Introduction

Terrestrial waters play a critical role in the release and transfer of geologically-bound carbon to the oceans (Butman and Raymond 2011). This process may be enhanced in regions where the legacy of coal mining has generated sulfate-rich waters. When sulfuric acid drives carbonate dissolution, it enhances carbonate weathering because, per mole of acid, sulfuric acid can dissolve twice as many moles of calcite than carbonic acid (Table 1; Eqs. 1, 2).

Table 1 Controlling reactions and equations for carbonate geochemistry
Full size table

The reaction of sulfuric acid with carbonate rocks has been increasingly recognized as an important process in terrestrial weathering (Calmels et al. 2007; Hercod et al. 1998; Lerman and Wu 2006; Millot et al. 2003; Rice and Herman 2012; Telmer and Veizer 1999) and in creating some karst systems (Palmer 2007). Natural sources of sulfuric acid include oxidation of deep-sourced H2S, volcanic activity, and dissolution of aluminum and iron sulfides. According to Calmels et al. (2007), sulfate-driven weathering of carbonates is “significantly underestimated” in the Canadian Mackenzie River Basin. In a study of a watershed in Texas, Hercod et al. (1998) concluded that pyrite oxidation accounted for nearly half of the flux of dissolved inorganic carbon (DIC) from the watershed although only trace pyrite was present. Sulfuric acid-driven dissolution of carbonates is a well-studied process of cave formation: well-known caves attributed to this process include Carlsbad and Lechuguilla Caves in New Mexico and Frasassi Cave in Italy (Palmer 2007).

Coal mining can be considered an anthropogenic form of mechanical and chemical weathering that exposes sulfide minerals to the atmosphere and consequently generates sulfuric acid (Calmels et al. 2007; Rice and Herman 2012). If the sulfuric acid interacts with carbonate rocks, either insitu or added as exogenous material for treatment, DIC is added to the water (Geroni et al. 2012; Younger et al. 2002). Depending upon the pH of the solution, DIC exists primarily as dissolved CO2 (H2CO *3 ), HCO3 −, or CO −23 . The governing chemical reactions between these species and a mass balance statement for DIC are provided on Table 1. If the pH of the mine water is low enough to produce dissolved CO2 from the DIC, and the partial pressure of CO2 in the water (PCO2) exceeds the atmospheric background pressure, then CO2 can degas and transfer carbon from the aqueous to the atmospheric reservoir.

The mine water–carbonate interaction can occur inside the mine, in waste piles, or in streams and retaining ponds. Field and laboratory studies have shown that the addition of low-pH mine waters to neutral-pH streams can lower the stream pH sufficiently to convert DIC into CO2 resulting in degassing and a loss of CO2 from the system (Atekwana and Fonyuy 2009; Fonyuy and Atekwana 2008a; Hercod et al. 1998; Webb and Sasowsky 1994). The outcome depends on the DIC concentration, the buffering capacity of the receiving stream, and the capacity of the mine water to generate acidity via precipitation of metal hydroxides. The degassing of CO2 and ensuing rise in pH are sometimes incorporated into the treatment design to enhance metal removal (Cravotta 2007; Cravotta and Trahan 1999; Geroni et al. 2012; Younger et al. 2002). On a more regional scale, the long-term impacts of mining can be seen over decadal scales in larger watersheds (Raymond and Oh 2009).

This study demonstrates the detailed release, speciation and evolution of DIC from two coal-mine drainage discharges. The flux of carbon released from the sites was determined along with means of export (CO2 evasion flux to the atmosphere versus export of dissolved carbonate species). These fluxes are also compared to values estimated from site and regional data reported in the literature. The results illustrate the importance of carbon release associated with mine-waters and current limitations in the estimation of carbon flux in these settings.

Methods and procedures

Sample collection and analysis

The pH, electrical conductivity (EC), and temperature were measured using a YSI hand meter (YSI, Inc. Yellow Springs, OH). The meter was calibrated daily using two buffers for pH (4 and 7) and a 1.417 mS/cm EC standard. A limited number of samples were analyzed in the field for ferrous iron (Fe2+) using a Hach DR2800 field spectrophotometer following the ferrozine method (Stookey 1970). Using this procedure, Fe2+ was determined four times for each sample; the relative standard deviations (RSD, the standard deviation divided by the mean expressed as a percentage) for the replicate measurements were always less than 9 %. The mean Fe2+ concentrations were used for assessing acidity as discussed below.

