United States
Environmental Protection
Agency
Office of Water
(4303)
EPA-821-B-00-001
March 2000
DRAFT
Coal Remining Statistical Support
Document
§ ^-*&t~~'r £&& ^:-''^.'^-'-
''
is-i.-"1- - - i . .--..% ..-.' . 'C' --' *--' -".-l^V=-' .'.- - ,. -.VJ---''' - I^L;:--- '.-* . .,1 , -"-.x-::-.1" ^- "v""?v£%'NV1-;"-'si~''' '-s^~ ' ' - i " *
3 " . \ ./-. .y' ,-f* ',-.,*-' . .'+<''' - " /' >*-."-- ^-'~*-"".-' -.:-?* .-. ^'^.-.-*-3-s . ^^^^^'ii-V.'''-^?.-^^'-'..-77~i.a / 7 S
I' "* / '. < ' ""'""' .".-^f''' "'. '"~\ >>'**'-'- £'-. '*" 'V-.'-'':--'. .*' '^.'-~~*-:/ ..''t-2.^--^---- .. '"'^"-^-'^- ''si';'.~'r---^ *S\ i"t ""*. rK-W^""".1^^l-11 V".'X " " ^^Z
a" **v i*' '" : * '"^.-- "-...-'', '*^v ! f ! ' ' ' ' ' ' ill I __ _ . . ;
-'-.-_ ,- '-^-s-^^i.' " -;-*"'~TjrTw--- '. Vv> tL^C^- -'-*"^w. ' " '- & '' " ~' ''-Z. A7-"--'»*: ,-.,-pllrr^^r -~/~T\
ij *_..;__^;>t^';:-- ' ^~V^ ".y-^' V : '',- '_.'.y- '%w?gy^l>r. " VI-:'',';L-'"I: % =^r"' "" .'.-' .V" "'-."* *" """" *'* ~ ~" *.----- ~". . '~--^f;."'~ ;^'-' " ..- ~.'.~*^~- .- '.fC^'^r^k
-------
-------
EPA82J-B-00-001
COAL REMINING
STATISTICAL SUPPORT DOCUMENT
MARCH 2000
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 of the Engineering and
Analysis Division (BAD) within the U.S. Environmental Protection Acency's (EPA's) Office of
Science and Technology (OST). EPA gratefully acknowledges the determined efforts of Roger
Hornberger, Michael W. Smith, Daniel J. Koury, and J. Corey Cram of Pennsylvania's
Department of Environmental Protection (PADEP) for their contributions in making this
document possible. EPA also wishes to thank DynCorp Information and Enterprise Technology
for its invaluable support.
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 EPA may decide to follow the draft guidance provided in this document, or to act
at variance with the guidance, based on its analysis of the specific facts presented. This draft
guidance is being issued in connection with the proposed amendments to the Coal Mining Point
Source Category. EPA has solicited public comment on the information contained in the
proposal. This guidance may be revised to reflect changes in EPA's approach. The changes in
EPA's approach will be presented in a future public notice.
The primary contact regarding questions or comments on this document is:
William A. Telliard
Engineering and Analysis Division (4303)
U.S. Environmental Protection Agency
Ariel Rios Building, 1200 Pennsylvania Avenue, N.W.
Washington, DC 20460
Phone: 202/260-7134
Fax: 202/260-7185
email: telliard.william@epamail.epa.gov
Acknowledgments
-------
-------
Coal Reminins Statistical Document
Acknowledgments
-------
-------
Coal Remining Statistical Support Document
TABLE OF CONTENTS
Page
LIST OF TABLES ..... Hi
LIST OF FIGURES zv
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 their Behavior 2-13
2.3 Distributional Properties of AMD Discharges 2-18
References 2-21
Section 3.0 Statistical Methodology for Establishing Baseline Conditions and Setting
Discharge Limits at Remining Sites 3-1
3.1 Objectives, Statistical Principles, and Issues 3-1
3.2 Statistical Procedures for Calculating Limits from Baseline Data 3-4
3.2.1 Procedure A (State of Pennsylvania Procedures) 3-4
3.2.2 Procedure B 3-6
3.2.3 Use of Triggered or Accelerated Monitoring in Procedures A and B . . 3-7
3.3 Statistical Characterization of Coal Mine Discharge Loadings 3-13
References 3-14
Section 4.0 Baseline Sampling Duration and Frequency .. 4-1
4.1 Power and Sample Size 4-1
4.2 Sampling Plan 4-4
References 4-5
Table of Contents
-------
Coal Remining Statistical Support Document
Section 5.0 " Long-term Monitoring Case Studies 5-1
5.1 A Comparison of Seven Long-term Water Quality Data Sets 5-2
5.1.1 Sampling Interval 5-8
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 Me Wreath Discharge 5-50
5.3.3 Trees Mills Site 5-54
5.3 Conclusions 5-66
References 5-68
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
Table 3.2a: Exceedance Probabilities for Given Numbers of Samples 3-12
Section 4.0 Baseline Sampling Duration and Frequency
Table 4.1a: Power of One-Sided Two-Sample t-test
4-2
Section 5.0 Long-term Monitoring Case Studies
Table 5.1a: 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 Sole: Comparison of Median Acidity and Iron Loads by Baseline Sampling Year .. 5-7
List of Tables
in
-------
Remining Statistical Support Document
IV
List of Tables
-------
Coal Re mining 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 Schuylkill County 2-16
Figure 2.2d: Streamflow and Acidity in Coal Refuse Pile 2-17
Figure 2.3a: Frequency Distribution of Sulfate at Dunkard Creek 2-19
Figure 2.3b: Frequency Distribution of Acidity in Mercer Coal Seam 2-20
Figure 2.3c: Frequency Distribution of Iron in Mercer Coal Seam . . . . '. 2-20
Figure 2.3d: Frequency Distribution of Sulfate in Mercer Coal Seam 2-21
Section 3.0 Statistical Methodology for Establishing Baseline Conditions and
Setting Discharge Limits at Remining Sites
Figure 3.2a: Procedure A 3-8
Figure 3.2b: Procedure B 3-9
Figure 3.2c: Accelerated (Triggered) Monitoring 3-11
Section 4.0 Baseline Sampling Duration and Frequency
List of Figures
-------
Coal Remining Statistical Support Document
Section 5.0
Figure 5.1a:
Figure 5.1b:
Figure S.lc:
Figure 5.1d:
Figure 5.1 e:
Figure 5. If:
Figure 5.1g:
Figure 5.1h:
Figure 5.1i:
Figure S.lj:
Figure 5.1k:
Figure 5.11:
Figure 5.1m:
Figure S.ln:
Figure S.lo:
Figure S.lp:
Figure S.lq:
Figure S.lr:
Figure 5.1s:
Figure S.lt:
Figure 5.1u:
Figure S.lv:
Figure S.lw:
Figure S.lx:
Figure 5.1y:
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:
Long-term Monitoring Case Studies
Amot-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
Amot-3 Monthly Acidity Load Comparison
Amot-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)
Fisher Iron Load Data (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 (Flow & 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)
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
List of Figures
-------
Coal Remining Statistical Support Document
Section 5.0 Long-term Monitoring Case Studies (cont.)
Figure 5.2m:
Figure 5.2n:
Figure 5.2o:
Figure 5.2p:
Figure 5.2q:
Figure 5.2r:
Figure 5.3a:
Figure 5.3b:
Figure 5.3c:
Figure 5.3d:
Figure 5.3e:
Figure 5.3f:
Figure 5.3g:
Figure 5.3h:
Figure 5.3i:
Figure 5.3j:
Figure 5.3k:
Figure 5.31:
Figure 5.3m:
Figure 5.3n:
Figure 5.3o:
Figure 5.3p:
Figure 5.3q:
Figure 5.3r:
Figure 5.3s:
Figure 5.3t:
Figure 5.3u:
Swatara Creek Flow and Sulfate Data 5-41
Swatara Creek Flow and Suspended Solids Data 5-41
Swatara Creek Flow and Iron Data , 5-42
Wapwallopen Creek Flow Data 5.43
Jeddo Tunnel Flow Data 5.44
Precipitation Data From Hazleton, PA .. . 5-44
Fisher Mining MP1 (Flow, Iron, Manganese, Sulfate) 5-46
Fisher Mining MP1 (Net Acidity) 5-47
Fisher Mining MP1 (Acid Load) 5-47
Fisher Mining MP1 (Iron Load) 5-48
Fisher Iron Load Box Plot 5.49
Fisher Net Alkalinity Box Plot 5-49
McWreath Dl (Flow, Net Acidity, Iron, Manganese, Sulfate) 5-50
Me Wreath D3 (Flow & Net Acidity) 5-51
McWreath D3 (Flow & Iron) 5-51
McWreath D4 (Flow & Net Acidity) 5-53
Trees Mills Site Map 5-55
Trees Mills Drill Hole Data 5-56
Trees Mills MP1 (Flow, Manganese, Aluminum, Net Acidity, Iron, Sulfate) 5-57
Trees Mills MP1 (Acid, Iron, Manganese Load) 5-59
Trees Mills MP2 (Flow, Iron, Manganese, Aluminum, Net Acidity, Sulfate) 5-60
Trees Mills MP2(Acid, Iron, Manganese Load) 5-61
Trees Mills MP3 (Flow, Iron, Manganese, Aluminum, Net Acidity, Sulfate) 5-62
Trees Mills MP3 (Acid, Iron, Manganese Load) 5-63
Trees Mills MP6 (Acid, Iron, Manganese Load) 5-64
Porter Run (Alkalinity: Upstream & Downstream) 5-65
Beaver Run (Alkalinity: Upstream & Downstream) 5-66
List of Figures
VII
-------
Coal Remining Statistical Support Document
List of Figures
-------
Coal Remininz Statistical Support 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 \I\
-------
Coal Remining Statistical Support Document
areas. If a pre-existing pollutional discharge of acid mine drainage was occurring within the area
or on an area hydrologically 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 Reminmg 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
rernine 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
i ' '
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 which 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. J. C. Griffiths of the Pennsylvania, State University. Since these unpublished statistical
Introduction
1-3
-------
Coal Remining Statistical Support Document
analyses of mine drainage datasets are relevant to baseline pollution load statistics, they are
presented in an abridged form in a companion volume to this report, prepared by Griffiths,
Homberger, and Smith (in preparation).
1.2 Pennsylvania DEP Remining Permitting Procedures
Since 1985, PADEP has issued approximately 300 renaming 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
i1
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 from multiple 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 Remitting Statistical Support Document
In many cases, pre-existing pollutional discharges may occur in the form of numerous discharge
points, all 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 lozid 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).
Introduction
1-5
-------
Coal Reminins Statistical Support Document
The statistical components of establishing baseline pollution load include characterizing the
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 summgiry 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
MINITAB statistical software package1, which includes statistical and graphical methods to
perform all of the steps in the algorithm presented in Figure 1.2a.
'MINITAB is a commercial software package from Minitab, Inc., © 1986, 3081 Enterprise
Drive, State College, PA 16801
1-6
Introduction
-------
Coal Remining Statistical Support 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
S. Univariate statistics.
DESCRIBE - Summary Statistics.
HISTOGRAM: symmetry?
Yes
No
7. Describe Histogram
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.
1O. 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 by Simulation
1. Choose samples (1S 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
1-7
-------
Coal Remininz 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
1 i'
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 from McGill, Tukey, and Larsen (1978).
Table 1.2a: Baseline Pollution Load Summary
Mine ID: Mine Name:
Hydrologic Unit ID:
# of Samples:
Statistical Results
1. Range Low:
High:
2. Median
3. Quartiles Low:
High:
4. Approximate 95% Low:
tolerance limits High:
5. 95% Confidence Int. Low:
about median* High:
Flow
(gpm)
Loading in Pounds Per Day
Acidity
Iron
Manganese
Aluminum
Sulfates
*Note: 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 et. al., 1978).
1-8
Introduction
-------
Coal Remining 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
1-9
-------
Coal Remining Statistical Support Document
Figure 1.2b: The Quick-Trigger Process
Monthly Sample
Is 95% tolerance limit violated?
Yes
NO
Is this the second
consecutive month
that the 95%
tolerance limit
exceeded?
-NO-
Continue monthly
sampling.
Yes
Commence weekly
sampling.
Do 4 weekly samples
exceed 95% tolerance
limit?
-No-
Yes
Notify Dept. and commence
treatment unless determined
that increase in pollution load
is not attributed to the mining
operation.
Yes
Are 4 consecutive weekly
samples in compliance?
No
A.
Continue Treatment.
1-10
Introduction
-------
Coal Remining Statistical Support Docziment
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 percentile 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-
Introduction
1-11
-------
Coal Remitting Statistical Support Document
1965 data (N = 14); and the range of the manganese concentrations in the 1983-1997 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), McGill,
Tukey, and Larsen (1978), Veleman and Hoaglin (1981) and Helsel (1989).
Figure 1.2c: Dunkard Creek pH
9.0n
8.1 =
7.2=
6.3=
5.4=
4.5=
3.6=
2.7=
1.8=
0.9
0.0
8
o
o
%
X
X
1950-1965
1966-1982
1983-1997
1-12
Introduction
-------
Coal Remining Statistical Support Document
Figure 1.2d: Dunkard Creek Manganese (mg/L)
10000.0;;
9000.0=
8000.0=
7000.0=
6000.0=
5000.0=
4000.0=
3000.0=
2000.0=
1000.0=
X
* ^":
^r i- s
,,
**,
X
X
55 X
X
5 g
-T- T
I [-. ^v^^^J>;^.''| |?^^%^i;;;iv' ^1
1950-1965
1966-1982
1983-1997
Monitoring and compliance inspections are conduct 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 IMCC Remining Task Force is a discussion paper on water quality issues
related to coal remining, for which EPA, OSMRE, and IMCC jointly solicited comments in
Introduction
1-13
-------
Coal Retaining Statistical Support Document
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 remining 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, remining 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
minimum of 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, June 1969.
Griffiths., J.C., Report No. 1: Reconnaissance into the use of MINITAB using Hamilton
01, 08 Data. Prepared for the U.S. Environmental Protection Agency, Office of Water,
Washington DC, (No Date).
Griffiths, J.C., 1987. Report No. 2: Time Series Analysis of Data from Hamilton 08. Prepared
for the U.S. Environmental Protection Agency, Office of Water, Washington DC, April
1987.
Introduction
1-15
-------
Coal Remining Statistical Support Document
Griffiths, J.C., 1987. Report No. 3: Analysis of Data from Arnot 001, 003, 004. Prepared for
the U.S. Environmental Protection Agency, Office of Water, Washington DC, June 1987.
Griffiths, J.C., 1987. Report No. 4: Analysis of Data from the Clarion Site. Prepared for the
U.S. Environmental Protection Agency, Office of Water, Washington DC, September,
1987.
Griffiths, J.C., 1987. Report No. 5: Analysis of Data from Ernest Refuse Pile, Indiana County,
Pennsylvania. Prepared for the U.S. Environmental Protection Agency, Office of Water,
Washington DC, October 1987.
Griffiths, J.C., 1987. Report No. 6: Analysis of Data from the Fisher Deep Mine. Prepared for
the U.S. Environmental Protection Agency, Office of ' Water,'AVashmgfohDC,'Decemb"er
1987.
Griffiths, J.C., 1988. Report No. 7: Analysis of Data from the Markson Site. Prepared for the
U.S. Environmental Protection Agency, Office of Water, Washington DC, January 1988.
Griffiths, J.C., 1988. Report No. 8: Synopsis of Reports 1 - 7. Prepared for the U.S.
Environmental Protection Agency, Office of Water, Washington DC, April 1988.
