge
Minepool, Anthracite
Region
Note: Extreme values (>100 mg/L) of alkalinity and acidity are highlighted for emphasis
* data from Schuek et aL (1996)
** data from DugasetaL (1993)
Since the alkalinity-production process has a dramatically different set of controls, the resultant
maximum alkalinity concentrations found in mine environments are typically one or two orders
of magnitude less than the maximum acidity concentrations. Examples of relatively high
alkalinity concentration in mine drainage, ground water, and surface water associated with
Pennsylvania bituminous and anthracite coal mines are presented in Table 2.0a. The highest
natural alkalinity concentration found in PA DEP mining permit file data (and reported in Table
2.0a) is 626 mg/L in a spring located near the cropline of the Redstone Coal in Fayette County.
Thick sequences of carbonate strata, including the Redstone Limestone and the Fishpot
Limestone underlie and overlie the Redstone Coal.
Carbonate minerals (e.g., calcite and dolomite) play an extremely important role in determining
post-mining water chemistry. They neutralize acidic water created by pyrite oxidation, and there
is evidence that they also inhibit pyrite oxidation (Hornberger et aL, 1981; Williams et aL, 1982;
Perry and Brady, 1995). Brady et aL (1994) concluded that the presence of as little as 1 to 3
percent carbonate (on a mass-weighted basis) at a mine site can determine whether that mine
produces alkaline or acid water. Although pyrite is clearly necessary to form acid mine drainage,
the relationship between the amount of pyrite present and water-quality parameters (e.g., acidity)
was only evident where carbonates were absent (Brady et al., 1994).
The paleoclimatic and paleoenvironmental influences on rock chemistry in the northern
Appalachians resulted in the formation of coal overburden with greatly variable sulfur content (0
percent to >15 percent S) and calcareous mineral content (0 percent to >90 percent CaCO3) as
Characteristics of Coal Mine Drainage Discharges
2-5
-------
Coal Remining Statistical Support Document
shown on figures of overburden drill hole data in Brady et al. (1998). The wide variations in
rock chemistry contribute to the wide variations in water quality associated with surface coal
mines. Figures 2.0a and 2.0b show the frequency distributions (i.e., range) of pH in mine
discharges in the bituminous and anthracite coal regions of Pennsylvania. The origin and
significance of this bimodal frequency distribution for mine drainage discharges are described in
Brady et al. (1997, 1998) and Rose and Cravotta (1998). Brady et al. (1997) explained that
although pyrite and carbonate minerals only comprise a few percent (or less) of the rock
associated with coal, these acid-forming and acid-neutralizing minerals, respectively, are highly
reactive and are mainly responsible for the bimodal distribution. Depending on the relative
abundance of carbonates and pyrite, and the relative weathering rates, the pH will be driven
toward one mode or the other.
Variations in the chemical composition of mine drainage discharges are principally related to
geologic and hydrologic factors. The hydrologic factors that cause individual mine drainage
discharges to vary in flow and concentrations of acidity, alkalinity, sulfates and metals (e.g., Fe,
Mn, Al) throughout the water year are discussed in the following sections of this chapter.
2-6
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Figure 2.0a: Distribution of pH in Bituminous Mine Drainage
40 -r
30 1
0
Dunkard
Monongahda
Cbnemaugh
Upper Allegheny
'I ~ 1 LwverAllegjieriy
ill 11IIII Pottsville
1 2 34
5
PH
6 7 8 9
Characteristics of Coal Mine Drainage Discharges
2-7
-------
Coal Remining Statistical Support Document
Figure 2.0b: Distribution of pH in Anthracite Mine Drainage
60-r
C
CD
=3
CT
O>
U_
Northern
Eastern Middle
T ~ f Western Middle
Southern
10
0
2.0 3.0 4.0
5.0
PH
6.0
7.0 8.0
2-8
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remitting Statistical Support Document
2.1 Impact of Stream Flow Variation on Water Quality Parameters
Annual variations in streamflow and surface water quality degraded by AMD discharges can be
very significant as shown in Hornberger et al. (1981) for water quality network stations including
small streams and large rivers in western Pennsylvania These water quality network stations are
closely monitored by PADEP. The streams are sampled several times yearly and analyzed for a
wide array of water quality parameters, and usually are located in close proximity to U.S.
Geological Survey (USGS) stream hydrograph stations for which extensive streamflow data area
compiled and published. The data from the network stations is contained in the STORET
database maintained by EPA.
The water quality network station with the greatest range in streamflow and concentration of
AMD related water quality parameters is the Dunkard Creek Station, in Greene County,
Pennsylvania (Hornberger et al., 1981). This compilation includes greater than 150,000 lines of
STORET data. Streamflow varied between 2 cubic feet per second (cfs) and 4,020 cfs in
approximately 100 samples collected between 1950 and 1976, while the concentration of sulfates
ranged from 40 to 4000 mg/L. The annual cycles of streamflow variations from October 1960 to
September 1970 for Dunkard Creek are shown in Figure 2. la, which was plotted by Hornberger
et al. (1981) from monthly means of discharge data compiled by the U.S. Geological Survey.
Characteristics of Coal Mine Drainage Discharges
2-9
-------
Coal Reminins Statistical Support Document
Figure 2.1a Annual Variability in Streamflow at Dunkard Creek
2-10
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remitting Statistical Support Document
In order to examine the relationship between variations in streamflow and corresponding
variations in a reliable water quality indicator parameter, a logarithmic plot of sulfate
concentration versus discharge was constructed using procedures described in Gunnerson (1967),
Hem (1970), and Hornberger et al. (1981). The sulfate concentrations in Dunkard Creek tend to
systematically decrease with increasing flow as shown by the approximately linear inverse
relationship on Figure 2. Ib. However, the relationship between streamflow and concentration
may be more appropriately defined by a general elliptical progression of monthly flow and water
quality relationships surrounding a least squares line fitted to the data points, similar to that
found-by Gunnerson (1967) and Hornberger et al. (1981). The tendency for high flow
accompanied by low sulfate concentration in January, February, March, April, and May and low
flow accompanied by high sulfate concentration in July, August, September, and October, and
other flow-quality relationships throughout the water year may be observed in Figure 2.1b.
Figure 2. Ib includes almost 50 years of data (1950-1997) that show a stronger inverse linear
relationship between sulfate concentration and streamflow than was shown in the first 26 years of
data (Hornberger et al., 1981). The correlation coefficient (r) between sulfate concentration and
streamflow data in Figure 2. Ib is -0.887 (for logarithmically transformed data), which is
statistically significant at the 1 percent level (N=307). The coefficient of determination [r2] for
this dataset is 0.787; therefore, 78.7 percent of the variations in sulfate concentration of the
Dunkard Creek are accounted for by variations in streamflow. Similar patterns of variation in
sulfate concentration and flow of a major AMD-impacted river were found (Hornberger et al.,
1981) for the West Branch Susquehanna River at Renovo, Pennsylvania.
Characteristics of Coal Mine Drainage Discharges
2-11
-------
CoalRemininx Statistical Support Document
Figure 2.1b: Sulfate Concentration vs. Streamflow at Dunkard Creek
Dunkard Creek at Shannopin, PA
10,000
1,000
tnn
1UU
10
IStfU ~ I09f
»
DA"
B • • * *
H + 1 * 8 »° D A»n
0 * *+* a *D
**+ ** A *°A
* + S* "^ » o 4. * — +
o + 5^*~ ^" ^
* * ^dU «ff^*t+ °" &
+ n A^O ^^ *^* ^
* ^ ^JSfetac* <^ J A £
+ *° ""« * ^C^15^:^*x » °
9 >P J| J|p A^ w-f*& M>&
* X ^* O '"Vx ^
* • Ji A
A
^
, 10 100 1,000
«januaiy
• februsry
A march
xapr»
xmay
• June
-1-july
A august
0 septenter
a October
••noventer
o decenter
A
X
;
1
;
10,
STREAMFLOW (c.f.s.)
2-12
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
2.2 AMD Discharge Types and Behaviors
Discharges of acid mine drainage (AMD) can exhibit very different behavior depending upon the
type of mine involved and its geologic characteristics. The hydrologic characteristics of a pre-
existing AMD discharge can have important ramifications for documenting baseline pollution
load - affecting the frequency and duration of sampling required to obtain a representative
baseline. Braley (1951) was among the first to study the hydrology of AMD discharges. He
noted that, much like a stream, flow rates vary dramatically in response to precipitation events
and seasons, and that acid-loading rates are chiefly a function of flow. The greater the flow, the
greater the load. Smith (1988), looking at long-term records of AMD discharges in
Pennsylvania, classified discharges based on three fundamental behaviors: 1) High flow - low
concentration / low flow - high concentration response, where the flow rate varies inversely with
concentration; 2) Steady response where changes in flow rate and chemistry are rninirnal or
damped; and 3) "Slug" response where large increases in discharge volumes are not accompanied
by corresponding reductions in concentrations, resulting in large increases in pollution loading.
Figure 2.2a presents the discharge and acidity hydrograph of a mine discharge exhibiting the first
(high flow - low concentration / low flow - high concentration) behavior. This discharge drains
from a relatively small underground mine complex (Duffield, G.M., 1985). Typical for this type
of discharge, the flow rate varies greatly and is subject to seasonal flow variations as well as
individual precipitation events. Acidity concentrations vary inversely with the discharge rate,
with the highest acidity occurring during the low-flow months of September, October, and
November. The inverse log-linear relationship between discharge and acidity is shown in Figure
2.2b. Acidity steadily decreases with increasing flow, reflecting dilution of the mine drainage
during periods of abundant ground-water recharge. Nonetheless, the pollution loading (i.e., the
total acidity produced from the discharge in pounds per day) increases during high-flow events,
as the, decrease in acidity is not commensurate with a given increase in flow. In this sense, the
discharge behaves very much like a stream and is subject to large increases in flow which dilute
the concentration of dissolved chemical constituents. However, concentration decreases are not
enough to offset flow increases. Pollution loading tends to parallel the flow rate but in a more
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remininz Statistical Support Document
subdued manner. The majority of pre-existing AMD discharges in Pennsylvania exhibit this type
of behavior. It is most common for surface mine discharges and discharges from small to
medium size underground mines where the capacity for ground water storage is relatively small
and ground water flow paths are short.
Some discharges, particularly large-volume discharges from extensive underground mine
complexes, show comparatively little fluctuation in discharge rate and only minor variation in
chemical quality. Figure 2.2c presents such an example from a Schuylkill County, Pennsylvania,
anthracite underground mine. In this case, the exceptionally large recharge area and volume of
water in the mine pool, and the stratification of water quality within the mine pool, are causing a
steady-response behavior of the discharge. Short-term fluctuations in flow and quality are
subdued, because of the large amount of stored ground water acting as a reservoir and dampening
fluctuations due to individual recharge events.
Figure 2.2a: Acidity and Streamflow of Arnot Mine Discharge
Arnot Mine Discharge
10 V-
1981
1982
1983
2-14
Characteristics of Coal Mine Drainage Discharges
-------
rrnj »-..•..*., Statistical Support Document
Figure 2.2b: Inverse Loglinear
60T "~~~
between Acidity and Streamflow
20
O
§
i
-20
10
LOG DISCHARGE (gal/min)
^ Mt!e
, AMD cHscha.ges are Subject ,o extte« variations ta flow
flow and acidity editing "slug" beha^or » a d.scha.ge
3
-------
Coal Remining Statistical Support Document
Figure 2.2c: Streamflow and Acidity in SchuylkiU County
MARKSON AIRWAY - 1985
2000 —
JAN FES MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Figure 2.2d: Streamflow and Acidity in Coal Refuse Pile
Ernest Refuse Pile
•E-10*
1981 1982 1983 1984 1985
2-16
Characteristics of Coal Mine Drainage Discharges
-------
Coal Reminins Statistical Support Document
phenomenon occurs on a large scale, potentially disastrous increases in acid loading can
adversely affect downstream water uses and aquatic life.
The Arnot, Markson, and Ernest mine drainage discharges described in the preceding paragraphs
were originally studied and graphically presented in Smith (1988) and Hornberger et al. (1990).
These three mine drainage discharges are also the subject of three of the eight water quality
reports completed by Griffiths (1987, 1988) as part of the cooperative EPA/PADEP remining
project and included in the abridged volume by EPA (2001, EPA-821-B-01-014).
For remining operations that will reaffect a pre-existing pollutional discharge, knowledge of
discharge behavior is critical to the establishment of a representative baseline. Allthree
discharge types exhibit some seasonal behavior, with highest flows during seasonal high ground-
water conditions and the lowest flows and loadings during low ground-water conditions. For
most of Appalachia, high ground-water conditions occur during late winter or spring. Low
ground-water conditions occur during late summer and early fall. The baseline sampling period
must cover the full range of seasonal conditions. Exactly when these extremes will occur is
unpredictable, as storm events may occur over relatively short time intervals. Accordingly, to
properly characterize an AMD discharge, it is usually necessary to monitor the discharge over at
least an entire water year with a sufficiently narrow sampling interval to capture short-term
extreme events. Slug-response discharges may require more frequent sampling due to their
flashy hydrologic response with large variations in pollution load over short time intervals.
Conversely, less frequent baseline sampling may be adequate for damped-response discharges.
Because the baseline is based on loading rates, accurate flow measurements are as important as
contaminant concentration measurements. Previous studies by Smith (1988), Hornberger et al.
(1990), and Hawkins (1994) have emphasized the strong relationship between flow rate and
contaminant load. Hawkins (1994) analyzed pre-and post-reiraning hydrologic data from 24
remining sites in Pennsylvania and noted that flow was the dominant factor in changes in
post-mining pollution loads. Most remining operations that reduced baseline pollution load did
so by reducing the flow of the pre-existing discharge. In view of this, Smith (1988) points out
Characteristics of Coal Mine Drainage Discharges
2-17
-------
Coal Remining Statistical Support Document
that proper flow measurement is of overriding importance in determining the baseline pollution
load.
2.3 Distributional Properties of AMD Discharges
Water quality parameters of many AMD discharges and AMD impacted streams are not normally
distributed. In most cases these frequency distributions are highly skewed because there are
many samples with relatively low concentrations and a few samples with very high
concentrations due to low-flow drought conditions or slugs of pollution in response to major
storm events. Plotting these data on a logarithmatic scale (as shown on Figure 2. Ib), or
logarithmically transforming the data produces a much closer approximation of the normal
frequency distribution.
Figure 2.3a: Frequency Distribution of Sulfate at Dunkard Creek (mg/L)
400
800
1200
1600
2000 2400
Sulfate (mg/L)
2800
3200
3600
4000
More
2-18
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Numerous variables with continuous data on the interval or ratio level of information exhibit log
normal behavior in the natural environment (Aitchison and Brown, 1973; Krumbein and
Graybill, 1965; Griffiths, 1967), and logarithms are frequently used in the analysis and graphical
expression of water quality data (Gunnerson, 1967; Hem, 1970). The log normal distribution is
also very common in previous EPA work with wastewater discharges. Figure 2.3a shows the
skewed frequency distribution for the sulfate data for the Dunkard Creek dataset used in Figure
2.1b.
Examples of the distributional properties of data from AMD discharges at remining sites in
Pennsylvania are shown in Figures 2.3b to 2.3f from the EPA publication Statistical Analysis of
Abandoned Mine Drainage in the Assessment of Pollutant Load (EPA-821-B-01-014), which is a
companion volume to this report. The figures show frequency distributions of data using stem-
and-leaf diagrams. For additional information on stem-and-leaf diagrams, see Hoaglin et al.
1983.
Figure 2.3b shows a nearly normal frequency distribution of pH of the Arnot 003 discharge
(N=82). An example of a highly skewed frequency distribution is given in Figure 2.3c for flow
of the Clarion discharge. Following logarithmic transformation, the frequency distribution
becomes more symmetrical, approaching normality, as seen in Figure 2.3d. However, some
caution must be exercised in applying log transformations to data sets because overcorrection
may occur. Such overcorrection is seen in the irregular frequency distribution of acidity
concentration in the Clarion discharge. In Figure 2.3e, the untransformed data are somewhat
positively skewed. Following transformation, these data become highly negatively skewed
(Figure 2.3f).
Characteristics of Coal Mine Drainage Discharges
2-19
-------
Coal Remining Statistical Support Document
Figure 2.3b: Stem-and-leaf Diagram of pH (Arnot 003)
N=82
Leaf Unit=0.010
1
2
5
17
36
(15)
31
19
10
6
2
1
1
1
30
30
31
31
32
32
33
33
34
34
35
35
36
36
37
4
7
134
555667888999
0011111112234444444
555666777789999
001111222234
556666778
1122
679
0
7
2-20
Characteristics of Coal Mine Drainage Discharges
-------
Coal Reminirig Statistical Support Document
Figure 2.3c: Stem-and-leaf Diagram of Discharge
N = 77
Leaf Unit = 1.0
(53) 0
24 1
13 2
7 3
4 4
3 5
2 6
2 7
2 8
1 9
1 10
1 11
1 12
1 13
1 14
1 15
1 16
1 17
00000000001111222333333344444555555556666667788999999
00222244444
001288
066
0
0
3
2 . . .
Figure 2.3d: Stem-and-leaf Diagram of Log Discharge
N = 75
Leaf Unit = 0.10
4 -1
7 -0
11 -0
19 0
(35) 0
21 1
6 1
1 2
3000
766 ...-•'
4330
01124444
55555566666777777777788888899999999
000111111333344
55679
2
Characteristics of Coal Mine Drainage Discharges
2-21
-------
Coal Retaining Statistical Support Document
Figure 2.3e: Stem-and-leaf Diagram of Acidity
N = 96
LeafUnit=10
8
21
30
38
(14)
44
34
26
20
18
8
6
4
2
1
1
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
00114568
0234466667788
033356899
01245778
01114456788899
0112778899
03344678
014499
15
1455568899
38
09
04
8
Figure 2.3f : Stem-and-leaf Diagram of Log Acidity
N=97
Leaf Unit =0.10
1
1
2
3
3
5
9
31
(58)
8
-1
-0
-0
0
0,
1
1
2
2
3
0
0
3
22
6679
0011122222222333444444
5555555566666666666666777777777777888888888889999999999999
00000011
2-22
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
References
Aitchison, J. and J.A.C. Brown, 1973. The Lognormal Distribution. Cambridge University Press:
London. 176 p.
Barnes, H.L. and S.B. Romberger, 1968. Chemical Aspects of Acid Mine Drainage. Journal of
the Water Pollution Control Federation, Vol. 40, No. 3, pp. 371 - 384.
Brady, K.B.C., E.F. Perry, R.L. Beam, D.C. Bisko, M.D. Gardner, and J.M. Tarantino, 1994.
Evaluation of Acid-base Accounting to Predict the Quality of Drainage at Surface Coal
Mines in Pennsylvania, U.S.A. Presented at the International Land Reclamation and Mine
Drainage Conference on the Abatement of Acidic Drainage, Pittsburg, PA, April 24-29,
1994.
Brady, K.B.C., A.W. Rose, C.A. Cravotta, m, and W.W. Hellier, 1997. Bimodal Distribution of
pH in Coal Mine Drainage. Abstracts with Programs, 27th Northeast Section Meeting,
Geological Society of America, Vol. 24, No. 3.
Brady, K.B.C., RJ. Hornberger, and G. Fleeger, 1998. Influence of Geology on Postmining
Water Quality: Northern Appalachian Basin. Chapter 8 in Coal Mine Drainage
Prediction and Pollution Prevention in Pennsylvania. Edited by K.B.C. Brady, M.W.