A carbonation meter (CarboQC; Anton Paar GmbH, Graz, Austria) was used to measure dissolved CO2 in the field (Vesper and Edenborn 2012). The meter is a commercially-available instrument used in the beverage industry for the determination of dissolved CO2 in water, soda and beer (Anton Paar 2010). The meter works by sealing a water sample in a volumetric chamber and changing the chamber volume with a piston. The CO2 concentration is calculated based on the differences in temperature and pressure at two volumes. DIC was also measured in the field using the carbonation meter. In this method the DIC was converted into CO2 through the addition of a high ionic-strength, low-pH citrate solution, and then CO2 was determined in the buffer-sample mixture. The DIC was calculated based on the measured CO2, the pH of the buffer-sample mixture, and the dilution caused by adding the buffer (Vesper et al. 2015). The samples for CO2 and DIC were collected in 140-mL syringes that were filled under water so that no air bubbles were introduced in the sampling process.

Water samples for laboratory analysis were collected as grab samples and filtered into 125-mL bottles through a Whatman GD/X 0.045-μm syringe filter. Samples for elemental analysis were preserved with trace-metal grade nitric acid and analyzed by inductively coupled plasma optical emission spectroscopy (ICP-OES) on a Perkin Elmer Optima 3000 Radial View spectrometer following US EPA method 200.7. Samples for anion analysis were kept cold prior to analysis on a Dionex ion chromatography system.

The discharge from the source was measured following the method of Buchanan and Somers (1969) at stream locations where the water was deep enough to allow the measurement to be made. A Swoffer propeller-meter (Velocity Meter 3000) was used to measure velocity at 60 % of depth over a cross-section perpendicular to flow direction. The RSDs for replicate measurements were <5 % for the higher discharge values and <20 % for the lower discharge values. Discharge was assumed to be constant throughout the site over each sampling period.

Calculation of acidity, alkalinity and saturation indices

Acidity is generated in mine waters when Al, Fe and Mn precipitate as hydroxides (Younger et al. 2002). The estimated potential metal-acidity produced by hydrolysis can be calculated as follows:

$${\text{Acid}}_{\text{MA}} = { 2}\left[ {{\text{Fe}}^{ + 2} } \right] \, + { 3}\left[ {{\text{Fe}}^{ + 3} } \right] \, + { 2}\left[ {{\text{Mn}}^{ + 2} } \right] \, + { 3}\left[ {{\text{Al}}^{ + 3} } \right]$$
(7)

where AcidMA is the calculated metal acidity in eq/L, and the concentrations of Fe2+, Fe+3, Mn2+, and Al+3 in filtered water samples are in mol/L. This value accounts for potential acidity generated if and when the metals precipitate as hydroxides and is commonly used in assessment of mine waters (Skousen et al. 2002). The equation as modified above does not account for acidity in other forms (e.g., protons) which is typically small in comparison to the metal acidity. Iron was not speciated in all of the collected samples, but in the 10 samples that were, Fe2+ accounted for at least 83 % of the iron detected. For consistency, AcidMA was calculated for the samples assuming that all iron was in the Fe2+ species; there was less than 5 % error in the calculated acidity from this assumption.

The primary acid-consuming reaction in the high-DIC waters is the dissolution of carbonate rocks which creates carbonate alkalinity (ALKCARB); this value was calculated based on results for the speciation of DIC:

$${\text{ALK}}_{\text{CARB}} = \Sigma {\left[ {\text{bicarbonate species}} \right]} + 2\Sigma {\left[ {\text{carbonate species}} \right]}$$
(8)

where ALKCARB is in eq/L, and the bracketed molar concentrations of the bicarbonate and carbonate species are calculated using Minteq (Allison et al. 1991) based on full water chemistry and the DIC. The calculated ALKCARB is used because titrated alkalinities may include colloidal hydroxides (Adams 2012). Saturation indices for calcite (SIC) and dolomite (SID) are also obtained from the Minteq model.

