1043Acidity of water from abandoned underground mines [611836]

1043Acidity of water from abandoned underground mines
decreases over time, and the rate of decrease can help formulate remediation approaches and treatment system designs. Th e objective of this study was to determine an overall
acidity decay rate for above-drainage underground mines in northern West Virginia from a large data set of mines that were closed 50 to 70 yr ago. Water quality data were obtained from 30 Upper Freeport and 7 Pittsburgh coal seam mines in 1968, 1980, 2000, and 2006, and acidity decay curves were calculated. Th e mean decay constant, k, for Upper Freeport
mines was 2.73 × 10
−2 yr−1, with a 95% confi dence interval of
± 0.0052, whereas the k value for Pittsburgh mines was not
signifi cantly diff erent at 4.26 × 10−2 yr−1 ± 0.017. Acidity from
the T&T mine, which was closed 12 yr ago, showed a k value
of 11.25 × 10−2 yr−1. Th is higher decay rate was likely due to
initial fl ushing of accumulated metal salts on reaction surfaces
in the mine, rapid changes in mine hydrology after closure, and treatment. Although each site showed a specifi c decay
rate (varying from 0.04 × 10
−2 yr−1 to 13.1 × 10−2 yr−1), the
decay constants of 2.7 × 10−2 yr−1 to 4.3 × 10−2 yr−1 are useful
for predicting water quality trends and overall improvements across a wide spectrum of abandoned underground mines. We found fi rst-order decay models improve long-term prediction
of acidity declines from above-drainage mines compared with linear or percent annual decrease models. Th ese predictions
can help to select water treatment plans and evaluate costs for these treatments over time.Acidity Decay of Above-Drainage Underground Mines in West Virginia
B. Mack Water Research Institute
L. M. McDonald and J. Skousen* West Virginia University
Extensive underground mining has taken place in West
Virginia since the late 1800s (West Virginia Geological and
Economic Survey, 2007), and Bennett (1991) estimated an area of about 610,000 ha with underground mining beneath the sur-face in West Virginia alone. Th is legacy of mining has changed
groundwater quality and quantity due to intercepting and chang-ing underground water fl ow paths (Da Silva et al., 2006). In
areas of northern Appalachia where high sulfur coal exists and no limestone units are present for neutralization, the greatest envi-ronmental impact from underground mines has been on surface water quality from acid mine drainage (AMD) (Herlihy et al., 1990). Acid mine drainage is produced when sulfi de minerals
associated with coal seams react with oxygen and water to form low-pH, sulfate-rich, and high-iron solutions. Th e eff ects on sur-
face water include high levels of acidity and metals that have detri-mental eff ects on aquatic organisms (Gray, 1995; Monterroso and
Macias, 1998; Stewart and Skousen, 2003), low pH conditions that accelerate weathering and release of aluminum and other toxic elements from minerals (Bigham et al., 1996; Kittrick et al., 1982), and orange-colored stream sediments from iron hydrox-ide precipitation (Rosseland et al., 1992; Winland et al., 1991; Younger, 1998).
Previous studies of water quality changes from underground
mines have shown that acidity declines over time (Wood et al., 1999; Younger, 2000). Demchak et al. (2004) observed that changes in water chemistry over time diff er between below-drain-
age (fl ooded) and above-drainage (not fl ooded) underground
mines, with fl ooded mines rebounding to much better water
quality within a decade and unfl ooded mines remaining acid for
much longer. Lambert and Dzombak (2000) found that fl ooded
underground mines in Pennsylvania change from very acid water to neutral or net alkaline water shortly after complete fl ooding
(see also Brady et al., 1998; Capo et al., 2001; Donovan et al., 2000; Jones et al., 1994; Younger, 1997). Borch (2009) found similar results in the fl ooded Meigs mine in Ohio and suggested
the following reasons for the dramatic water quality improvement within a few years after fl ooding at Meigs: (i) Pyrite oxidation
ceased in the fl ooded sections; (ii) after the initial fl ush, there
was less readily available iron sulfate salts to dissolve; (iii) alkaline
Abbreviations : AMD, acid mine drainage; CD, cumulative diff erence.B. Mack, Water Research Institute; L.M. McDonald and J. Skousen, Division of Plant and
Soil Sciences; West Virginia Univ., Morgantown, WV, 26506. Scientifi c article number
3067, West Virginia Agriculture and Forestry Experiment Station, Morgantown, WV. Assigned to Associate Editor Robert Darmody.Copyright © 2010 by the American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America. All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including pho-tocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.
J. Environ. Qual. 39:1043–1050 (2010)
doi:10.2134/jeq2009.0229Freely available online through the author-supported open-access option.Published online 26 Mar. 2010.Received 17 June 2009.*Corresponding author (jskousen@wvu.edu).© ASA, CSSA, SSSA5585 Guilford Rd., Madison, WI 53711 USATECHNICAL REPORTS: SURFACE WATER QUALITY

1044 Journal of Environmental Quality • Volume 39 • May–June 2010strata in the roof rock of the mine pool provided some neu-
tralization; (iv) dilution and infl ux of alkalinity occurred from
groundwater infl ows; (v) the groundwater fl ow path exhibited
some short circuiting, so areas of rapid transport or fl ow exhib-
ited better water quality than areas of restricted water move-ment; and (vi) geochemical reactions, such as sulfate reduction and cation exchange, occurred along the underground water fl ow path, thus improving the quality before discharge.