Smith, and J.Schueck. Department of Environmental Protection: Harrisburg, PA. pp. 8-
1 to 8-92.
Braley, S.A., 1954. Summary Report on Commonwealth of Pennsylvania, Department of Health
Industrial Fellowship (Nos. 1,2,3,4,5,6, and 7). Fellowship 326B, Mellon Institute of
Industrial Research.
Cravotta, C.A., m, 1994. Chapter 23: Secondary Lran-sulfate Minerals as Sources of Sulfate and
Acidity - the Geochemical Evolution of Acidic Ground Water at a Reclaimed Surface
Coal Mine in Pennsylvania. US Geological Survey: Lemoyne,PA.
Duffield, G.M., 1985. Intervention Analysis Applied to the Quantity an Quality of Drainage
from an Abandoned Underground Coal Mine in North-Central Pennslvania. Masters of
Science Thesis, Pennsylvania State University, Department of Geology.
Dugas, D.L., C.A. Cravotta, m, and D.A. Saad, 1993. Water Quality Data for Two Surface Coal
Mines Reclaimed with Alkaline Waste or Urban Sewage Sludge, Clarion County,
Pennsylvania, May 1983 through November 1989. U.S.Geological Survey Open File
Report 93-115: Lemoyne, PA.
Characteristics of Coal Mine Drainage Discharges
2-23
-------
Coal Remininx Statistical Support Document
Emrich, G.H., 1966. Tests for Evaluating the Quality of Mine Drainage Characteristics of Coal
Seams. Pennsylvania Department of Health, Division of Sanitary Engineering, Mine
Drainage Technical Bulletin No. 2., pp. V-6 thru V-14.
Evangelou, V.P., 1995. Pyrite Oxidation and its Control. CRC Press: New York, NY. 293 p.
Griffiths, J. C., 1967. Scientific Method in Analysis of Sediments. McGraw Hill Book Co.: New
York, NY. 508 p.
Griffiths, J.C., (No date). Report No. 1: Reconnaissance into the use of MINITAB using
Hamilton 01, 08 Data. U.S. Environmental Protection Agency, Office of Water:
Washington, DC.
Griffiths, J.C., 1987a. Report No. 2: Time Series Analysis of Data from Hamilton 08. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987b. Report No. 3: Analysis of Data from Arnot 001, 003,004. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987c. Report No. 4: Analysis of Data from the Clarion Site. U.S. Environmental
Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987d. Report No. 5: Analysis of Data from Ernest Refuse Pile, Indiana County,
Pennsylvania. U.S. Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987e. Report No. 6: Analysis of Data from the Fisher Deep Mine. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1988a. Report No. 7: Analysis of Data from the Markson Site. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1988b. Report No. 8: Synopsis of Reports 1-7. U.S. Environmental Protection
Agency, Office of Water: Washington, DC.
Gunnersqn,.C.G., 1967. Streamflow and Quality in the Columbia River Basin. Journal of the
Sanitary Engineering Division, Proceedings of the American Society Civil Engineers,
Vol. 93, No. SA6, pp. 1 - 16.
Hawkins, J.W., 1994. Statistical Characteristics of Coal Mine Discharges on Western
Pennsylvania Remining Sites. Water Resources Bulletin, Vol. 30, No. 5, pp. 861 - 869.
Hem, J.D., 1970. Study and Interpretation of the Chemical Characteristics of Natural Water. US
Geological Survey Water Supply Paper No. 1473, 2nd ed. U.S. Government Printing
Office: Washington, DC. 363 p.
2-24
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Hoaglin, D.C., F. Mosfeller, and J.W. Tukey, 1983. Understanding robust and exploratory data
analysis. John Wiley & Sons: New York, NY. pp. 7 -32.
Hornberger, R.J., R.R. Parizek, and E.G. Williams, 1981. Delineation of Acid Mine Drainage
Potential of Coal Bearing Strata of the Pottsville and Allegheny Groups in Western
Pennsylvania. Final Report, Department, of Interior, Office of Water Research and
Technology Project B-097-PA. Pennsylvania State University: University Park, PA.
Hornberger, R.J., M.W. Smith, A.E. Friedrich, and H.L. Lovell, 1990. Acid Mine Drainage from
Active and Abandoned Coal Mines in Pennsylvania. Chapter 32 in Water Resources in
Pennsylvania: Availability, Quality and Management. Edited by S.K. Majumdar, E.W.
Miller and R.R. Parizek. The Pennsylvania Academy of Science: Easton, PA. pp. 432 -
451.
Hornberger, R.J., and K.B.C. Brady, 1998: Kinetic (Leaching) Tests for the Prediction of Mine
Drainage Quality. Chapter 7 in Coal Mine Drainage Prediction and Pollution Prevention
in Pennsylvania. Edited by K.B.C. Brady, M.W. Smith and J. Schueck. Department of
Environmental Protection: Harrisburg, PA. pp. 7-1 to 7-54.
Kleinmann, R.L.P., D. A. Crerar, and R.R. Pacelli, 1981. Biogeochemistry of Acid Mine
Drainage and a Method to Control Acid Formation. Mining Engineering, pp. 300 - 305,
March 1981.
Krumbein, W.C., and F.A. Graybill, 1965. An Introduction to Statistical Models in Geology.
McGraw Hill Book Company, Inc: New York, NY. 475 p.
Lovell, H.L., 1983. Coal Mine Drainage in the United States - An Overview. Water Science
and Technology, Vol. 15, pp. 1 - 25.
McKibben, M.A. andHJL. Barnes, 1986. Oxidation of Pyrite in Low Temperature acidic
Solutions - Rate Laws & Surface Textures. Geochimica et Cosmochimica Acta, Vol. 50,
pp. 1509 -1520.
Moses, C.O. and J.S. Herman, 1991. Pyrite Oxidation at Circumneutral pH. Geochimica et
Cosmochimica Acta, Vol. 55, pp. 471 - 482.
Nordstrom, D.K., 1982. Aqueous Pyrite Oxidation and the Consequent Formation of Secondary
Iron Minerals. Acid Sulfate Weathering, Special Publication No. 10, Soil Science Society
of America, pp. 37-56.
Characteristics of Coal Mine Drainage Discharges
2-25
-------
Coal Reminins Statistical Support Document
Perry, E.F. and K.B.C. Brady, 1995. Influence of Neutralization Potential on Surface Mine
Drainage Quality in Pennsylvania. In: Proceedings 16th Annual West Virginia Surface
Mine Drainage Task Force Symposium, Morgantown, WV, April 4-5,1995.
Rimstidt, J.D. and W.D. Newcomb, 1993. Measurement and Analysis of Rate Data - The Rate of
Reaction of Ferric Iron with Pyrite. Geochimica et Cosmochimica Acta, Vol. 57, pp. 1919
-1934.
Rose, A.W. and C.A. Cravotta, m, 1998. Geochemistry of Coal Mine Drainage. Chapter 1 in
Coal Mine Drainage Prediction and Pollution Prevention in Pennsylvania. Edited by:
K.B.C. Brady, M.W. Smith and J. Schueck Department of Environmental Protection:
Harrisburg, PA. pp. 1-1 to 1-22.
Schueck, J., M. DiMatteo, B. Scheetz, and M. Silsbee, 1996. Water Quality Improvements
Resulting from FBC Ash Grouting of Buried Piles of Pyritic Materials on a Surface Coal
Mine. Proceedings of the 13th Annual Meeting - American Society for Surface Mining
and Reclamation, Knoxville, TN, May 19-24, 1996.
Singer, P.C., and W. Stumm, 1970. Acidic Mine Drainage: The Rate Determining Step.
Science, Vol. 167, pp. 1121 - 1123.
Smith, M.W., 1988. Establishing Baseline Pollution Load from Pre-existing Pollutional
Discharges for Remining in Pennsylvania. Paper presented at the 1988 Mine Drainage
and Surface Mine Reclamation Conference, Pittsburg, PA, April 17-22, pp. 311 - 318.
U.S. EPA, 2001. Statistical Analysis of Abandoned Mine Drainage in the Establishment of the
Baseline Pollution Load for Coal Remining Permits. Prepared for US Environmental
Protection Agency, Office of Water by Pennsylvania Department of Environmental
Protection and DynCorp I&ET. EPA-821-B-01-014.
Watzlaf, G.R., 1992. Pyrite Oxidation in Saturated and Unsaturated Coal Waste. In: Proceedings,
1991 National Meeting of the American Society for Surface Mining and Reclamation,
Duluth, MR pp. 191 - 205, June 14-18,1992.
Williams, E.G., A.W. Rose, R.R. Parizek, and S.A. Waters, 1982. Factors Controlling the
Generation of Acid Mine Drainage. Final Report on U.S. Bureau of Mines Research
Grant No. G5105086. Pennsylvania State University: University Park, PA.
2-26
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Section 3.0 Statistical Methods for Establishing Baseline
Conditions and Setting Discharge Limits at
Remining Sites
3.1 Objectives & Evaluation of Statistical Methods
The Rahall amendment, CWA Section 301(p), states in part:
(2) LIMITATIONS. - The Administrator or the State may only issue a permit pursuant to
paragraph (1) if the applicant demonstrates to the satisfaction of the Administrator or the
State, as the case may be, that the coal remining operation will result in the potential for
improved water quality from the remining operation but in no event shall such a permit
allow the pH level of any discharge, and in no event shall such a permit allow the
discharges of iron and manganese, to exceed the levels being discharged from the
reminedarea before the coal remining operation b'egins."
EPA has promulgated the Coal Remining Subcategory (40 CFR Part 434.70) consistent with the
requirements and intent of the Rahall amendment. The regulations for the Coal Remining
Subcategory establish a standardized statistical procedure for determining baseline pollutant
loadings and pollutant loadings during remining for net acidity, solids, iron, and manganese in
pre-existing discharges. These statistical procedures are codified in Appendix B to Part 434 and
are intended to identify increases (during remining) of discharge pollutant loadings above the
baseline levels.
EPA has interpreted "levels" to mean the entire probability distribution of loadings, including the
average, the median, and the extremes. It follows that if P percent of loadings did not exceed
some number Lp during baseline, then no more than P percent should exceed Lp during and after
remining. For example, if during the baseline period, 95 percent of iron loadings are ^ 8.1
Statistical Methodology 3-1
-------
Coal Reminins Statistical Support Document
Ibs/day and 50 percent are < 0.3 Ibs/day, then during and after remining the same relationships
should hold true.
The objective of Section 3 is to provide statistical procedures for deciding when the pollutant
loadings in a discharge exceed the levels of baseline. These procedures are intended to provide a
good chance of detecting a substantial, continuing state of exceedance, while reducing the
likelihood of a "false alarm." The procedures (or the numbers calculated from them) are also
referred to here as "triggers."
In developing these procedures, EPA considered the statistical distribution and characteristics of
discharge loadings data from pre-existing discharges, the suitability of parametric and
nonparametric statistical procedures for such data, the number of samples required for these
procedures to perform adequately and reliably, and the balance between false positive and false
negative decision error rates. EPA also considered the cost involved with sample collection as
well as delays in permit approval during the establishment of baseline, and considered the
potential that increased sampling could discourage remining. In order to sufficiently characterize
pollutant levels during baseline determination and during each annual monitoring period,
Appendix B to Part 434 requires that the results of a minimum of one sample be obtained per
month for a period of 12 months.
*
The procedures described below will provide limits for both single observations and annual
averages. This is intended to provide checks on both the average and extreme values. There is a
need to take into account the number of observations used to determine compliance when setting
a limit or when otherwise determining compliance with baseline. For example, the collection of
a greater number of samples from a discharge will reduce the variability of the average level
(provided that samples are distributed randomly or regularly over the sampling year).
Accordingly, the statistical procedures described here take into account the amount of data in an
appropriate fashion.
3-2
Statistical Methodology
-------
Coal Remitting Statistical Support Document
Use of a statistical decision procedure should result in suitable error rates. Technically these are
usually referred to as the rate alpha (a) for Type I errors and the rate beta (P) for Type II errors.
The error of concluding that an exceedance has occurred when the discharge is exactly matching
the baseline condition is intended to happen with probability a. Alpha can be characterized as
the maximum "false alarm rate." When the discharge level is substantially less than baseline, the
probability of making this error is expected to be very low. The error of concluding that no
exceedance has occurred, when the discharge has in fact exceeded baseline levels, is intended to
happen with probability p. Power (TC) equals 1-p. Power can be defined as the probability that a
statistical decision procedure will declare that remining loadings exceed baseline loadings when
there really has been an increase, or as the rate of giving correct alarms.
When many decisions will be made, the overall error rate is a concern. For example, the single-
observation triggers described below will be applied every month during remining; the annual
triggers will be applied every year. In evaluating statistical methods, EPA considered the overall
or cumulative decision error rates during a five-year period of compliance monitoring.
The degree of serial correlation of the data will influence the decision error rates of statistical
procedures. There is significant, positive serial correlation of flow, concentration, and loading in
mine discharges over periods of 1 to 4 weeks, that is, sequential samples are correlated with each
other (U.S.E.P.A., 2001a, 200Ib). Also, estimates of the variance, used in parametric statistical
procedures, are inaccurate unless adjusted for autocorrelation. (Loftis and Ward, 1980;
U.S.E.P. A., 1993). Such adjustments require an estimate of the autocorrelation coefficient.
However, one cannot reliably estimate site-specific autocorrelation from small samples (e.g.,
n=12). Using long-term datasets for pre-existing discharges at abandoned mines and at remining
sites, EPA estimated the first-order serial correlations (at a monthly interval) for flow and for
iron, manganese, and acidity loadings. The estimates fell mostly in the range 0.35 to 0.65, with a
central tendency just below 0.50 (U.S.E.P.A., 2001b).
EPA evaluated parametric and nonparametric statistical procedures for characterizing the
baseline level and determining compliance with the baseline level (U.S.E.P.A., 2001c). For the
Statistical Methodology
3-3
-------
Coal Reminins Statistical Support Document
evaluation, EPA simulated discharge loadings data. These data had realistic statistical properties,
resembling actual discharge loadings in terms of distribution and serial correlation (U.S.E.P.A.,
2001b, 2001c, 2001d). The data simulated a 1- year (12 month) baseline period followed by a 5-
year (60-month) remining period, with loadings measured once every month (also weekly, when
the procedure required a period of accelerated monitoring). The evaluation examined the ability
of a number of statistical procedures to react to various degrees of decrease and increase in
loadings after baseline. The parametric procedures employed appropriate adjustments to the
estimated variance to account for first-order serial correlation (assumed to be 0.5). The
evaluation assumed that a minimum of 12 measurements of pollutant loads were made every
year, once each month.
The ideal statistical procedure would always declare "not larger" when remining pollutant
loadings are less than or equal to baseline loadings, and would always signal "larger" when
remining loadings exceeded baseline. No such ideal procedure exists. Instead, the rate of
signalling "larger" will increase as the average difference between baseline and remining
loadings increases in magnitude. Statistical triggers may be "tuned", by choosing their
numerical constants, so that a compromise is achieved between false alarms, that is, signalling
"larger" when remining loadings are not larger than baseline loadings, and correct alarms, when
remining loadings truly are greater.
The evaluations led to a choice of procedures and of numerical constants that achieve a
reasonable balance between false alarms and correct alarms. This reasonable balance was
considered to be achieved when a trigger produced the following results:
(a) when there was no change in loadings from the baseline to the remining time
period, the "false alarm rate" (type-I error rate) was not larger than that for the
triggers used by the Commonwealth of Pennsylvania. Pennsylvania's trigger was
used as a benchmark because of the demonstrated success of this approach
(Hawkins 1994).
3-4
Statistical Methodology
-------
Coal Remining Statistical Support Document
(b) when the mean pollutant load increased by one standard deviation after the
baseline period, statistical power (probability of detecting the increase) was at
least 0.75. ,
(c) when there was a decrease of (X5 standard deviations in the meati loading after
the baseline period, the"false alarm rate" was smaller than 5%.
(d) when the mean loading increased by 1 to 2 standard deviations after the
baseline period, the "correct alarm rate" (power) was maximized (compared with
other procedures).
Details of EPA's evaluation and comparisons of statistical procedures are provided in a separate
document (U.S.E.P. A., 2001c). EPA reached the following conclusions about the statistical
triggers based on these evaluations.
(1) The magnitude of serial correlation has a substantial effect on power. Statistical triggers
that have reasonable power when there is no serial correlation could be unreasonable
when there is substantial serial correlation, because they could then have very high rates
of type I errors (false alarms). It was necessary to select numeric constants for the
statistical triggers that are appropriate to data having autocorrelation. For evaluating and
comparing statistical methods and triggers, EPA assumed a first-order autocorrelation
coefficient of 0.5.
(2) To avoid false alarms, EPA determined that sequential exceedances of the Single
Observation Trigger and accelerated monitoring were necessary. This method has long
been used successfully in Pennsylvania's Remining Program. Specifically, the Single
Observations Trigger requires the following: "If two successive monthly monitoring
observations both exceed L, immediately begin weekly monitoring for four weeks (four
weekly samples). If three or fewer of the weekly observations exceed L, resume monthly
monitoring. If all four weekly observations exceed L, the baseline pollution loading has
been exceeded."
Statistical Methodology • 3-5
-------
Coal Reminins Statistical Support Document
(3) In Method 2, the Annual Comparison was set such that tables for the 99.9% level (alpha =
0.001) rather than the 95% level (alpha = 0.05) are to be used for the Wilcoxon-Mann-
Whitney Test. When Type I Error rates of alpha = 0.05 or 0.01, the Wilcoxon-Mann-
Whitney test in Method 2 had a high rate of declaring loadings to be larger than baseline
when if fact, they were not larger.
(4) Method 1 and Method 2 were both designed as nonparametric rather than parametric
procedures, with power comparable to that of a parametric procedure. Unlike a
parametric method which would require log-transformation, the nonparametric methods
accommodate zero flows (which may occur during remining) and negatively valued data
(which may occur for net acidity) without requiring additional or complex modifications.
(5) EPA believes that the error rates and power of these triggers are acceptable in practice
because best management practices (BMPs) reduced discharge loadings substantially.
Hawkins (1994) reviewed the application of these triggers to remining operations in
Pennsylvania, and concluded that the rates of triggering were low because remining
usually reduced loadings substantially. EPA's BMP guidance manual includes an
extensive analysis of remining discharges that supports this conclusion (EPA, 2001a).
EPA concluded that the statistical triggers that Pennsylvania uses in its remining program
are acceptable and effective and has used them as the basis for Method 1 with minor
modifications to meet the criteria in (a) to (d). Method 1 herein follows the Pennsylvania
triggers exactly except that a constant (1.815 = 1.96 * 1.25 / 1.35) is used in the formula
for the Annual Procedure (see McGill, Tukey, and Larsen, 1978). Pennsylvania uses a
more stringent number (1.58 = 1.7 * 1.25 / 1.35).
(6) The evaluation of the false alarm rate applies to a worst-case situation. The rate of
declaring loadings to be larger than baseline when they are not is overstated by the
evaluations (U.S.E.P. A., 2001c). It is evaluated in terms of the percentage of mines that
would experience at least one determination that loadings exceed the baseline level over a
period of five years (60 months), when in fact there has been no change from baseline. In
practice, the area contributing to a discharge should be remined and regraded in less time,
after which the discharge flow and loading will be substantially reduced. Thus the time
period during which pollutant loadings are monitored for each discharge will usually be
•jTg"! ' Statistical Methodology
-------
Coal Remining Statistical Support Document
shorter than five years. This in turn will mean lower percentages of false positives and
false negatives than reported in Table 3. la.