CO2 calculations

Excess CO2 is the ratio of the measured concentration of dissolved CO2 to the dissolved concentration in equilibrium with atmospheric background. Estimation of the CO2 lost to the atmosphere (evasion) cannot be calculated solely from the difference in upstream and downstream CO2 concentrations because some of the CO2 may be lost due to conversion to HCO3 − as the pH rises. Therefore, flux and evasion are estimated based on DIC so that conversions between carbonate species do not affect the calculation.

The carbon mass balance for the studied sites has two inputs and two outputs (Fig. 1); the DIC lost due to CO2 evasion can be determined by rearranging the mass balance as follows:

Fig. 1
figure 1

The mass balance of DIC for the study sites. The DIC discharged at the mine portal (DICIN) and the dissolved species exported from the site (DICEX) are a combination of dissolved CO2 gas (H2CO *3 ) and the bicarbonate ion (HCO3 −) with the relative concentrations of the two species controlled by the pH; DIC lost to gas evasion (DICEVAS) is in the form of dissolved gas; and the DIC generated from the dissolution of limestone in the discharge channel (DICRX) is in the form of the carbonate ion (CO −23 )

Full size image
$${\text{DIC}}_{\text{EVAS}} = {\text{ DIC}}_{\text{IN}} + / - {\text{ DIC}}_{\text{RX}} \,{-}{\text{ DIC}}_{\text{EX}}$$
(9)

where DICEVAS is the flux of DIC from evasion of dissolved CO2 to the atmosphere, DICIN is the DIC flux from the portal, DICRX is the flux of DIC into and out of the system due to in-stream reactions, and DICEX is the DIC exported from the study site in the dissolved form (Fig. 1).

Site 1: Iron Falls

Site description

The Iron Falls site is approximately 32 km southeast of Pittsburgh, Pennsylvania near the town of Blythedale. The mine is at the base of the Pittsburgh Formation of the Monongahela Group, a Pennsylvanian rock unit (Pullman-Swindell Pullman-Swindell 1977) in the western limb of the Irwin Syncline (Capo et al. 2000).

The coal mine portal is partially sealed but water continues to discharge from around the portal seal and, to a lesser degree, from several adjacent pipes. Five meters from the portal is a waterfall with a drop of approximately 10.5 meters. At the base of the falls the stream flows down slope for ~ 180 m before discharging into the Youghiogheny River. Water samples were collected at the portal and over the entire distance from the portal to the confluence with the river (Fig. 2). A set of longitudinal samples were collected at Iron Falls in July 2012 and again in June 2013 (Tables 2, 3). Additional CO2 and/or DIC data were obtained between 2009 and 2014.

Fig. 2
figure 2

Iron Falls site. a Location of sites in eastern US; b the discharge portal (sample location F000); c approximate sampling locations on aerial imagery. Location labels indicate meters from portal. Base image from Google Maps

Full size image
Table 2 Field measurements and calculated carbonate alkalinity
Full size table
Table 3 Major element chemistry and calculated parameters
Full size table

Water chemistry at Iron Falls

Water flowing from the Iron Falls portal (F000) in 2012 and 2013 had slightly acidic pH (5.9 and 6.0) and contained dissolved CO2 approximately 240 times in excess of atmospheric equilibrium concentrations (Table 2). Periodic samples collected between 2009 and 2014 indicate that the mine has been consistently discharging water high in CO2 since at least 2009 (Fig. 3). Based on δ13C for three samples obtained from the Iron Falls portal in 2009 and 2010, Sharma et al. (2013) estimated that the carbon source was primarily from limestone dissolution.