Above-drainage mines did not show the same dramatic
improvement as below-drainage mines; they tended to improve slightly in water quality but remained acidic (Demchak et al., 2004; Lambert and Dzombak, 2000).
Acid mine drainage is unsightly, and treatment costs for
abandoned underground mines are a public fi nancial burden.
Th erefore, the length of time that these discharges continue to
be a burden (aesthetically and fi nancially) is important, and
predicting their longevity is necessary to determine poten-tial remediation strategies and cost projections. However, the physical setting in underground mines is diffi cult to study
because abandoned mine maps may not be accurate and con-ditions below ground are always changing and unpredictable. Some sections or voids of abandoned above-drainage mines are fl ooded or partially fl ooded, which virtually removes those
pyrite reaction surfaces from contributing acid products. Many other areas within the mine remain open to oxygen and water exchange and are susceptible to reaction. Th ese exposed pyrite
surfaces produce less acidity over time due to (i) weather-ing products forming an iron hydroxy sulfate coating, which reduces air and water contact and release of acid products (Younger, 1998), and (ii) the more morphologically reactive pyrite (framboidal) is depleted fi rst, thereby leaving the less
reactive pyrite (massive) for subsequent oxidation. Th erefore,
changes in pyrite reaction rate and availability of surfaces in these areas can result in drainage quality improvement. Only during roof or pillar collapse are fresh pyrite surfaces exposed to the mine atmosphere and water. Once mines are closed, ven-tilation systems cease, which greatly reduces the availability of oxygen for pyrite oxidation. Land surfaces over underground mines can be compacted or altered to reduce the amount of infi ltration, or surface cracks can be clogged, thereby inhibit-
ing direct infl ow of surface water into the mine. Roof or pillar
collapses within the mine can change fl ow paths or create pools
of water in the mine. Although all of these factors presum-ably decrease acidity with time, most are diffi cult or impossible
to validate. We are therefore left with empirical predictions of decline based on long-term data sets. Although site-specifi c
models are unlikely, a regionally valid model would signifi –
cantly improve our ability to plan and budget for treatment and could be useful for watershed-based water quality model-ing and classifi cation (Merovich et al., 2007).
Water quality improvements with time have been observed
for parameters other than acidity. Typically, these improvements are modeled using percent decreases in one or more water qual-ity parameters. For example, Demchak et al. (2001) found a linear relationship between sulfate and acidity with an R
2 value
of 0.67 using data from Upper Freeport above-drainage under-ground mines. Using sulfate as an indicator, they calculated a 2.2% decrease in acidity per year for 40 mines between 1968 and 2000. Ziemkiewicz (1994) used a similar rate of 2% acid-ity decrease per year to estimate changes in AMD discharges
over time (see also Koryak et al., 2004). Wood et al. (1999) calculated a slightly higher 3.3% acidity decrease per year in coal mine discharge chemistry over time in Scotland. Mack and Skousen (2008) found a 2.1% decrease per year in acidity from 40 underground mines in West Virginia.
Th e objective of this research was to determine the rate of
acidity decline from above-drainage underground coal mines. T wo distinct datasets were used: (i) a long-term data set for mines closed more than 50 yr ago for which data were only sporadically available and (ii) a 12-yr-old, recently closed Upper Freeport underground mine for which yearly data were available.
Materials and Methods
From a data set of previously sampled abandoned underground mines (Demchak et al., 2004), 37 sites were selected for sam-pling in 2006. Th ese sites were selected because water quality
data from 1968, 1980, and 2000 were available for these sites and approximate sizes and opening dates were known (Table 1). All sampling sites were located in Preston and Monongalia counties of West Virginia, and all sites discharged water from abandoned, above-drainage underground mines. Th is area in
northern West Virginia receives an average of 115 cm of pre-cipitation, which is somewhat evenly distributed throughout the year, and the average temperature is 11 °C. Based on pump-
ing and discharge rates of surrounding above-drainage mines, an average of about 20% of the precipitation on a year-round basis is discharged from underground mines in this area (Bruce Leavitt, Consulting Hydrologists, personal communication, 2009; GAI Consultants, 2001).
Th e Pittsburgh coal seam is the lowest stratum of the
Monongahela Group in the Pennsylvanian System. Th e seam
has 1.5 to 2% sulfur and an ash content of 6%, with seam thickness of about 3 m (Hennen and Reger, 1914). In this region, few overlying limestone materials are available within 30 m above the coal seam to neutralize the high amounts of acid-producing material in this coal and associated rocks.