The power of statistical triggers for the final regulation is shown in Table 3. la. The results show
that Method 1 and Method 2 have comparable power to detect large increases (see columns 'la'
and '+2a'). The main difference sterns from the Monthly (Single Observation Limit) Test, which
has higher false alarm percentages (see columns labeled '-0.5o' and '0') when Method 1 is used.1
Note that the Annual Comparison used without the Single Observation Limit Test would not
have a high rate of detecting an increase of one standard deviation above baseline. Used in
combination, the single observation and annual triggers provide power over 90% to detect
substantial increases above baseline at least once during five years (Table 3. la), although in
practice the power may be smaller for reasons discussed above under (6).
*As explained in (5), EPA believes that the error rates and power of these triggers will be
acceptable in practice because best management practices (BMPs) reduce discharge loadings
substantially.
Statistical Methodology
-------
Coal Reminine Statistical Support Document
Table 3.1a. Statistical Triggers as Modified for Final Regulation: Percentage of
Discharges Declared to Exceed Baseline Level (at least once during 5 years of
simulated monthly monitoring) 1
Annual Trigger
NA
Method 1
(Multiplier 1.96)
Method 1
(Multiplier =1.96)
Method 1
(Multiplier = 1.96)
NA
Method 2
(a = 0.001)
Method 2
(a = 0.001)
Method 2
(a = 0.001)
Single-Observation
Trigger
Method 1
NA
Method 1
Method 2
Method 2
NA
Method 2
Method 1
Shift from Baseline to Remtning Period 2
-0.5 o
10%
3%
12%
7%
5%
2%
7%
12%
0
33%
11%
39%
29%
22%
11 %
28%
38%
1 0
89%
59%
93%
91%
86%
65%
91%
93%
+2o
99%
94%
100%
100%
100 %
97%
100%
100%
1 Assumes monthly serial correlation of 0.5 for log(x), with x distributed lognormally. Percentages were
rounded to the nearest 1%.
2 The shift was scaled in terms of standard deviation units ( a = standard deviation)
3-8
Statistical Methodology
-------
Coal RemininK Statistical Support Document
3.2 Statistical Procedures for Calculating Limits from Baseline Data
The procedures to be used for establishing effluent limitations for pre-existing discharges at coal
remining operations, in accordance with the requirements set forth in 40 CFR part 434, Subpart
G; Coal Remining are presented below. The requirements specify that pollutant loadings of total
iron, total manganese, total suspended solids, and net acidity in pre-existing discharges shall not
exceed baseline pollutant loadings. The two alternative procedures described (Method 1 and
Method 2) are applied to determine site-specific, baseline pollutant loadings/and to determine
whether discharge loadings during coal remining operations have exceeded baseline loading. For
each procedure, both a monthly (single-observation) test and an annual test are applied. In order
to sufficiently characterize pollutant loadings during baseline determination and during each
annual monitoring period, the regulations require that at least one sample result be obtained per
month for a period of 12 months.
The calculations described are applied to pollutant loadings. Each loading value is calculated as
the product of a flow measurement and pollutant concentration taken on the same date at the
same discharge sampling point, using standard units of flow and concentration (to be determined
by the permitting authority). For example, flow may be measured in cubic feet per second,
concentration in milligrams per liter, and the pollutant loading calculated in pounds per year.
In the event that a pollutant concentration in the data used to determine baseline is lower than the
daily maximum limitation established in Subpart C for active mine wastewater, the statistical
procedures should not establish a baseline more stringent than the BPT and BAT effluent
standards established in Subpart C. Therefore, if the total iron concentration in a baseline sample
is below 7.0 mg/L, or the total manganese concentration is below 4.0 mg/L, the baseline sample
concentration should be replaced with 7.0 mg/L and 4.0 mg/L, respectively, for the purposes of
some of the statistical calculations. The substituted values should be used for all methods
described in this section with the exception of the calculation of the interquartile range (R) in
Method 1 for the annual trigger, and in Method 2 for the single observation trigger. The
interquartile range (R) is the difference between the quartiles M^ and Mt; these values should be
Statistical Methodology
-------
Coal Reminins Statistical Support Document
calculated using actual loadings (based on:measured concentrations) when they are used to
calculate R. This should be done in order to account for the full range of variability in the data.
3.2.1 Method 1
Method 1 is a modification of the methodology used by the Commonwealth of Pennsylvania.
Computational details appear in Figure 3.2a. Pennsylvania's monthly and annual average checks
can be described as follows:
Monthly for single-observation maximum) cheek: A tolerance interval is estimated for the
baseline loadings (for n < 17, the smallest and largest observations define the interval endpoints).
The baseline upper bound (usually the maximum baseline loading) is the value of interest. Two
consecutive exceedances of the upper bound trigger weekly monitoring. Four consecutive
exceedances during weekly monitoring trigger a treatment requirement. Thus, six exceedances
must occur consecutively before a treatment requirement is triggered.
Annual average check: A robust, asymptotic estimator2 of a 95 percent confidence interval for
the median is calculated for the baseline period and post-baseline periods; if the post-baseline
interval exceeds the baseline interval, an exceedance is declared. This estimate is based upon
McGill, Tukey, and Larsen (1978).
2Because loadings data for pre-existing discharges are highly asymmetric, and annual
means and medians are likely to be somewhat asymmetrically distributed, EPA used an
asymptotic approximation to develop the confidence intervals for the annual averages. However,
the approximation results in confidence intervals that are symmetric rather than asymmetric.
Thus, this approximation is expected to be accurate only for very large samples, because their
means are approximately normally distributed (by the Central Limit Theorem). EPA has used
this approximation for smaller samples because it provides reasonably good performance as
demonstrated in simulations (see U.S.E.P.A., 2001c).
3-10 Statistical Methodology
-------
Coal Remining Statistical Support Document
3.2.2 Method 2
Similarly to Method 1, Method 2 consists of two checks: an upper limit on single observations
and an annual test of the mean or median. Computational details of Method 2 are provided in.
Figure 3.2b. The single-observation limit is a nonparametric estimate of the 99th percentile of
loadings, developed using baseline data. The annual test of the average or median employs the
nonparametric WUcoxon-Mann-Whitney test.
3.2.3 Accelerated Monitoring
For Methods 1 and 2, triggered or accelerated monitoring is applied after two consecutive
exceedances of the Single Observation Trigger L. If this occurs, weekly sampling must be
commenced immediately. After four weekly samples are collected, the results should be
compared to the Single Observation Trigger L. If three or fewer of the weekly observations
exceed L, then monthly sampling can be resumed. However, if all four weekly observations
exceed L, the baseline pollution loading has been exceeded.
Accelerated monitoring (if used as a condition or option for determining non-compliance) guards
against a declaration of non-compliance on the basis of a transient exceedance, and provides a
means to demonstrate continuing exceedances. It guards against the possibility of instituting
expensive remedial measures when there was no continuing exceedance of baseline conditions.
Statistical Methodology
3-11
-------
Coal Reminins Statistical Support Document
Figure 3.2a: Method 1: The single-observation trigger is applied to each new
measurement; the annual test is applied once a year, using all measurements
for the past year
x = pollutant loading measurement (product of flow, concentration, and
conversion factor)
n = number of Xj results in the taseline dataset
1. Single-observation trigger
Order all n baseline measurements such that x^ is the lowest value, and x16then:
Calculate the sample median (M) of the baseline events:
If n is odd, then M equals x(n/2+1/2).
If n is even, (hen M equals 0.5*(x(n/2)+X(n/2+i))-
Calculate Mt as the median between M and the maximum X(n).
Calculate M2 as the median between Mt and x T , conclude that the median loading during re-mining has exceeded the
median loading during the baseline period, and declare an exceedance.
3-12
Statistical Methodology
-------
Coal Remining Statistical Support Document
Figure 3.2b: Method 2: All three tests or limits are applied. The single-
observation trigger is applied to each new measurement; the annual
test is applied once a year, using all measurements for the past year
1. Single-observation trigger
Calculate M and Mj as described in Method 1 (Figure 3.2a).
Calculate M.j as the median between the minimum. x(1) and the sample median.
Calculate R= (Mj - M.j).
Calculate the Single Observation Trigger as L = Mj + (3 * R)
If, during remining, two successive monthly observations exceeds L, proceed immediately to weekly
monitoring for four weeks (four weekly samples). If, during weekly monitoring, all four observations
exceed L, declare exceedance of the baseline distribution.
2. Annual comparison'
Compare baseline year loadings with current annual loadings using the Wilcoxon-Mann-
Whitneytest2 for two independent samples. Use a one-tailed test with alpha 0.001.
'Hirsch, R.M., and J.R. Stedinger. 1987. Plotting Positions for Historical Hoods and Their
Precision. Water Resources Research. Vol. 23, No.4:715-727.
2 See Conover, W.J., 1980, Practical Nonparametric Statistics, 2nd ed., and other textbooks.
Statistical Methodology
3-13
-------
Coal Reminins Statistical Support Document
References
Brady, K.B.C., M.W. Smith, and J. Schueck (editors), 1998. Coal Mine Drainage
' Prediction and Pollution Prevention in Pennsylvania. Pennsylvania Department of
Environmental Protection, publ. no. 5600-BK-DEP2256, October 1998.
Hawkins, J.W. 1994. A statistical evaluation ofremining abandoned coal mines to reduce
the effluent contaminant load. International Journal of Surface Mining,
Reclamation and Environment 8: 101-109.
Hornberger, RJ. et al., 1990. Acid Mine Drainage from Active and Abandoned Coal Mines
in Pennsylvania. Chapter 32 of Water Resources in Pennsylvania: Availability,
Quality, and Management, Pennsylvania Academy of Science, pp. 432-451.
Loftis, J.C. and R.C.Ward, 1980. Sampling Frequency Selection for Regulatory Water
Quality Monitoring. Water Resources Bulletin, VoL 16, No. 3, pp.501-507.
McGill, R., J.W. Tukey, and W.A. Larsen, 1978. Variations of Box Plots. The American
Scientist, Vol. 32, No. 1, pp. 12-16.
U.S.E.P.A., 1993. Statistical Support Document for Proposed Effluent Limitations
Guidelines and Standards for the Pulp, Paper, and Paperboard Point Source
Category. EPA publ. no. 821/R-93-023.
U.S.E.P.A., 1999. Office of Water. Coal Remining Database: 61 State Data Packages,
March 1999. (docket number DCN 3054)
U.S.E.P. A., 2001a. Statistical Analysis of Abandoned Mine Drainage in the Assessment of
Pollution Load. EPA-821-B-01-014.
U.S.E.P. A., 2001b. "Serial Correlation of Coal Mine Discharge Loadings," memorandum
in the rulemaking record (docket number DCN 3050).
U.S.E.P. A., 2001c. "Evaluation of Statistical Triggers," memorandum in the rulemaking
record (docket number DCN 3051).
U.S.E.P. A., 2001d. "Distribution & Variability of Coal Mine Discharge Loadings,"
memorandum in the rulemaking record (docket number DCN 3049).
3-14
Statistical Methodology
-------
Coal Remitting Statistical Support Document
Section 4.0 Baseline Sampling Duration and Frequency
4.1 Power and Sample Size
Power, in the context of this discussion, quantifies the probability that a particular statistical test
and sample size will indicate that the mean or median loading has increased over the baseline,
given that it truly has increased some specified amount (see Table 4. la). The test is designed to
guard against incorrectly concluding that the mean or median has increased by setting alpha at a
low value. The probability is less than or equal to a that a statistical test and sample size will
incorrectly indicate that the mean or median loading has increased over the baseline, given that it
has not increased. If there has been a decrease in loadings, the risk of such an incorrect decision
will be considerably less than alpha.
EPA evaluated the power of the statistical triggers by simulating a 60-month monitoring program
for 5000 discharges, and recording the frequency with which the triggers indicated that the
remining loadings exceeded baseline (see Section 3.1). The evaluations led to a choice of
statistical procedures that achieve acceptable power and a reasonable balance between rates of
false alarms and correct alarms.
The error rates of statistical decision procedures will depend upon the number of measurements
("sample size") used. If the false positive rate (alpha) is held constant, the power (the ability to
detect an increase in pollutant load) will necessarily decrease as sample size decreases. EPA's
evaluation assumed monthly sampling, using twelve samples taken over one year to characterize
the baseline level, and using twelve samples taken over each year to monitor pollutant loads
during remining. The performance of the evaluated statistical procedures was shown to be just
adequate to meet the detailed objectives set out in Section 3.1 (see also U.S.E.P.A., 2001c) when
based upon measurements taken once a month. Therefore, if these procedures are applied to
measurements taken less frequently than once a month, or are applied to fewer than twelve
Baseline Sampling Duration and Frequency
-------
ilRe
Statistical Support Document
measurements per year (for annual triggers), the ability to detect an increase in pollutant load will
necessarily be lower than intended.
4.2 Necessary Duration and Frequency of Sampling
Without an adequate duration and frequency of sampling, the statistical procedures could
establish baseline levels that are either too low or too high. Baseline sample collection
requirements protect both the renaming operator and the environment. If baseline
characterization of pre-existing pollutant discharges is inadequate (for example, if it is based on
too few samples), there is a chance that an operator could consistently face noncompliance by
discharging pollutant loadings above an underestimated baseline. In addition, there is the chance
that environmental improvement could be jeopardized by allowing for pollutant loading
discharges at high levels that still fall below an overestimated baseline. EPA believes that 12
monthly samples are the minimum to derive a statistically sound estimate of baseline (U.S.E.P.A.
2001b).
EPA has determined that the smallest acceptable number and frequency of samples is 12 monthly
samples, taken consecutively over the course of one year. Twelve samples may provide less than
the required power if autocorrelation is very high, if sampling duration is less than a year, or if
the sampling interval is shortened (e.g., to one week) while the number of samples is not
increased above 12. Therefore, EPA has required a minimum of 12 monthly samples to establish
baseline.
One of the criteria for sample size is the ability to detect a change of one standard deviation
above baseline loadings with reasonably high power. Discharge flows, concentrations, and
loadings vary remarkably among monthly or weekly samples over the course of 1-4 years (Brady
et al., 1998; EPA, 2001b). Sample coefficients of variation (CV, the ratio of standard deviation
to mean) for iron loadings range from0.62 to 2.7 for 80% of discharges (U.S.E.P.A., 2001d). .
—— ——— Baseline Sampling Duration and Frequency
-------
Coal Remining Statistical Support Document
Sample CVs for manganese loadings ranged from 0.54 to 1.7 for 80% of discharges (U.S.E.P.A.,
2001d). The median CV is about 1.0, thus the standard deviation is as large as the mean.
Assuming that the CV remains constant at 1.0 from baseline to post-baseline, an increase of one
standard deviation above baseline means that the mean loading has doubled. Thus, it is
important to have a sampling frequency and duration that will permit the statistical procedures to
detect increases in loadings with high probability when the standard deviations increase.
A permitting authority may want to consider requiring more than twelve samples per year during
and after the baseline year in order to increase power and in order to provide a fair chance of
observing a representative sample of discharge flows and loadings.
It is possible that one year of sampling may not adequately characterize baseline pollutant levels,
because discharge flows can vary among years in response to inter-year variations in rainfall and
ground water flow. There is some risk that the particular year chosen to characterize baseline
flows and loadings will be a year of atypicaUy high or low flow or loadings. Permitting
authorities should be aware of this risk and may want to inform permittees of this risk in order to
encourage multi-year characterization of baseline. To design a procedure to evaluate inter-year
variations, EPA evaluated correlations between discharge flow and various parameters of
existing mine discharge data and indices for which data spanning over many years are available
to the public (i.e., Palmer Indices, Standardized Precipitation Index, Crop Moisture Index,
Surface Water Supply Index, and USGS Current and Historical Daily Streamflow). EPA
concluded that historical stream flow data from a USGS gage station associated with a discharge
could be used to test whether the given baseline year was significantly different from the
previous years. This would be done by comparing the mean stream flow for the baseline year to
the 2.5th and 97.5th percentiles of annual mean stream flows prior to the baseline year. If the
mean stream flow for the baseline year falls below the 2.5th percentile or above the 97.5th
percentile, the year may have unusually low or high flow, respectively. In such cases, it may be
best to continue baseline sampling for another year.
Baseline Sampling Duration and Frequency
4-3
-------
Coal Remitting Statistical Support Document
A sampling plan should be designed to prevent biased sampling. Sampling, during and after
baseline, should systematically cover all periods of the year during which substantially high or
low discharge flows can be expected. Unequal sampling of months could bias the baseline mean
or median toward high or low loadings by over sampling of high-flow or low-flow months.
However, unequal sampling of different time periods can be accounted for by using statistical
estimation procedures appropriate to stratified sampling. Stratified seasonal sampling, possibly
with unequal sampling of different time periods, is a suitable alternative to regular monthly
sampling, provided that correct statistical estimation procedures for stratified sampling are
applied to estimate the mean, median, variance, interquartile range, and other quantities used in
the statistical procedures, and provided that at least one sample is taken per month over the
course of one year.
How measurement methods also should accurately measure flows during high-flow events. If the
discharge overflows or bypasses the weir or flume, or if a measurement is not made as scheduled
on a high-flow day, statistical characterizations of flow and loading will be inaccurate. The
sampling location and methods should be designed as much as possible to permit access and
sampling on all scheduled days, and to avoid the need to reschedule sampling because flow is
extremely high.
References
Brady, K.B.C., R.J. Hornberger, and G. Reeger, 1998. Influence of Geology on Postmining
Water Quality: Northern Appalachian Basein. Chapter 8 in Coal Mine Drainage
Prediction abnd Pollution Prevention in Pennsylvania. Edited by K.B.C. Brady, M.W.
Smith, and J. Schueck, Department of Environmental Protection, pp. 8-1 to 8-92.
Griffiths, J.C., 1990. Letter to M. Smith, Pennsylvania Department of Environmental Resources
dated January 14, 1990, 3 pages, with attachment of 54 pages titled "Chapter 3, Statistical
Analysis of Mine Drainage Data".
Hawkins, J.W., 1994. Statistical Characteristics of Coal Mine Discharges on Western,
Pennsylvania Remining Sites. Water Resources Bulletin, Vol. 30, No. 5, pp. 861-869.
4.4 Baseline Sampling Duration and Frequency
-------
Coal Remining Statistical Support Document
McGill, R., J.W. Tukey, and W.A. Larsen, 1978. Variations of Box Plots. The American
'scientist, Vol. 32, No. 1, pp. 12-16.
Mfflard, S.P., 1998. Environmental Stats for S-Plus: Users Manual for Windows and Unix.
Springer-Verlag: New York, NY.
Sanders T.G., R.C. Ward, J.C. Loftis, T.D. Steele, D.D. Adrian, and V. Yevjevich, 1983. Design
of Networks for Monitoring Water Quality. Water Resources Publications, Littleton, CO.
U.S. EPA, 2001a. Coal Remining Best Management Practices Guidance Manual.
EPA-821-B-01-010.
U.S.E.P. A., 200 Ib. Statistical Analysis of Abandoned Mine Drainage in the Assessment of
Pollution Load. EPA-821-B-01-014.