Griffiths, J.C., R.J. Hornberger, and M.W. Smith, (in preparation). Statistical Analysis of
Abandoned Mine Drainage in the Establishment of the Baseline Pollution Load for Coal
Renaming Permits. Details available from U.S. Environmental Protection Agency
Sample Control Center, operated by Dyncorp I&ET, 6101 Stevenson Avenue,
Alexandria, VA 22304.
Helsel, D., 1989. Boxplots: A Graphical Method for Data Analysis. US Geological Survey,
Bureau of Systems Analysis Technical Memorandum No. 89.01, Reston, VA, June 1989.
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.
Kohlmann Ruggiero 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.
1-16
Introduction
-------
Coal Remining Statistical Support Document
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., and M.A. Thomas, 1986. Assessment of Technological and Economical Feasibility
of Surface Mine Reclamation, Stage One Research. Pennsylvania Department of
Environmental Resources Contract No. 898621.
Tukey, J.W., 1977. Exploratory Data Analysis. Reading, Massachusetts: Addison Wesley
Publishing Company.
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
Protection Agency, Office of Water, Industrial Technology Division Energy and Mining
Branch, Washington D.C. by Kohhnann Ruggiero Engineers, P.C.
Velleman, P.P., and D.C. Hoaglin, 1981. Applications, Basics, and Computing of Exploratory
Data Analysis. Boston, MA: Duxbury Press.
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.
Introduction
1-17
-------
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 (FeS2), 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 (FeS2), 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+, followed 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
neutralizes 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
2-2
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
examined the significance of pyrite morphology, especially the framboidal form with high
surface area.
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 Kittanning 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 Remining Statistical Support Document
Table 2.0a: High Alkalinity Examples in Pennsylvania Mine Discharges
Site Name
Willow Tree
Susan Ann
Bertovich
Smith
Brown
Trees Mills
State Line
Cover Hill
Hager
Fruithill
Laurel Hill
#1
Morrison
Stuart
Clinger
Leechburg
* Fran
Swiscambria
Albert #\
Snyder#l
Lawrence
Graff Mine
Philipsburg
** Old 40
Stratigraphic
Interval
Waynesburg
Waynesburg
Sewickley
Redstone
Redstone
Pittsburgh
Upper &
Lower
Bakerstown
Lower
Bakerstown
Brush Creek
Upper &
Lower
Freeport
U. Freept. to
U. Kittng.
Upper
Kittanning
Upper
Kittanning
Middle
Kittanning
Lower
Kittanning
Lower
Kittanning
Lower
Kittanning
Lower
Kittanning
Lower
Kittanning
Lower
Kittanning
L. Kittng. &
Vanport Ls
Clarion
Clarion
pH
7.8
3.3
3.1
7.7
7.4
2.5
8.1
3.6
6.8
7.8
8.1
7.0
2.8
6.8
2.4
2.2
4.2
3.1
6.9
2.2
7.8
2.7
2.2
Alkalinity
mg/L
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
Acidity
mg/L
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
Fe
mg/L
0.12
324.40
74.80
1.47
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
0.01
1959.80
3200.00
Mn
mg/L
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
260.00
SO4
mg/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
14000.0
Flow
gpm
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
Comments
Seep at deep mine,
pre-mining
Seep near sealed deep
mine entry.
Deep mine discharge
Pit water at lowwall
sump.
Spring near cropline.
Deep mine discharge
Post-mining seep from
backfilled spoil
Discharge from
abandoned pit below
site
Logan spring
Deep mine discharge.
Toe 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 paleoenvt.
Pit water, marine
paleoenv.
Pit water, sandstone
overburden
Seep above road
Spoil discharge
Monitoring well in
backfilled spoil
2-4
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Site Name
Orcutt
Cousins
Zacherl
Horseshoe
Wadesville
Stratigraphic
Interval
Clarion
Clarion
Clarion
Mercer
Llewellyn
pH
3.9
7.6
2.3
2.3
6.7
Alkalinity
mg/L
0.0
130.0
0.0
0.0
414.0
Acidity
mg/L
5179.6
0.0
9870.0
1835.0
0.0
Fe
mg/L
2848.00
7.15
2860.00
194.00
3.61
Mn
mg/L
349.00
0.30
136.60
27.00
3.37
SO.,
mg/L
11120.0
71.0
7600.0
2510.0
1038.0
Flow
gpm
0.0
0.0
no data
700.0
no data
Comments
Spoil water from
piezometer
Pit water, glacial till
influence
Toe of spoil discharge
Abandoned deep mine
discharge
Minepool, Anthracite
Region
Note: .Extreme values (>100 mg/L) are highlighted for emphasis
* data from Schuek et al. (1996)
** data from Dugas et al. (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).
Characteristics of Coal Mine Drainage Discharges
2-5
-------
Coal Remining Statistical Support Document
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
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
Dunkard
Monongahela
Gonemaugh
Upper Allegheny
I ~ ~T Ixwer Allegheny
Pcttsville
0
Characteristics of Coal Mine Drainage Discharges
2-7
-------
Coal Remining Statistical Support Document
Figure 2.0b: Distribution of pH in Anthracite Mine Drainage
cr\ _
DU ~~ ~
U
o>
LL.
ll^fl Northern
§BHH Eastern Middle
1 T Western Middle
Southern
20
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 Remining 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 2.4 to 7.5 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 Remitting Statistical Support Document
Figure 2.1a Annual Variability in Streamflow at Dunkard Creek
2-10"
Characteristics of Coal Mine Drainage Discharges
-------
CoalRemining 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 2.1b. 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 2.1b 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 2.1b 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
-------
Coal Remitting Statistical Support Document
Figure 2.1b: Sulfate Concentration vs. Streamflow at Dunkard Creek
10.000
1.000
100
10
Dunkard Creek at Shannopin, PA
1950-1997
» n
Q A"
n * .
+ 1 * rj »° o A »n
6 *+*°«tn
* j+ * * »*A *DA
I ^a~f f^H--*- +x " +
* * frfL.^-A"+* J~ft *,
' ° * ^Hff'&S?*"** JA * X
;,- .*" ""^.?^gitf<;A.
x ° x ^o o "^ *i«. ^
» ° ° J.x,, t, A a
,; , , A
1 10 100 1,000
STREAMFLOW (c.f.s.)
o January
february
^ march
xapril
xmay
June
+july
4 august
* septerrber
n October
noverrber
o decerrber
A
X
\
I
10.C
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 minimal 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
Characteristics of Coal Mine Drainage Discharges 2-13
-------
Coal Remining Statistical Support Document
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
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
, i i
medium size underground mines where the capacity for ground water storage is relatively small
and ground water flow paths are short.
Figure 2.2a: Acidity and Streamflow of Arnot Mine Discharge
Arnot Mine Discharge
10*|
110 a-:
o
on
v_<
LU
CD
OL
<.
X
O
en
Q
10*:
10 -.-
1980
1981
1982
1983
2-14
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Figure 2.2b: Inverse Loglinear Relationship between Acidity and Streamflow
60__
40 4-
Q
b
10
-t-Ap^a.
. *^
10* , 10 \
LOG DISCHARGE (gai/min)
10.*
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.
Characteristics of Coal Mine Drainage Discharges
2-15
-------
Coal Remining Statistical Support Document
Figure 2.2c: Streamflow and Acidity in Schuylkill County
MARKSON AIRWAY - 1985
200Q-T
500
JAN FE3 MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Occasionally, AMD discharges are subject to extreme variations in flow rates with little change
in water quality. Figure 2.2d presents flow and acidity exhibiting "slug" behavior in a discharge
from a coal refuse pile. Flow rates vary dramatically in response to recharge events (from less
than 3 to 470 gpm). Concomitantly, acidity concentrations change very little, and result in large,
rapid variations in acid loading. This discharge behavior results where conditions favor the
accumulation of water-soluble, acid-bearing shales in the unsaturated zone. During recharge
events, infiltrating water permits rapid dissolution of salts producing additional acidity in the
discharge, rather than causing a dilution effect. The longer the time period between recharge
events, the more time is available for the build up of acid-bearing salts in the unsaturated zone.
Coal refuse piles, and surface mines with very high sulfur spoil in the unsaturated zone and
2-16
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
limited ground-water storage capacity, provide the most favorable environment for this discharge
behavior. In the most severe cases, increases in flow can be accompanied by increased
concentrations of acidity or metals, resulting in extreme increases in loading rates. When this
Figure 2.2d: Streamflow and Acidity in Coal Refuse Pile
Ernest Refuse Pile
1981
1 I I I I t I . I I I I I I 1 I ( 1 1 ,
1982 1983 1984 1985
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 Griffiths, Hornberger, and Smith (in preparation).
For remining operations that will reaffect a pre-existing pollutional discharge, knowledge of
discharge behavior is critical to the establishment of a representative baseline. All three
discharge types exhibit some seasonal behavior, with highest flows during seasonal high ground-
Characteristics of Coal Mine Drainage Discharges
2-17
-------
Coal Remitting Statistical Support Document
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-remining 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
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 2.1b), or
2-18
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
logarithmically transforming the data produces a much closer approximation of the normal
frequency distribution.
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; and Griffiths, 1967), and logarithms are frequently used in the analysis and
Figure 2.3a: Frequency Distribution of Sulfate at Dunkard Creek (mg/L)
250
200.
150.
50.
400
800
1200
1600
2000 2400
Sulfate (mg/L)
2800
3200
3600
4000
More
graphical expression of water quality data (Gunnerson, 1967, and Hem, 1970, pp 271-280). 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.
Characteristics of Coal Mine Drainage Discharges
2-19
-------
Coal Remining Statistical Support Document
Several datasets for remining sites in Cambria County, Pennsylvania were also arrayed to
examine the frequency distribution for AMD parameters. These datasets include a large
abandoned deep mine on the Mercer Coal seam (Page Mine). Figures 2.3b through 2.3d show the
positively skewed frequency distributions of acidity (Figure 2.3b), iron (Figure 2.3c) and sulfate
(Figure 2.3d) for the Page Mine discharge.
Figure 2.3b: Frequency Distribution of Acidity in Mercer Coal Seam (mg/L)
20
15
10
0
200 400 600 800
Figure 2.3c: Frequency Distribution of Iron in Mercer Coal Seam (mg/L)
14
12
10
8
6
4
2
0
40
70
2-20
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Figure 2.3d: Frequency Distribution of Sulfate in Mercer Coal Seam (mg/L)
20
15
10
5
0
300 600 900
References
Aitchison, J., and J.A.C. Brown, 1973. The Lognormal Distribution. London: Cambridge
University Press. 176 p.
Barnes, H.L., and S.B. Romberger, 1968. Chemical Aspects of Acid Mine Drainage. Journal
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 Int'l 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, III, and W.W. Hellier, 1997. Bimodal Distribution of
pH in Coal Mine Drainage. Abstracts with Programs, 27th Northeast Section Meeting,
Geological Soc. of America, Vol. 24, No. 3.
Characteristics of Coal Mine Drainage Discharges
2-21
-------
Coal Remining Statistical Support Document
Brady, K.B.C., R.J. 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, February., 1954.
Cravotta, C.A., III, 1994. Chapter 23: Secondary Iron-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, Dept. of Geology.
Dugas, D.L., C.A. Cravotta, III, 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.
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. June 1966, pp. V-6 thru V--14.
Evangelou, V.P., 1995. Pyrite Oxidation and its Control. New York: CRC Press, 293 p.
Griffiths, J. C., 1967. Scientific Method in Analysis of Sediments. New York: McGrawHill
Book Co. 508 p.
Griffiths, J.C., Report No. 1: Reconnaissance into the use of MINITAB using Hamilton
01, 08 Data. Prepared for the U.S. Environmental Protection Agency, Office of Water,
Washington DC, (No Date).
Griffiths, J.C., 1987. Report No. 2: Time Series Analysis of Data from Hamilton 08. Prepared
for the U.S. Environmental Protection Agency, Office of Water, Washington DC, April
1987.
Griffiths, J.C., 1987. Report No. 3: Analysis of Data from Arnot 001, 003, 004. Prepared for
the U.S. Environmental Protection Agency, Office of Water, Washington DC, June 1987.
2-22
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Griffiths, J.C., 1987. Report No. 4: Analysis of Data from the Clarion Site. Prepared for the
U.S. Environmental Protection Agency, Office of Water, Washington DC, September
1987.
Griffiths, J.C., 1987. Report No. 5: Analysis of Data from Ernest Refuse Pile, Indiana County,
Pennsylvania. Prepared for the U.S. Environmental Protection Agency, Office of Water,
Washington DC, October 1987.
Griffiths, J.C., 1987. Report No. 6: Analysis of Data from the Fisher Deep Mine. Prepared for
the U.S. Environmental Protection Agency, Office of Water, Washington DC, December
1987.
Griffiths, J.C., 1988. Report No. 7: Analysis of Data from, the Markson Site. Prepared for the
U.S. Environmental Protection Agency, Office of Water, Washington DC, January 1988.
Griffiths, J.C., 1988. Report No. 8: Synopsis of Reports 1 - 7. Prepared for the U.S.
Environmental Protection Agency, Office of Water, Washington DC, April 1988.
Gunnerson, C.G., 1967. Streamflow and Quality in the Columbia River Basin. Journal of the
Sanitary Engineering Division, Proc. 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., Washington: U.S. Government
Printing Office, 363 p.
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, Dept. of Interior, Office of Water Research and Technology
Project B-097-PA, University Park: Pennsylvania State University.
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, Easton: The Pennsylvania Academy of Science, 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.
Characteristics of Coal Mine Drainage Discharges
2-23
-------
Coal Remining Statistical Support Document
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.
New York: McGraw Hill Book Company, Inc. 475 p.
Lovell, H.L., 1983. Coal Mine Drainage in the United States - An Overview. Wat. Sci. Tech.,
Vol. 15, pp. 1-25.
McKibben, M.A., and H.L. 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
Irqn Minerals. Acid Sulfate Weathering, Spec. Pub. No. 10, Soil Science Society of
America, pp. 37 56.
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, III, 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 &
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.
2-24
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
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.
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, MN. 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, University Park: Pennsylvania State University.
Characteristics of Coal Mine Drainage Discharges
2-25
-------
Coal Remining Statistical Support Document
2-26
Characteristics of Coal Mine Drainage Discharges
-------
Coal Remining Statistical Support Document
Section 3.0 Statistical Methodology for Establishing Baseline
Conditions and Setting Discharge Limits at Remining
Sites
3.1 Objectives, Statistical Principles and Statistical Issues
The Rahall amendment (CWA Section 301(p)) states:" "... 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
c
and manganese, to exceed the levels being discharged from the remined area before the coal
remining operation begins."
"Levels" is interpreted here to mean the entire probability distribution of loadings, including the
average, the median, and the extremes. If P percent of loadings are < some number Lp during
baseline, then no more than P percent should be < Lp during and after remining. For example, if
during the baseline period, 95 percent of iron loadings are <; 8.1 Ibs/day and 50 percent are < 0.3
Ibs/day, then during and after remining the same relationships should hold true. This should hold
true for pH, and for loadings of acidity, iron, and manganese.
The objective of Section 3.0 is to provide statistical procedures for deciding when the pollutant
levels in a discharge exceed the levels at 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." To do this, it is essential to have a sufficiently large number of
samples during and after baseline. The methods provided here may be applied to either pH or
pollutant loadings.
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
Statistical Methodology
3-1
-------
Coal Remitting Statistical Support Document
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).
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 ((3) 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 no more than 0.05. When the
discharge level is substantially less than baseline, the probability of making this error is expected
to be very low. This probability could be controlled for each decision, or it could be controlled
for all decisions for one analyte (e.g., 48 tests of a maximum daily limit and four tests of an
annual average over a four-year period). When many decisions will be made, the overall error
i . -," . . " i."'
rate is a concern. 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 no more than 0.25.