Th e Upper Freeport coal seam is the topmost stratum of
the Allegheny Formation in the Pennsylvanian System. Upper Freeport coal contains <1.5% sulfur and an ash content from 8 to 12% and averages 2 m in thickness (Hennen and Reger, 1914). Th e strata above the Upper Freeport coal contain mas-
sive sandstones and some shales, and no limestone or alkaline-bearing rock units are found within 50 m above the Upper Freeport coal in this area.
1968 Sampling
During June though September of 1968 through 1970, researchers sampled all mine discharges in the Monongahela River basin. In the Cheat River subbasin from Parsons, West Virginia to Pt. Marion, Pennsylvania, 555 AMD sources were found, with 315 of these being underground mines (USEPA, 1971). Maps and fi eld sheets were completed for each site.
A 1-L bottle was fi lled with discharge water, put on ice, and
analyzed in the laboratory for acidity, alkalinity, conductivity, sulfate, and pH. Water samples were delivered to the labora-tory each Friday and were analyzed using methodology from

Mack et al.: Acidity Decline in Underground Mines 1045standard methods (American Public Health Association,
1965). Water analyses were monitored for accuracy and pre-cision by running periodic samples of reference standards.
1980 Sampling
Th e West Virginia Division of Water Resources also conducted
sampling and analyses of underground mine discharges in this area during 1980 (West Virginia Division of Natural Resources, 1985). We accessed their data and found that some of their sample sites matched discharges sampled in 1968. Th erefore,
where suffi cient information was available, we used their water
quality analyses in 1980 to aid in estimating the rate of change in water quality. Water samples were collected, placed on ice, and taken to the laboratory, where acidity was measured by
titration (Sheila Vukovich, West Virginia Division of Mining and Reclamation, personal communication, 2005).
2000 Sampling
We acquired the maps and fi eld sheets from the 1968 study and
located 40 of the underground mine discharge sites in 2000 (Demchak et al., 2004). A 250-mL water sample was taken at each sample point. Th e samples were not acidifi ed. Th ey
were placed on ice and analyzed by West Virginia University’s National Research Center for Coal Energy laboratory to deter-mine pH, total acidity, and alkalinity by titration.
Table 1. Discharge name, the year the mine opened, coal seam mined, size of mine, and acidity values for each discharge.
Discharge Year opened Coal seam SizeAcidity
1968 1980 2000 2006
h a —————————— m g L−1 as CaCO3 ——————————
Bull 1 1955 UF† 21 2805 1401 780
Bull 2 1957 UF 923 1905 756 540Bull 3 1957 UF 923 640 214 78Bull 4 1955 UF 86 250 360 530 478Bull 5 1955 UF 58 1370 336 334Fickey 1 1945 UF 28 3270 961 486Fickey 5‡ 1950 UF 38 515 460 697 461Fickey 6 1950 UF 75 1300 425 118 94Fickey 7 1950 UF 60 1670 1086 490Fickey 8 1952 UF 78 1505 625 390 420Fickey 9 1945 UF 47 1920 498 636Glade 1 1955 UF 26 1705 151 90Glade 2 1950 UF 52 390 179 31Glade 3 1950 UF 69 675 412 266Glade 4 1950 UF 156 1660 1250 230 450Glade 5 1950 UF 156 1765 1330 283 239Greens 1 1945 UF 33 945 455 702 188Greens 2§ 1945 UF 42 8 4 6Greens 3‡ 1950 UF 88 1504 830 1732 1214Martin 2 1955 UF 11 2315 545 135 110Martin 3 1955 UF 11 490 253 35Middle 1 1952 UF 310 917 515 291 290Muddy 2 1940 UF 72 687 410 86 198Muddy 3 1935 UF 278 170 110 45 72Muddy 5 1950 UF 148 20 30 71Muddy 6 1945 UF 98 4400 492 192Muddy 7 1945 UF 86 520 57 27Muddy 9 1952 UF 78 1515 1225 1050 800Muddy 10 1940 UF 121 1440 487 414Muddy 11 1943 UF 35 2140 634 550 444Cheat PA 1 1935 P 63 2457 563 424Cheat 2 1935 P 112 1061 1033 1048Cheat 5 1935 P 55 1825 210 104 446Cheat 6 1952 P 311 1450 488 214Lynn 1 1943 P 34 1368 605 102 170Lynn 2 1935 P 448 4690 3800 434 360
Lynn 3 1935 P 448 4988 1930 537 810
† P , Pittsburgh; UF, Upper Freeport.
‡ Not included in analyses because of surface disturbance.§ Not included in analyses because of low acidity.

1046 Journal of Environmental Quality • Volume 39 • May–June 20102006 Sampling
Using 37 of the sites from the 2000 data set (three sites from
that study ceased to discharge water by 2006), sampling was performed quarterly in 2006 to establish water chemistry con-ditions across seasons. Although four samples were taken in 2006 for each site (Mack and Skousen, 2007), only the acid-ity values from summer 2006 were used to keep the sampling season consistent with all other sampling years. Th e sample col-
lection procedure was the same as the 2000 sampling. Water samples were not acidifi ed. Th ey were placed on ice and ana-
lyzed by the laboratory as described previously.