U.S.E.P.A., 2001c. "Evaluation of Statistical Triggers," memorandum in the rulemaking record
(docket number DCN 3051).
U.S.E.P. A., 2001d. "Distribution & Variability of Coal Mine Discharge Loadings,"
memorandum in the rulemaking record (docket number DCN 3049).
Baseline Sampling Duration and Frequency
-------
Coal Remining Statistical Support Document
4-6
Baseline Sampling Duration and Frequency
-------
Coal Remining Statistical Support Document
Section 5.0 Long-term Monitoring Data and Case Studies
A remitting study conducted by EPA and PA DEP from 1984 through 1988 involved the
statistical analysis of long-term abandoned mine discharge data from six sites in Pennsylvania.
This study is described in Section 1.0 of this document. These sites and corresponding discharge
data were selected because they contained a sufficient number of samples for examining mine
drainage discharge behavior with univariate, bivariate, and time series statistical analyses
following the algorithm shown in Figure 1.2a. The results of the statistical analyses are included
in a series of eight unpublished reports prepared for EPA and PA DEP by Dr. J. C. Griffiths of
the Pennsylvania State University. These reports are discussed in EPA's Statistical Analysis of
Abandoned Mine Drainage in the Establishment of the Baseline Pollution Load for Coal
Remining Permits (USEPA, 2001; EPA-821-B-01-014).
Sections 3.0 and 4.0 of this document describe the statistical methodology for establishing
baseline for pre-existing discharges, and determined that the minimum baseline sampling
duration and frequency is twelve samples in one year at approximately monthly intervals. Some
discharge datasets in Pennsylvania contain more than twelve samples. These additional samples
represent pre-mining baseline conditions of more than one year, and in some cases, discharges
were monitored more frequently than monthly (e.g., weekly). In this Section, data from seven
discharges at six sites previously studied by Griffiths are further examined by varying the
baseline sampling interval and the number of samples used to establish baseline.
A benefit of further evaluation of the EPA/PA DEP remining study is that for some of the six
sites, there are now approximately ten years of additional monitoring data. In addition, PA DEP
has issued approximately three hundred remining permits since 1985, and for many completed
sites there are complete historical records of discharge data from pre-mining baseline conditions,
through active surface mining phases (open pit), to post-mining reclamation. Several examples
of these long-term monitoring case studies are presented in this Section. These studies provide
Long-term Monitoring Data and Case Studies
-------
ilRe
Statistical Support Document
additional information on the magnitude and variability of natural seasonal variations, and show
mining-induced changes in water quality and pollution load.
The long-term monitoring case studies in this Section can be used as examples of the application
of the baseline statistical test described in Section 3.0 to actual remining datasets. The quality
control approach to long-term baseline monitoring data is presented in the figures included in
Section 5.3. Further examples of application of the baseline statistical test are presented in
Appendix A.
5.1 A Comparison of Seven Long-term Water Quality Datasets
As part of the investigation documenting the baseline pollution load of pre-existing acid mine
drainage (AMD) discharges, seven individual discharges with long-term water quality records
were studied. Each of these seven discharges had datasets of at least 3 years duration, and were
sampled at least monthly and as frequently as weekly. The seven discharges represent the three
principal discharge behavior types (typical, slug, steady) discussed in Section 2.0. Table 5. la
lists the discharge behavior type, location, period of record, and number of samples for each of
the seven long-term discharges evaluated.
Table S.la: Long Term Acid Mine Drainage Datasets
Dataset
Arnot-3
Arnot-4
Clarion
Ernest
Fisher
Hamilton
Markson
Discharge
Behavior Type
typical
typical
typical
slug
typical
typical
steady
Location
Tioga County, PA
Tioga County, PA
Clarion County, PA
Indiana County, PA
Lycoming County, PA
Centre County, PA
Schuylkffl County, PA
Period of
Record
1980-1983
1980 - 1983
1982 - 1986
1981 - 1984
1982 - 1985
1981 - 1985
1984 - 1986
Number of
Samples
82
81
79
189
36
109
99
5-2
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
The discharge behaviors discussed in Section 2.0 are summarized as:
1)
2)
Typical Discharge Response: Typical discharge response exhibits lower pollutant
concentrations during high-flow periods and higher concentrations during low-flow
periods Most pre-existing discharges exhibit this type of behavior. These discharges
tend to vary significantly, both seasonally and in response to individual recharge events.
Five of the seven discharges listed in Table 5. la exhibit this characteristic flow-response
behavior (Arnot-3, Arnot-4, Clarion, Hamilton, and Fisher). All but the Clarion
discharge are from relatively small (less than one square mile) underground mine
complexes. The Clarion discharge emanated from a previously surface mined area.
Slug Response: The Ernest discharge emanates from an extensive unreclaimed coal
refuse pile and exhibits highly variable behavior responding to individual precipitation
events. It exhibits "slugger response" behavior, Increases in flow are not necessarily
offset by decreased concentration and at times may even exhibit increased concentration
due to the build-up of water-soluble acid salts in the unsaturated zone during periods of
decreased precipitation or little recharge. These discharge types are extremely variable in
flow and pollutant loading rates. They present the biggest challenge for accurate
documentation of baseline pollution load.
Steady response: The Markson discharge illustrates steady response behavior typical of
discharges from very large underground mine pools. These discharges vary seasonally,
but because of their large ground water storage capacity, respond in a damped fashion and
do not exhibit large changes in pollutant concentrations. These types of discharges are
the least variable in terms of baseline pollution load. However, because loading rates
change slowly, they are also the most susceptible to year-to-year variation in pollution
load.
Baseline pollution load statistical summaries were calculated for each dataset using the
exploratory data analysis approach discussed in Section3.0. .It is rare, however, that datasets of
Long-term Monitoring Data and Case Studies
3)
-------
Coal Remining Statistical Support Document
this duration and with as great a number of samples are available. Coal remining operations tend
to be relatively small and run by small mining companies. The time available for a small coal
operator to lease a reminable reserve, gather permit application information, obtain a permit, and
actually mine and reclaim a site is frequently very short, making long-term baseline sampling
periods infeasible. Moreover, because these operations tend to be economically marginal, large
sample sizes with frequent sampling intervals can be cost-prohibitive. In view of these
constraints, the primary concern with establishing a valid baseline is to determine the minimum
sampling period and sampling interval which will yield statistically valid results.
The problem of determining the minimum number of samples and minimum sampling period
that would yield statistically valid results was examined using the long-term datasets listed in
Table 5. la. It was first assumed that the baseline pollution load determined using all of the
available samples over the entire period of record represents the most accurate baseline
achievable. Reduced datasets (subsets) were then used to recalculate the baseline, and
comparisons to the full dataset were made using the following data subsets: monthly sample
collection, quarterly sample collection, and nine-month sample collection (February through
October, excluding November, December, and January). The nine-month sample collection
subset was used to test the possibility that excluding three months (typically November,,
December, and January are average flow months) could adequately represent the full water year.
This comparison is presented in Table 5. Ib. In addition, baselines were calculated for each full
calendar year (full data baselines) to examine the extent of year-to-year variability in baseline
pollution load (Table 5. Ic).
For simplicity, this evaluation looks at net acidity (the principal parameter of concern and
indicator of pH) and iron (the most prevalent metal present in AMD). Tables 5. Ib and 5. Ic list
median loads and calculated approximate 95 percent confidence intervals (C.I.) around each
median load. Assuming that the full data baseline best represents the true population median
load, the percent error for each data subset is calculated as the difference between the full data
baseline value and the subset baseline value, divided by the full data baseline value. Percent
errors are presented in Table 5. Ib. While no particular percent error is considered to be
5-4
Long-term Monitoring Data and Case Studies
-------
Coal Remitting Statistical Support Document
acceptable or unacceptable, these percentages are useful for examining which subsets provide the
closest approximations of the full data baseline. Percent errors less than 10 are highlighted.
Table 5. Ic presents the median baseline pollution loads and 95 percent confidence intervals for
each of the seven discharges studied. Years that do not show overlapping 95 percent confidence
intervals are considered to be statistically different at the 95 percent significance level.
Table S.lb: Comparison of Median Acidity and Iron Loads by Sample Period
and
U11C1V*
Parameter 1
Arnot-3
Number of Samples
Median Acid Load
Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% C.I.
Arnot-4
Number of Samples
Median Acid Load
Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% C.I.
Clarion
Number of Samples
Median Acid Load
Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% C.I.
Ernest
Number of Samples
Median Acid Load
Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% C.I.
Fisher
u
Pull Data
82
72.3
86.53
58.07
0.96
1.26
0.66
81
194
232 31
155.69
2.70
3 35
2.05
75
39.50
49.11
29 89
5.51
7 27
3.75
189
1456
1991.91
92009
229
342.83
115.17
9 Month 1
Data
66
84.1
101.59
66.61
1.17.
1.55
0.79
66
221
263.22
178.78
3.00
3.78
2.22
53
40.00
51.45
28.55
4.26
6.07
2.45
146
1682
2410.88
953.12
264
412.68
115.32
Percent j
Error |
16.32 %
21.88 %
13.92 %
11.11 %
S^fpNlte.
-22.69 %
15.52 %
15.28%
Monthly
Samples
43
73.9
91.09
56.71
0.95
1.27
0.63
43
193
233.41
152.59
2.50
3.22
1.78
41
39.00
52.01
25.99
4.45
7.02
1.88
53
2048
2923.35
1172.65
304
474.61
133.39
Percent
Error
2-§i •w* •>
:!::S$yl!S:lfe$
„ ~&$% *$£•
ssss&H-fesifes
v ~l-W$? *&•:
-19.24 %
40.66 %
32.75 %
Quarterly
Samples
14
72.3
102.71
41.89
0.96
1.46
0.46
14
185
248.05
121.95
2.60
3.62
1.58
40.00
54.74
25.26
7.37
10.51
4.23
19
1882
3805.81
-41.81
348
662.48
33.52
Percent
Error
" &«y?L.M
V"4$Mfe;
**"&«*"$??>.
%c 1 £?%,;•;:;
33.76 %
29.26 %
51.97 %
Long-term Monitoring Data and Case Studies
5-5
-------
Coal Remining Statistical Support Document
Parameter
Number of Samples
Median Acid Load
Upper 95% C.I.
Lower 95% C.I.
Median ton Load
Upper 95% C.I.
Lower 95% C.I.
Full Data
35
'72
95.00
49.00
1.4
1.74
1.06
9 Month
Data
24
82
109.47
54.53
1.4
1.75
1.05
Percent
Error
13.89 %
CM&1& .
Monthly
Samples
24
85
121.73
48.27
1.4
1.66
1.14
Percent
Error
18.06 %
&00"$k ,
THTpmiUnn-R
Number of Samples
Median Acid Load
Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% C.I.
109
59.00
67.92
50.80
2.66
3.19
2.13
85
66.86
77.16
56.56
3.12
3.76
2.48
13.32 %
17.29 %
52
58.70
68.90
48.50
2.63
3.45
1.81
liiillilli
•ji-& <$
Markson
Number of Samples
Median Acid Load
Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% C.I.
98
1467
1575.47
1358.53
408
430.76
385.24
77
1452
1597.55
1306.45
402
428.18
375.82
• *stM'%
••&&}&"
30
1491
1624.02
1357.98
402
428.14
375.56
TL$4<®\
^M?&
A yergprp nf All Discharges
Median Acid Load
Median Iron Load
10.75 %
12.82%
! &£%.%*
^9&l%!"
Quarterly
Samples
10
102
165.38
38.62
1.4
1.63
1.17
38
55.70
72.60
38.30
1.81
2.77
0.85
22
1546
1816.11
1275.89
402
434.13
369.87
Percent
Error
41.67 %
C.tM>8#>
« *5ji& *i$
-31.95 %
.^f^-,-"
r-1^7%
12.55 %
17.55 %
5-6
Long-term Monitoring Data and Case Studies
-------
Coal .Remining Statistical Support Document
Table S.lc: Comparison of Median Acidity and Iron Loads by Baseline Sampling Year
Median Acid Load
Upper 95% C.I.
Fisher
Number of Samples
Median. Acid Load
Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% Cj.
Hamilton-8
—.„ —
Number of Samples
1981 |
21
63.8
76.79
46.62
0.98
1.37
0.59
20
159
209.32
108.68
1.6
1.76
1.44
20
56.5
75.80
37.20
10.78
14.49
7.07
38
615
1141.38
88.62
85
169.49
0.51
8
101
202.9C
-0.9C
2.1
3.6f
0.5^
2'
69.1
1982
27
83.9
110.69
43.72
1.44
2.04
0.84
29
208
256.19
159.81
3.0
4.11
1.89
11
27.0
46.41
7.59
7.65
18.00
-2.70
47
574
1295.70
-147.70
60
142.37
-22.37
11
8C
) 119.3S
) 40.6]
i i.:
5 1.5<
5 0.8
*l 2
3| 37.6
1983
15
86.5
128.53
39.51
1.17
2.12
0.22
15
242
368.52
115.48
3.0
5.20
0.80
16
14.0
27.99
0.01
1.66
3.13
0.19
49
5193
6551.26
3834.74
1069
1346.84
791.16
21
) 36
) 45.26
L 26.74
I O.S
3 1.11
L 0.6J
7j ' 2i
0| 54.5'
1984
9
42.0
61.26
22.74
5.69
9.68
1.70
39
1697
2906.26
487.74
216
448.62
-16.62
12
26
, 41.2C
1. 10.8C
) 0.^
> 0.41
5 -0.0
5 V
\ 77.4<
1985
8
42
) 60.10
) 23.90
> 0.2
I 0.36
L 0.04
7
D
Long-term Monitoring Data and Case Studies
5-7
-------
Coal Remitting Statistical Support Document
Parameter
Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% C.I.
Full Data
67.92
50.80
2.66
3.19
2.13
1980
79.01
34.39
4.35
6.74
1.96
1981
87.02
51.18
3.50
4.98
2.02
1982
60.69
14.51
1.07
1.72
0.42
1983
71.41
37.67
1.53
2.16
0.90
1984
91.27
63.53
3.34
4.21
2.47
1985
Markson
Number of Samples
Median Acid Load
Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% C.I.
98
1467
1575.47
1358.53
408
430.76
385.24
15
1502
1726.37
1277.63
336
406.26
265.74
49
1327
1445.08
1208.92
403
423.90
382.10
34
1888
2366.21
1409.79
449
512.73
385.27
5.1.1 Sampling Interval
For net acidity loads, two (Ernest and Fisher) out of the seven discharge datasets exceeded 10
percent error when both monthly and quarterly sample intervals were used. The average error for
net acidity load using monthly sample collection was 9.27 percent. The average error for net
acidity load using quarterly samples was 12.55 percent. For iron loads, 10 percent error was
exceeded for monthly sampling on the Clarion and Ernest discharges. Ten percent error was
exceeded with quarterly sampling for the Clarion, Ernest, and Hamilton discharges. The average
error for iron load was 9.01 percent for monthly samples and 17.55 percent for quarterly samples.
Monthly sampling yielded results closer to the full baseline than quarterly sampling. The effect
of quarterly sampling would likely be even more pronounced if a shorter sampling period (e.g.,
one year) had been used.
The difference in baselines calculated for each discharge using monthly samples versus quarterly
samples is illustrated in Figures 5. la through 5. lab. These figures also present yearly
comparison of baseline pollutant loadings. In the data comparison (monthly versus quarterly)
figures, the short horizontal lines represent the median values. The parallel vertical lines
represent the range of the 95 percent confidence intervals around the median. The left-hand line
shows the 95 percent confidence interval calculated based on the actual number of samples (N)
as listed in Table 5. Ib. However, because each sample subset contains a different number of.
5-8
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
samples, the confidence intervals are affected by different N values. A smaller number of
samples results in a wider confidence interval. For purposes of comparison between datasets, the
right-hand line shows the 95 percent confidence interval based on an arbitrarily set value for N
equal to 12.
Long-term Monitoring Data and Case Studies
5-9
-------
il Remining Statistical Support Document
Figure S.la: Arnot-3 Acidity Loading (1980-1981)
140-a
120:
100:
60:
40
•20
i.'YEARLY GdMPAftlSQt-i
1980 1931 19S2 1883
Figure S.lb: Arnot-3 Flow Data Comparison
2£H
5-10
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure 5.1c: Arnot-3 Iron Load Data Comparison
• 3ti ARNOT. 3 IRON LOAD: DATA- TYPE COMPARISON
"*•-.
&
3 ,i
Q
(£.
NINE MONTH
4UARTERW
Figure 5.1d: Arnot-3 Acidity Load Data Comparison
ARNQT 3 ACID LOAD :EAtA- TY^E
140-a
100 1
:? \
T» :
Q -• E
g '60^
9 i
-» !
-.
•
,
'-
ALL mm. MONTH ' MONTHLt; QUARTERLY
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure 5.1e: Arnot-3 Monthly Flow Comparison
250 q
•too-
•50
ARN'OT 3 EWtf MONTHLY COMPARISON
•JAN "FEB MAR APR -iv(At .d'UN- jut, • AU0 SEP OGT ..tw
Figure S.lf: Arnot-3 Monthly Acidity Load Comparison
J
Q
5-12
Long-term Monitoring Data and Case Studies
-------
Coal Remitting Statistical Support Document
Figure 5.1g: Arnot-4 Acidity Load Data Comparison
*RNCff 4. ACID LOAD DATA TYPE COMPARISONS:
1
L
NINE MONTH MONTHLY
Figure 5.1h: Arnot-4 Acidity Load (1980-1983)
•400:
300-j
-§ 2.QQ-
cr .-
0
. —.isti
-2SQ
ARNPT .4 AG1D UOAD- I960 - 1SS3 ;
1.9:80 • 1S«.1 •• '«*« "^
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure S.li: Clarion Acidity Load Data Comparison
CtARIQN ACID.LOAD DATATYPE COMPARISON
o_
s-
•o
ALL -NINE MONTH MONTHLY . QUARTERLY
Figure 5.1j: Clarion Iron Load Data Comparison
CLARION IRON .LOAD DATA TYPE COMPARISON
10-
I *
o
CE
•ALL NIME '.MONTH ' 'MGNtHLY' QUARTERLY.
5-14
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure 5.1k: Clarion Acidity Load (1982-1986)
H ACIO LOAD YIAKLT'COMPARISON
•SO-:
'
1932
1'9SS
Figure 5.11: Clarion Iron Load (1980-1983)
S-,
I
M IRON LOAD YEARLY 'COMPARISON
!••
Long-term Monitoring Data and Case Studies
5-15
-------
Coal Remining Statistical Support Document
Kgure 5.1m: Ernest Acidity Load Data Comparison
6600-a
4000-
>20QOr
ERNST AGIO, U<3Al> DATA lYRE
-2000-
ALL NINE
BlWEEKCY MONTHLY QUARTER!.? ,
Figure S.ln: Ernest Iron Load Data Comparison
ERNST 1ROM L©AD DATA TYPE COMPARISON
1200-]
H S°°c
co ;
1 =
•2. :
••
•
••
ALL MII-E
:-
,t
-!T
HQNTH BIWEEKliy MONTHLY QUARTERLY
5-16
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure S.lo: Ernest Acidity Load Data Comparison
1200C-3
I
<4DQ0-3
-40.00
T ACID LOAD; IMQ^THLY 'esi
JAfct FEB- WAR APR MAY JUN JUL AUQ- ' SEP . OCt' M0V ' DEC
Figure S.lp: Ernest Acidity Load (1981-1985)
ERNSlt A01& .LOAD YEARLY
,6000
. 50QO
4000
*3,
•a
.or 3®Q0
§5006
o
.U
:;
^
-
;
;
~
-•-
''"'•£
':-
-jsnj^^>-""
—
,•':•*'
' :
Long-term Monitoring Data and Case Studies
5-17
-------
Coal Remining Statistical Support Document
Figure S.lq: Fisher Monthly Acidity Load
FISHER ACID LOM3 DATA TYPE COMPARISONS
o-
3 B-
o_
10.
