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
—A—Manganese .
—x—Aluminum  '
 m Net Acidity I
—a—Iron
—o—Sulfate
                                                                                   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
                 __o—Sulfate  ,
                                                                                             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

-------