Again this could be controlled for each decision or over all decisions during the life of a permit.
,n' , ,''!'
There is significant, positive autocorrelation of flows, concentrations, and loadings in mine
discharges over periods of 1-4 weeks (Griffiths, in preparation). Sample estimates of the
variance, used in the statistical procedures proposed here, are inaccurate unless adjusted for
autocorrelation. Without adjustment, variance is underestimated. Such adjustments are
discussed by Loftis and Ward (1980), and EPA (1993) has accounted for autocorrelation in a
previous effluent guideline. Such adjustments require an estimate of the autocorrelation
/ :
coefficient. However, one cannot reliably estimate site-specific autocorrelation from small
sample sizes (e.g., n=12). Thus, there is a need for default adjustment factors for sample
variance. These factors would be developed from estimates of autocorrelation, using data from
coal mine discharges having sufficient data obtained over the course of at least two years. Such
factors may or may not be specific to all types of mine discharge. EPA has provided a default
value of first-order autocorrelation pj = 0.5 to be used in calculating an adjustment. EPA may
3-2
Statistical Methodology
-------
Coal Remining Statistical Support Document
provide a different value in a final rulemaking and in the final version of this Statistical Support
Document, after analysis of data and comments provided by states and other sources.
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 atypically 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. There is a need to evaluate the differences among baseline
years in loadings and flows, based on further analysis of data now available to the Agency.
Using such information, EPA may provide optional statistical procedures in a final rulemaking
that could account for the uncertainty in characterizing baseline from one year, or that could
account for the unrepresentative character of a baseline sampling year. Such procedures could
employ modifications of the proposed statistical procedures that use estimates of the variance
among baseline years in loadings, developed from long-term datasets, or could employ
adjustments to the baseline sample statistics to account for a baseline sampling year that was
atypical in rainfall or discharge flow. Such an adjustment could be a factor (multiplier) or a
statistical equation estimated by regression.
The proposed statistical procedures are intended to provide environmental protection and to
ensure compliance with the Rahall amendment. EPA has not yet evaluated the error rates of
these decision procedures. EPA intends to evaluate the decision error rates of each procedure by
computer simulations. Depending upon comments and associated evidence, and depending upon
EPA's further evaluations, EPA may modify or reject any of these procedures, or may change the
recommended number of samples, in order to provide suitable error rates.
Statistical Methodology
3-3
-------
Coal Remining Statistical Support Document
3.2 Statistical Procedures for Calculating Limits from Baseline Data
Two alternative statistical procedures are described, A and B. These may be modified to require
accelerated monitoring (Procedure C) after a warning level or condition is exceeded.
3.2.1 Procedure A
Procedure A is a modification of the methodology used by the State of Pennsylvania.
Computational details appear in Figure 3.2a. Pennsylvania monthly and annual average checks
are defined as follows:
Monthly for single-observation maximum') check: 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.
;': ' ; ' ' ''', I1
Annual average check: A robust, asymptotic estimator of an 95 percent confidence interval for
the median is calculated for the baseline period and a post-baseline period; if the post-baseline
interval exceeds the baseline interval, an exceedance is declared.
1 " ,,i ,, i,.
EPA strongly recommends that Pennsylvania's procedure be modified so that corrective actions
are triggered after three exceedances of the single-observation maximum have occurred within a
two-month period. Weekly monitoring should be initiated promptly, within 7-10 days, after a
single exceedance. Occurrence of two more exceedances (whether consecutive or not) during
this weekly monitoring should then result in appropriate and effective corrective actions. In a
final rule and in the final version of this Statistical Support Document, EPA may evaluate other
3-4
Statistical Methodology
-------
Coal Remining Statistical Support Document
decision rules and may identify decision rules that provide a reasonable balance between the two
decision error rates.
This modification would increase the ability to detect exceedance of the baseline pollutant level,
while providing sufficient protection against unnecessary monitoring. Pennsylvania's single-
observation check requires six consecutive exceedances to trigger treatment. Even when the
probability of an exceedance during remining is high, say 0.5, the chance of seeing six
consecutive exceedances is still low. For example, after one exceedance, the odds of seeing five
more consecutive exceedances is about 1:31 (assuming no serial correlation; in practice the odds
may well be higher). The modification provides substantially better protection. The risk of
unwarranted monitoring is still low. When the observed maximum of twelve baseline loadings is
used as the single-observation maximum limit, the probability is 0.95 that this value is at least as
large as the true baseline 77.91-th percentile, and the probability is 0.90 that this value is at least
as large as the true baseline 82.54-th percentile. Therefore, when the baseline pollutant level
distribution is not being exceeded during remining, then with 95 percent confidence the
probability is no more than 0.214 of observing two exceedances in four weekly observations, and
with 90 percent confidence the probability is no more than 0.143 of observing two exceedances.
Comments: (a) The annual intervals are robust but are not non-parametric. A suitable non-
parametric procedure would apply the Wilcoxon-Mann-Whitney test to compare baseline to post-
baseline periods; use of this test would not greatly decrease, and could increase, statistical
power, (b) The annual intervals depend upon an asymptotic approximation and the intervals are
symmetric. Loadings data for pre-existing discharges are highly asymmetric, and annual means
and medians are likely to be somewhat asymmetrically distributed. Therefore, suitability of the
approximation needs to be evaluated for small samples (e.g., n = 12). (c) The single-observation
check uses a non-parametric estimate of a percentile, equivalent to a 1-sided tolerance bound.
For n = 12, and using the maximum baseline observation as the upper bound, the probability that
a proportion P of the baseline distribution is not greater thaa this maximum is 1 - P12. For
example, the probabilities are 0.93, 0.72, and 0.46 that 80 percent, 90 percent, and 95 percent,
Statistical Methodology
.3-5
-------
Coal Remitting Statistical Support Document
respectively, of the baseline distribution lies below the observed maximum. Therefore, twelve
baseline observations provide a rather imprecise tolerance interval, (cl) As previously stated, the
single-observation check requires six exceedances before treatment is required. It might be
better if one exceedance triggered prompt follow-up monitoring for confirmation, and one or two
additional exceedances lead to a site investigation and remedial or corrective actions.
3.2.2 Procedure B
Procedure B consists of three checks: an upper limit on single observations, a yearly test of the
mean or median, and a cumulative monthly evaluation with a cusum test. Computational details
of Procedure B are provided in Figure 3.2b.
The single-observation limit is a parametric estimate of the 99th percentile of loadings,
developed using baseline data. The method of calculating this limit assumes that loadings are
approximately log-normally distributed (using natural logarithms). The methodology is similar
to that used by EPA to calculate "maximum daily" limits for effluent discharges. It would be
acceptable to substitute a non-parametric estimate of a high percentile if at least 50 baseline data
results were available. That estimate would be the k-th largest of n data, and would estimate the
100k/(n+l) percentile; e.g., the largest of 50 observations is a non-parametric point estimate of
the 98th percentile.
The annual test of the average or median employs either the t-test (using the natural logarithms of
the loadings) or the non-parametric Wilcoxon-Mann-Whitney test.
The cumulative monthly evaluation employs a cusum test. This is expected to detect an increase
in the mean loading somewhat sooner than the annual test in many cases. More important, it is
expected to provide better detection of a long-term gradual increase and a long-lasting step
increase than would be provided by the annual test. The cusum test as employed here, requires
that data be approximately normally distributed; for that reason, transformation of loadings
3-6
Statistical Methodology
-------
Coal Remining Statistical Support Document
would be required. Computational details appear in Figure 3.2b. Procedure B also calls for
reporting time-series plots of monthly data and the monthly cusum statistic, along with the
baseline statistics.
Comments: Accuracy of the parametric estimates depends on approximate log-normality of the
data. If the data are even more heavy-tailed (e.g., gamma distribution), this procedure could
under-estimate the percentiles; if less heavy-tailed, it could over-estimate percentiles. The data
are highly asymmetric, and annual means are likely to be somewhat asymmetrically distributed.
Therefore, suitability of the approximation needs to be evaluated for small samples (e.g., n=12).
Estimates of p, should not be made for individual discharges using n=12 data. Reliable estimates
require a larger number of data, possibly from a different sample than that used to estimate the
mean and variance (np> 30 is necessary). It should also be possible to use the average of P!
values calculated for a number of discharges having similar flow and concentration relationships.
3.2.3 Use of Triggered or Accelerated Monitoring in Procedures A and B
Triggered or accelerated monitoring (Figure 3.2c) can be applied with Procedure A or B. This
consists of accelerated monitoring after a single exceedance of baseline, or after an exceedance or
continued exceedances of the cusum warning level. Triggered or accelerated monitoring
provides a way to confirm an exceedance seen during routine monitoring. A decision is based on
the new monitoring results. Accelerated monitoring would begin promptly (within 7-10 days of
the exceedance) and would be conducted weekly or more frequently, for 3-6 sampling occasions.
Accelerated monitoring (if used as a condition or option for determining non-compliance) could
guard against a declaration of non-compliance on the basis of a transient exceedance, and would
provide a means to demonstrate continuing exceedances. It could guard against the possibility of
instituting expensive remedial measures when there was no continuing exceedance of baseline
conditions, or when simpler remedial measures can quickly be implemented.
Statistical Methodology
3-7
-------
Coal Remining Statistical Support Document
Figure 3.2a: Procedure A: both tests 1 and 2 are applied
Xj = pollutant loading measurement (product of flow and concentration measurements)
n = number of Xj results in the baseline dataset
1. Single-observation trigger
Order all n baseline measurements such that x(1) is the lowest value, and x(n) is the highest.
If n<17, then:
The single-observation trigger will equal the maximum baseline value, x(n).
Ifn>16then:
Calculate the sample median (M) of the baseline events:
If n is odd, then M equals x(n/2+I/2).
If n is even, then M equals 0.5*(x(n/2)+ x(n/2J.,)).
Calculate M, as the median between M and the maximum x(n).
Calculate M2 as the median between M, and x(n).
Calculate M3 as the median between M2 and x(n).
Calculate M4 as the median between M3 and x(n).
The single-observation trigger L equals M4.
If, during remining, an observation exceeds L, proceed immediately to weekly monitoring for four
weeks (four weekly samples). If, during weekly monitoring, any two observations exceed L, declare
exceedance of the baseline distribution.
2. Annual test
Calculate M and M, as described above.
Calculate M-, as the median between the minimum x(I) and the sample median.
Calculate R= (M, - M-,).
The subtle trigger (T) is calculated as:
M+
1.5 8* [(1.25 *jg)]'
Calculate M' and R' similarly for a year's data during re-mining. Calculate T' = M* -
(1.58* 1.25*R')/(1 -35 \/n'). If T' > T, conclude that the median loading during re-mining has
exceeded the median loading during the baseline period, and declare an exceedance.
3-8
Statistical Methodology
-------
Coal Remining Statistical Support Document
Figure 3.2b: Procedure B: All three tests or limits are applied.
X;
y\
n
s/
Z95
Z99
- pollutant loading measurement (product of flow and concentration measurements)
= number of observations, V;, in the baseline dataset
= D(yj)/n l
-------
Coal Remining Statistical Support Document
Figure 3.2b (continued): Procedure B
2. Cumulative Sum (Cusum) test (decision interval) 5
Limit value: C, must be less than H - declare an exceedance if Ct reaches or exceeds H.
Set C0 = 0. Index each new datum sequentially (e.g., t = 1 to 60), and calculate
C*. = MAX{0,Ct+(Y, -K)},
that is, the new sum CtH., is C, + (Y, - K), if that is positive or else it is zero.
Calculate K = Ey + fSy, using f= 0.25.
Calculate H = h Sv, using h = 8.0.
Cusum Warning level: Also calculate W, = MAX{0, Wt +(Y, -Kw) }, as done for Ct, using
K,v = Ey'+ fSy, using f= 0.5.
Hw = hSy, using h = 3.5.
A warning is indicated if W, reaches or exceeds Hw.
Keep and report charts showing C, and W, vs. month or successive observation number, and
showing the Cusum Limit and warning levels H and H^,. Consider making an investigation and
taking action when the warning level is reached.
3. Annual comparison 4
Compare baseline year loadings with current annual loadings using the Wilcoxon-Mann-Whitney test6
for two independent samples. Alternatively, use the two-sample t-test with log-transformed data (using
natural logarithms), y,. Use a one-tailed test with alpha" 0.025 to 0.05.
5 Wetherill, BG & Brown, DW, 1991, Statistical Process Control, Sections 7.1.7 and 7.2.1, and Table 7.6.
6 See Conover, W.J., 1980, Practical Nonparametric Statistics, 2nd ed., and other textbooks.
3-10
Statistical Methodology
-------
Coal Remining Statistical Support Document
Figure 3.2c: Accelerated (Triggered) Monitoring.
The following could be implemented as a permit condition that requires weekly monitoring after one or
more exceedances of a daily maximum level or a Cusum warning level during routine monitoring
(monthly monitoring is assumed to be the routine). Accelerated monitoring is intended to confirm the
continuation and the magnitude of a state of non-compliance, not to confirm the first exceedance (which
is in fact an exceedance). Thus, accelerated monitoring is best triggered after a warning level is exceeded,
and before a legally enforceable limit is exceeded. Professional judgement may be exercised in
interpreting the data. For example, did the exceedance coincide with record flows and rainfall ? Do the
accelerated monitoring data suggest a return to baseline levels, or a trend of rapid decrease ?
Accelerated (Triggered) Monitoring.
Promptly, within 5-10 days after an exceedance of a maximum level or a warning level (below),
begin weekly monitoring. Require monitoring for three to six weeks in all; six observations are
recommended to quantify the loading. Declare confirmation after observing one exceedance. The
probabilities of false positives and false negatives can be Inferred from Table 3.2a (which applies to
uncorrelated data; we expect the probabilities to be higher for positively correlated data).
Application to Procedure A.
Single-Observation Maximum Level. Apply this procedure after the single-observation maximum
limit is exceeded.
Application to Procedure B.
Single-Observation Maximum Level. Apply this procedure after the single-observation maximum
limit is exceeded.
Cusum warning level. Apply this procedure after the warning level Hw in the Cusum test is
exceeded once, or twice in succession. Apply it before the Cusum limit H is exceeded or is likely to
be exceeded by the next observation. In addition to checking for an exceedance of the single-
observation maximum during weekly monitoring, use the weekly data to continue the Cusum, and
observe whether the Cusum limit H is exceeded.
Statistical Methodology
3-11
-------
Coal Remining Statistical Support Document
Table 3.2a: Exceedances Probabilities for Given Numbers of Samples
Probability (for independent, uncorrelated data) of observing data exceeding L at least X times when the
true probability is P that data will exceed L. In routine or accelerated monitoring, some level L is set so
that L should be exceeded only rarely (P = 0.95 to 0.99). Thus, only rarely (with low probability) will
there be one or more observations out of N for which L is exceeded. If the true probability P that data
may exceed L is larger, there is a higher probability of observing X = 1, 2, or 3 exceedances of L when
taking N observations.