Fig. 3
figure 3

CO2 at the portal water at Iron Falls. All data based on direct readings using carbonation meter. Numbers by symbols indicate n for replicates. Error bars based on standard deviations from replicates. The dashed line is for dissolved CO2 in equilibrium with atmospheric background (based on a log PCO2 = −3.5 atm). The solid line is the method detection limit (0.23 mmol/L)

Full size image

Downstream from the portal, DIC, CO2, metal concentrations, and ACIDMA decreased while the pH increased (Table 2; Fig. 4). The loss of DIC and increase in pH are attributed to CO2 evasion. The decrease in iron concentrations and metal acidity occur as oxygen in-gasses, iron is oxidized, and ferric hydroxide precipitates:

Fig. 4
figure 4

Changes in measured and calculated parameters downstream from the portal at Iron Falls site (left) and LRM (right). 2012 data in solid squares, 2013 data in open black triangle. Physical features discussed in the text are shaded. The horizontal line for pH is at 6.33 (the pK1 for carbonic acid): above the line HCO3 − is the dominant DIC species, below the line dissolved CO2 is the dominant DIC species

Full size image
$${\text{Fe}}^{ 2+ } + \, 0. 2 5 {\text{O}}_{ 2} + { 2}. 5 {\text{H}}_{ 2} {\text{O}} \to {\text{Fe}}\left( {\text{OH}} \right)_{{ 3({\text{solid}})}} + {\text{ 2H}}^{ + } .$$
(10)

As the mine discharge flows over the 10.5-m high waterfall (Figs. 2, 4), agitation and mixing with air enhances the degassing process; in the 2012 and 2013 sample sets, approximately 90 % of the CO2 was lost over the waterfall. In-gassing by oxygen also increases in waterfalls causing the ferrous iron to precipitate as Fe(OH)3 which causes Fe concentrations to decrease between the portal and the base of the waterfall (Table 3). The ACIDMA continued to decrease downstream at Iron Falls as iron precipitated from solution. Although the precipitation of hydrolysable metals releases protons, the pH remained approximately constant over the downstream reaches; this is attributed to the pH increases with CO2 degassing and the pre-existing carbonate alkalinity buffering the system against the free protons generated during metal hydrolysis. In 2012, the system remained net-acid throughout the stream reach while in 2013 it was generally net-alkaline (Fig. 4). The water was undersaturated with respect to calcite and dolomite over the entire distance (Table 3); the SIC values varied from −1.18 to −2.32.

The estimated CO2-evasion fluxes at Iron Falls were 67–272 kg-C/day (Table 4). The CO2 concentrations at Iron Falls were relatively constant; the range in flux was controlled by variations in discharge.

Table 4 Calculated fluxes of DIC loss to CO2 evasion
Full size table

The distribution of CO2 and HCO3 − is controlled by pH. At Iron Falls there is no carbonate rock source external to the portal; however, past attempts at treating the mine water may have included adding carbonates to the mine system. Therefore, it is not certain if the DIC generated within the mine is from dissolution of carbonates within the geologic strata or from the addition of limestone during or after mining.

Site 2: Lamberts Run (LRM)

Site description

The LRM site is located approximately 4.5 km north of Clarksburg, WV (Figs. 2, 5). Lamberts Run flows northeast from the site until it joins the West Fork River near the town of Spelter, WV (Fig. 5). The Lamberts Run watershed is underlain by the upper Monongahela and upper Conemaugh Series, both of which are of Upper Pennsylvanian in age. The mine drainage likely comes from mining of the Pittsburgh coal which forms the base of the Monongahela Series. The Lamberts Run area was mined for at least 100 years (Hennen 1912); however, limited historic information is available regarding this site.

Fig. 5
figure 5

LRM site. a The discharge portal (sample location LRM000); b approximate sampling locations on aerial imagery—location labels indicate meters from portal. Site location shown on Fig. 2a. Base image from Google Maps

Full size image

A passive treatment system was installed at this site in 2006 and included a portal seal, a limestone-lined channel, an altered natural wetland, a newly-constructed wetland, and a steel slag leech pit (Fig. 5). The natural wetland was altered using a series of baffles to increase the travel time of the drainage to promote precipitation of metals before discharge into West Fork River. The slag leech pit was designed to add alkalinity to the mine water, but little to no water flows from that unit. Water samples were collected from the portal and downstream locations (Fig. 5) in August 2012 and July 2013.