T&T Data Set
Th e West Virginia Department of Environmental Protection
began treating the mine discharge at the T&T Mine in 1996. Water samples were collected weekly by agency personnel, and the samples were analyzed by WVU’s National Research Center for Coal and Energy for pH, total acidity and alkalinity by titration, and sulfate. Mean acidity for each year was used in our analysis.
Data Analysis
Of the 30 Upper Freeport sampling sites, one site (Greens 2) was discarded because the 2006 acidity was <10 mg L
−1, and
two (Fickey 5 and Greens 3) were discarded because of surface disturbance. Of the 27 remaining sites, 15 had three sampling dates (1968, 2000, 2006), and 12 had four sampling dates (1968, 1980, 2000, 2006). A calibration subset of 18 sites and nine validation sites was randomly selected six times, such that sites with four samplings and sites with three samplings were equally represented. Each calibration subset was fi t to the fol-
lowing fi rst-order decay model:
C
t = C0e−kt [1]
where Ct is acidity (mg L−1) at time t, C0 is acidity (mg L−1) at
time = 0, and k is the fi rst-order decay constant (1/time), using
nonlinear regression (PROC NLIN; SAS Institute, 2005). A single mean ( n = 6) decay constant, k (yr
−1), was used to pre-
dict 2006 acidity at each validation subset site ( n = 9) for each
group. Cumulative diff erence (CD) was calculated as
()92006 2006
,predicted ,actual
1CD Acidity Acidityvv
v==−∑ [2]
and percent error (% Error) was calculated as
2006 20069,predicted ,actual
2006
1 ,actualAcidity Acidity100
Acidity
% Error9vv
v v =⎛⎞ −⎜⎟⎜⎟⎝⎠=∑
[3]
where v is a validation site, and summary statistics (mean,
median, min, max) were calculated ( n = 6). Upper and lower
confi dence intervals for k were calculated. Th is procedure was
repeated with the calibration and validation subset reversed. In addition, fi rst-order decay constants were adjusted so that the
mean CD was zero and was calculated for all sites combined and for all sites individually. Because there is no physical reason to fi t a fi rst-order model to the data and because the 1968 data are potentially high infl uence points, sites with 1980 data were
used to fi t a linear model (1980–2006),
t
01
1980Acidity()Aciditybb t=+ [4]
where t is time and a fi rst-order decay model. Th e results were
compared as described previously. First-order decay constants were also calculated for the Pittsburgh and T&T data sets, but because there were insuffi cient sample numbers, all sites
(Pittsburgh) and years (T&T) were used to fi t a single fi rst-
order constant for each site.
Results
Acidity at the Upper Freeport sites decreased with time but not at all sites and not uniformly (Table 1). Four sites (Bull 4, Fickey 5, Greens 3, and Muddy 5) had higher acidity in at least one sampling compared with 1968, only two of which could be explained by surface reclamation (Fickey 5 and Greens 3, which were excluded from the analysis). Acidity increased from 2000 to 2006 at Fickey 8, Fickey 9, Glade 4, Muddy 2, Muddy 3, and Muddy 5.
General patterns of decline were diffi cult to discern at sites
with only three sampling dates. Of the sites with four sampling dates, acidity decreased nearly linearly at Muddy 3 and Muddy 9 and concavely at Glade 4 and Glade 5. Th e largest decrease
for many sites was between 1968 and 1980 (Fickey 6, Fickey 8, Martin 2, Middle 1, and Muddy 2), whereas others showed sharp declines from 2000 to 2006 (Glade 3, Muddy 9, and Bull 1).
Pittsburgh sites showed much more variable declines with
time than did the Upper Freeport sites (Table 1). Acidity at Cheat 2 was essentially unchanged from 1968 to 2006; three sites had consistently decreasing acidity with time (Cheat PA 1, Cheat 6, and Lynn 2), whereas at three sites acidity decreased from 1968 to 2000 but increased between 2000 and 2006 (Cheat 5, Lynn 1, and Lynn 3). Th erefore, acidity changes with
time at the Pittsburgh sites were inconsistent.
Th e range of acidity within any sampling year from these
sites was large, with relative standard deviations of 60 to 82% for Upper Freeport sites and 64 to 98% for Pittsburgh sites (Table 2). For Upper Freeport sites, the trend was for mean acidity to decrease with time, with the largest absolute decrease (54%) occurring between 1968 and 1980; for the Pittsburgh sites, the largest absolute decrease (72%) occurred between 1980 and 2000, although the sample number for 1980 was small. Annual percent decreases varied between 1.7 and 4.9% per year, with an increase of 1.1% yr
−1 between 2000 and 2006
at the Pittsburgh sites (Table 2). Th e overall annual percent
decrease of 2.1% for both sites was similar to the 2% decrease per year used by Ziemkiewicz (1994) and the 3.3% decrease per year found by Wood et al. (1999).