P
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
0.95
0.975
0.99
N = 3
X* 1
1.00
0.99
0.97
0.94
0.88
0.78
0.66
0.49
0.27
0.14
0.07
0.03
Xs2
0.97
0.90
0.78
0.65
0.50
0.35
0.22
0.10
0.03
0.01
0.00
0.00
Xs3
0.73
0.51
0.34
0.22
0.12
0.06
0.03
0.01
0.00
0.00
0.00
0.00
N = 4
Xs 1
1.00
1.00
0.99
0.97
0.94
0.87
0.76
0.59
0.34
0.19
0.10
0.04
Xa2
1.00
0.97
0.92
0.82
0.69
0.52
0.35
0.18
0.05
0.01
0.00
0.00
Xs 3
0.95
0.82
0.65
0.48
0.31
0.18
0.08
0.03
0.00
0.00
0.00
0.00
N = 5
Xs 1
1.00
1.00
1.00
0.99
0.97
0.92
0.83
0.67
0.41
0.23
0.12
0.05
Xs2
1.00
0.99
0.97
0.91
0.81
0.66
0.47
0.26
0.08
0.02
0.01
0.00
X>3
0.99
0.94
0.84
0.68
0.50
0.32
0.16
0.06
0.01
0.00
0.00
0.00
N = 6
Xs 1
1.00
1.00
1.00
1.00
0.98
0.95
0.88
0.74
0.47
0.26
0.14
0.06
X;>2
1.00
1.00
0.99
0.96
0.89
0.77
0.58
0.34
0.11
0.03
0.01
0.00
Xs3
1.00
0.98
0.93
0.82
0.66
0.46
0.26
0.10
0.02
0.00
0.00
0.00
Tabled probabilities were calculated from binomial cumulative probabilities, as 1 - Pr( X-l; N, P).
In selecting a rule ("X in N") for accelerated monitoring, the goal is to balance the chance of a false positive
against that of a false negative. Because limits or warning levels are intended to represent 95th to 99th
percentiles, the chance of a false positive is given in the rows for P = 0.95, 0.975, and 0.99. The chance of a
correct confirmation, when P < .95 is given in the other rows; thus the probability of a false negative when P is
truly 0.7 (which means that exceedances are expected to occur 30% of the time) is 0.22 for the rule (N=3, X=2).
3-12
Statistical Methodology
-------
Coal Remining Statistical Support Document
3.3 Statistical Characterization of Coal Mine Discharge Loadings
Permit data submitted for EPA's use in development of a coal renaming database (EPA, 1999)
were used to characterize variability of eastern coal mine discharges. This collection of permit
data is not a random or statistically designed sample. The survey solicited data for mine sites
that would provide examples of BMP performance. Unpublished reports (Griffiths, in
preparation) and published sources (Brady et al., 1998, Hornberger et ah, 1990) were also used.
Discharge flows, concentrations, and loadings vary remarkably among monthly or weekly
samples, over the course of 1-4 years (Brady et al., 1998, Griffiths, in preparation). Sample
coefficients of variation (CV) for iron loadings range approximately from 0.25 to 4.0. Sample
CVs for manganese loadings ranged from 0.24 to 3.8. These sample statistics were calculated
for a selection often mine sites having 42 discharges. The CV for the least variable discharge at
each site ranged from 0.24 to 1.5 across the ten sites; the CV for the most variable discharge at
each site ranged from 0.85 to 4.7.
There is significant, positive autocorrelation of flows and concentrations in mine discharges
over periods of 1-4 weeks (Griffiths, in preparation). From Griffiths' reports, the 4-week
autocorrelation of flow is 0.4-0.8. Griffiths did not report statistics for loadings.
Loadings of iron, manganese, and sulfate appear to be approximately distributed lognormally.
Normality of log (loading) is often rejected by the Shapiro-Wilk test for the larger samples of 80
to 100 data results, but skewness and kurtosis are reduced considerably by log transformation.
For log transformed loadings, the sign of the kurtosis estimates is not consistently positive or
negative; about 54 percent of samples have skewness estimates less than one in absolute value,
and about 83 percent, less than two. For log transformed loadings, the skewness estimate is
more often negative than positive, and about 80 percent of samples have skewness less than one
Statistical Methodology
-------
Coal Retaining Statistical Support Document
in absolute value. For untransformed loadings, non-normality is often confirmed in small
samples and the skewness and kurtosis estimates are usually large and positive.
More important than normality of individual data for parametric hypothesis testing is
homogeneity of variances and approximate normality of averages. Log transformation
eliminates the relation between variance and mean that is apparent in untransformed loadings
data, and so provides a variance stabilizing transformation. T-statistics, calculated for a test of
the difference between baseline and post-baseline periods, are normally distributed in aggregate,
indicating that the standardized difference in averages is very nearly normal. However, data are
still asymmetric after transformation, so suitability of the t-test needs to be evaluated for small
samples (e.g., n = 12).
These observations indicate that the usual parametric hypothesis tests can be used to compare
the average loadings during and after baseline, if one accounts for autocorrelation.
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.
Griffiths, J.C., RJ. Hornberger, and M.W. Smith (in preparation). Statistical Analysis of
Abandoned Mine Drainage in the Establishment of the Baseline Pollution Load for Coal
Remining Permits. U.S. Environmental Protection Agency, Washington, D.C.
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.
U.S. EPA, 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.
3-14
Statistical Methodology
-------
Coal Remining Statistical Support Document
U.S. EPA, 1999. Office of Water. Coal Remining Database: 61 State Data Packages, March
1999. Details provided by the U.S. Environmental Protection Agency's Sample Control
Center, operated by DynCorp I&ET, 6101 Stevenson Avenue, Alexandria, VA 22304.
Statistical Methodology
3-15
-------
Coal Remining Statistical Support Document
3-16
Statistical Methodology
-------
Coal Remining Statistical Support Document
Section 4.0 Baseline Sampling Duration & Frequency
4.1 Power and Sample Size
To determine an adequate sample size (number of samples), EPA will require approximate
statistical power of at least 75 percent (TT = 0.75) for a statistical decision procedure used to
detect a difference of one standard deviation over a period of one year, while requiring a Type I
error rate (significance level) of 5 percent (a = 0.05). Lower power is acceptable on a monthly
(or single-observation) basis, but cumulative and annual decision procedures are required to have
at least 75 percent power over the course of one year.
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 alpha 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.
Power was calculated for the one-sided, two-sample t-test with alpha = 0.05 (Table 4.1 a). This
does not apply exactly to alternative statistical decision procedures, but EPA believes that it
provides a reasonable basis for calculating a minimum sample size. The effect of autocorrelation
upon power of the t-test was examined (using an approximate method) and was found to be
small.
Baseline Sampling Duration and Frequency
-------
Coal Remining Statistical Support Document
Table 4.1a: Power of one-sided, two-sample t-test with equal n in each
group, a = 0.05
N
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Difference in means divided by sigma, (m - m) /
0.5
0.07
0.10
0.13
0.16
0.18
0.21
0.23
0.25
0.27
0.29
0.31
0.33
0.35
0.37
0.39
0.41
0.42
0.44
0.46
0.47
0.49
0.51
0.52
Values were calculated
(Millard, 1998)
1.0
0.10
0.21
0.31
0.39
0.47
0.53
0.59
0.64
0.69
0.73
0.76
0.80
0.82
0.85
0.87
0.88
0.90
0.91
0.93
0.94
0.95
0.95
0.96
using function
1.5
0.15
0.39
0.57
0.69
0.77
0.84
0.88
0.91
0.94
0.96
0.97
0.98
0.98
0.99
0.99
0.99
1.00
1.00
1.00
1.00
1.00
1.00
1.00
"t.test.power()" from
2.0
0.23
0.62
0.79
0.89
0.94
0.96
0.98
0.99
0.99
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1,00
Environmental
2.5
0.36
0.80
0.92
0.97
0.98
0.99
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
Stats
0
3.0
0.53
0.90
0.97
0.99
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
1.00
for S-Plus
4-2
Baseline Sampling Duration and Frequency
-------
Coal Remining Statistical Support Document
To estimate the Type I and Type II error rates for more complex decision procedures, simulations
or resampling of a large baseline dataset would be needed. The Type I and Type II error rates of
the statistical procedures proposed in this document have not yet been fully evaluated by EPA.
To approximate the power likely to be provided (in the least favorable case) by a most-powerful
non-parametric test, EPA note's that the Wilcoxon-Mann-Whitney test has asymptotic relative
efficiency (ARE) at least 0.864 compared to the t-test, and may have ARE >1 for heavy-tailed
distributions.
The criterion for sample size applied above was the ability to detect a change of one standard
deviation above baseline loadings with high power. An increase of one standard deviation can
represent a large increase in loading, given the large variability of flows and loadings observed in
mine discharges. The coefficient of variation (CV) is the ratio of standard deviation to mean.
Sample CVs for iron loadings range approximately from 0.25 to 4.00, and commonly exceed 1.
Sample CVs for manganese loadings range approximately from 0.24 to 5.00. When the CV
equals 1, an increase of the average loading by one standard deviation above baseline means a
doubling of the loading.
Loadings from mine discharges appear to be approximately distributed lognormally. Thus,
logarithms of loadings may be approximately distributed normally, justifying use of a t-test.
Based on these considerations and the power of the t-test (Table 4. la), EPA has determined that
the smallest acceptable number and frequency of samples is 12 monthly samples, taken
consecutively over the course of one year. This number is approximate and represents the
absolute minimum. 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. In such cases, more
samples should be taken, enough to provide the required statistical power.
Baseline sampling Duration and Frequency
-------
Coal Remining Statistical Support Document
A permitting authority may want to consider the statistical power appropriate for environmental
protection, and may decide to require 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.
4.2 Sampling Plan
Based on considerations described above, EPA requires a minimal sampling plan for establishing
baseline and post-baseline pollution load that consists of taking one sample per month for one
year. The duration, frequency, and seasonal distribution of sampling are important aspects of a
sampling plan, and can affect the precision and accuracy of statistical estimates as much as can
the number of samples.
Sampling, during and after baseline, should systematically cover all periods of the year during
which substantial discharge flows can be expected. Sampling should not bias the baseline mean
toward high loadings by over-sampling the high-flow months. Unequal sampling of different
time periods can be accounted for using statistical estimation procedures appropriate to stratified
sampling.
Twelve samples are unlikely to provide a representative sample of discharges for the year, given
the variability observed for coal mine discharges. Eighteen to twenty-four samples seem much
more likely to adequately characterize a baseline year.
EPA proposes a minimal sampling plan for baseline and post-baseline that consists of taking one
sample per month for one year.
There may be acceptable alternatives to the proposed minimum duration and frequency of one
sample per month for twelve months. The merits of alternative sampling plans have notbeen
evaluated thoroughly. Alternative plans could be based upon subdivision of the year into distinct
4.4 Baseline Sampling Duration and Frequency
-------
Coal Remining Statistical Support Document
time periods that might be sampled with different intensities, and other types of stratified
sampling plans (e.g., Sanders et al, 1983, and Griffiths, 1990) that attempt to account for
seasonal variations. Seasonal stratification has the potential to provide a basis for more precise
estimates of baseline characteristics, if the sampling plan is designed and executed correctly and
if results are calculated using appropriate statistical estimators.
Sampling should be designed to prevent biased sampling. In particular, flow measurement
methods 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
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".
Millard, 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.
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 remining 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 contained in an abridged form in Griffiths
et al. (in preparation).
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
5-1
-------
Coal Remitting 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
, nil ,
As part of title 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
Amot-3
Arnot-4
Clarion
Ernest
Fisher
Hamilton
Markson
Discharge
j Behavior Type
typical
| typical
typical
| slugger
; typical
typical
steady
Location ;
Tioga County, PA !
Tioga County, PA
Clarion County, PA
Indiana County, PA
Lycoming County, PA :
Centre County, PA
Schuylkill County, PA
Period of
Record
1980 - 1983
1980-1983
1982 - 1986
1981 - 1984
1982 - 1985
1981 - 1985
Number of
Samples
82
81
79
189
36
109
1984-1986 ! 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) 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.
2) 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.
3) 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.
Long-term Monitoring Data and Case Studies
5-3
-------
Coal Remitting Statistical Support Document
Baseline pollution load statistical summaries were calculated for each dataset using the
exploratory data analysis approach discussed in Section 3.0. It is rare, however, that datasets of
this duration and with as great a number of samples are available. Coal remimng 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 to
yield statistically valid results, was examined using the long-term datasets listed in Table 5.la. It
was 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
were then used to recalculate the baseline. The baseline, recalculated with fewer samples, was
compared to the "full data baseline." Several comparisons 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-
' ' ; ,' :i . . '',;.,: , 'ii;' '- .! V , . * "'- " ,- ... .'..{ !':'.
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. 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.
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). Baseline pollution load
summaries for each dataset are presented in Tables 5.1b and 5.1c. The tables 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
5-4
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
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 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 5.1c 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 5.1b: Comparison of Median Acidity and Iron Loads by Sample Period and
Interval
i Parameter iFuII Data j 9 month %
. j Data 1 Error
Arnot-3
I Number of Samples 82 ' 66 j
Median Acid Load 72.3 | 84.1 16.32%
Upper 95% C.I. ; 86.53 ! 101.59
i Lower 95% C.I. , 58.07 [ 66.61 :
'Median Iron Load 0.96 ' 1.17 ;21.88%
1 ! i
Upper 95% C.I. \ 1.26 [ 1.55 j
Lower 95% C.I. j 0.66 | 0.79 i
Arnot-4
^Number of Samples 81 66
'Median Acid Load 194 ; 221 \ 13.92 %
I Upper 95% C.I. < 232.31 j 263.22 j
Lower 95% C.I. ! 155.69 1 178.78 : i
Median Iron Load 2.70 | 3.00 11.11%
Upper 95% C.I. 3.35 j 3.78 j
Lower 95% C.I. 2.05 ' 2.22 j
Clarion
Number of Samples ; 75 53 | j
Median Acid Load ! 39.50 | 40.00
Upper 95% C.I. ; 49.11 ; 51.45 j
Lower 95% C.I. 29.89 ; 28.55 j
Median Iron Load 5.51 i 4.26 -22.69 %j
Upper 95% C.I. 7.27 6.07
Lower 95% C.I. 3.75 i 2.45
Ernest
Number of Samples 189 146
Median Acid Load 1456 1682 15.52%
Monthly % ; Quarterly 1 % Error
Samples Error Samples
43 i 14
73.9 .2.21%;, 72.3 iTlWO,%' ~»
91.09 102.71 ^ \
56.71 : 41.89 : l
0"95 0.96 -«<^1
1.27 . . . . i 1.46 ;
0.63 j 0.46 . ;
43 i . i ..14... ! ... , i
233.41 248.05
152.59 i 121.95 !
2.50 b^f41'%. 2.60
3.22 | 3.62 j ;
1.78 | 1.58 i j
I
41 ; 28 i
39-°° 63^S:'?i 40-00
.52.01 54.74
25.99 j 25.26 :
4.45 I -19.24% 7.37 i 33.76%
7.02 10.51
1.88 4.23 ;'
53 ' 19
2048 40.66% 1882 29.26%
Long-term Monitoring Data and Case Studies
5-5
-------
Coal Remining Statistical Support Document
Parameter
| Upper 95% C.I.
Lower 95% C.I.
Median Iron Load
Upper 95% C.I.
Lower 95% C.I.
JFull Data j
i i
' 1991.91 '
; 920.09 ;
229
! 342.83 i
! 115.17 1
1 f
9 month %
Data Error
2410.88 !
953.12 ; j
264 ! 15.28 % !
412.68 ' !
115.32 j j
Monthly | %
Samples j Error
2923.35 I
1172.65
304 ; 32.75 %
474.61
133.39 j
. ; - f «. . ,:: , .. , .1
Fisher
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.
| 35 !
i 72 j
[ 95.00 !
! 49.00 !
; 1-4 i
' 1.74 ;
1 IM I
24 ; ;
82 : 13.89 % !
109.47 : ;
54.53 | j
1.4 H|00fi«|
1.75 : i
1.05 j |
24
85 18.06 %
121.73
Quarterly % Error
Samples ;
3805.81
-41.81 :
348 1 51.97%
662.48 :
... , " , , ', f , t . , u
10 ; i
102 , 41.67% i
165.38
48.27 j" .' 38.62 !