Water chemistry at LRM

Mine water samples were collected directly at the discharge point and had pHs of 3.13 and 5.28 with respective DIC concentrations of 3.7 and 5.2 mmol/L C (Table 2). At the lower pH, the DIC and CO2concentrations were equal within the error of the measurement. At the higher pH, the bicarbonate ion contributed a greater proportion of the DIC. The DIC, CO2, metal concentrations, and ACIDMA decreased downstream of the portal. Over the same reach, the pH increased (Table 2; Fig. 4). The loss of DIC and rise of pH are attributed to CO2 evasion while the decrease in iron concentrations and metal acidity are attributed to hydroxide precipitation.

The mine water flows through a limestone-lined channel in the 50 meters downstream from the portal (Fig. 5). Based the increase in Ca and Mg concentrations over this distance (approximately 18 %), the limestone is dissolving into the mine water. SIC values ranged from −7.48 (LMR000) to −0.8 (LRM172) which is indicative of waters that are undersaturated and aggressively dissolving calcite. Between the portal (LRM000) and the end of the limestone channel (at LRM050), the SIC increased which further supports that limestone is being actively dissolved. This dissolution releases CO −23 and acts as an in-stream source of DIC (+DICRX in Eq. 9). The contribution of this CO −23 source was estimated as equal to the sum of the Ca and Mg molar concentrations over the same stream reach. Dissolution from the limestone-lined channel contributed 27 % of the DIC in 2012 and 7 % in 2013. The greater contribution in 2012 may be driven by the lower pH of the portal water at that time (Table 2). The ALKCARB increase in this stream reach is a combination of carbonate released from rock dissolution plus a shift in the DIC from CO2 to HCO3 − as the pH increases. The rock dissolution should result in an increase in DIC over the first 50 meters; however, the DIC decreased over that distance because the loss of DIC by CO2 evasion exceeds the addition of DIC from rock dissolution.

In the wetland reach (80–160 m; Figs. 4, 5), alkalinity may also be generated by biological processes such as sulfate reduction (Younger et al. 2002); further limestone dissolution is unlikely in this reach because the Ca and Mg concentrations do not increase. The solution was undersaturated with respect to calcite indicating that precipitation of calcite is unlikely to account for changes in the Ca and Mg concentrations.

The CO2-evasive flux at LRMin August 2013 was 13.3 kg-C/day (Table 4). Due to the treatment channel, DIC sources included all three carbonate species: dissolved CO2, HCO3 − and CO −23 (Fig. 6). The DIC left the site via CO2 evasion and the downstream export of both dissolved CO2 and HCO3 −; CO2 evasion dominated the output.

Fig. 6
figure 6

The balance between input and output of DIC graphed in terms of fraction of the total source mass. The IN includes the source at the portal (DICIN) and in-stream reactions (DICRX) as per Eq. (5). Output includes export of dissolved DIC from site and CO2 evasion. Calculations assume a constant discharge throughout the site

Full size image

Discussion and interpretation

Comparison of carbon evasion and export between Iron Falls and LRM

The concentrations of dissolved CO2 measured at the mine portals were similar at Iron Falls (F000) and LRM (LRM000) (3.7–4.5 mM); these concentrations are up to 300 times greater than what would be expected for water in equilibrium with atmospheric CO2 (Table 2). The flux of DIC and loss of DIC via CO2 evasion is higher at Iron Falls due to its higher discharge rate (Table 4).

The physical characteristics of the two sites govern the location of the DIC loss. At Iron Falls, the greatest loss of DIC (ca. 90 %) took place as the mine drainage flowed over a waterfall (Fig. 4). This can be attributed to enhanced degassing at waterfalls; a process recognized at karst and travertine-depositing sites (Lorah and Herman 1988; Zhang et al. 2001). In contrast, at LRM the slope is more gradual and the loss of DIC was more evenly distributed from the portal downstream (Fig. 4).

Additional alkalinity sources may contribute to the DIC at LRM: dissolution of limestone in the drainage channel and sulfate reduction in the wetland. It is unlikely that carbonate minerals will precipitate at either site because the water was undersaturated with respect to calcite and dolomite. Although the two sites differ in physical configuration and the location of maximum flux, in all of the datasets obtained, the majority of the DIC lost was due to CO2 evasion (Table 4; Fig. 7).