Th e best-fi t fi rst-order decay constant, k, for the Upper
Freeport sites was 2.73 × 10
−2 yr−1 (95% confi dence interval,
± 0.0052). Th is was comparable to the average k for all sites
(2.62 × 10−2 yr−1 ± 0.0035) but led to large prediction errors
in mean CD and % Error for acidity in 2006 in the validation data sets (Table 3). Although the prediction errors were large, they were positive and thus are very conservative (Fig. 1). Th e

Mack et al.: Acidity Decline in Underground Mines 1047resulting constants were not diff erent when the sample num-
bers in the calibration and validation data sets were reversed, suggesting that the results are robust. A k of 3.95 × 10
−2 yr−1
resulted in a mean CD of zero and considerably better predic-tion errors (Table 3) and was closer to the mean k of all sites
taken individually of 4.11 × 10
−2 yr−1 (range, 2.68 × 10−2 yr−1 to
11.56 × 10−2 yr−1). However, it also led to two sites
(Muddy 5 and Bull 4) having large negative pre-diction errors (Fig. 2). Th e best fi t k for the seven
Pittsburgh sites was 4.26 × 10
−2 yr−1 ± 0.017 (Fig.
3), a value not signifi cantly diff erent from either
of the best-fi t Upper Freeport constants. Although
a better constant for Pittsburgh mines could be obtained with more data, a k of 0.03 to 0.04 yr
−1
seems to be a reasonable estimate for both mines.
Th e fi rst-order decay constant from 1980 for-
ward was 2.62 × 10−2 yr−1 ± 0.008, which was not
diff erent from when the 1968 data were included.
Th is suggests that the 1968 data were not potential
high infl uence points. When simple linear regres-
sion from 1980 was used to predict acidity in 2006 in the Upper Freeport mines, the resulting predic-tion errors were smaller (Table 4) than when the 1968 data were used in a fi rst-order model (Table
3). However, when extrapolating backward to 1940, the approximate time these mines opened, the simple linear regression predicts an acidity of approximately 1200 mg L
−1, a 2% decline per year
predicts approximately 1700 mg L−1, and the fi rst-
order model predicts between 2800 and 4000 mg L
−1. Communication with T&T company person-
nel (Larry Harris, personal communication, 2005) at the closure of the Upper Freeport T&T mine confi rmed that water acidity was 5000 to 6000
mg L
−1. Another above-drainage Upper Freeport
mine, Omega, was closed in 1998, and average water acidity after closure was 3800 to 5050 mg L
−1 (GAI Consultants, 2001). Th erefore, the fi rst-
order model predicted more accurately the origi-nal acidity at mine closure than either the linear or percent
annual decrease models.
Acidity declines were greater at the T&T site ( k = 11.25 ×
10−2 yr−1) (Fig. 4) than that predicted from the subset regres-
sion (Table 3). Th is could be because T&T is not representative
of the other Upper Freeport mines. Th e decay constant from Table 2. Mean acidity, summary statistics, percent decrease, and annual percent decrease for Upper Freeport and Pittsburgh coal seam underground
mines in 1968, 1980, 2000, 2006, and overall.
Coal seam Year(s) n Mean SD Min. Median Max. Decrease Annual decrease
—————————————— m g L−1 —————————————— % % y r−1
UF† 1968 27 1422 1003 20 1440 4400
1980 12 657 394 110 530 1330 54‡ 4.5‡2000 27 434 357 30 336 1401 34‡ 1.7‡2006 27 306 228 27 266 800 30‡ 4.9‡
overall 93 712 20 486 4400 78§ 2.1§
P 1968 7 2548 1627 1061 1825 4988
1980 4 1636 1619 210 1268 3800 36‡ 3.0‡2000 7 466 316 102 487 1033 72‡ 3.6‡2006 7 496 320 170 424 1048 −6.5‡ −1.1‡
overall 25 1245 102 605 4988 80§ 2.1§
† UF, Upper Freeport; P , Pittsburgh.
‡ From previous data collection time.§ From 1968 to 2006.
Table 3. Prediction errors for fi rst-order decay models.
Subset regression
(k = 2.73 × 10−2 yr−1)Mean CD† = 0
(k = 3.95 × 10−2 yr−1)
CD % Error CD % Error
mg L−1% mg L−1%
Mean 174 151 0 58
Median 159 138 −13 50
Min. 86 73 −80 9
Max. 296 224 77 104
† CD, cumulative diff erence.
Fig. 1. Acidity (mg L−1) for the Upper Freeport sites, best-fi t fi rst-order decay function ( k =
0.0273), upper and lower confi dence intervals, and, when adjusted so that mean cumula-
tive diff erence = 0 ( k = 0.0395), extrapolated to the year 2050. Horizontal dashed line
represents the cutoff acidity to begin passive treatment (100 mg L−1). CAD, cumulative
average diff erence; LCL, lower confi dence limit; UCL, upper confi dence limit.