1.4
1-4
1.66' ' [ 1.63 | , ;.
1.14
" ;..;. .LI? , j. . f ir i.
Hamilton-8
Number of Samples
Median Acid Load
Upper 95% C.I.
i Lower 95% CJ.
Median Iron Load
! Upper 95% C I
Lower 95% C.I.
I 109.00 j
! 59.00 !
i 67.92 i
J 50.80 !
2.66 ;
i 3.19 |
: 2.13 !
85.00 , j
66.86 | 13.32 % j
77.16 i !
56.56 1 :
3.12 1 17.29 %
3.76 ! i
2.48 : ;
52.00 38.00 j
58-70 55.70
68.90 1 72.60
48.50 j 38.30 ;.
2-63 . I:81 -31-95..°/0 ,.
3.45 ; 2.77
1.81 j 0.85
Markson
Number of Samples
Median Acid Load
Upper 95% C.I.
Lower ?5%C.I.
: Median Iron Load
Upper 95% C.I.
' Lower 95% C.I.
"> 98 ' ;
! 1467 1
! 1575.47 |
1358.53 |
; 408 |
: 430.76 j
; 385.24 |
77 j
1452 ||[Gl2l|,j
1597.55 i !
1306.45 i i
402 ' lUmjij
428.18 ' !
375,82 ; i
30 ! 22
1491
1624.02
1 546 fffm^ffwf~w'"
L->^°
1816.11
1357.98 | 1275.89
402 402
428.14 434.13 ;
375.56 369.87 ;.
^Average of All Discharges
'Median Acid Load
Median Iron Load
i i
i i
i !
' 10.75 % i
I 12.82%
12.55% ;
: 17.55% ;
5-6
Long-term Monitoring Data and Case Studies
.ill!!!!
-------
Table 5.1c: Comparison of Median Acidity and Iron Loads by Baseline Sampling Year
| Parameter Full Data 1980 1981 , 1982 1983 i 1984 1985
jArnot-3 ;
.Number of Samples 82
i Median Acid Load j 72.3;
17| 21 27| 15 I
66.9! 63.8! 83.9! 86.5: i
1 Upper 95% C.I. j 86.53 94.23 76.79: 110.69! 128.53 i ' i
; Lower 95% C.I. 58.07 34.37 46.62; 43.72: .39.51
j Median Iron Load 0.96 j
i Upper 95% C.I. i 1.26
i Lower 95% C.I. , 0.66
0.57 0.98; 1.44; 1.17]
0.90 1 1.37; 2.04 i 2.12]
0.24J 0.59; 0,841 . 0.22 i '
;Arnot-4
Number of Samples 8 1 i
j Median Acid Load j 194;
17 201 29! 15 i ! ' '
157 159: 208| 242' !
j Upper 95% C.I. ,! 232.31 253.83 209.32 256.19! 368.52 '
; Lower 95% C.I. 155.69 60.17 108.68 159.811 115.48 j
Median Iron Load 2.7 j
Upper 95% C.I. 3.35
| Lower 95% C.I. 2.05
1.5 1.6; 3.0| 3.0] i
2.95 1.76, 4.111 5.20, '
0.05! 1.44, 1.89! O.SOl .]
[Clarion
Number of Samples 75 i
Median Acid Load 3 9.5 !
17 20; HI 16) 9
41.0 56.5| 27.0J 14.0] 42.0
Upper 95% C.I. ' 49.11 ; 74.00J 75.80J 46.41 1 27.99 61.26] !
Lower 95% C.I. 29.89'
Median Iron Load ' 5.51 :
Upper 95% C.I. ; 7.27
Lower 95% C.I. ' 3.751
Ernest
'Number of Samples 1 89 i
Median Acid Load 1456; 1
8.00; 37.20! 7.59] 0.01 22.74]
4.13 10.78i 7.65] 1.66 5.69^
7.19 14.49! .18-°0| 3.13 9.68
1.07J 7.071 -2.70| 0.19] 1.70 \
i
16 381 47 49! 39 i
736] 615] 574J 5193 1697 !
Upper 95% C.I. 1991.91; 2742.88] 1141.38! 1295.70J 6551.26 2906.26J
Lower 95% C.I. j 920.09: 72
Median Iron Load ; 229!
9.12 88.62! -147.70 3834.74] 487.74!
225 85| 601 1069! 216
Upper 95% C.I. 342.83 327.04 169.49: 142.371 1346.84 448.62
Lower95%C.I. , 115.17. 122.96] 0.5 Ij -22.37 791.161 -16.62
Fisher
Number of Samples ! 35 i
Median Acid Load j 72 i
9 8! 171 21! 12 8|
49! 101! 80! 36 26| 42j
Upper 95% C.I. i 95.00 86.76 202.90: 119.39J 45.26) 41.20! 60.101
Lower 95% C.I. ! 49.00J 11.24 -0.90' 40.61 26.74! 10.801 23.90]
Median Iron Load ; 1.4|
1-5 2.1 ' 1.2i 0.9: 0.2' 0.2:
Upper 95% C.I. ; 1.74) 2.13! 3.65: 1.59] 1.12 0.41 ! 0.36i
Lower 95% C.I. : 1.06 0.87 0.55; 0.81 0.68i -0.01; 0.04:
Hamilton-8
Number of Samples 109!
i
16] 24' 27i 25i 17! i
Long-term Monitoring Data and Case Studies
5-7
-------
Coal Remitting Statistical Support Document
\ Parameter
\ Median Acid Load
1 Upper 95% C.I.
| Lower 95% C.I.
; Median Iron Load
j Upper 95% C.I.
i Lower 95% C.I.
Markson
jNumber of Samples
(Median Acid Load
1 Upper 95% C.I.
! Lower 95% C.I.
i Median Iron Load
; Upper 95% C.I.
| Lower 95% C.I.
Full Data
59.00'
67.92!
50.80;
2.66 !
: 3.10;
i 2.131
i 98i
i 1467 i
' 1575.471
: 1358.53!
: 408
i 430.76
385.24
1980
56.70
79.01
34.39
4.35;
6.74.
1.96:
15!
1502!
1726.371
1 277.63 i
336!
406.26 1
265.74J
1981
69.10
87.02
51.18
3.50
4.98
2.02
49
1327
1445.08
1208.92
403
423.90
382.10
1982 |
37.60!
60.69!
14.51 ;
1-071
1.72J
0.42 i
i
34|
1888J
2366.2 1!
1409.79 1
449 i
..5 12.73 j
385.27:
1983 !
54.54,
71.41;
37.67:
1.53
2.16;
0.90;
'
!
" !
1
!
i
1984
77.40
91.27
63.53
3.34
4.21
2.47
1985 '
1
»
«
i
! ..
, , i
!
I-
;
r
5.1.1 Sampling Interval
"'I ' \ " !. i ' ' . | V ' ,; ' ,,, ' ij !"
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
n, '"",. .. , ' ','",, i '"' ' '! ]1 ! , " ,; ' 'Si
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
i'ir , , , :' r1 , iiir
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 verses 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)
5-8 Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
as listed in Table 5.1b. However, because each sample subset contains a different number of
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
-------
Coal Remining Statistical Support Document
Figure 5.1a: Arnot-3 Acidity Loading (1980-1981)
140-3 ARN07 3 ACID LOAD YEARLY COMPARISON
120 !
100:
80-
60-i
40-
20-
1980 1981 1982 1983
Figure 5.1b: Arnot-3 Flow Data Comparison
ARNOT 3 FLOW DATA TYPE COMPARISON
160.-=.
140-E
II i -
120|
I100]
3 -
1 1
LL.
60 i
,' . ... 40-=
",i , 20 i
6^
/
-
-
-
"
**
ALL NINE MONTH ,',, MONTHLY .QUARTERLY ,',
5-10
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure 5.1c: Arnot-3 Iron Load Data Comparison
ARNOT 3 IRON LOAD DATA TYPE COMPARISON
2-
o
o
ALL NINE MONTH MONTHLY QUARTERLY
Figure 5.1d: Arnot-3 Acidity Load Data Comparison
ARNOT 3 ACID LOAD DATA TYPE COMPARISON
1407
120 i
100-E
a
f 803
S 60-E
g
< 40-E
20 -.
ALL NINE MONW MONTHLY QUARTERLY
Long-term Monitoring Data and Case Studies
5-11
-------
Coal Remining Statistical Support Document
Figure S.le: Arnot-3 Monthly Flow Comparison
250-1
200 :
150-
100-
50-
ARNOT 3 FLOW MONTHLY COMPARISON
JAN FEE MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Figure 5.1f: Arnot-3 Monthly Acidity Load Comparison
250-1
200-
150-
I
100-
50-
ARNOT 3 ACID LOAD MONTHLY COMPARISON
h.^|.
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
5-12
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure 5.1g: Arnot-4 Acidity Load Data Comparison
ARNOT 4 ACID LOAD DATA TYPE COMPARISONS
§
o
o
ALL . NINE MONTH MONTHLY QUARTERLY
Figure S.lh: Arnot-4 Acidity Load (1980-1983)
400
3.00--
4[ 200-
_Q
° 100-
Q
-100:
2QQ3.
ARNOT 4 ACID LOAD 1980 - 1983
-
iy«o iysi 19S2 1933
Long-term Monitoring Data and Case Studies
5-13
-------
Coal Remining Statistical Support Document
Figure 5.1i: Clarion Acidity Load Data Comparison
CLARION ACID LOAD DATA TYPE COMPARISON
o_
ol
3
,' ALL NINE MONTH MONTHLY QUARTERLY
Figure 5.1j: Clarion Iron Load Data Comparison
CLARION IRON LOAD DATA TYPE COMPARISON
10-
-8
o
o;
5-
0-
-5-
ALL NINE MONTH MONTHLY QUARTERLY
5-14
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure 5.1k: Clarion Acidity Load (1982-1986)
TOO-,
80-
'60-
40-
20-
0-
-20-
CLARION ACID LOAD YEARLY COMPARISON
1982 1983 1984 19S5 19S5
25-3
20-
Q 53
§
i o-.
-10-
Figure 5.11: Clarion Iron Load (1980-1983)
CLARION IRON LOAD YEARLY COMPARISON
1980 1981
1982 1983
Long-term Monitoring Data and Case Studies
5-15
-------
Coal Remitting Statistical Support Document
Figure 5.1m: Ernst Acidity Load Data Comparison
6000 -I
4000 :
2000 :
ERNST ACID LOAD DATA TYPE COMPARISON
-2000:
-4000-
ALL NINE MONTH BIWEEKLY MONTHLY QUARTERLY
Figure 5.1n: Ernst Iron Load Data Comparison
ERNST IRON LOAD DATA TYPE COMPARISON
1200-
-i
800-
§ . oq
ALL NINE MONTH BIWEEKLY MONTHLY QUARTERLY
5-16
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure S.lo: Ernst Acidity Load Data Comparison
12000 n
8000-
4000 -
0-
-4000-
-8000:
ERNST ACID LOAD MONTHLY COMPARISON
I-
JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC
Figure 5.1p: Ernst Acidity
Load (1981-1985)
7000 a
6000 \
5000
4000'
3000-
O 200D-E
1000-E
0-
-1000-
-2000^
ERNST ACID LOAD YEARLY COMPARISON
1931 1962 1983 . 1984 . 1985
Long-term Monitoring Data and Case Studies
5-17
-------
Coal Remining Statistical Support Document
Figure S.lq: Fisher Monthly Acidity Load
RSHER ACID LOAD DATA TYPE COMPARISONS
B-,
CM-
o "~
>
2
§ o
-1 O-
0_
BEFORE AFTER NINE MONTH MONTHLY QUARTERLY
Figure 5.1r: Fisher Iron Load Data Comparison
FISHER IRON LOAD DATA TYPE COMPARISONS
.CN-^
13
J <*E
f
_J I
' ' 1 ^
-
-
-
- ;, .
-
-
-
- -
|j 'I'lll' 111.
BEFORE AFTER NINE MONTH MONTHLY QUARTERLY
5-18
Long-term Monitoring Data and Case Studies
f Uw , !, ",',.,1,1 LAI!:!;,! Illililllli f'hhiii,: / ;'',: "" I1'!'1 in,' I, 'Ti';!i ,'.i .1-
-------
Coal Remining Statistical Support Document
Figure 5.1s: Fisher Acidity Load Data (1982-1987)
200-
150-
100-
50-
0-
-50-
-100-
FISHER ACID LOAD 1982 - 1987
I-
1982 1983 1984 1985 19S6 1987
Figure S.lt: Fisher Iron Load
Data (1982-1987)
4-
3-
2-
1-
0-
-1-
FISHER IRON LOAD YEARLY COMPARISONS
1982 1983 1384 1985 1986 19S7
Long-term Monitoring Data and Case Studies
5-19
-------
Coal Remitting Statistical Support Document
Figure S.lu: Hamilton-8 Acidity Load Data Comparison
HAMILTON 8 ACID LOAD DATA TYPE COMPARISON
₯
o
p
O;
03.
ALL NINE MONTH MONTHLY QUARTERLY
Figure 5.1v: Hamilton-8 Iron Load Data Comparison
HAMILTON 8 IRON LOAD DATA TYPE COMPARISON
,'. "' . ^ 4-i
t 1
. " tn 3 r
. a -
§ 2i
, t ~
1 z' =
i 1i
', . . .. 0|
-
-
''i ; - " ALL NINE MONTH MONTHLY QUARTERLY
5-20
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
80-
60-
40-
20-
HAMILTON 8 ACID LOAD YEARLY COMPARISON
Figure S.lw: Hamilton-8 Acidity
Load (1981-1985)
1981 1982 1983 1984 1985
Figure S.lx: Hamilton-8
Iron Load (1981-1985)
7-
^ 4-3
§
§
1-E
HAMILTON 8 IRON LOAD YEARLY COMPARISON
19S1 1982 1983 1984
1985
Long-term Monitoring Data and Case Studies
5-21
-------
Coal Remining Statistical Support Document
Figure 5.1y: Markson Acidity Load (1984-1986)
2600^
2400
2200
^2000-i
1 i
1 1800:
1 \
, 3 1600 r
s i
1400^
1200^
1000
" «nn^
MAKKSON AC
ID LOAD YEARLY
_
JQMPAKISGN
-
1984
1985
1986
Figure 5.1z: Markson Iron Load (1984-1986)
550-.
400 -E
'
300-
250 -_
200 -_
150
MARKSON IRON LOAD YEARLY COMPARISON
1934
1985
1986
5-22
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure S.laa: Markson Figure S.laa: Acidity Load Data Comparison
MARKSON AC!D LOAD DATA TYPE COMPARISON
Q
2000-q
18005
16005
14005
12005
1000r
8005
600-
ALL NINE MONTH MONTHLY QUARTERLY
Figure S.lab: Markson Iron Load Data Comparison
MARKSON IRON LOAD DATA TYPE COMPARISON
500-i
450-
a
a
300-
250-
ALL NINE MONTH MONTHLY QUARTERLY
Long-term Monitoring Data and Case Studies
5-23
-------
Coal Remining Statistical Support Document
5.1.2 Duration of Baseline Sampling
Previous study of these datasets (Griffiths, in preparation) 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
1 ' ili" ' ' ' ' ' '! ''' " ' ' ' " ' ' ' '' " " '": ;
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 subserted 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 5.1 b.
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 witli using the full 12-month dataset. Acidity loading rates, which are dominated by
flows, parallel this effect (Figure 5.1 d).
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 follow 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 5.1e and 5.1f).