BEFORE AFTER NINE MPNTH MONTHLY QUARTERLY
Figure S.lr: Fisher Iron Load Data Comparison
IT>^
O ;
o """:
£ •-
*•»•• •
•
rra.t-
!
itK iKyi:-t. -uw
-
\w
-
J*»ir* I iF*- A'VWrw^i'Owra^
..'
BEFORE AFTER NINE iMQNTH MOI^HULY CjyARTERLX
5-18
-------
Coal Remining Statistical Support Document
Figure 5.1s: Fisher Acidity Load Data (1982-1987)
WQ-
-50-
FlSHER ACttJ. 'EMQ- 1982 -
1982: 1983 1984 1985 1:?!S§ 3987
Figure S.lt: Fisher Iron Load Data (1982-1987)
*-
I 1
o-
FISHER- IRON LOAD YEARLY COMPARISONS
1982 10S5 t§84
19B6 1987
Long-term Monitoring Data and Case Studies
5-19
-------
Coal Remining Statistical Support Document
Figure S.lu: Hamilton-8 Acidity Load Data Comparison
HAMILTON 3 AGIO LOAD DATA TYPE COMPARISON
f
in
ALL NINE: MONTH MONTHLY
Figure S.lv: Hamilton-8 Iron Load Data Comparison
HAMILTON 8 IRON LOAD DATA,TYPE COMPARISON
6;=
r*. *"S
w 4
' I *J
§ 1f
• -2-1 —
••
ALL ' NIKE
-
-•
'
--
-
-
W#m M0WHLY • aUARTERLY
5-20
-------
Coal Remining Statistical Support Document
Figure S.lw: Hamilton-8 Acidity Load (1981-1985)
80-
'^5*
•a
§
'49-
•20-
.g. ACID LOAD YEARLY
19S1. 1SS2 1S83 • 1984 1385
Figure S.lx: Hamilton-8 Iron Load (1981-1985)
1
1
HAMILTON & IROM \X)W YEARLY
1981
1S82
1983
1985
Long-term Monitoring Data and Case Studies
5-21
-------
Coal Remining Statistical Support Document
Figure S.ly: Markson Acidity Load (1984-1986)
26'n MARKSON AG!D LOAD YEARLY COMPARISON
,2400-:
,2200-j
.g.2000
J. 1800
3 160.0
14QO
1200
100R
aoo
1986;
Figure S.lz: Markson Iron Load (1984-1986)
50'0't
4SD*
4QO-.-E
.
Q
' — 30.01
.•250'-
: -LOAD. YEARLY CO.WPARIS.(SN
r
•H
•1984.
rase
5-22
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure S.laa: Acidity Load Data Comparison
:MARK$ON AGIO -LOAD DATA TYPE COMPARISON:
200:0-3 •••:'• ' . ••
•T2Q0'-r
*
10001
• BOO I
6.QQ-
ALL NINE MONTH MONTHLY .QUARTERLY
Figure S.lab: Markson Iron Load Data Comparison
ARKSPW IRON. LOAD DATA TYP£ CiOMPARISON, •
500-.
4SO-
jf 40.0-;
ALL - NINE MONTH WONTHLY QUARTERLY
Long-term Monitoring Data and Case Studies
5-23
-------
Coal Retaining Statistical Support Document
5.1.2 Duration of Baseline Sampling
Previous study of these datasets (Griffiths, no date; 1987a-e; 1988a, b) observed that the months
of November, December, and January typically exhibited behavior characteristic of median
values and that extreme high and low flows and low and high concentrations were represented by
the late winter/early spring and late summer/early fall months, respectively. This indicates the
possibility that it may be acceptable to limit sample collection to a nine-month period that
excludes the months of November, December, and January. To test this hypothesis, the long-
term datasets were subsetted by eliminating all of the data from these three months, recalculating
the baselines, and comparing the baseline median values for the full dataset and the nine-month
subset. The results are shown in Table 5. Ib.
Again using a 10 percent error criterion as a threshold for comparison, only two of the seven
datasets (Clarion and Markson) showed less than 10 percent error in baseline median net acidity
loads. Baseline iron loads showed similar results, with only two datasets (Fisher and Markson)
showing less than 10 percent error. The average error for median acidity load was 10.75 percent.
The average error for median iron load was 12.82 percent. The source of this error may be
because even though the three excluded months typically have average flows, the median yearly
flow may be greatly over or under estimated by excluding these months. For example, the
median flow of the Arnot-3 discharge (Figure 5.1b) is much higher when using the nine-month
data than with using the full 12-month dataset. Acidity loading rates, which are dominated by
flows, parallel this effect (Figure 5. Id).
Based on this analysis, exclusion of the months of November, December, and January (in
Pennsylvania and for areas with similar climates) poses a significant risk of not being
representative of the entire water year and skewing the baseline loading rates, either higher or
lower. Similarly, a sampling period of less than a full water year should be applied very
cautiously before the results can be relied upon to develop a representative and statistically valid
baseline.
5-24
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
5.1.3 Effects of Discharge Behavior on Baseline Sampling
Five of the seven discharges studied represent typical discharge behavior. These discharges
exhibited relatively large seasonal fluctuations in flow rates with pollutant concentrations
inversely proportional to flow. However, because changes in flow tend to be much greater than
corresponding changes in concentration, flow tends to be the dominant factor in determining
pollutant loading in these discharges. The result is a flow-dominated system with pollution
loading rates that tend to closely fallow the flow rate, although perhaps in a damped manner.
This typical behavior is illustrated by the monthly flow and loading data from the Arnot-3
discharge (Figures 5.1 e and 5. If).
The remaining two discharges, Ernest and Markson, reacted very differently to changes in the
baseline sampling period and interval. The Ernest discharge (a "slug response" discharge),
yielded large percent errors for virtually every data subset. This discharge varied greatly in flow
rate, concentration, and load, and responded very quickly to recharge events. These variations
make representative monitoring very difficult. A baseline monitoring sampling interval that is
too long (e.g., greater than monthly), can easily cause extreme events to be missed, or can over-
represent extreme events if one happens to be sampled. Therefore, where this type of discharge
behavior is evident, it would be prudent-to use a shorter sampling interval (e.g., at least monthly)
and/or expand the baseline sampling period.
The Markson discharge was the least affected by increasing the sample interval or using only
nine months of data. Percent errors were relatively low regardless of the data subset used. This
suggests that for discharges with typical steady-response behavior, it may be possible to obtain a
suitable baseline using less frequent and possibly shorter sampling intervals. However,
examination of the data on a yearly-year basis (Table 5. Ic) indicates reason for caution. High
volume discharges with very large storage reservoirs may be the most vulnerable to slow, long-
term changes in flow caused by long-term or yearly variations in precipitation.
Long-term Monitoring Data and Case Studies 5-25
-------
Coal Re
Statistical Support Document
5.1.4 Year-to Year Variability
Annual median pollution loads and 95 percent confidence intervals for each of the seven
discharges studied are presented in Table 5. Ic. Although virtually all of the discharges showed
some variability in confidence intervals from year-to-year, most of this variability was not
statistically significant. There was only one discharge which exhibited statistically significant
differences in baseline loading. The Ernest discharge, which has "slug response" behavior, tends
to show extreme variability in both flow rate and load. As illustrated in Figure 5. lp, this was
particularly the case in 1984, when the median acidity load was in excess of 5,000 Ibs/day. This
median is in contrast to all other years which had median acidity loads less than 2,000 Ibs/day.
5.2 The Effects of Natural Seasonal Variations and Mining Induced
Changes in Long-term Monitoring Data
A primary reason for establishing a baseline pollution load prior to remining is to distinguish
between natural seasonal variations and mining-induced changes in flow and water quality that
may occur during remining and following reclamation. The reasons for using a sufficient number
of samples, an adequate duration of sampling, and an acceptable sampling interval for
establishing baseline pollution load are discussed throughout this document, and in EPA's
Statistical Analysis of Abandoned Mine Drainage in the Establishment of the Baseline Pollution
Load for Coal Remining Permits (USEPA, 2001; EPA-821-B-01-014).
The purpose of Section 5.2 is to: (1) depict the magnitude of natural seasonal variations of flow
and water quality in several large abandoned underground mine discharges that were closely
monitored for numerous years, and (2) provide examples of significant mining-induced changes
in baseline pollution load at remining sites in Pennsylvania. Abandoned underground mine
discharges (Markson, Tracy Airway, and Jeddo Tunnel) from the Pennsylvania Anthracite Coal
Region are used to demonstrate the magnitude and patterns of natural seasonal variations. These
5-26
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
discharges have been equipped with continuous flow recorders, and water quality analysis (at
monthly or lesser sampling intervals) is available.
The Markson discharge is located approximately 1.2 miles (2 km) upstream from the Rausch
Creek Treatment Plant operated by PA DEP and Schuylkill County (Figure 5.2a). This discharge
emanates from an airway of an abandoned colliery at an elevation of 865 feet, and is a principal
contributor to the acid load treated at the Rausch Creek Treatment Plant. The Tracy Airway
discharge from another abandoned colliery is located 5.1 miles (8.3 km) east of the Markson
discharge, and emanates from a mine-pool at an elevation of 1153 feet. The Tracy Airway
discharge accounts for the largest iron load of all mine drainage discharges within the Swatara
Creek watershed. The extensive data that were collected for both Markson and Tracy (Section
5.2.2) discharges is not typical of remining operations. These data were collected as the result of
interest in diverting the discharges to a nearby treatment facility.
The U.S. Geological Survey (USGS) operates several gauging stations within the Swatara Creek
Watershed as part of an EPA Section 319 National Monitoring Program (NMP) Project (the first
of these projects in the United States focused on coal mine drainage problems) in cooperation
with PA DEP, Schuylkffl County, and other cooperators. The USGS station at Ravine shown in
Figure 5.2a is the principal downstream gauge of the NMP project and is equipped with
continuous flow and water quality recorders. .The Markson discharge, Tracy discharge, and
Ravine Station are located within a 5 mile radius, and therefore, should have been subjected to
nearly equivalent amounts of precipitation, and duration and intensity of storm events during the
period of record.
Long-term Monitoring Data and Case Studies
5-27
-------
Coal Remining Statistical Support Document
Figure 5.2a: Mine Discharge Map
5-28
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
5.2.1 Markson Discharge
The Markson discharge is characterized as a "steady response" type of discharge, where flow rate
may vary seasonally, but changes in acidity concentrations or other water quality parameters are
minimal or damped (Smith, 1988; Hornberger et al., 1990; and Brady, 1998). In a 1988 study of
Markson data containing approximately 100 samples collected at weekly intervals from 1984 to
1986, Griffiths (1988a) found a lack of wide variation in all variables except flow, and found a
lack of any strong relationship between pairs of variables (e.g., flow and acidity) except for an
inverse correlation between iron and flow.
Monthly and annual variations in flow and concentrations of sulfate, acidity, pH, iron, and
manganese are shown for an eight year period (1992-1999) in Figure 5.2b. The data were plotted
on a logarithmic scale to demonstrate the range of variations in all of these variables on a single
plot. Large annual variations in flow are apparent and appear to be inversely related to variations
in sulfate and iron concentrations. Variations in acidity and manganese concentrations are more
subtle, and do not readily show a strong relationship to flow variations.
Figure 5.2c depicts the relationships between the same flow and sulfate concentration data for the
Markson discharge plotted on linear scales, while Figure 5.2d depicts the relationships between
the flow and acid concentration on linear scales. Both Figures 5.2c and 5.2d show a generally
strong inverse relationship between flow and pollutant concentration.
Long-term Monitoring Data and Case Studies
-------
Coal Remitting Statistical Support Document
10000
1000-
100
Figure 5.2b: Markson Time Plot
Flow, pH, Acidity, Iron, Manganese, Sulfate
Flow (gpm)
pH
Acidly (mg/l)
Iron (mg/l)
Manganese (mg/l)
•Su ate(mci/l|
Figure 5.2c: Markson Time Plot
Flow & Sulfate
5-30
Long-term Monitoring Data and Case Studies
-------
Coal Remitting Statistical Support Document
8000
. 5000-
1000
Figure 5.2d: Markson Time Plot
Flow & Acidity
120
Figure 5.2e shows the relationship between monthly flow measurements and iron concentration.
Figure 5.2f shows the corresponding relationships between flow and manganese concentration on
linear scales. Both of these figures indicate a general inverse relationship between flow and
metals concentrations in the Markson discharge. On a logarithmic scale (Figure 5.2b),
manganese concentration did not appear to vary substantially in response to flow variations. This
can be attributed to the relatively small range in manganese concentrations (2.2 to 8.9 mg/L) as
compared to the range in iron concentrations (5.6 to 30.2 mg/L).
Long-term Monitoring Data and Case Studies
5-31
-------
CoalRemining Statistical Support Document
Figure 5.2e: Markson Time Plot
Flow & Iron
8000 n
7000-
6000-
Figure 5.2f: Markson Time Plot
Flow & Manganese
&
I
5-32
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
In comparing the monthly measurements of flow for the eight year period shown in Figures 5.2b,
5.2c, and 5.2d, the following observations of annual variations can be made:
• There is a fairly regular annual pattern (the highest flow generally occurring in early to
mid-March and the lowest flow generally occurring in mid to late September).
• Additional yearly peaks may occur (e.g., January 1996 and September 1999).
• The highest recorded monthly flow within water years can vary significantly (from a low
of 3,500 gallons per minute in 1995 to a high of 7,500 gallons per minute in 1994).
• The lowest recorded monthly flow measurements are similar (ranging from 600 to 900
gallons per minute).
• The duration of high flow periods can vary substantially (e.g.> 1996 compared to 1995).
The flow measurements presented in Figures 5.2b, 5.2c, and 5.2d represent the instantaneous
flow recorded at the time monthly grab samples were collected for water quality analysis. Figure
5.2g shows the full range of continuous flow measurements for the three year period from March
1994 to March 1997, compared to the plot of the monthly data used in Figures 5.2b, 5.2c, and
5.2d. In compiling the continuous flow data, all of the continuous flow gauge recorder charts
were evaluated to best define the extremes and duration of high and low flow events.
Long-term Monitoring Data and Case Studies
5-33
-------
Coal Remining Statistical Support Document
I
10000
9000-
8000
7000
6000
5000
4000 J
3000
2000
1000
Figure 5.2(g): Markson Flow Measurement Comparison
Flow 1994-1997
09/03/94
03/05/95
09/03/95
03/04/96
09/02/96
03/04/97,
In comparing the continuous flow line to the instantaneous monthly flows, the following
observations can be made:
• Continuous flow measurements exhibit much more variability than instantaneous monthly
flow measurements.
• Monthly measurements missed some major storm events (July 1995, October 1995,
October 1996, and November 1996).
• Although the highest annual continuous flow measurement usually corresponded to the
same month as the highest annual monthly measurement, the differences between these
measurements were very large (3,500 to 6,700 for 1995; 5,300 to 8,100 for 1996; and
6,600 to 8,900 for 1997).
• Monthly flow measurement may have occurred somewhat after the peak of a high flow
event (February 1995) or somewhat before the peak (January 1996). This may explain
most of the variations mentioned in the previous item.
5.34 Long-term Monitoring Data and Case Studies
-------
Coal Remitting Statistical Support Document
• The differences between low flow events on. the continuous flow plot and on the
instantaneous monthly flow plots are relatively small for these three water years. This
may imply that it is probably not difficult to define low flow periods with monthly
samples.
5.2.2 Tracy Discharge
Monthly and annual variations in discharge flow and concentrations of sulfate, acidity, pH, iron,
and manganese in the Tracy Airway discharge for the eight year period for 1992 through 1999 are
shown in Figure 5.2h.
Figure 5.2h: Tracy Airway
Flow, pH, Iron, Manganese, Sulfate
100001
1000-
100-
Long-term Monitoring Data and Case Studies
5-35
-------
Coal Remining Statistical Support Document
The range and patterns of annual and long-term variations in flow and in concentrations of
sulfate, iron, and manganese concentrations are similar to those for the same variables for the
Markson discharge (Figure 5.2V). However, the water quality characteristics are fundamentally
different in terms of pH, acidity, and alkalinity. The pH of the Tracy discharge ranged between
5.7 and 6.5 during the eight year period, and generally had an alkalinity concentration exceeding
that of acidity. The pH of the Markson discharge ranged between 3.2 and 3.7 during the eight
year period, had no net alkalinity (i.e., its pH is less than the titration end point), and generally
had acidity concentrations around 100 mg/L. These distinct chemical differences in two
discharges, emanating from similar mines in the same geologic structure and coal seam (the
Donaldson Syncline), are attributable to stratification of large and deep anthracite minepools.
The Tracy discharge is a "top-water" discharge from a relatively shallow groundwater flow
system (at an elevation of 1153 feet), while the Markson discharge emanates from '"bottom
water" at a much lower elevation in the minepool (865 feet). The chemistry of stratified
anthracite mine-pools is described by Brady et al. (1998) and Barnes et al. (1964). However,
these discharges are similar in the relationship of flow and water quality to natural seasonal
variations. .
The monthly flow pattern of the Tracy discharge (Figure 5.2h) is very similar to that of the
Markson discharge (Figure 5.2b), except the Tracy discharge flows appear to be somewhat more
variable or peaked. When the annual patterns of high and low flows are compared, the Tracy
discharge has two flow peaks (November 1995 and October 1996) that do not occur for the
Markson discharge. These two peaks indicate storm events undetected by monthly sampling.
The plot of continuous and monthly flow records for the Tracy discharge (Figure 5.2i) reveals
that the Tracy discharge was sampled a short time before the November 1995 flow peak and well
after the flow peak for October 1996 (continuous flow peak equals 6700 gpm, monthly flow
equals 2700 gpm).
5-36
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
10000
9000-
Figure 5.2i: Tracy Airway
Row 1994-97
Monthly Row
-Continuous Flow
09/03/94
03/05/95
09/03/95
03/04/96
09/02/96
03/04/97
10000
Figure 5.2j: Tracy Airway
Flow & Sulfate
T 400
- • 350
50
Long-term Monitoring Data and Case Studies
5-37
-------
Coal Remining Statistical Support Document
Figure 5.2k: Tracy Airway
Flow & Iron
10000
In comparing continuous flow data (Figure 5.2i) to monthly flow data, most of the Markson
discharge results were also apparent in the Tracy discharge. While the monthly samples
correspond well during the first two major high flow events in 1994 (April and September),
several major storm events go undetected in the succeeding data (e.g., October and November
1994, November 1995, and November 1996). In addition, there are several major storm events
where the monthly sample was collected after (February, July, and August 1995, and October
1996) or before (January 1996) the peak recorded by continuous monitoring. The differences
between the monthly and continuous recorder data peaks are most significant in February 1995
(3700 and 6700 gpm), January 1996 (6500 and 9900 gpm), and October 1996 (2700 and 6700
gpm). One interesting characteristic of the Tracy flow data is that the continuous flow monitor
results for 1995 "bottoms out" several hundred gallons per minute below the monthly data, but
corresponds well with the low flow continuous recorder data for other water years.