1048 Journal of Environmental Quality • Volume 39 • May–June 2010the subset regression is meant to be an overall representation
of 27 sites; individual sites had decay constants greater and less than the mean. It is also possible that treatment and reme-diation eff orts by the company and state agencies at T&T are
improving water quality faster than would be predicted from untreated mines. Th e remediation eff orts at T&T involve (i)
pumping limestone slurry into the mine in 2000 to neutral-ize the acidity and (ii) continual injection of AMD treatment sludge back into the mine since 2002. Finally, it is possible that acidity declined faster during the fi rst decade after mine
closure than in subsequent decades (Borch, 2009). Younger (1997) stated that this rapid decline in acidity after closure is related to initial fl ushing of stored acid products (vestigial acid-
ity), with lower acidity emanating from the mine with time. Th e underground mines we sampled were closed more than 10
yr before our fi rst sample date in 1968, so presumably much
of their stored acid products were fl ushed before our fi rst sam-pling. Th is may also explain why the
backward prediction to 1940 was a little lower than expected.
Discussion
It is probable that the discharges from underground mines would not follow a consistent decay rate throughout their history due to initial rapid changes in the mine environment immediately after mine closure. Th e time between mine
closure and the fi rst sampling could
be very important to the assessment of water quality changes from the mine. Although we were able to determine the mine opening dates (Table 1), it is nearly impossible to determine the closure date because there was no requirement for operators to report closure dates.
We do, however, know that these mines were closed in the 1950s to early 1960s
because the USEPA fi eld sheets, which were fi lled out in the
late 1960s, denoted that the mines were already closed. Pyrite oxidation rate, availability of pyrite surface area, and mine geo-chemistry could change rapidly once the mine is closed due to a lack of new pyrite exposure from further mining. It is also possible that the accumulated and stored metal salts within the mine could be fl ushed out soon after mine closure, which
would show an initial high acidity with rapid declines. Periodic changes within the mine, like random physical alterations due to high rainfall and infl ow, make prediction of water chemistry
at any given time diffi cult.
Although studies on the longevity of acid mine drainage
have calculated annual percent decreases, this approach is not the best for prediction because it is sensitive to the time interval used (i.e., it decreases as time increases if acidity changes are small or zero). Depending on the initial conditions and time period, a 2% rate of acidity decline could be essentially indis-tinguishable from a fi rst-order decay constant of 0.04, but the
forward and backward predictions would be vastly diff erent.
Given that similar fi rst-order decay constants were obtained
with and without the 1968 data and that the backward predic-tions using the fi rst-order model were closer to what would
be expected at mine closure, we conclude that the fi rst-order
approach is preferred over simple linear or percent decline functions. Although there is no a priori reason to fi t a fi rst-
order function to the data, the advantages are that it allows for simple forward and backward estimation, and, if a valid constant is known, similar predictions could be made for other areas with limited data sets. Th e disadvantage of fi rst-order
(and percent decline) functions is that they asymptotically approach zero acidity. We are not aware of any above-drainage acid mine drainage site that has naturally attenuated to zero acidity; therefore, some modifi ed version of the fi rst-order
function could be considered, for example:
C
t = C∞ + (C0 − C∞)e−kt [5]
Fig. 2. Percent error for 2006 predicted acidity as a function of actual acidity in 2006 for Upper
Freeport sites.
Fig. 3. Acidity (mg L−1) for the seven Pittsburgh sites, best-fi t fi rst-
order decay function ( k = 0.0.0426), and upper and lower confi dence
intervals. Horizontal dashed line represents the cutoff acidity to
begin passive treatment (100 mg L−1). LCL, lower confi dence limit;
UCL, upper confi dence limit.

Mack et al.: Acidity Decline in Underground Mines 1049where C∞ is the long-term, steady-state acidity. Much
more research is needed to determine what to use for C
∞, but presumably it is a function of the geol-
ogy, including sulfur content, pyrite forms, limestone layers, and perhaps the area of disturbance.
An acidity of 100 mg L
−1 is often used as the
cutoff for where passive treatment becomes a viable
treatment option. Predictions from the fi rst-order
model indicate that acidity will be 100 mg L−1 from
the Upper Freeport and Pittsburgh sites somewhere between 2030 and 2060 (Fig. 1 and 3) and between 2015 and 2020 at the T&T site (Fig. 4). We are not suggesting that acidity at every site will be <100 mg L
−1; rather, we suggest that the combined contri-
bution from all these sites to the region would be <100 mg L
−1.
Conclusions
Given that the average acidity of Upper Freeport
underground mines was 1422 mg L−1 in 1968 and
306 mg L−1 in 2006, a signifi cant improvement in
regional water quality from above-drainage under-ground mines has occurred in the last 50 to 70 yr. A similar trend was found for Pittsburgh underground mines: acidity was 2548 mg L
−1 in 1968 and 496 mg
L−1 in 2006. Even with this large decline in discharge
acidity, there is still a large quantity of stored acidity in these watersheds, which apparently will continue to be gradually released over time. First-order decay curves show an improvement in prediction over the long-term compared with linear declines and percent annual decreases and provide an important predic-tion tool for future water quality. Such a tool could be extremely benefi cial when a discharge is being considered
for passive or active treatment. By more accurately estimat-ing future acidity, treatment systems can be designed to more effi ciently neutralize AMD discharges and predict when active
treatment systems are necessary and when less costly passive treatment techniques may be used as the acidity declines over time. In addition, cost projections for treatment based on acidity and fl ow could be evaluated, and future benefi ts for
improved water quality conditions in receiving streams could be assessed. Th e analysis of decay curves could be improved by
more data sets where water samples were taken from mines at a greater frequency, at least annually, as well as beginning to take samples at the time of mine closure to obtain baseline acidity.