The remaining two discharges, Ernest and Markson, reacted very differently to changes in the
baseline sampling period and interval. The Ernest discharge (a "slugger 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 year-by-year basis (Table 5.1c) 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 Remining Statistical Support Document
5.1.4 Year-to Year Variability
III I I' ' I ,1 'I , I , llfl
Annual median pollution loads and 95 percent confidence intervals for each of the seven
discharges studied are presented in Table 5.1c. 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
dife The Ernest discharge, which has "slugger response" behavior,
tends to slipw extreme variability in both flow rate and load. As illustrated in Figure 5. Ip, this
It , " ,'. ' ' : "" ,' i,1 1 , ",, i , , ;.! ' ' "! ...!:' . r, 'V 'I.. ' ' i,/i' ' ' 'V.1' 1*
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 MiningInduced
Changes in Long-term Monitoring Data
A primary reason for establishing a baseline pollution load prior to rernining is to distinguish
between natural seasonal variations and mining-induced changes in flow and water quality that
may occur during rernining 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 Griffiths et al.
1, ' i , -"' '! '.. ". ., , ' . ' ""..., .. '. ' , . ',1. '":' i .»
(in preparation).
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 baseling pollution load at rernining 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, Schuylkill 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 Reminmg 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; Griffith et al, in preparation; and
Brady, 1998). In a 1988 study of Markson data containing approximately 100 samples collected
at weekly intervals from 1984 to 1986, Griffith 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 hi flow and concentrations of sulfate, acidity, iron, and manganese
are shown for an eight year period (1992-1999) in Figure 5.2b. The data was 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
5-29
-------
Coal Remining Statistical Support Document
\ !* 'iil;:
10000
1000.-,
100 .»=;
10
Figure 5.2b: Marksbn time Plot
Flow, pH, Acidity, Iron, Manganese, Sulfate
-Flow (gpm)
I&cidt
-Iron (mgfl)
-Manganese (rng/l) ;
_ Sulfate (mg/1)
s
u?
i
to CD
5> S 3>
'! " l| ill
Figure 5.2c: Markson Time Plot
Flow & Sulfate
Flow ,
Sulfatei
5-30
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
8000 ..
7000
Figure 5.2d: Marks on 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 attributable 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
-------
Coal Remining Statistical Support Document
Figure 5.2e: Markson Time Plot
Flow & Iron
8000
7000
6000
t FlnW
i_n_lron
Figure 5.2f: Markson Time Plot
Flow & Manganese
8000
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
Figure 5.2g: Markson Flow Measurement Comparison 71-iMonthiyF|OW
Flow 1994-1997 : _Con«nuous Flowi
03/05/94
03/05/95
09/02/96
In comparing the continuous flow line to the instantaneous monthly flows, the following
observations can be made:
1!! i, :, ' ' 'i ' . i, ,, i i 'I- ' ' i !,
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
,:'; ' :" '"'i , !' ' ' : :' . '-' , . ' . ,' >yr. :; ' Vr ""- ; .;' '; ,i,r i" . "-:.'",, , :. , "i'!';;, i! 'iiif
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 Time Plot
Flow, pH, Iron, Manganese, Sulfate
10000
1000
100
Row (gpm) "
pH |
Iron (rng/l)
Manganese (rrg/l)
Sulfate (rr^/l) ;
CM
^_
CO
o
25
^^
O>
o
CO
CD
^~
CO
o
CO
^~
o
s
m
^~
CO
0
sr
CO
^
O)
o
10
en
T
CO
o
If)
03
f\J
T
55
o
CD
gj
CM
T
CO
0
CO
o>
o
T~"
55
o
1
T
CO
o
9?
o
o
CO
OJ
o
T
CO
0
co
C5
CO
o
o
O5
O3
55
0
CO
o
O)
o
o
Long-term Monitoring Data and Case Studies
5-35
-------
Coal Remitting 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
; , , *.. :.,. , ., ., , ; . ./ , , )v
Markson discharge (Figure 5.2b). 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 1.153 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 for 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
'*" " : ', ill i1,;!!1: ; ,1; ,: .5. :ivi.,; ' ' ;, ' 1 li I ill I '" ,;"'i ': ' f ii riri" '"!: "'""l " ' ' :' "'r i
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
ii '.: , ' , ' , - ' .1,1 i. - , " ii i-S''
equals 2700 gpm).
,'I i 1
5-36
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
10000
Figure 5.2i: Tracy Airway
Flow 1994-97
; » _ Monthly Flow
Continuous Flow
03/05/94
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
400
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 (FigureS .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
,;|I|, , , , , ^ , i , ' , '
(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 Remining Statistical Support Document
10000
Figure 5.21: Tracy Airway
Flow, pH, Manganese
-Row
-PH i
.Manganese |
CM
92
CM
25
CO
o>
3
CO
92
^j-
25
a-
£
in
$
m
CO
CM
O3 O5
co 35 P:
g g g
O5 CO O>
o o o
A strong inverse relationship between flow and pollutant concentration in the Tracy discharge is
shown in Figures 52], 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 5.2]), iron (Figure 5.2k), and manganese (Figure 5.21). There appears to be a trend over
time, where the range of median values for sulfate and manganese are diminished from 1992
through 1997.
Monthly flow and water quality relationships of the Markson and Tracy discharges, throughout
the eight year period shown in Figures 5.2b through 5.21, indicate a general inverse relationship
between flow and concentration, but also show that the distribution, magnitude, and duration of
high flow events is not uniform from water year to water year. In fact, sometimes 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
Long-term Monitoring Data and Case Studies 5-39
-------
,,(1 ,, ;,» r»'i':;':!! t
'
Coal Remining Statistical Support Document
greater than one month, and a sampling duration of at least a water year (12 monthly samples) is
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
"\; "., ,]: .* ' ' - v :. A.---.' : r'&'i *>VV! .:,!>:"..'.'';..' t(V: ;,. : -::i ,:. if"'* i
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 less 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
''!' , , ' ' , "''liilili! I'll1 , .,!"'. 'i,i''i ,,,i ." 1 T, '''!', ' ,,:! ' , !: ' ,,,,;; i,], ,: "' ',, ,; >;,;; , ,;.,, «<«< ,,, , , , ". n, i.<|i' ^j i;
and Lebanon Counties, Pennsylvania since before 1960. The results of this data collection are
included in numerous publications including McCarren et al. (1961), and Fishel & 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
1 VII ,",;,, 11;,,,, . ^ -,-1 /V._j " , /,; " 'Viii,1;, i: '.,?.:$;. . ' . i ,-" ,, ' i ; I ',; ;.''
Station, the Swatara Creek watershed changes to a more agricultural land use without acid mine
' 'i'j ,r if . ' ».,: ' , i|in ,, 'i'" "' ' ;, '" ' ". , " ; ' i, " .i,yi1'.!.,.,!,!" I1 | i, .;, i ' i:|,j,,,, ,, ,,,'. J i,, ' :. J aj!1,, .W,,,,,
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
'"i|», ' : , ';,:'!!"' I1" nii ,» ' 'n i . .1 m i., ,n i,,,,.,j\ < ,,«. M, , i1 " v " .y., <» » M j 't i»"
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
if i, i "!|. .i mill " ",", '', n . ', ''i?;,,;1 , '. , u1';,;"' , " "' ' i ' !' , . 'i! *,. M
-------
Coal Remining Statistical Support Document
Figure 5.2m: Swatara Creek Flow and Sulfate Data
Figure 5.2n: Swatara Creek Flow and Suspended Solids Data
U5QO
5 §00
**i
** 800
I 700
S 600
w 500
S
400
«
2CO
100
-7,0
B
B
2
11
10
9
8
-7
6
5
4
-3
2
1
O
Pt
w
i
Long-term Monitoring Data and Case Studies
5-41
-------
e Coal Remitting Statistical Support Document
Figure 5.2o: Swatara Creek Flow and Iron Data
5JO
. 4J9
. 3JD
Ite.total
f 1JD
0
13
14
17
5.2.4 Jeddo Tunnel Discharge
The Jeddo Tiinneln^e 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.
.i ' The results of that monitoring for the water year from October 1, 1974 ttirough September 30,
1975 are shown in Figure 5.2p (Growitz et al., 1985). During that year, the discharge ranged
from 36 to 230 cfs (16,157 to 103,224 gpm). The Jeddo Tunnel discharge flows are compared to
the 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 had 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 S.2p: Jeddo Tunnel Discharge and Wapwallopen Creek Flow Data
10.000
5000
2000
1000
500
200
100
50
20
10
~i r
O
O
HI
CO
a:
LU
a.
o
m
O
g
LU"
a
cc
o
co
D
WAPWALLOPEN CREEK
JEDDO TUNNEL
WAPWALLOPEN CREEK
gfj JEDDO TUNNEL
OCT. NOV DEC.
1974
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 Balleron 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) hi October 1995 and 482 cfs (216,322 gpm) in
November 1996, following 3.89 inches of rainfall ( Balleron, 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 Balleron (1999).
Long-term Monitoring Data and Case Studies
5-43
-------
Coal Remitting Statistical Support Document
Figure 5.2q: Jeddo Tunnel Flow Data
500
: ' 450
400
350
I300
§
§ 250
«
Q 200
150,
100
50
I
6 i S £ 4. *. S
S 3 £ i « i 3
q> CD cr> oi as
Figure 5.2r: Precipitation Data From Hazleton, Pennsylvania
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,
Me Wreath, 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 principle 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 thousand 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 from remining 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 aquifier materials. The most significant
change in pollutant concentration was in net acidity (Figure 5.3b). Prior to activation of the
remining permit, the acidity concentration was typically in the range of 100 to 200 mg/L. The
(I,;. !|1'' !>ii.'.ir!i ., .-. '"'".:' ; y':;'i: j'^v "J' ':, \ ^'^V'''.'^."^ V:-! ' < . j''., ;": i ''; :v: ,': ''^V'M
effect of remining was to turn a distinctly acidic discharge into one that is now distinctly alkaline
(i.e., post-mining net acidity concentrations of 6 through -75 mg/L).
Figure 5.3a: Fisher Mining MP1
Flow, Iron, Manganese, Sulfate
3500
5-46
Long-term Monitoring Data and Case Studies
!l I
-------
Coal Remining Statistical Support Document
250
Figure 5.3b: Fisher Mining MP1
Net Acidity
-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
ann
800
700
600.
CO
1 500.
s
a
CO
3 400.
*o
300.
200 .
100.
0 .
Baseline
Established
I
/,
' ,V
, J >
13
2
~
£
s
1
o
CJ
_c
-------
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
I
part of the baseline pollution load computation, quality control limits are established 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 acid load (67.9 pounds per day). The median acid 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 5.3d: Fisher Mining MP1
Iron Load
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
Figure 5.3e: Iron Load Boxplot
Fisher M1 Discharge
10
33
*5)
.Q
D
8 5
c:
o
Premining
During
Postmining
Figure 5.3f: Net Alkalinity Boxplot
Fisher M1 Discharge
100 -H
D)
cc
^ -100 -|
s
-200
Premining
During
Postmining
Long-term Monitoring Data and Case Studies
5-49
-------
Coal Remining Statistical Support Document
5.3.2 Me Wreath Remining Site
The Me Wreath 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 best 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 Pittsburg 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.
35
25
20
15
10
Figure 5.3g: McWreath D1
Flow, Net Acidity, Iron, Manganese, Sulfate
.Flow
- Iron
- Manganese
.Net Acidity
_ Sulfate
2000
1800
1600
1400
1200 _
"5>
E.
1000 j?
s
3.
800 £
z
*f
600 £
3
tn
400
200
0
-200
8/11/87
5W90
5-50
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Figure 5.3h: McWreath D3
Flow & Net Acidity
._ Flow
Net Acidity
0
-500
7/7/1986 11/19/19 4/2/1989 8/15/199 12/28/19 5/11/199 9/23/199 2/5/1996 6/19/199 11/1/199
87 0 91 3 4 78
S 10
Figure 5.3i: McWreath D3
Flow & Iron
. Flow,
.Iron
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
5-51
-------
Coal Retaining Statistical Support Document
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
i ." i ' !,!'!" j"i|i "ill, ,,i"' ,,.;, ,i" , " , ' " i i,;,.1 !' " ij 11 "hiij1.., i ,' ,,4 i' i .Mil,' i"" , , '"i" ,i|'«,,:
Me Wreathsite. 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
rnonitoring data, the discharge has otherwise gone dry as a result of rernining 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.
>i'! , '.".' " . : ' ', ,,' -I1 ' ', ' ' . -'> »; - i ' ' i <|J
overburden strata during remining and reclamation operations.
5-52
Long-term Monitoring Data and Case Studies
-------
r
Coal Remining Statistical Support Document
Figure 5.3j: Me Wreath D4
Flow & Net Acidity
Flow |
Net Acidity j
500
-300
3/29/86 8/11/87 12/23/88 5/7/90 9/19/91 1/31/93 6/15/34 10/28/95 3/11/97 7/24/98
Long-term Monitoring Data and Case Studies
5-53
-------
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
: i '."''If ' " . ' ' '.,' . ,: 't" 1 ' i" ' ,1 ,-, .< V .',.' t',;:n:,!i,'
sandstone samples in OB-2 have appreciable neutralization potential. The overburden quality of
, 'nil';,, !iii'f: ' ' | i '' "',:: , ' -i ': .. ' ' " ": I .' ': '">'.. ' ' ' ' } §='
the Pittsburgh Coal at the Trees Mills site is much different (i.e., less calcareous strata, less
' liii ,' I:,!, ' ,, , ' ,, I ' ' ! ' ', I ' ' " " ;> " : ' li .1"
alkalinity production potential) than at the Me Wreath 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.
5-54
Long-term Monitoring Data and Case Studies
-------
r
Coal Remitting Statistical Support Document
Figure 5.3k: Trees Mills Site Map
₯(P T^^^'T^-^r^^m^
yjs v, -T-'/,j;\&\. F*3*r «>
>«<7^S^.i-sS-^v. ,>'.'.<<£; .4?i«s?
?5-.- v* 'vjs^.*&.,*--f'/,-w£?iff.-,.- ,."!'vl]'iir2«;-si''l tl*»"'-'if.-J
Long-term Monitoring Data and Case Studies
5-55
-------
Coal Remining Statistical Support Document
Figure 5.31: frees Mills Drill Hole Data
( ' j ..... -*i !, ' "/ ..... ' '..inlif '!" ..... ,j
, ! , t,,,:,; i 'L ' ! !"'l, " : ,, l! , .', Kill,!
'i" .iii1'"" ....p.
OB -2
2.00
55
15
J
163
122
'll'l',1l v1"' :J" !
5-56
Long-term Monitoring Data and Case Studies
-------
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 remining, 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. Remining operations commenced on the Trees Mills
site on October 1991, thus 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.
180
Figure 5.3m: Trees Mills MP1
Flow, Manganese, Aluminum, Net Acidity, Iron, Sulfate
AManganese .
xAluminum '
m Net Acidity I
aIron
oSulfate
10000
Long-term Monitoring Data and Case Studies
5-57
-------
Coal Remining Statistical Support Document
Figure 5.3m shows the variations in flow and concentrations of net acidity, suifate, 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
if1'1 - , ' Iiii" ni,'11:11!, i1 ' ,' ..I1 ' ''j ,:«''i i1", . ' : liiii :»,'' "i ,,i ':. :",' ,"'i "i ] ff'M
'i i . lln I1!1,;,; " ' ;Hi!ii. " . . i ,. ' ' . ., "I'.i, , ' .: ' "i": ' ,, '> ' ,; " . ",,, >i, } ;.' ' <; , , < , ' / ' I ^vs
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.
f " " i,,H " i ,, >" ' ' ." ., ":,< i | 1,,'i |. -I I" ' 'i- ' ' " " , ', " '' ii; ".,' ' I ;!;';
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
,iE'"i' i '" i" I, '"i.i, ', ,1" " i|.. , ',' '1,1 ' , "' ' '"i ,,,L 'fill',,,, :,!! ' T i" ' n'|i in ',, , ',i,i i ' ', ,l| "if,
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
I
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 remining
1 i
operation at the Trees Mills site removed a significant acid load (439 pounds per day =160
thousand pounds per year) and iron load (71.8 pounds per day = 26 thousand pounds per year)
from the Beaver Run tributary and the Beaver Run public water supply reservoir.