5-38
Long-term Monitoring Data and Case Studies
-------
Coal Reminin? Statistical Support Document
Figure 5.21: Tracy Airway
Flow, pH, Manganese
A strong inverse relationship between flow and ponntan. concentration in the Tracy discharge .
shown in Figures 5.2J, 5.2k, and 5.21, with the highest flows corresponding to the lowest
concentrations and the lowest flows corresponding to the highest concentrations for sulfate
(Figure 5.2J), iron (Figure 5.2k), and manganese (Figure 5.21). Tnere appears to he a trend over
time, where the range of median values for sulfate and manganese are diminished from 1992
through 1997.
Monthly flow and water quanty relationships of the Markson and Tracy dfecharges, throughout
the eight year period shown in Figures 5.2b through 5.21, indicate a general inverse relanonshrp
between flow and concentration, but also show that the distribution, magnitude, and duration of
highfloweventsisnotuntformfromwateryeartowateryear. In fact, somettaes the highest
flow events appear during what is traditionally the low flow period of the water year (e.g.,
October 1996 and September 1999). These data suggest that a sampling interval length of not
Lang-term Monitoring Dam and Cm Studiel
-------
Coal Remining Statistical Support Document
greater than one month, and a sampling duration of at least a water year (12 monthly samples) are
necessary to document baseline flow and water quality variations, particularly if high flow events
are important in establishing the baseline pollution load. The monthly and continuous flow data
for the Markson and Tracy discharges (Figures 5.2g and 5.2i) show that representative sampling
of storm events can be tricky, as isolated sampling events may not always capture the range and
pattern of natural seasonal variations. This problem is illustrated in Figures 5.2g and 5.2i, where
monthly flow measurements indicate high flows that are much lower than those measured by the
continuous flow monitors, or where significant high-flow periods detected by continuous
monitoring were undetected by the monthly measurements.
5.2.3 Swatara Creek Monitoring Station
USGS has been sampling water quality and flow characteristics of Swatara Creek in Schuylkill
and Lebanon Counties, Pennsylvania since before 1960. The results of this data collection are
found in numerous publications including McCarren et al. (1961) and Fishel and Richardson
(1986). The USGS Ravine Station shown in Figure 5.2a has been a key station because it is
located on the main stem of Swatara Creek immediately below the confluence of several
tributaries draining the coal operations in the Swatara Creek headwaters. Below the Ravine
Station, the Swatara Creek watershed changes to a more agricultural land use without acid mine
drainage contributions to water quality. Figures 5.2m, 5.2n, and 5.2o contain a series of plots of
the storm-flow hydrograph and continuous measurements of specific conductance and pH for a
five day period in December 1996 (Cravotta, personal communication). These figures also show
water quality data for sulfate, suspended solids, and iron (total and dissolved) that were collected
by automatic samplers for flows resulting from this storm flow period. These figures indicate
that water quality data peaks for suspended solids and iron precede the peak for flow. According
to Cravotta (1999), the occurrence of these concentration peaks prior to peak flow are the result
of scour and transport of stream bed deposits.
5-40
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure 5.2m: Swatara Creek Flow and Sulfate Data
Figure 5.2n: Swatara Creek Flow and Suspended Solids Data
Long-term Monitoring Data and Case Studies
5-41
-------
Coal Remining Statistical Support Document
Figure 5.2o: Swatara Creek Flow and Iron Data
5.2.4 Jeddo Tunnel Discharge
The Jeddo Tunnel mine discharge near Hazleton Pennsylvania is the largest abandoned
underground mine discharge in the Eastern Middle Field of the Anthracite Region, and is among
the largest mine drainage discharges in Pennsylvania. The Jeddo Tunnel has a total drainage area
of 32.24 square miles, and its underground drainage system collects and discharges more than
half of the precipitation received in the drainage area (Balleron et al., 1999). The flow of this
discharge was monitored with a continuous recorder from December 1973 through September
1979 by the USGS in cooperation with Pennsylvania Department of Environmental Resources.
The results of that monitoring for the water year from October 1, 1974 through September 30,
1975 are shown in Figure 5.2p (Growitz et al., 1985). During that year, the discharge ranged
from36 to 230 cfs (16,157 to 103,224 gpm). The Jeddo Tunnel discharge flows are compared to
the stream-flow of Wapwallopen Creek (approximately 10 miles north of the Jeddo Tunnel).
5-42
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
The Wapwallopen Creek drains an area of 43.8 square miles and has a measured mean discharge
of 78 cfs (35,008 gpm) (Growitz et al., 1985). Growitz et al. found that the response of the Jeddo
Tunnel discharge to precipitation events is considerably less than that of the Wapwallopen Creek,
and that during large storm events, the Jeddo Tunnel data peaked later than the stream discharge.
Figure 5.2p: Jeddo Tunnel Discharge and Wapwallopen Creek Flow Data
OCT. . NOV DEC.
J974
JAN. . FEB. MAR. APR MAY JUNE JULY AUG. SEPT.
1975
The continuous flow recording station at the mouth of the Jeddo Tunnel was reconstructed and
operated by USGS from October 1995 through September 1998 in cooperation with PA DEP, the
Susquehanna River Basin Commission, US EPA, and other project cooperators. Figure 5.2q
(from BaUeron et al., 1999) shows variations in the flow of this discharge during this period. The
average annual discharge flow was 79.4 cfs (35,635 gpm) and the range of recorded flow
measurements was between 20 cfs (8,976 gpm) in October 1995 and 482 cfs (216,322 gpm) in
November 1996, following 3.89 inches of rainfall (BaUeron, 1999). Figure 5.2r shows a graph of
precipitation data from Hazleton Pennsylvania for the period from October 1995 through
September 1998. This graph was plotted from data contained in BaUeron (1999).
Long-term Monitoring Data and Case Studies
5-43
-------
Coal Remining Statistical Support Document
Figure 5.2q: Jeddo Tunnel Flow Data
500
5.0-
4.5-
4.0-
3.5'
| 3.0.
o
EC 2.0
1.5
1.0
0.5
0.0
Figure 5.2(r): Precipitation Data From Hazleton, Pa
I
5-44
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
5.3 Case Studies
Baseline pollution loading in pre-existing discharges must be measured and monitored accurately
to determine to what extent polluting conditions are affected by remining operations. The Fisher,
McWreath, and Trees Mills remining sites in the Pennsylvania Bituminous Coal Region are
presented as case studies in Section 5.3 to demonstrate significant changes in flow, water quality,
and pollution load resulting from remining and reclamation activities. The case studies also
demonstrate how a regular monitoring program can be used to evaluate and document both pre-
and post-remining water quality from a pre-existing discharge. In each of these cases, monitoring
was conducted at monthly intervals and proved to be adequate to document baseline conditions
and to demonstrate post-remining changes in water quality. These case studies also illustrate the
water quality and quantity changes that are typical of remining operations
5.3.1 Fisher Remining Site
The Fisher remining site is located in Lycoming County, Pennsylvania. Prior to remining, the
surface of the site was extensively disturbed by abandoned surface mine pits and spoil piles. A
large abandoned underground mine known as the Fisher deep mine occupied much of the sub-
surface. The principal discharge (monitoring point M-l) was the main concern during the
remining permit process. Baseline pollution load data collection took place between 1982 and
1985. The original remining permit was issued on November 5, 1985, and remining operations
commenced by February 1986. Final coal removal occurred in June 1995 and backfilling was
completed within the permit area by February 1996.
The best management practices employed on the Fisher remining site include: (1) daylighting the
abandoned Fisher deep mine, (2) regrading abandoned spoils and backfilling abandoned pits, (3)
alkaline addition, and (4) biosolids used for revegetation enhancement. Alkaline addition was
accomplished with 140,000 tons of limestone fines on the two most recently permitted
Long-term Monitoring Data and Case Studies
5-45
-------
Coal Remining Statistical Support Document
areas
;, resulting in an alkaline addition rate of approximately 400 tons per acre over an area of
approximately 350 acres. Biosolids were applied to approximately 500 acres.
The relationships between the sulfate, iron, and manganese concentrations, and flow in the M-l
discharge for the period between 1982 and 1999, are shown in Figure 5.3a. There are several
trends in the relationships resulting fromremining activities. While iron concentrations
decreased over time, sulfate and manganese concentrations increased. Discharge flow increased
following backfilling (1996), probably because this point became the down gradient drain for
greater than 500 acres of unconsolidated mine spoil aquifer materials. The most significant
change in pollutant concentration was in net acidity (Figure 5.3b). Prior to activation of the
remitting permit, the acidity concentration was typically in the range of 100 to 200 mg/L. The
effect of remaining was to turn a distinctly acidic discharge into one that is now distinctly alkaline
(i.e., post-mining net acidity concentrations of 0 through-75 mg/L).
5-46
Long-term Monitoring Data and Case Studies
-------
(l/Bui)
'(vudB)
00/91YSO
86/9I-/SO
98/61-/SO
178/61/80
- Z8/OS/SO
08/02/EO
LO
I
a
13
I
I
(l/Suj) asaueBuei/y 'uoj|
-------
Coal Remining Statistical Support Document
Figure 5.3b: Fisher Mining MP1
Net Acidity
250
200
150
100
-100
01/01/80 12/31/81 12/31/83 12/30/85 12/30/87 12/29/89 12/29/91 12/28/93 12/28/95 12/27/97 12/27/99
Figure 5.3c: Fisher Mining MP1
Acid Load
900
800- -™.—^-——•—£—•
01/01/80 12/31/81 12/31/83 12/30/85 12/30/87 12/29/89 12/29/91 12/28/93 12/28/95 12/27/97 12/27/99
5-48
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
The Fisher permit (and most remining permits in Pennsylvania) was written in a format that
evaluates remining performance on a pollution load basis rather than a concentration basis. As
part of the baseline pollution load computation, quality control limits are based upon the
approximate 95 percent tolerance limits around the frequency distribution, and the median is
used as the measure of central tendency of the frequency distribution (see Section 1.0, Table
1.2a). Acidity loading prior to permit activation (baseline establishment), during open pit
mining, and following backfilling and reclamation for the M-l discharge are shown in Figure
5.3c. The upper and lower of the three horizontal dashed lines correspond to the upper and lower
95 percent tolerance limits for the pre-mining baseline pollution load. The middle dashed line is
the baseline median acidity load (67.9 pounds per day). The median acidity load for the three
year period following backfilling (1996 through 1999) is 0 pounds per day, showing
improvement in water quality. Figure 5.3d shows corresponding improvement in the iron load
(pre-mining baseline median iron load was 1.36 pounds per day; median of the three years
following backfilling is 1.04 pounds per day). The differences in iron load and in net alkalinity
concentration during these time periods are presented in Figures 5.3e and 5.3f respectively.
Figure 5.3(d): Fisher Mining MP1
Iron Load
0°/'l/80 12/31/81 12/31/83 12/30/85 12/30/87 12/29/89 12/29/91 12/28/93 12/28/95 12/27/97 12/27/99
— :
Long-term Monitoring Data and Case Studies
5-49
-------
Coal Remining Statistical Support Document
Figure 5.3e: Iron Load Boxplot
10 -
* 5
o
0 -
Fisher M1 Discharge
Premining
During
-^3-
(
Postmining
Figure 5.3f: Net Alkalinity Boxplot
Fisher M1 Discharge
100 -i
a ° H
.&•
^ -100 -
-------
Coal Remining Statistical Support Document
5.3.2 McWreath Remining Site
The McWreath remining site is located in Robinson Township, Washington County,
Pennsylvania. The initial surface mining permit for this remining operation was issued on July
21, 1987 for 112.1 acres. The principal hest management practice in the remining plan was
daylighting of an abandoned underground mine. In this area of Washington County, the
overburden of the Pittsburgh Coal includes extensive calcareous strata which produce alkaline
mine drainage when disturbed. A similar daylighting example for pH changes at the Solar mine
of the Pittsburgh Coal seam, in Allegheny County, Pennsylvania is included in Brady, 1998. The
McWreath site had three pre-existing pollution discharges emanating from abandoned
underground mine workings prior to remining (monitoring points D-l, D-3, and D-4). The
remining operation mined through these discharge locations, and the effects on the flow and
water quality are shown in Figures 5.3g through 5.3j.
The flow and concentrations of net acidity, sulfate, iron, manganese, and aluminum in the D-l
discharge are shown in Figure 5.3g. This was the largest of the three deep mine discharges at the
McWreath site. This discharge had four sampling events during baseline data collection and
following permit issuance when the flow was between 35 and 40 gallons per minute. In April of
1990, the discharge dried up, only briefly reappearing as a 1.2 gallon per minute flow in
December 1990, and as a 10 gallon per minute flow in December 1992, According to monthly
monitoring data, the discharge has otherwise gone dry as a result of remining from 1990 to
present. Flow and concentrations of iron, manganese, acidity, sulfate, and aluminum in the D3
and D4 discharges are shown in Figures 5.3h through 5.3J.
Long-term Monitoring Data and Case Studies
-51
-------
§
o
CM
Q o
£ 8
To =
>
c
•• 2
D> —
o
3 <
LL.
1
U.
I:
to
O
O
co
o
o
CD
o
o
(I/Bui) AllPPV 19N
§o
o o
CM O O
T- T- CO
o
o
CO
8
o
o
CM
_L.
J_
_1_
o
I
o
o
CM
O---
te!
co
CM
I"-
1
CM
O
I
a
I
a
I
a
-------
Coal Remining Statistical Support Document
Figure 5.3H: McWreath 03
Row & Net Acidity
7/7/86
11/19/87 4/2/89 8/15/90 12/28/91 5/11/93 9/23/94
2/5/96
500
400
-500
6/19/97 11/1/98
Figure 5.3i: McWreath D3
Flow & Iron
20 T
7/7/86 . 11/19/87
4/2/89
8/15/90 12/28/91
5/11/93
9/23/94
2/5/96
6/19/97
11/1/98
Long-term Monitoring Data and Case Studies
5-53
-------
Coal Remining Statistical Support Document
Figure 5.3h shows a dramatic change in net acidity concentration which was in the range of 200
to 400 mg/L in 1986 and 1987, but since 1989, has dropped to predominantly less than 0 and as
low as -350 mg/L. Figure 5.3J shows a substantial reduction in iron concentration since 1990.
Most of the flow measurements from 1996 through 1999 in Figures 5.3h and 5.3i are less than 2
gallons per minute.
The D-4 discharge had a pre-mining flow intermediate between discharge D-3 and D-1, and
recent flow measurements for this discharge are several times higher than the D-3 discharge.
Figure 5.3J depicts a large change in net acidity concentration as a result of remining on the
McWreath site, where the net acidity prior to 1990 was always greater than 100 mg/L and as high
as 500 mg/L, and the net acidity since 1990 is always less than 100 mg/L and as low as -250
mg/L. Therefore, remining transformed two distinctly acidic discharges into distinctly alkaline
discharges through daylighting the abandoned deep mine and exposing naturally alkaline
overburden strata during remining and reclamation operations.
20 n
18
Figure 5.3j: McWreath D4
Flow & Net Acidity
8/11/87
12/23/88
5/7/90
9/19/91
1/31/93
6/15/94 10/28/95 3/11/97
7/24/98
600
-300
5-54
Long-term Monitoring Data and Case Studies
-------
Coal Remitting Statistical Support Document
5.3.3 Trees Mills Site
The Trees Mills remining site is situated in Salem Township, Westmoreland County,
Pennsylvania (Figure 5.3k). The remining permit boundary is shown in this figure as a bold line.
The surface mining permit for the 325 acre site was issued on May 25, 1990. Surface water
drainage from the permit area flows to Beaver Run to the West and Porter Run to the East.
Beaver Run is classified as a High Quality - Cold Water Fishery, and the Beaver Run Reservoir
(a public water supply impoundment for 100,000 people) is located less than 2500 feet
downstream from the Trees Mills remining site.
The primary best management practice in the pollution abatement plan for this site was the
daylighting of an abandoned underground mine on the Pittsburgh Coal seam. There were also
abandoned surface mine pits and highwalls that were regraded and reclaimed. As the result of
extensive mine subsidence overlying the abandoned underground mine, prior to remining, much
of the surface of the site resembled a waffle ground that promoted internal drainage to the
abandoned deep mine workings rather than overland surface runoff. The geochemical
characteristics of the overburden strata were more conducive to acidity production than alkalinity
production. Figure 5.31 (Brady et al., 1998) features drill hole data for this site. Geochemical
information listed on the left hand side of the bore holes in this figure represent percent sulfur;
information listed on the right hand side represent neutralization potential in CaCO3 equivalents.
Except for high sulfur shale strata immediately overlying the coal, the overburden strata are
characterized by a thick sandstone unit with several zones of relatively high sulfur. Only two
sandstone samples in OB-2 have appreciable neutralization potential. The overburden quality of
the Pittsburgh Coal at the Trees Mills site is much different (i.e., less calcareous strata, less
alkalinity production potential) than at the McWreath site. Hence, the success of the remining
pollution abatement plan for the Trees Mills site is more dependent on the hydrogeologic
characteristics than on the geochemical characteristics that were significant at the Fisher and
McWreath remining sites.
Long-term Monitoring Data and Case Studies
5OD
-------
Coal Remitting Statistical Support Document
Figure 5.3k: Trees Mills Site Map
^•^/^^^(K^^i^^
5-56
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure 5.31: Trees Mills Drill Hole Data
20 -I
10-
OB-l
- 6
_ 3
0 0
O.99
O.53
2.83
OB-2
2.OO
15
J 169
_|l22
Long-term Monitoring Data and Case Studies
5-57
-------
Coal Remining Statistical Support Document
There were numerous acid mine drainage discharges and seeps emanating from the abandoned
underground mine workings at the Trees Mills site prior to renaming, and baseline pollution load
statistical calculations were completed for ten of these monitoring points. The largest of these
pre-existing discharges was MP-1 with pre-mining flows as high as 139 gallons per minute
(Figure 5.3m and 5.3n). The effects of remining upon three other pre-existing discharges (MP-2,
MP-3, and MP-6) will also be discussed. Because, remining operations commenced on the Trees
Mills site on October 1991, water quality and flow data from 1987 through September 1991 can
be considered pre-mining data. According to mine inspection reports, backfilling was completed
by May 14, 1998, thus the intervening time from September 1991 through May 1998 includes the
phases of active open pit mining and reclamation activities.
Figure 5.3m shows the variations in flow and concentrations of net acidity, sulfate, iron,
manganese, and aluminum in the MP-1 deep mine discharge. The flow was highly variable prior
to the initiation of mining in October 1991, ranging from less than one gpm to 139 gpm, with a
median flow of 21.7 gpm and an average flow of 38.96 gpm. As the result of remining, the MP-1
discharge dried up by October 1992, reappearing in only one sampling event during the next
seven years (0.26 gpm flow on March 3, 1998). The range in acidity concentrations for the
period of July 1987 through October 1991 was 773 to 3,616 mg/L with a median of 1,336 mg/L
and an average of 1,417 mg/L. The corresponding range in iron concentrations for the MP-1
discharge was 104 to 430 mg/L, with a median of 211 mg/L and an average of 224 mg/L.