References
American Public Health Association. 1965. Standard methods for the exami-
nation of water and wastewater. 10th ed. APHA, Washington, DC.
Bennett, L. 1991. Abandoned mines: Report from WV. EPA J. 17:16–18.Bigham, J.M., U. Schwertmann, S.J. T raina, R.L. Winland, and M. Wolf.
1996. Schwertmannite and the chemical modeling of iron in acid sulfate waters. Geochim. Cosmochim. Acta 12:2111–2121.
Borch, M. 2009. Mine fl ooding in the Meigs Mine pool complex. In 30th
Annu. West Virginia Surface Mine Drainage Task Force Symp. 31 Mar.–1 Apr. 2009. Morgantown, WV.
Brady, K.B., R.J. Hornberger, and G. Fleeger. 1998. Infl uence of geology on
postmining water quality: Northern Appalachian Basin. p. 8-1 to 8-92. In Coal mine drainage prediction and pollution prevention in Pennsylva-
nia. Penn. Dep. of Environmental Protection, Harrisburg, PA.Capo, R.C., W.R. Winters, T.J. Weaver, S.L. Staff ord, R.S. Hedin, and B.W.
Stewart. 2001. Hydrologic and geochemical evolution of deep mine dis-charges, Irwin Syncline, Pennsylvania. p. 1–10. In 22nd Annual West
Virginia Surface Mine Drainage Task Force Symposium. 3–4 Apr. 2001, Morgantown, WV.
Da Silva, E.F ., P . Reis, E.C. Fonseca, J.X. Matos, J. Barrosinho, and J.M. San-
tos Oliveira. 2006. Interaction of acid mine drainage with waters and sediments at the Corona stream, Lousal mine (Iberian Pyrite Belt, South-ern Portugal). Environ. Geol. 50:1001–1013.
Demchak, J., J. Skousen, and L.M. McDonald. 2001. Water quality improve-
ments over time and longevity of acid mine discharges from under-ground mines in northern West Virginia. p. 174–182. In 18th Annu.
National Meeting of the Am. Soc. of Mining and Reclamation. 3–7 June 2001. Albuquerque, NM.
Demchak, J., J. Skousen, and L.M. McDonald. 2004. Longevity of acid dis-
charges from underground mines located above the regional water table. J. Environ. Qual. 33:656–658.
Donovan, J., B. Leavitt, E. Werner, E. Perry, and K. McCoy. 2000. Long-term
hydrogeological and geochemical response to fl ooding of an abandoned
below-drainage Pittsburgh coal mine. In Proc. 21st West Virginia Surface
Mine Drainage Task Force Symp.. 4–5 Apr. 2000. Morgantown, WV.
GAI Consultants. 2001. Omega Mine Injection Program: Monongalia
County, West Virginia. EPRI, Palo Alto, CA, Allegheny Energy Supply, Greensburg, PA, and USDOE, Pittsburgh, PA.
Gray, N.F . 1995. Field assessment of acid mine drainage contamination in
surface and ground water. Environ. Geol. (Berlin) 27:358–361.
Hennen, R.V., and D.B. Reger. 1914. West Virginia Geological Survey, county
reports, Preston, County. USGS, Morgantown, WV.
Herlihy, A.T., P .R. Kaufmann, M.E. Mitch, and D.D. Brown. 1990. Regional
estimates of acid mine drainage impact on streams in the mid-Atlantic and southeastern United States. Water Air Soil Pollut. 50:91–107.
Jones, P .M., S.M. Mulvay, and D. Fish. 1994. Th e role of sulfate and ionic
strength on the shift from acid to alkaline mine drainage in southwest Pennsylvania. p. 289–295. In Proc. Int. Land Reclamation and Mine Table 4. Prediction errors for simple linear decline models.
Subset regression
(b1 = −1.81 × 10−2 yr−1)Mean CD† = 0
(b1 = 2.04 × 10−2 yr−1)
CD % Error CD % Error
mg L−1% mg L−1%
Mean 115 119 0 58
Median 95 107 −13 50
Min. 29 51 −79 9
Max. 221 183 77 104
† CD, cumulative diff erence.
Fig. 4. Acidity (mg L−1) at the T&T site for the fi rst 11 yr after closure, best-fi t fi rst-
order decay function ( k = 0.1125), and upper and lower confi dence intervals.
Horizontal dashed line represents the cutoff acidity to begin passive treatment (100
mg L−1). LCL, lower confi dence limit; UCL, upper confi dence limit.