5-58
Long-term Monitoring Data and Case Studies
, it.it ..... it. ; ..... shi i ..... ........... "Is .......... i: ....... t ii:! ..... "ji' ...... ii|iii..iiaii||li: ;i1!!|ii|iiiiii ...... s^i,,!!;!!!!1!!! ...... i liili '.!:,:,:i,.iii!!!iiilii,ii i
-',. ii . ..... , i"!.:!!":: iilfc'i,
i; a:ia,i!j''t iiit'ii-iii. 8' - ijaiiitk. JIB ........ .:.'!.!! ,i; na;, ' .' I'li:.,,:.;!! ......
, ........ * ': <.!:, ....... M1' . ....... ...... ' '.ni'i ..... ; ...... i
-------
Coal Remining Statistical Support Document
Figure 5.3n: Trees Mills MP1
Acid, Iron, Manganese Load
2000
1800
1600
1400
1200
1000
-Iron j
. Manganese
375
25
3/29/1986 8/11/1987 12/23/1988 5/7/1990 9/19/1991 1/31/1993 6/15/1994 10/28/1995 3/11/1997 7/24/1998 12/6/1999
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 less 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-mining 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.
Long-term Monitoring Data and Case Studies
5-59
-------
Coal Remining Statistical Support Document
fit,"
Figure 5.3o: Trees Mills MP2
Flow, Iron, Manganese, Aluminum, Net Acidity, Sulfate
_a_lron
A Manganese
K A'lurrintim '
__ Net Acidity I
__oSulfate ,
10000
I li
The overall environmental impacts of these water quality changes are put in perspective by
examining 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 janged from 0.31 to 197.6 Ibs/day (median of 8.5§). Post-backfilling acid load
ranged from 3.89 to 54.95 Ibs/day. Hence, while extreme values were reduced, median acid load
iSi increasedBy approximately 5 Ibs/day. The range in pre-mining iron loads was 0.02 to 13.9
"!" ,! ' - '!" i>, Jin HUllliii, i ' '' ' : ,'!!n. ", , "I., Aiif; , ' - , mil1''i: .jij!1!" ,' ., jii! ,.' HW* \J' , j" .Ji'.l.'l1'1"11"" > ''»' , .''"'.'iWiiliif /, ,i .-ii « , .'f ,''"Pi l "i'Hl
Ibs/day (median of 0.26), while the post backfilling range was 0.3 to 3.36 Ibs/day (median of
6.46). 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
,, j ^ ,,
yalues, 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
5-60
Long-term. Monitoring Data and Case Studies
IL,; 5 1:1!?!1;,' llllln I:1!;,', II ;!:',!'III. M11",, I Ilini injlllijlillilliinilinil i''i| li.lliliil1L»1l/niilllll!« i;i:l I!1 I'ti.iill"" , ' II11 ;-v»i| Hi' i , ,11, jljiiiiiiiqi,',i||lii mlh i|.M,n;,i| L, ,i;,if,"';,iI'1 uMiiilH' i,' dim/ijl lulu;':' ,'|;,ijiijil Ill liiliii!,ii|i' .;i' 'Illili IIniilvnllllli .illiiiiimilii "' iiLlliMJililil i!i; 1 .,;.''
-------
Coal Remining Statistical Support Document
increase of 0.2 Ibs/day iron load from MP-2, with an elimination of 71.8 Ibs/day from MP-1.
Finally there was an increase of approximately 4.6 Ibs/day acid load from MP-2, offset by the
elimination of 439 Ibs/day from MP-1.
200
Rgure 5.3p: Trees Mills MP2
Acid, Iron, Manganese Load
.Acid :
.Iron !
. Manganese j
03/29/86 08/11/87 12/23/88 05/07/90 09/19/91 01/31/93 06/15/94 10/28/95 03/11/97 07/24/98 12/06/99
The net effect on the Beaver Run receiving stream is 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 from 7.9 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-
\
Long-term Monitoring Data and Case Studies 5-61
-------
Hi ,'.i "; i,1:. "' »; ".iiiliirJii'rinil "i , *, , .' ir
.. ', ,,r:i, i M ,11' '"ll'iii '" IJlih r
Coal Remining Statistical Support Document
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
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 of 7.97 Ibs/day post-backfilling. Extreme values of
iron loads were substantially reduced following backfilling.
Figure 5.3q: Trees Mills MP3
Flow, Iron, Manganese, Aluminum, Net Acidity, Sulfate
Flow
Iron
Manganesej
Aluminum
250
fifV;;' i I 'I'11
The MP-6 discharge is located below the outcrop of the Pittsburgh Coal seam, and varied in pre-
mining flow from 6.8 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
5-62 Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
(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 equal to the post-mining
median of 1.06 Ibs/day.
Rgure 5.3r: Trees Mills MP3
Acid, Iron, Manganese Load
Acid
Iron
_ Manganeseg0
0
0
3/29/86 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 12/6/99
Long-term Monitoring Data and Case Studies
5-63
-------
Coal Remining Statistical Support Document
Figure 5.3s: Trees Mills MP6
Acid, Iron, Manganese Load
-Acidity
-fron
1400
1200
1000
800
600
400 -
200
o,l__J
3/29/gg 8/W87 Va'23788"
'9/ig7&i 1/3Y/S3' WISISS TO/28'/S5 37YWW 71WS8 T2/6/99
'!,! ;.
Due to the cumulative effects of remitting upon the MP-1, MP-2, MP-3, and MP-6 discharges,
|H|I'I| I II I l|| I I ' i',,,, ' Ill ,' ,'Hlhii 1.1' ,' ' ," i - Jn ' " ," II i 'Hi", ,
the Trees Mills remitting 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
i iiii i i i ii ,. "V1!"11'!!' ;= ,,Wi hv'ij', n'lii'lv].*'1' '.' .(.'iiini-'r'i1 ' '","! "i1 ;\ ' . i ! ' i. !> "
Keservbir.To determine whether these pollution reduction effects could be detected in the water
chemistry of the receiving stream, permittee's self monitoring reports and PA DEP mining
;i!|!l"l'! ',' ::l!;, " ' IE ' |! " !!!!,," H" ,,' ,i, :M i ,iV ; i \ "'ii'!"!!"'.,,,'L r / i,11''! i1;!!,!. i,» '",' ' ,'' , Uli ' '" , :'». ,'|, : " !,,:, " < MIK nvi i , ,,,.' ": | ; i,i|,
inspector's monitoring data were evaluated from the same monitoring points located upstream
arid downstream of the Trees Mills operation on the Porter Run and Beaver Run tributaries. The
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), subtle changes in alkalinity concentration are observed
5-64
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
which 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 pollutional
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
140 _
120 . _
1/23/87 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
Long-term Monitoring Data and Case Studies
5-65
-------
I l ,
Coal Remining Statistical Support Document
/I L " SI, i! |,|'
140
120
100
80
Is"
s
eo
40
20
Figure 5.3u: Beaver/Run ;-
Alkalinity: Upstream & Downstream
. Upstream
. Dow nstream
1/23/87 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
* ,>,; > i. , , in . " , ;, ' i i
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.
: . "' "It " , , '"' ..,!""!' , +' n . ' "W ,:"' i."":; " ,i r i 1 '- '" ' , il Hi
Not all discharges behave in a similar fashion. Some discharges respond steadily, with
1 'H, ,;;;;,; ' ,;;;; ":;," , ,, , ,, , ; , ,, , , , , ,, , , ,,,, i, ,, ,
1 !nM'i"> .1 . « I. ' '..,' 'I1,, ',,: i, i' ''!',', i "ii:,,,!'iiir, '»ii I1 ' ,, "Mi"1 'i '"iftiij" 'i1,1 .''ill1 'i'n if i """ .if' "i . ' ' "' I
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-
; ' ! . ' i ,
case monitoring, most discharges exhibit fairly predictable behavior, and are
5-66
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
appropriately monitored using a monthly sampling interval and a one-year baseline
monitoring period.
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 adequately documented pre and post-rernining 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 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.
Long-term Monitoring Data and Case Studies 5-67
-------
":.I!'!I!!,P 'PL: J'lll" «!!:,:!'!!, I'll'W,!'!,
Coal Remining Statistical Support Document
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
(1 " ' it , ( ' . '"' ' "!' - .,..' "| r, 'i '' > , " Vllijf':
concentrations unchanged. Sources of alkalinity may increase pH and reduce acidity,
.I. '",. ,,,::, ' "' ; ;. ,. .,;,. , , ,.,, v - , .;,.!.. ,-.!, i,,
increase pne 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
|i ' ' ' "."
" ...... a
"ii
I*
Ballarpn, P.B., 1999. Water Balance for the Jeddo Tunnel Basin, Luzerne County, Pennsylvania.
Publication No. 208, PA DEP, Pottsville District Mining; Office under Grant ME97105,
t 1999^ '"'' ' "" ' """"" '' ' "' ' ""' '""" !
Ballarpn, P3-, 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, PA DEP, Bureau of Watershed Conservation under Grant ME96114,
My 1999.
I | » | flj.ii "' ''fij j!,'!' « ' '", : ' "!'i ' ' i' 1 ',»; '" ' ' ' :' ' ", MJI. ,.,i ' , , ' ,,| ^ ill! .'i 'I ' !! ,,jjii.; .,;.']' |r" ' ' '!""! V ,;, : ' ' i /*'
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., R.J. Hornberger, and G. Fleeger, 1998. Influence of Geology on Postmining Water
Quality: Northern Appalachian Basin. Chapter 8 in Coal Mine Drainage Prediction and
Pollution Preventipn 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 §-92, October 1988.
Cravptta, 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.
" ;':"'' :; '^'^
*; 1 :iit
...... 5-68
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Cravotta, C.A., M.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 DEP, Pottsville, PA.
Fishel, D.K., and J.E. Richardson, 1986. Results of a Preimpoundment Water-Quality Study of
Swatara Creek Pennsylvania. US Geological Survey Resources Investigations Report 85-
4023, Harrisburg, PA.
Griffiths, J.C., R.J. Hornberger, and M.W. Smith, (in preparation). Statistical analysis of
abandoned mine drainage in the establishment of the baseline pollution load for coal
remining permits. Details available from U.S. Environmental Protection Agency Sample
Control Center, operated by DynCorp I&ET, 6101 Stevenson Avenue, Alexandria, VA
22304.
Growitz, D.J., et al. 1985. Reconnaissance of Mine Drainage in the Coal Fields of Eastern
Pennsylvania. US Geological Survey, Water Resources Investigations Report 83-4274,
Harrisburg, PA.
Hornberger, R., et al., 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
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.
Long-term Monitoring Data and Case Studies
5-69
-------
,Cpal Remining Statistical Support Document
,!:! . ' lill.
j i:
*f:;
Til! 'Xi ,,T
.'111"1! I1 "' , , IF '
:»'' l'1'lt !'"<
5-70
Long-term Monitoring Data and Case Studies
-------
Coal Remining Statistical Support Document
Appendix A: Example Calculations of Statistical Methods
1.0 Example 1
Assume 12 baseline iron loading observations are collected by sampling once per month for a
year. Likewise, 12 iron loading 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 Procedures A and B were used. For all calculations in Example 1, assume the
following iron loading observations (Ibs/day).
Baseline
Monitoring
1.33
0.94
0.53
0.74
0.92
0.87
0.82
1.03
0.88
0.91
0.79
1.00
0.87
0.80
0.73
1.19
0.83
1.16
0.89
1.15
1.10
1.12
0.86
0.91
Procedure A (Figure 3.2a)
1.1.1 Single Observation Trigger:
1) Twelve baseline observations were collected, therefore n = 12.
2) The baseline observations were placed in sequencial order from smallest to largest.
[0.53, 0.73, 0.79, 0.82, 0.83, 0.86, 0.87, 0.88, 0.89, 0.92, 1.10, 1.33]
3) The number of observations, n,,is less then 16, therefore the Single Observation Trigger (L)
equals x(12)> (the maximum) = 1.33.
4) All monitoring observations are less than 1.33, therefore the Single Observation Trigger (L)
(1.33) was not exceeded.
1.1.2 Subtle Trigger
1) Twelve is an even number, therefore the median of the baseline observations is:
M = 0.5*(x(6) + x(7)).
M = 0.5 * (0.86 + 0.87) = 0.865
Appendix A
A-l
-------
Coal Remitting Statistical Support Document
In order to determine M,, calculate the median of the subset ranging from x(?) to x(12).
Because 12 - 6 = 6 is even, MI = 0.5 * (x(9) + x(10))
Mt = 0.5 * (0.89 + 0.92) = 0.905
2) In order to determine M.l5 calculate the median of the subset ranging from x(I) to x(6).
Because 6 is even,M., = 0.5 * (X(3) + x(4))
M., = 0.5* (0.79+ 0.82) = 0.805
3} To calculate R, subtract M., from
R = 0^05-0.805 = (U
4) The calculated value for R, is then substituted into the equation for T.
r" |. til';, , I
T =
5) The following monitoring observations are ordered from smallest to largest.
[0^470.80, 6.87, 0.911 0.91 ,* 0.94,1.00,1.03, 1.12, l.i"$,T.l6, L19]
d) There are 12 monitoring observations, therefore m = 12.
The number of observations is even, merefore M'= 6.5"'* (x(6) + x(7))
M' = 0.5* (0.94+1.00) = 0.97
This holds true for M/ and M.]1 as well.
M,1 = 0.5 * (x(9) + x(10)) = 0.5 * (1.12 + 1.15) = 1.135
M., '= 0.5 * (X(3) + x(4)) = 0.5 * (0.87 + 0.91) = 0.89
7) To calculate R, subtract MY from MI.'
"'"; " Rr=i.)35 -0.89 = 0.245. ' ' " \ ''"^' | ';
8) The calculated value for R' is then substituted in the equation for T'.
"I,;.
jii
T -- J925-1-58>[(L25 >0-245)1 - 0.867
'
,,
'"ill,
9) T (0.867) is less than T (0.907), therefore the median baseline pollution loading was not
'.' exceeded,
A-2
Appendix A
-------
Coal Returning Statistical Support Document
1.2
Procedure B (Figure 3.2b)
1.2.1 Calculation of Single Observation Limit
1) Again, the number of baseline observations, n = 12.
2) The following log-transformed (using natural logs) baseline observations are sequencially
ordered from smallest to largest. [-0.64, -0.31, -0.24, -0.20, -0.19, -0.15, -0.14, -0.13,
-0.12,-0.08, 0.10, 0.29]
3) The mean of the 12 log-transformed observations, Ev = -0.15.
4) An appropriate estimate of the first-order autocorrelation (p, ~) of the log-transformed data is
0.5. Given the number of observations and the auto-correlation estimate, the following
equation is used to calculate A.
A =
1
A-
"12'
= 1.09
5) The factor A is then used to calculate S2V.
[yr(-0.15)]
Sy2 = 1.09 *
n-1
= 0.0535
6) To find Ex the values for Ey and Sv2 into the following equation:
Ex=-exp[(Ey) + (0.5*Sv2)]'
Ex= exp[(-0.15) +(0.5 * 0.0535)] = exp (-0.12325) = 0.884
7) The Single Observation Limit (Lso) is defined as the following:
Lso = exp[(Ev)+(Z99*ySv2)]
LSO = exp [(-0.15) +(2.3263 * \/0.0535)]
LM = exp [0.388] = 1.47
8) Monitoring observations are below 1.47, therefore the Lso was not exceeded.