5-58
Long-term Monitoring Data and Case Studies
-------
(I/Bui)
'UQJI
o
q
o
I
1
a
66/90/21-
86/172/ZO
Z6/1- I/SO
S6/8Z/01
t6/9l-/90
LS6/IS/10
' 16/61760
06//0/90
- 89/S2/21-
Z8/U/80
98/62/SO
a
S3
R
I
'asaueBuB|/\i *(iudB)
-------
(Aep/sqi) peoi asaueBuei/y 'uoj|
u
o
Q
Q
I
"s
05
S?
•
a
a-
60
•S
(Aep/sq|) peoi p\o\f
-------
vo
ft/Bui)
'AliPPV iaN
=0
.f
.s
1
a
66/9/Sl-
96/8S/01-
• 88/SS/Sl-
• Z8/H-/8
98/6S/S
1
I
-------
Coal Remining Statistical Support Document
Figure 5.3n shows the variations in the pollution load of acidity, iron, and manganese prior to and
following permit activation. The three horizontal dashed lines represent the 95 percent tolerance
levels around the frequency distribution of acid loads and the median value of 439 pounds per
day acid load calculated in the baseline statistical analysis. The corresponding iron load was 71.8
pounds per day for the baseline sampling period. The median acid load during the period from
October 1991 through October 1992 (while the discharge was still flowing) was 128.5 pounds
per day and the corresponding iron load was 20.27 pounds per day. The remitting operation at
the Trees Mills site removed a significant acid load (439 pounds per day = 160,000 pounds per
year) and iron load (71.8 pounds per day = 26,000 pounds per year) from the Beaver Run
tributary and the Beaver Run public water supply reservoir.
Variations in flow and concentrations of acidity, sulfate, iron, manganese, and aluminum in the
MP-2 discharge are shown in Figure 5.3o. Corresponding variations in acidity, iron, and
manganese loads prior to permit activation, during mining, and following backfilling are shown
in Figure 5.3n. This discharge had substantially lower flow than the MP-1 discharge. The range
in flow prior to permit activation was 0.1 to 26.4 gpm (median of 1.3 gpm, average of 3.12 gpm).
During active mining, the flow of the MP-2 discharge ranged from 0.02 to 12.1 gpm (median of
1.75, average of 2.66 gpm), while the flow following backfilling ranged from 0.39 to 8.9 gpm
(median of 2.55, average of 2.97 gpm). Thus, while the range of flows decreased during mining
and post-mining, the median flow increased by approximately one gpm Net acidity, sulfate, and
iron concentrations increased following permit activation (Figure 5.3m). Aluminum
concentrations increased during mining but returned to pre-rnining levels following backfilling.
There also was a notable increase in manganese concentrations, from a median of 10.24 mg/L
pre-mining to a median of 171.2 mg/L following backfilling.
5-62
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
The overall environmental impacts of these water quality changes are put in perspective by
^mining the pollution load data for MP-2 (Figure 5.3n), in comparison to the pollution load
reduction achieved at the MP-1 location. The horizontal dashed lines represent the upper and
lower 95 percent tolerance levels and the median baseline acid load. Baseline (pre-mining) acid
load for MP-2 ranged from 0.31 to 197.6 Ibs/day (median of 8.55). Post-backfilling acid load
ranged from 3.89 to 54.95 Ibs/day. Hence, while extreme values were reduced, median acid load
increased by approximately 5 Ibs/day. The range in pre-mining iron loads was 0.02 to 13.9
Ibs/day (median of 0.26), while the post backfilling range was 0.3 to 3.36 Ibs/day (median of
0.46). The range of pre-mining manganese loads was 0.02 to 3.79 Ibs/day (median of 0.19),
while the post-backfilling range was 0.78 to 10.6 Ibs/day (median of 4.31). Based upon median
values, there was an increase of 4.31 Ibs/day of manganese from the MP-2 discharge, but an
elimination of 3.22 Ibs/day (average of 5.78 Ibs/day) from MP-1. There was a corresponding
increase of 0.2 Ibs/day iron load fromMP-2, with an elimination of 71.8 Ibs/day fromMP-1.
Finally there was an increase of approximately 4.6 Ibs/day acid load fromMP-2, offset by the
elimination of 439 Ibs/day fromMP-1.
The net effect on the Beaver Run receiving stream was a significant reduction in pollution loads.
Variations in concentrations of net acidity, sulfate, iron, manganese, and aluminum from MP-3
(Figure 5.3q) are similar to that from MP-2, except for a significant reduction in iron
concentration. Pre-mining iron concentrations in MP-3 ranged from7.9 to 226.4 mg/L (median
of 75), while the post-backfilling iron concentrations ranged from 6.55 to 84.72 mg/L (median
value of 29.77). Pre-mining median manganese concentration was 11.18 mg/L, and the post-
backfilling median was 194.55. Aluminum concentrations were 61.93 mg/L pre-mining and
63.38 mg/L post-backfilling. The flow of MP-3 ranged from 0.1 to 67 gpm pre-mining (median
of 6.95), while post-backfilling flow ranged from 1.5 to 21.7 gpm (median of 3.95). Variations
in acidity, iron, and manganese loads from MP-3 are shown in Figure 5.3r. Again, the horizontal
dashed lines represent the upper and lower 95 percent tolerance levels and median value for pre-
mining acid loads. The median baseline pollution load for acidity is 41.79 Ibs/day compared to a
post-backfilling median acid load of 34.79 Ibs/day. The corresponding medians for iron loads are
Long-term Monitoring Data and Case Studies
-------
(Aep/sqi) peon asaue6ue|/\| 'uoj|
CM O __
I
•£
c
,5
Cr.
&0
g
5
c
CM
Q.
J2 -3
r= o
2(0
0)
II
1-1
• • _
Q. C
CO O
0)
e
.5
I
(Aep/sqi) peon ppv
-------
vo'
(I/Bui) Aisprav V»N
u
a
IS
o
I
1
S
66/9/31-
1- 96/93/01-
88/S2/SI.
98/6S/S
K
I
'uoj| '(uidB) Moy
-------
Coal Remining Statistical Support Document
4.03 Ibs/day pre-mining and 1.95 following backfilling. The pre-mining median manganese load
was 0.7 Ibs/day, and increased to a median of7.97 Ibs/day post-backfilling. Extreme values of
iron loads were substantially reduced following backfilling.
The MP-6 discharge is located below the outcrop of the Pittsburgh Coal seam, and varied in pre-
mining flow from 0.8 to 62.4 gpm (median of 9.5). Following completion of backfilling, the
range in flow is 0.39 to 21.7 gpm (median of 3.8). The pre-mining range of acidity concentration
from the MP-6 discharge was 125 to 2,587 mg/L (median of 784.7), and the post-mining range
was 522 to 968 mg/L (median of 804.5). The range of the iron concentrations pre-mining was
11.54 to 161.0 mg/L (median of 74.7), while the post-mining range was 31.64 to 94.73 mg/L
(median of 55.62). The pre-mining range in manganese concentration was 6.62 to 19.24 mg/L
(median of 11.3), while the post-mining manganese range was 14.63 to 30.14 mg/L (median of
26.06 mg/L). Again, the horizontal dashed lines in Figure 5.3u represent the median acid load
and upper and lower 95 percent tolerance levels for baseline statistical calculations. The median
acid load was 73.13 Ibs/day pre-mining as compared to 38.08 Ibs/day post-mining. The
corresponding iron loads were a median of 7.5 pre-mining and 2.98 Ibs/day post-mining. The
pre-mining median load of manganese was 1.11 Ibs/day, and was nearly equalto the post-mining
median of 1.06 Ibs/day.
Due to the cumulative effects of remining upon the MP-1, MP-2, MP-3, and MP-6 discharges,
the Trees Mills remining operation has resulted in a significant reduction in the pollution load of
acidity and metals (iron, manganese, and aluminum) to the receiving stream and the Beaver Run
Reservoir. To determine whether these pollution reduction effects could be detected in the water
chemistry of the receiving stream, the permittee's self monitoring reports and PA DEP mining
inspector's monitoring data were evaluated from the same monitoring points located upstream
and downstream of the Trees Mills operation on the Porter Run and Beaver Run tributaries. The
downstream monitoring points are located immediately above the confluence of these two
tributaries (MP- 12a and MP- 12b). The upstream monitoring points are shown in Figure 5.3k.
5-66
Long-term Monitoring Data and Case Studies
-------
(Aep/sqi) peon ssauBBuei/y '
s
s
u
a
o
I
03
1
I
(Aep/sqi) peon
-------
(Aep/sqi) pecn asaueBuei/y 'uoj|
a
u
o
Q
I:
.§
•
.
a
a
60
I
(Aep/sqi) peon
oo
VO
-------
Coal Remining Statistical Support Document
These two tributaries had appreciable alkalinity concentrations during the entire monitoring
period (1987 through 1999), undoubtedly due to the presence of significant carbonate lithologic
units in the drainage basin (e.g., the limestone underlying the Pittsburgh Coal Seam, Figure 5.31).
However, by comparing upstream and downstream alkalinity concentrations in Porter's Run and
Beaver Run (Figures 5.3t and 5.3u), the subtle changes in alkalinity concentration observed are
believed to be due to the reduction in acid load from the Trees Mills remining operation. In
Figure 5.3t, the upstream alkalinity concentration in Porter's Run is consistently higher than the
downstream alkalinity concentration during the period of record. In Beaver Run (Figure 5.3u),
the upstream alkalinity concentration was higher than the downstream alkalinity pre-mining and
during the first year or two of remining. However, since 1994 the trend reversed, and the
downstream alkalinity concentrations are typically higher than the upstream alkalinity
concentrations. It is inferred from this data that the MP-1 discharge (and other positional
discharges) impacted the receiving stream, but the elimination or reduction of pollution loads
from these discharges during and following remining increased the downstream alkalinity. The
effect of this elimination, likely would be more dramatic without the presence of significant in-
stream alkalinity and flow.
Figure 5.3t: Porter Run
Alkalinity: Upstream & Downstream
120-
1/23/87 1/23/88 1/2289 1/22/90 1/22/91 1/22/92 1/21/93 1/21/94 1/21/95 1/21/96 1/20/97
1/20/98 1/20/99 1/20/OD
Long-term Monitoring Data and Case Studies
5-69
-------
Coal Remining Statistical Support Document
140
Figure 5.3u: Beaver Run
Alkalinity: Upstream & Downstream
1123187 1/23/88 1/22/89 1/22/90 1/22/91. 1/22/92 1/21/93 1/21/94 1/21/95 1/21/96 1/20/97 1/20/98 1/20/99 1/20/00
5.4 Conclusions
• Pre-existing discharges vary widely in flow and consequently, also in pollutant loading
rates. Because there is such a large seasonal component to flow variability, it is necessary
that baseline pollution load monitoring cover the entire range of seasonal conditions
(generally an entire water year). Use of a partial water year may significantly under or
over represent the baseline pollution load and therefore is not recommended.
• Not all discharges behave in a similar fashion. Some discharges respond steadily, with
relatively small variation, while others change rapidly and by several orders of magnitude.
While it is important to consider these behaviors, possibly requiring case-by-case
monitoring, most discharges exhibit fairly predictable behavior, and are appropriately
monitored using a monthly sampling interval and a one-year baseline monitoring period.
5-70
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
• Although it maybe possible to miss the most extreme high-flow events using a monthly
sampling interval, as long as a consistent sample interval is used for determining and
monitoring baseline, and the statistical test is not overly sensitive to extreme values, this
sampling protocol should be adequate. Low flow events occur over longer periods of
longer duration and should be adequately represented with monthly sampling.
• Extremely dry or extremely wet years may pose difficulties in establishing a
representative baseline pollution load, but significant year-to-year variations in pollution
load are rare and would be even more rare for multiple consecutive years. Seldom would
it be worth the additional time and expense to require a multi-year baseline period.
However, water quality monitoring should consider the possibility, though infrequent, of
year-to-year pollution load variations that rise to the level of statistical significance.
• Remining-induced changes in pollution load tend to be very dramatic and can result from
either significant changes in flow or significant changes in water quality. The fact that
these changes are rarely subtle makes it relatively easy to design a monitoring program
that can detect significant changes, while minimizing the incidence of false positives (i.e.,
indications of significant changes in water quality which may be due to seasonal changes
or changes due to weather patterns). The monthly monitoring interval used in the case
studies did adequately document pre and post-remining water quality, and was sufficient
to detect significant changes in pollution loading rates.
• Less frequent water monitoring intervals are much more likely to over or under represent
the baseline pollution load, and to inaccurately detect changes in pollution loading rates.
Monitoring intervals that are more frequent than monthly, are generally unnecessary and
may not be worth the added expense.
• Acidity and alkalinity, pH, metals, sulfates, and flow rates, often respond differently
depending on the BMP used. Some BMPs may reduce flows while leaving pollutant
Long-term Monitoring Data and Case Studies 5-7
-------
Coal Remining Statistical Support Document
concentrations unchanged. Sources of alkalinity may increase pH and reduce acidity,
increase one or more metals and decrease others, and increase or decrease sulfates.
Observing the response of individual parameters allows the analysis of BMP efficiency.
This is useful for applying particular BMPs to similar situations, in troubleshooting, and
in adding or modifying BMPs to achieve a desired result.
References
Ballaron, P.B., 1999. Water Balance for the Jeddo Tunnel Basin, Luzerne County, Pennsylvania.
Publication No. 208. Pennsylvania Department of Environmental Protection, Pottsville
District Mining Office under Grant ME97105.
Ballaron, P.B., C.M. Kocher, and J.R. Hollowell, 1999. Assessment of Conditions Contributing
Acid Mine Drainage to the Little Nescopeck Creek Watershed, Luzerne County,
Pennsylvania, and Abatement Plan to Mitigate Impaired Water Quality in the Watershed.
Publication No. 204. Pennsylvania Department of Environmental Protection, Bureau of
Watershed Conservation under Grant ME96114.
Barnes, I., W.T. Stuart, and D.W. Fisher, 1964. Field Investigation of Mine Waters in the
'Northern Anthracite Field, Pennsylvania. U.S. Geological Survey, Professional Paper
473-B. U.S. Government Printing Office: Washington, DC.
Brady, K.T., RJ. Hornberger, and G. Fleeger, 1998. Influence of Geology on Postmining Water
' Quality: Northern Appalachian Basin. Chapter 8 in Coal Mine Drainage Prediction and
Pollution Prevention in Pennsylvania. Edited by K.B.C. Brady, M.W. Smith and J.
Schueck. Pennsylvania Department of Environmental Protection: Harrisburg, PA. pp. 8-1
to 8-92.
Cravotta, C.A., 1998. Oxic Limestone Drains for Treatment of Dilute, Acidic Mine Drainage.
Paper presented at the 1998 Annual Symposium of the West Virginia Surface Mine
Drainage Task Force, Morgantown, WV, April 7-8.
Cravotta, C.A. andM.K. Trahan, 1999. Limestone Drains to Increase pH and Remove Dissolved
Metals from Acidic Mine Drainage. Applied Geochemistry, Vol.14, pp. 581-601,
Cravotta, C.A., 1999. Personal communication between Charles Cravotta and Roger Hornberger
of the Pennsylvania Department of Environmental Protection, Pottsville, PA.
Fishel, D.K., and J.E. Richardson, 1986. Results of a Preimpoundment Water-Quality Study of
5-72
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Swatara Creek Pennsylvania. Water Resources investigations Report 85-4023. US
Geological Survey: Harrisburg, PA.
Griffiths, J.C., (No date). Report No. 1: Reconnaissance into the use of MINITAB using
Hamilton 01, 08 Data. U.S. Environmental Protection Agency, Office of Water:
Washington, DC.
Griffiths, J.C., 1987a. Report No. 2: Time Series Analysis of Data from Hamilton 08. U.S.
Envkonmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987b. Report No. 3: Analysis of Data from Arnot 001, 003, 004. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987c. Report No. 4: Analysis of Data from the Clarion Site. U.S. Environmental
Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987d. Report No. 5: Analysis of Data from Ernest Refuse Pile, Indiana County,
Pennsylvania. U.S. Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1987e. Report No. 6: Analysis of Data from the Fisher Deep Mine. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1988a. Report No. 7: Analysis of Data from the Markson Site. U.S.
Environmental Protection Agency, Office of Water: Washington, DC.
Griffiths, J.C., 1988b. Report No, 8: Synopsis of Reports 1 - 7. U.S. Environmental Protection
Agency, Office of Water: Washington, DC.
Growitz, D.J., et al. 1985. Reconnaissance of Mine Drainage in the Coal Fields of Eastern
Pennsylvania. Water Resources Investigations Report 83-4274. US Geological Survey:
Harrisburg, PA
Hornberger, R.J., M.W. Smith, A.E. Friedrich, and H.L. Howell, 1990. Acid Mine Drainage from
Active and Abandoned Coal Mines in Pennsylvania. Chapter 32 in Water Resources in
Pennsylvania: Availability, Quality and Management. Edited by S.K. Majumdar, E.W.
Miller and R.R. Parizek, Easton: The Pennsylvania Academy of Science, pp. 432 - 451.
McCarren, E.F., J.W. Wark, and J.R. George, 1961. Hydrologic Processes Diluting and
Neutralizing Acid Streams of the Swatara Creek Basin, Pennsylvania. Geological Survey
Research, pp. D64-D67.
Smith, M.W., 1988. Establishing Baseline Pollutant Load from Pre-existing Pollutional
Long-term Monitoring Data and Case Studies
5-73
-------
Coal Remining Statistical Support Document
Discharges for Remining in Pennsylvania, Mine Drainage and Surface Mine Reclamation.
Paper presented at the 1988 Mine Drainage and Surface Mine Reclamation Conference,
Pittsburgh, PA, April 17-22, pp. 311-318.
US EPA 2001 Statistical Analysis of Abandoned Mine Drainage in the Establishment of the
Baseline Pollution Load for Coal Remining Permits. Prepared for US Environmental
Protection Agency, Office of Water by Pennsylvania Department of Environmental
Protection and DynCorp I&ET. EPA-821-B-01-014.
5-74
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Appendix A: Example Calculations of Statistical Methods
The following calculations are examples of the statistical procedures presented in Section 3.
Example 1 includes concentration and flow results for 12 baseline and monitoring
samples. This example presents application of both Method 1 and Method 2, using the
steps defined for 12 samples.
• Example 2 includes loading results for 18 baseline and monitoring samples. This
example presents application of both Method 1 and Method 2, and includes the
calculation of the Single Observation Trigger specified for 17 samples or greater in
Method 1.'
• Example 3 includes concentration and flow results for 12 baseline and monitoring
samples. This example presents application of both Method 1 and Method 2 and
demonstrates a recommended approach when replacing baseline concentrations below the
limits established in 40 CFR part 434, Subpart C.
1.0 Example 1
Assume 12 baseline flow and iron concentrations are collected by sampling once per month for a
year. Likewise, 12 flow and iron monitoring observations are obtained by sampling once per
month for a period of one year. Determination of exceedances are presented using both Methods
1 and 2. For all calculations in Example 1, assume the following flows (in gpm) and iron
concentrations (in mg/L).