1050 Journal of Environmental Quality • Volume 39 • May–June 2010Drainage Conf. and the Th ird Int. Conf. of the Abatement of Acid Mine
Drainage. Vol. 2. Mine Drainage. U.S. Bureau of Mines Spec. Publ. SP 06B-94. 24–29 Apr. 1994, Pittsburgh, PA.
Kittrick, J., D. Fanning, and L. Hossner. 1982. Acid sulfate weathering. SSSA
Spec. Publ. 10. SSSA, Madison, WI.
Koryak, M., L. Staff ord, and R. Reilly. 2004. Declining intensity of acid mine
drainage in the northern Appalachian bituminous coal fi elds: Major Al-
legheny River tributaries. J. Am. Water Resour. Assoc. 40:677–689.
Lambert, D.C., and D.A. Dzombak. 2000. Evaluation of natural amelioration
of acidic deep mine discharges in the Uniontown Syncline, Pennsylvania. In Proceedings, 21st West Virginia Surface Mine Drainage Task Force
Symposium, 4–5 Apr. 2000, Morgantown, WV.
Mack, B., and J. Skousen. 2007. Changes in water quality of 34 above-drain-
age mines in West Virginia. p. 413–423. In Proceedings, 2007 American
Society of Mining and Reclamation, 3–6 June 2007, Gillette, WY.
Mack, B., and J. Skousen. 2008. Acidity decay curves of 40 above drainage
mines in West Virginia. p. 612–627. In Proceedings, 2008 American
Society of Mining and Reclamation, June 14–19, 2008, Richmond, VA.
Merovich, G.T., J.M. Stiles, J.T. Petty, P .F . Ziemkiewicz, and J.B. Fulton.
2007. Water chemistry-based classifi cation of streams and implications
for restoring mined Appalachian watersheds. Environ. Toxicol. Chem. 26:1361–1369.
Monterroso, C., and F . Macias. 1998. Drainage waters aff ected by pyrite oxi-
dation in a coal mine in Galicia (NW Spain): Composition and mineral stability. Sci. Total Environ. 216:121–132.
Rosseland, B.O., I.A. Blaker, A. Bulgar, F . Kroglund, A. Kvellstad, E. Lydersen,
D.H. Oughton, B. Salbu, M. Staurnes, and R. Vogt. 1992. Th e mixing
zone between limed and acidic river waters: Complex aluminum chemis-try and extreme toxicity for salmonids. Environ. Pollut. 78:3–8.
SAS Institute. 2005. SAS user’s guide. Version 9.1. SAS Inst., Cary, NC.Stewart, J., and J. Skousen. 2003. Water quality changes in a polluted stream over a twenty-fi ve-year period. J. Environ. Qual. 32:654–661.
USEPA. 1971. Summary report, Monongahela River mine drainage remedial
project. p. 1–146. In Proceedings: Th e Matter of Pollution of the Inter-
state Waters of the Monongahela River and Its T ributaries: Pennsylvania, West Virginia, and Maryland. 24 Aug. 1971, Pittsburgh, PA.
West Virginia Division of Natural Resources. 1985. Cheat River subbasin
abandoned mine drainage assessment. WV DNR, Division of Water Re-sources. Charleston, WV.
West Virginia Geological and Economic Survey. 2007. History of West Vir-
ginia mineral industries: Coal. WVGES, Department of Commerce, Morgantown, WV.
Winland, R.I., S.J. T raina, and J.M. Bigham. 1991. Chemical composition of
ochreous precipitates from Ohio coal mine drainage. J. Environ. Qual. 20:452–460.
Wood, S.C., P .L. Younger, and N.S. Robins. 1999. Long-term changes in the
quality of polluted minewater discharges from abandoned underground coal workings in Scotland. Q. J. Eng. Geol. 32:69–79.
Younger, P .L. 1997. Th e longevity of minewater pollution: A basis for decision-
making. Sci. Total Environ. 194–195:457–466.
Younger, P .L. 1998. Coalfi eld abandonment: Geochemical processes and hy-
drochemical products. p. 1–29. In Nicholson, K. (ed) Energy and the environment: Geochemistry of fossil, nuclear, and renewable resources. MacGregor Sci., Univ. of Newcastle, UK.
Younger, P .L. 2000. Predicting temporal changes in total iron concentrations
in groundwaters fl owing from abandoned deep mines: A fi rst approxima-
tion. J. Contam. Hydrol. 44:47–69.
Ziemkiewicz, P . 1994. AMD/time: A simple spreadsheet for predicting acid
mine drainage. p. 395–400. In Proc. Int. Land Reclamation and Mine
Drainage Conf. and the Th ird Int. Conf. of the Abatement of Acid Mine
Drainage. Vol. 2. Mine Drainage. U.S. Bureau of Mines Spec. Publ. SP 06B-94. 24–29 Apr. 1994, Pittsburgh, PA.

Reproduced withpermission ofthecopyright owner. Further reproduction prohibited without permission.