Appendix A
-------
Coal Remining Statistical Support Document
1.2.2 Calculation of Single Observation Warning Level
1) The Single Observation Warning Level (WLSO) is determined by using the following equation:
WL =exp [(-0.15) + (1:6449 V0.0535)] = exp[0.230] - 1.26
SO
2) Allof the monitoring observations are below 1.26, therefore the Single Observation Warning
Level was not exceeded.
', . -. ,
1.2.3 Calculation of Cusum test
1) The number of monitoring observations, n = 12.
. , '|l |||,J,| ., , , ' ,i , , ," , , , ,, ,,, h
2) The log-transformed (using natural logs) monitoring observations are listed and labeled
sequencially, in order of collection.
Obs.
Y,
-0.06
Y2
-0.30
Y3
-0.14
Y4
0.03
Y5
-0.09
Y6
0.00
Y7
-0.22
Y8
0.17
Y9
0.15
Y10
0.14
Y,,
0.11
YI2
-0.09
3) Using the values for Ey and S;2 from the Lso calculations, the value for K can be determined
using the following equation:
ilil " K = (-0.>15) + 0.25* (0.229) = -0.092
1 J;
4) The values for C(t), can be determined using the following equation:
~ = C, = Ct., + (Yn - K) for example
C, = 0 + (-0.06 - (-0.093)) = 0.033
A-4
Appendix A
-------
Coal Remining Statistical Support Document
The values for Ct are given in the following table for each collection time t. Negative C,
values are replaced with 0, as shown in parentheses.
t
1
2
3
4
5
6
c,
0.033
-0.174(0)
-0.047 (0)
0.123
0.126
0.219
t
7
8
9
10
11
12
c,
0.092
0.355
0.598
0.831
1.034
1.037
5) The baseline pollution Cusum Single Observation Limit, H, can be determined using the
following equation:
H = 8.0 *
-------
ll l":;i'ilf"
Coa/ Remitting Statistical Support Document
The limit values, Wt, are given in the following table for each collection time t. Negative
W, values were replaced with 0, as shown in parentheses.
.li"1,*!,, I:
t
1
2
3
4
5
6
W,
-0.0257 (0)
-0.2657 (0)
-0.1057(0)
0.0643
0.0086
0.0429
t
7
8
9
10
11
12
w,
-0.1428 (0)
0.2043
0.3886
0.5629
0.7072
0.6515
I , ' ,1! '"I
5) The baseline pollution Cusum Warning Level, H^, can be determined using the following
equation:
Hw = 3.5*(VSy2)
''X>'= 3.5* (0.229) = 0.8'096
6) All values for Wt are below 0.8096, therefore the baseline pollution Cusum Warning Level
was riot reached or exceeded.
!;,' ' ! ' 'I11! !; fl ' ' I" . 1 ',:> it
"1,3 Annual Comparisons
1.3.1 Wilcoxon-Mann-Whitney Test
1,,i!l,,'"..i ni1' !. ,ii :, Jiti'i: iiii '"'" ' '"'V .'" ". " ,' 'i""1 "" " ' ''"",, ,'. "'. !«:: ' '.,i. 'u! ' , ,.'i" ' .'ft!'" :,;' :'!;' ',?"
Instructions fp| the Wilcoxon-Mann-Whitney test are given in Conover (1980), cited in Figure
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.
Baseline
Observations
Baseline
Rankings
0.53
1
0.73
2
0.79
4
0.82
6
0.83
7
0.86
8
0.87
9.5
0.88
11
0.89
12
0.92
15
1.10
19
1.33
24
. , ,: . i,,;;1 ,;. ,;;:; ,, ,; ., :. ,:;, ,; , ., ,.h, , - ,
" '!> ',nr i1", A
is. i 'ii' ; i ' i. , "ii"
A-6
Appendix A
-------
Coal Remining Statistical Support Document
Monitoring
Observations
Monitoring
Rankings
0.74
3
0.80
5
0.87
9.5
0.91
13.5
0.91
73.5
0.94
16
1.00
17
1.03
18
1.12
20
1.15
21
1.16
22
1.19
23
Due to the fact that values of 0.87 and 0.91 were each obtained for more than one
observation. The average rankings are obtained for these values. For 0.87, the average of
9 and 10 is 9.5. For 0.91, the average of 13 and 14 is 13.5.
3) The sum of the twelve baseline ranks, Sn = 118.5.
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 121.
Critical Values (C) of the Wilcoxon-Mann-Whitney Test
(for a one-sided test at the 95 percent level)
n
m
10
11
12
13
14
15
16
17
18
19
20
10
83
87
90
93
97
100
104
107
111
114
118
11
98
101
105
109
113
117
121
124
128
132
136
12
113
117
121
126
130
134
139
143
147-
151
156
13
129
134
139
143
148
153
157
162
167
172
176
14
147
152
157
162
167
172
177
183
188
193
198
15
165
171
176
182
187
193
198
204
209
215
221
16
185
191
197
202
208
214
220
226
232
238
244
17
205
211
218
224
231
237
243
250
256
263
269
18
227
233
240
247
254
260
267
274
281
288
295
19
249
256
263
271
278
285
292
300
307
314
321
20
273
280
288
295
303
311
318
326
334
341
349
5) Sn (118.5) is less than C (121). Therefore, according to the Wilcoxon-Mann-Whitney test, the
monitoring observations did exceed the baseline pollution loading.
Appendix A
A-7
-------
, 1*
ii
' Coal Remining Statistical Support Document
I
2.0
Example 2
Assume 18 baseline iron loading determination observations are collected by sampling twice per
month for nine months. Likewise, 18 iron loading monitoring observations are obtained by
Sampling twice per month for a period of nine months. In order to determine whether baseline
pollution loading has been exceeded, examples of Procedures A and B are presented below. For
all calculations in Example 2, assume the following iron loading observations (in Ibs/day):
.'" ,' ,[:;-:, i ""ifl ,>,:' '<".' i, ' ;:;"' ,r>'i ' ' '!i'"'\ '' '' r... I I I i > " ' ,"' |.,'iki ,v
ii*;!"
"* >
i: * ii' ,:
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
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
A-8
Appendix A
; si:::,
-------
Coal Remining Statistical Support Document
2.1
Procedure A
2.1.1 Single Observation Trigger
1) The number of base.line observations collected, n = 18.
2) The baseline observations are ordered sequencially 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, M,, 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 Ml5 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).
18 - 13 = 5 is odd, therefore M2 = x(16) = 1.803.
6) To determine M3, calculate the median of the subset ranging from x(I6) to x(18).
18-15 = 3, which is odd, therefore M3 = x(17) =1.915.
7) To determine L, calculate the median of the subset ranging from x(17) to x(js).
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 Subtle Trigger:
1) As determined in section 2.1.1, M = 0.346, and Mj = 0.900.
2) To find M.,, calculate the median of the subset ranging from x(1) to x(9).
9 is odd, therefore M., = x(5) = 0.063.
3) The value for R is found by subtracting M., from M
R= 0.900-0.063 = 0.837
Appendix A
A-9
-------
Coal Remining Statistical Support Document
n i;
4) To find T, the value for R is inserted in the following equation:
'
1.58.K1.25.0.837)] .
,' ..... (1.35
: .4;'
5) The monitoring observations are placed in order from lowest to highest.
0530?b:d33, 6.040, d.0465 0.230, 0.240, 0370, 0.390, &.51Q, 0.530, 0.580, 0.63(1,
' '
;i/':;t. '"' ^djlo," 0.830^ 1.040, ilSo, 3.050]
6) The number of monitoring observations, m = 18.
18 is even, making M1 = 0.5 * (x(9) + x(10))
M1 = 0.5* (0.51 0 + 0.530) = 0.520
. .
..'W'ii ..... /,!, '[ ......
7) To determine M,1, calculate the median of subset x(10) to x(18).
Because 18 - 9 = 9 is odd, M,1 = (x(I4)) = 0.710
8) To determine M^', calculate the median of subset x(,) to x(9).
Because 9 is odd, ML'^ = (x(5)) = 0^230
... ^ ' ., , ... .... ,,
i i1;; '> i ., i1 i "' - . . , . ,. . ; .'' 'Vv .-
9) The value for R1 is found by subtracting M./ from M/.
R' = 0.710-0.230 = 0.48
10) To find-T, the value for R is inserted into the following equation:
ill'» "'... " « '.i... >.(!& ,. ii1"1:!1! " ,.."',. '.'.." i ; ,. .*./, .. -'it "i:: : ti
ii , ; :.: «;'
ili- ', Ji
T, = 0.520 - '-^Kl.25.0.48)] . 0.354
\:: :..'..': : (1.35
11) T1 (0.354) is less than T (0.635), therefore the median baseline pollution loading is not
I: i:'" "",' exceeded.
:i 'i. f -,'
A-10
Appendix A
-------
Coal Remining Statistical Support Document
2.2
Procedure B
2.2.1 Calculation of Single Observation Limit
1) The number of baseline observations, n = 18.
2) The natural log-transformed baseline observations are place in order from smallest to largest.
[-5.30, -3.51, -3.22, -3.10, -2.76, -2.75, -2.24, -1.83, -1.25, -0.90, -0.59, -0.50, -0.40,
-0.11,0.37,0.59,0.65,1.02]
3) The mean of the 18 log-transformed observations, Ey, is -1.44.
4) Given the number of observations, the following equation is used to find A.
A =
= 1.06
5) The factor A is then used to calculate Sy2.
2 ^
Sy2 = 1.06 * E - - 3.24
6) To find Ex , the following equation is used:
,Ex=exp[(Ey)+(0.5(Sy2))]
Ex= exp [(-1.44)+ (0.5 * 3.24)] = exp [0.18] = 1.20
7) Using the values that have been calculated the Single Observation Limit is found.
= exp[2.75] -= 15.60
Lso=QXp [(-1.44) + (2.3263'
8) All of the monitoring observations are below 15.60, therefore the Single Observation Limit
was not exceeded.
Appendix A
-------
r", ' . 'uLS'fi "> ,r' jr.' "! i HI ,. fill..!',,: |!"H|
,. I'" ! ' ,:!f!'' "'i' , rl' i,'1 S'iV,; ,.".C
Coal Remining Statistical Support Document
2.2.2 Warning Level
1) The Single Observation Warning Limit (WLSO) is determined using the following equation:
WL =exp [(-1.44) + (1. 6449*^334)] = exp[l. 52] = 4.58
i IIB mi, M ii ,
w1:1'1!!:* " «!!! I!1:, . '",[
,: ";",',!"; '''' ;ii: '' i ' '*! ' '!;,!"'- -'i vi. "'l|l;l i.'i! S: "' . 'if'"!'- ; '.-! f "'"'' ;'; :!l i
2) All of the mpnitoring observations are below 4.58, therefore the Single Observation Warning
Level is not exceeded
air si t,
2.2.3 Calculation of Cusum limit
' C
' iiBlill . I
i1', - " ':*?,,''I;!1' ; «.:"'!:"',,:: « .*,":!', >.": *'
t- '
1) The number of monitoring observations, n = 18.
2) The log-transformed monitoring observations are listed and labeled, in order of collection.
I' '«' >: :>, ,!,«< ".iI Jii I . ./I'll
Natural Log-Transformed
Monitoring Observations
Y,
Y,
Y3
Y4
Y5
Y6
Y7
Y8
Y9
-0.63
-3.22
0.04
-3.41
-3.51
-1.47
-0.34
-1.43
-0.94
Natural Log-Transformed
Monitoring Observations
Y(!0)
Y(ii>
Y(12)
Y(13)
Y(14)
Y(is)
Y(16)
Y(17)
Y(is)
-0.19
1.12
-0.54
0.17
-0.67
-3.08
-0.37
-0.46
-0.99
3) Using t£e values for Ey and Sy2 from the Lso calculations, the value for K can be determined
using the following equation:
...... ' ......................................
K = (-1.44) + 0.25 * 1.8 = -0.99
A-12
Appendix A
-------
Coal Remining Statistical Support Document
4) The values for Ct, can be determined using the following equation:
Q = "Q.! + (Yn - K) for example
C] = -0.63 - (-0.99) = 0.36
The values for Ct are given in the table for each collection time t. Negative Ct values are
replaced with 0, as shown in parentheses.
t
1
2
o
o
4
5
6
7
8
9
C«
0.36
-1.87(0)
1.03
-1.39(0)
-2.52 (0)
-0.48 (0)
0.65
0.21
0.26
t
10
11
12
' 13
14
15
16
17
18
Q
1.06
3.17
3.62
4.78
5.10
3.01
3.63
4.16
4.16
5) The baseline pollution Cusum Single Observation Limit, H, can be determined using the
following equation:
H = 8.0*(v/Sy2)
.H = 8.0* (1.8) =14.4
6) All values for Ct are below 14.4, therefore the baseline pollution Cusum Single Observation
Limit was not exceeded.
2.2.4 Cusum Warning Level
1) The number of monitoring observations, n = 18.
2) The following equation can be used to determine Kw:
Kw = [(-L44) + 0.5(1.8)] = -0.54
3) The values for Wt, can be determined using the following equation:
Wt = Wt., + (Yn - Kw) for example
W, = 0 + (-0.06 - (-0.0355)) = -0.0245 (0)
Appendix A
A-13
-------
Hill
Coal Remining Statistical Support Document
The limit values, Wt, are given in the table for each collection time t. Negative Wt values
were "replaced with 0, as shown in parentheses, and in the equation above.
t
1
2
3
4
5
6
7
8
9
Wt
-0.09 (0)
-2.68 (0)
0.58
-2.29 (0)
-2.97 (0)
-0.93 (0)
0.20
-0.69 (0)
-0.40 (0)
t
10
11
12
13
14
15
16
17
18
w«
0.35
2.01
2.01
2.72
2.59
0.05
0.22
0.30
-0.15(0)
4) The baseline pollution Cusum Warning Level, Hw, can be determined using the following
equation:
Hw = 3.5* (1.8) = 6.3
5) All values for Wt are below 6.3, therefore the baseline pollution Cusum Warning Level was
, ; " ,'." nf '"exceeded.
it;:
|," (! ,"" f ,'JL . ;,''' ; .-., '"5 " " ;, ; .; ;. r.'-
i i ""'lii .liliiii ' < 'nif ' i<< IK . ;:ii' , >:': ni : si i i
Annual Comparisons
I1"1!!1::, ' " jll
'' ; ''"2.3'.l ' n Wiicpxpn-Mann-Whitney test
Instructipns for, the Wiicoxpn-Marj^\^tney test are given in Conover (1980), cited in Figure
1) When using both baseline and monitoring data, n = 18 and m = 18.
'liMA-14
Appendix A
-------
Coal Remining Statistical Support Document
2) The baseline and monitoring observations are listed in order of collection, and ranked as
follows:
Baseline Observations
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
2.5
1
34
25
10
9
23
21
15
11
18
32
29
5
35
*t *">
3.5
12
7
Monitoring Observations
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
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.
4) From the table in section 1.3.1 of this appendix, the critical value (C) for 18 baseline and 18
monitoring observations is 281.
5) Sn (322.5) is greater than the critical value for C (281). Therefore, according to the Wilcoxon-
Mann-Whitney test, the monitoring observations did not exceed the baseline pollution
loading.
Appendix A
A-15
-------
Coal Remining Statistical Support Document
: '(;:, "V-"" j
A-16
Appendix A
------- |