Flow
Baseline
Monitoring
5.0
8.0
14.0
14.0
42.0
15.0
35.0
32.0
26.0
43.0
22.0
28.0
12.0
16.0
11.0
14.0
11.0
16.0
6.0
9.0
11.0
9.0
6.0
11.0
Trrm Concentration
Baseline
Monitoring
14.2
16.2
14.0
17.8
20.6
16.3
13.6
15.1
13.2
20.9
12.4
15.7
13.2
15.8
13.4
16.7
13.6
15.4
14.3
15.6
15.2
16.7
13.4
15.3
The resulting iron loads (in Ibs/day = flow * concentration * 0.01202) are given below:
Baseline
Monitoring
0.85
1.56
2.36
3.00
10.40
2.94
5.72
5.81
4.13
10.80
3.28
5.28
1.90
3.04
1.77
2.81
1.80
2.96
1.03
1.69
2.01
1.81
0.97
2.02
Appendix A
A-l
-------
Coal Remitting Statistical Support Document
1.1 Single Observation Trigger:
1.1.1 Method 1 (See Figure 3.2a):
1) Twelve baseline observations were collected, therefore n = 12.
2) The baseline loading observations are placed in sequential order from smallest to largest.
[0.85, 0.97, 1.03, 1.77, 1.80, 1.90, 2.01, 2.36, 3.28, 4.13, 5.72, 10.40]
3) The number of observations, n, is less than 16, therefore the Single Observation Trigger
(L) equals x^ (the maximum) = 10.40.
4) One monitoring load (10.80) is greater than 10.40, therefore the Single Observation
Trigger (L) was exceeded.
1.1.2 Method 2 (See Figure 3.2b):
1) Twelve is an even number, therefore the median of the modified baseline observations is:
M = 0.5 * (x,6) + XOT).
M = 0.5 * (1.90 + 2.01) = 1.955
In order to determine M15 calculate the median of the subset ranging from x(7) to x^.
Because 12 - 6 = 6 is even, Mj = 0.5 * (x(9) + x,-10))
M! = 0.5 * (3.28 + 4.13) = 3.705
2) In order to determine M.:, calculate the median of the subset ranging from x(1) to x(6).
Because 6 is even, M.j = 0.5 * (X(3) + X(4))
M_!= 0.5* (1.03+1.77) = 1.40
3) To calculate R, subtract M.J from ML
R= 3.705 -1.40 = 2.305
4) L = M! + (3 * R) = 3.705 + (3 * 2.305) = 10.62
5) One monitoring observation (10.80) is greater than 10.62, therefore the Single
Observation Trigger (L) was exceeded.
1.2 Annual Comparison
1.2.1 Method 1 (See Figure 3.2a)
1) From 1.1.2 Step I,M!= 3.705
A-2
Appendix A
-------
Coal Remining Statistical Support Document
2) From 1.1.2 Step 2, M.x= 1.40
3) From 1.1.2 Step 3, R = 2.305
4) The calculated value for R is then substituted into the equation for T.
5)
6)
7)
8)
The following monitoring observations are ordered from smallest to largest.
[1.56, 1.69, 1.81, 2.02, 2.81, 2.94, 2.96, 3.00, 3.04, 5.28, 5.81, 10.80]
There are 12 monitoring observations, therefore m = 12.
The number of observations is even, therefore M1 = 0.5 * (x(6) +
M1 = 0.5 * (2.94 + 2.96) '= 2.95
This holds true for Mj' and MV as well.
M; = 0.5 * (X(9) + x^) = 0.5 * (3.04 + 5.28) = 4. 16
MY = 0.5 *(x(3) + x(4)) = 0.5 * (1.81 + 2.02) = 1.915
To calculate R, subtract M^1
R' = 4. 16 -1.9 15 = 2.245.
The calculated value for R1 is then substituted in the equation for T '.
9) T (1.77) is less than T (3.16), therefore the median baseline pollution loading was not
exceeded.
1.2.2 Method 2 (Wttcoxon-Mann-Whitney Test) (See Figure 3.2b)
Instructions for the Wilcoxon-Marm-Whitney test are given in Conover (1980), cited in Figure
3.2b. ,
1) When using both baseline and monitoring data, n = 12 and m=12
2) The baseline and monitoring observations are listed with their corresponding rankings.
Appendix A
A-3
-------
Coal Re
Statistical Support Document
Baseline
Observations
(Ibs/day)
Baseline
Rankings
r
Monitoring
Observations
(Ibs/day)
Monitoring
Rankings
0.85
1
1.56
4
0.97
2
1.69
5
1.03
3
1.81
8
1.77 .
6
2.02
11
1.80
7
2.81
13
1.90
9
2.94
14
2.01
10
2.96
15
2.36
12
3.00
16
3.28
18
3.04
17
4.13
19
5.28
20
5.72
21
5.81
22
10.40
23
10.80
24
3) The sum of the twelve baseline ranks (Sn) = 131.
4) In order to find the appropriate critical value (C), match the column with the correct n
(number of baseline observations) to the row with the correct m (number of monitoring
observations). As found in the table, the critical value C for 12 baseline and 12
monitoring observations is 99.
Critical Values (C) of the Wilcoxon-Mann- Whitney Test
(for a one-sided test at the 99.9 percent level)
n
m
10
11
12
13
14
IS
16
17
18
19
20
10
66
68
70
73
75
77
79
81
83
85
88
11
79
82
84
87
89
91
94
96
99
101
104
12
93
96
99
102
104
107
110
113
116
119
121
13
109
112
'115
118
121
124
127
130
134
137
140
14
125
128
131
135
138
142
145
149
152
156
160
IS
142
145
149
153
157
161
164
168
172
176
180
16
160
164
168
172
176
180
185
189
193
197
202
17
179
183
188
192
197
201
206
211
215
220
224
18
199
204
209
214
218
223
228
233
238
243
248
19
220
225
231
236
241
246
251
257
262
268
273
20
243
248
253
259
265
270
276
281
287
293
299
5) Sn (131) is greater than C (99). Therefore, according to the Wilcoxon-Mann-Whitney
Test, the monitoring observations did not exceed the baseline pollution loading.
A-4
Appendix A
-------
Coal Remining Statistical Support Document
2.0 Example 2
Assume 18 baseline iron loading determination observations are collected by sampling twice per
month for nine months. Likewise, 18 iron load monitoring observations are obtained by sampling
twice per month for a period of nine months. Examples of both Methods 1 and 2 are presented
below. For all calculations in Example 2, assume the following iron load observations (in
Ibs/day):
Observation
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Baseline
0.030
0.005
1.915
0.673
0.064
0.063
. 0.607
0.553
0.286
0.106
0.406
1.447
0.900
0.040
2.770
1.803
0.160
0.045
Monitoring
0.530
6.040
1.040
0.033
0.030
0.230
0.710
0.240
0.390
0.830 ,
3.050
0.580
1.180
0.510
0.046
0.690
0.630
0.370
2.1 Single Observation Trigger
2.1.1 Method 1 (See Figure 3.2a)
1) The number of baseline observations collected, n = 18.
Appendix A
A-5
-------
Coal Re
Statistical Support Document
2) The baseline observations are ordered sequentially from smallest to largest.
[0.005, 0.030, 0.040, 0.045, 0.063, 0.064, 0.106, 0.160, 0.286, 0.406, 0.553, 0.607, 0.673,
0.900/1.447,1.803,1.915,2.770]
3) The number of observations is greater than 16, therefore M, Mt, M2 and M3 must be
calculated. The number of observations is even, which means the median of the baseline
observations must be calculated using the following equation:
M = 0.5*(x(9) + x(10)).
M = 0.5 * (0.286 + 0.406) = 0.346
4) To determine M15 calculate the median of the subset ranging from x(10) to X(18).
18 - 9 = 9 is odd, therefore M! = X(14) = 0.900.
5) To determine M2, calculate the median of the subset ranging from x(14) to x^.
18 - 13 = 5 is odd, therefore M2 = X(16) = 1.803.
6) To determine M3, calculate the median of the subset ranging from x(16) to x(18).
18 -15 = 3, which is odd, therefore M3 = x^ = 1.915.
7) To determine L, calculate the median of the subset ranging from x(17) to x(18).
18 - 16 = 2, which is even, therefore L = 0.5 * (x(17) + X(18)) = 0.5 * (1.915 + 2.770) =
2.343.
8) One monitoring observation (3.050) is above L (2.343), therefore the Single Observation
Trigger was exceeded.
2.1.2 Method 2 (See Figure 3.2b)
1) From 2.1.1 Step 4, MJ (the third quartile of the baseline data) is equal to 0.900.
2)
3)
To find M.15 calculate the median of the subset ranging from x^ to x(9).
9 is odd, therefore M.! = x(5) = 0.063.
The value for R is found by subtracting M.! fromMj
R = 0.900-0.063 = 0.837
4) L = M: + (3 *R)= 0.900+ (3* 0.837) = 3.411.
5) All monitoring observations are less than 3.411, therefore the Single Observation
Trigger (L) was not exceeded.
A-6
Appendix A
-------
Coal Remitting Statistical Support Document
2.2 Annual Comparison
2.2.1 Method 1 (See Figure 3.2a)
1) As determined in Section 2.1.2 step 1, M = 0.346, and M! = 0.900.
2) As determined in Section 2.1.2 step 2, M.! = 0.063.
3) As determined in Section 2.1.2 step 3, R = 0.837.
4) To find T, the value for R is inserted in the following equation:
1.815*0.837
5)
6)
7)
8)
9)
T = 0.346 +
Vl8
- = 0.704
The monitoring observations are placed in order from lowest to highest.
[0.030,0.033, 0.040,0.046,0.230,0.240, 0.370, 0.390, 0.510,0.530,0.580, 0.630,0.690,
0.710,0.830,1.040,1.180,3.050]
The number of monitoring observations (m) =18.
18 is even, making M' = 0.5 * (X(9) + x(10))
M1 = 0.5 * (0.510 + 0.530) = 0.520
To determine M/, calculate the median of subset X(10) to x(18).
Because 18 - 9 = 9 is odd, M^ = (x(14?) = 0.710
To determine M.j', calculate the median of subset x^ to x(9).
Because 9 is odd, My = (x(5)) = 0.230
The value for R1 is found by subtracting M.j1 from M^.
R' = 0.710-0.230 = 0.48
10) To find T ', the value for R is inserted into the following equation:
1.815* 0.48 =0315
11) T' (0.315) is less than T (0.704), therefore the median baseline pollution loading is not
exceeded.
Appendix A
A-7
-------
Coal Remining Statistical Support Document
2.2.2 Method 2 (Wilcoxon-Mann-Whitney Test) (See Figure 3.2b)
Instructions for the Wilcoxon-Mann-Whitney test are given in Conover (1980), cited in Figure
3.2b.
1)
2)
When using both baseline and monitoring data, n = 18 and m = 18.
The baseline and monitoring observations are listed in order of collection, and ranked as
follows:
Baseline Observations
(Ibs/day)
0.030
0.005
1.915
0.673
0.064
0.063
0.607
0.553
0.286
0.106
0.406
1.447
0.900
0.040
2.770
1.803
0.160
0.045
(Ranking)
2.5
1
34
25
10
9
23
21
15
11
18
32
29
5
35
33
12
7
Monitoring Observations
(Ibs/day)
0.530
0.040
1.040
0.033
0.030
0.230
0.710
0.240
0.390
0.830
3.050
0.580
1.180
0.510
0.046
0.690
0.630
0.370
(Ranking)
20
6
30
4
2.5
13
27
14
17
28
36
22
31
19
8
26
24
16
The value of 0.030 was obtained for more than one observation. The ranking displayed is
the average of 2 and 3 (2.5). , .
3) The sum of the 18 baseline ranks (Sn) = 322.5.
A-8
Appendix A
-------
Coal Remining Statistical Support Document
4) From the table in section 1.2.2 of this appendix, the critical value (G) for 18 baseline and
18 monitoring observations is 238.
5) Sn (322.5) is greater than the critical value C (238). Therefore, according to the
Wacoxon-Mann-Whitney test, the monitoring observations did not exceed the baseline
pollution loading.
3.0 Example 3
Assume 12 baseline flow and iron concentrations are collected by sampling once per month for a
year. Likewise, 12 flow and iron monitoring observations are obtained by sampling once per
month for a period of one year. In order to determine whether baseline pollution loading has
been exceeded, both Methods 1 and 2 were used. For all calculations in Example 3, assume the
following flows (in gpm) and iron concentrations (in mg/L).
Flow
Baseline
5.0
7.0
12.0
11.0
15.0
17.0
34.0
29.0
21.0
22.0
11.0
12.0
16.0
13.0
9.0
14.0
10.0
10.0
11.0
12.0
9.0
11.0
13.0
9.0
Iron Concentration
Baseline
Monitoring
11.4
12.3
8.2
13.5
6.0
9.8
11.1
7.9
6.4
5.8
10.3
7.5
12.1
8.2
14.2
9.3
6.1
8.4
8.3
12.5
10.0
14.1
13.5
15.3
Because there are three baseline concentrations (6.0 mg/L, 6.4 mg/L and 6.1 mg/L) below the
Subpart C effluent limit for iron (7.0 mg/L), two separate sets of loading results are calculated.
The first calculates iron loading using all the unmodified concentrations, and the following
standard equation:
Load (in Ibs/day) = Flow (in gpm) * Concentration (in mg/L) * 0.01202.
The resulting iron loads are, given below:
Baseline
Monitoring
0.69
1.03
1.18
1.78
1.08
2.00
4.54
2.75
1.62
1.53
1.36
1.08
2.33
1.28
1.54
1.57
0.73
1.01
1.10
1.80
1.08
1.86
2.11
1.66
The second set of calculated iron loads are calculated after replacing the baseline iron
concentration below 7.0 mg/L with 7.0 mg/L. This set is given below, with the three modified
loads in bold:
Appendix A
A-9
-------
Coal Remininj? Statistical Support Document
Iron Load (using modified concentrations)
Baseline
Monitoring
0.69
1.03
1.18
1.78
1.26
2.00
4.54
2.75
1.76
1.53
1.36
1.08
2.33
1.28
1.54
1.57
0.84
1.01
1.10
1.80
1.08
1.86
2.11
1.66
3.1 Single Observation Trigger
3.1.1 Method 1 (See Figure 3.2a):
1) Twelve baseline observations were collected, therefore n = 12.
2) The modified baseline observations were placed in sequential order from smallest to
largest.
[0.69, 0.84, 1.08, 1.10, 1.18, 1.26, 1.36, 1.54, 1.76, 2.11, 2.33, 4.54]
3) The number of observations, n, is less then 16, therefore the Single Observation Trigger
(L) equals x(12)) (the maximum) .= 4.54.
4) All monitoring observations are less than 4.54, therefore the Single Observation Trigger
(L) (4.54) was not exceeded.
3.1.2 Method 2 (See Figure 3.2b):
1)
Twelve is an even number, therefore the median of the modified baseline observations is:
M = 0.5 *(x(6) + x(7))= 1.31.
In order to determine M15 calculate the median of the subset ranging from x(7) to X(12).
Because 12 - 6 = 6 is even, Mx = 0.5 * (X(9) + x(10))
M! = 6.5 * (1.76 + 2.11) = 1.935
Because Mj is needed to calculate R, which must be based on unmodified concentrations,
M! must also be calculated based on unmodified concentrations. The unmodified loads
are ordered sequentially below:
[0.69, 0.73, 1.08, 1.08, 1.10, 1.18, 1.36, 1.54, 1.62, 2.11, 2.33, 4.54]
The median of unmodified baseline loads is:
M = 0.5 *(x(6) + X(7))= 1.27.
The third quartile Mx of the unmodified baseline loads is:
Mj = 0.5 *(Xc9) + x(io)) = 1.865.
2) The first quartile M.j of the unmodified baseline loads is:
A-10
Appendix A
-------
Coal Remining Statistical Support Document
M.! = 0.5 * (x(3) + x<4)) = 1.08.
3) Using the values of M^ and Mt calculated using the unmodified baseline loads,
R= 1.865- 1.08 = 0.785.
4) L = Mi + (3 *R)= 1.935 + (3* 0.785) = 4.29
4) All monitoring observations are less than 4.29, therefore the Single Observation Trigger
(L) was not exceeded.
3.2 Annual Comparison
3.2.1 Method 1 (See Figure 3.2a)
1)
2)
3)
4)
Twelve is an even number, therefore the median of the modified baseline observations is:
M = 0.5 * (X(6) + XOT). • . • '
M = 0.5* (1.26+1.36) =1.31
The following steps are needed to calculate R. Therefore, the unmodified baseline loads
must be used. These load observations are listed in listed from sequential order from
smallest to largest:
[0.69, 0..73, 1.08, 1.08, 1.10, 1.18, 1.36, 1.54, 1.62, 2.11, 2.33, 4.54]
hi order to determine M1? calculate the median of the subset ranging fromx^ to xci2).
Because 12 - 6 = 6 is even, M! = 0.5 * (x,.9) + x^)
Mj = 0.5 * (1.62+2.11) = 1.865
hi order to determine M.1? calculate the median of the subset ranging from x^ to X(6).
Because 6 is even, M.I = 0.5 * (x(3) + x(4))
M.t = 0.5 * (1.08 + 1.08) = 1.08
To calculate R, subtract M.! from Mj.
R= 1.865-1.08 = 0.785
The calculated value for R is then substituted into the equation for T.
1.815* 0.785 = 172
5) The following monitoring observations are ordered from smallest to largest.
[1.01, 1.03, 1.08, 1.28, 1.53, 1.57, 1.66, 1.78, 1.80, 1.86, 2.00, 2.75]
Appendix A
A-ll
-------
Coal Remitting Statistical Support Document
6) There are 12 monitoring observations, therefore m= 12.
The number of observations is even, therefore M' = 0.5 * (X(6) + X(7))
M1 = 0.5 * (1.57 + 1.66) = 1.615
This holds true for Mj1 and M.!1 as well.
M,1 = 0.5 * &M + x(10)) = 0.5 * (1.80 + 1.86) = 1.83
1^'= 0.5 * (X(3) + X(4)) = 0.5 * (1.08 + 1.28) = 1.18
7) To calculate R, subtract M.j1 fromMj'
R'= 1.83 -1.18 = 0.65
8) The calculated value for R1 is then substituted in the equation for T.
= 1.274
9) T' (1.274) is less than T (1.72), therefore the median baseline pollution loading was not
exceeded.
3.2.2 Method 2 (WUcoxon-Mann-Whitney Test) (See Figure 3.2b)
1) When using both baseline and monitoring data, n = 12 and m=12
2) The modified baseline and monitoring observations are listed with their corresponding
rankings.
Baseline
Observations
(Ibs/day)
Baseline
Ranlcincs
Monitoring
Observations
(Ibs/day)
Monitoring
Rankings
0.69
1
1.01
3
0.84
2
1.03
4
1.08
5.5
1.08
5.5
1.10
7
1.28,
10
1.18
8
1.53
12
1.26
9
1.57
14
1.36
11
1.66
15
1.54
13
1.78
17
1.76
16
1.80
18
2.11
21
1.86
19
2.33
22
2.00
20
4.54
24
2.75
23
Due to the fact that the value of 1.08 was obtained for two observations, an average
ranking is used for this value. For 1.08, the average of 5 and 6 is 5.5.
3) The sum of the twelve baseline ranks (Sn)= 139.5.
4) From the table in section 1.2.2 of this appendix, the critical value (C) for 12 baseline and
12 monitoring observations is 99.
A-12 Appendix A
-------
Coal Remining Statistical Support Document
5) Sn (139.5) is greater than C (99), Therefore, according to the Wilcoxon-Mann-Whitney
test, the monitoring observations did not exceed the baseline pollution loading.
Appendix A
A-13
-------
------- |