Fig. 7
figure 7

Comparison of CO2 fluxes from this study to (1) total flux estimated from 140 sites in Pennsylvania (Cravotta 2008; Kirby and Cravotta 2005); (2) the average annual flux for the Susquehanna River Basin over the past 100 years (Raymond and Oh 2009); and, (3) 2014 fluxes reported for West Virginia power plants (USEPA Air Markets Program Data 2014). The blue squares are for the Pennsylvania mine sites with pHs ≥4.5; the red triangles are for the mine sites with pHs <4.5. The WV power plant annual data from 2014 are in gray and circled as a group

Full size image

Comparison of CO2 fluxes

Few studies have reported CO2 evasion estimates associated with coal mine drainage although related studies have focused on DIC loss in streams receiving mine waters (Atekwana and Fonyuy 2009; Fonyuy and Atekwana 2008a), CO2 degassing effects as part of mine drainge treatment (Cravotta 2007; Geroni et al. 2012; Kirby et al. 2009), and long-term basin-scale estimates on CO2 released due to mining (Raymond and Oh 2009). Although the goal of those studies was not to calculate the annual CO2 flux, their data can be used as a comparison with the fluxes calculated in this study.

Fonyuy and Atekwana (2008a) measured 50 to >98 % loss of DIC in streams that received mine drainage. The greatest loss of DIC occurred in streams containing high concentrations of Fe and low concentrations of alkalinity because these parameters controlled the acidity-alkalinity balance, the system’s ability to consume generated protons, and ultimately the pH of the water. In both Fonyuy and Atekwana’s (2008a) study and the data presented in this study, the pH of the discharge was a principle control on the DIC fate. At pHs below the first pKa for carbonic acid (~6.3), the DIC is predominantly in the dissolved CO2 species and may either degas or be exported as an aqueous species. Above that pH, dissolved export of HCO3 − is the dominant fate of the DIC.

The loss of DIC via CO2 degassing, and the consequent rise in pH, can cause the rate of oxidation of ferrous iron in mine waters to increase. Geroni et al. (2012) compared degassing via mechanical aeration and a cascades for the treatment of net-acid and net-alkalinity mine waters. The flux of CO2 can be estimated from the discharge and CO2 concentrations provided in that paper: using the change in dissolved CO2 for each cascade and an average discharge of 29.5 L/s, a CO2 flux of 0.0001–0.0002 Mt-CO2/year can be estimated (Table 5). Geroni et al. caveat their CO2 concentrations because they are based on colorimetric alkalinity data; however, the estimated fluxes from those data are within the range measured at the Iron Falls and LRM sites in this study (Table 5).

Table 5 Estimates of annual CO2 fluxes
Full size table

Cravotta (2008) conducted an extensive survey of coal mine water chemistry in Pennsylvania and provided data for 140 samples as supplemental material. In his 2008 analysis of the data, Cravotta focused on trace-element distributions and fluxes; this dataset included pH, alkalinity, and acidity that had previously been evaluated by Kirby and Cravotta (2005). The CO2 flux was estimated for these data as follows: (1) for sites with pHs >4.5, the dissolved CO2 concentration was estimated in PhreeqC (Parkhurst and Appelo 1999) using the full site data; and (2) for sites with pHs ≤4.5, the PCO2 was assumed to be 10−1 atm and the equilibrium dissolved concentration of CO2 was estimated using a temperature-corrected Henry’s Law constant. CO2 concentrations can be estimated using either alkalinity or DIC data; an assumed concentration was necessary for the low-pH samples because they contain no alkalinity and no DIC data were obtained. The average PCO2 calculated for the high-pH samples was 10−1 atm.

The CO2 evasive flux was then estimated by combining the CO2 data with the discharge and assuming that degassing continued until CO2 reached equilibrium with atmospheric concentrations. It is recognized that the degassing is unlikely to proceed to this concentration (Cravotta 2008; Geroni et al. 2012), but it provides a consistent endpoint for the calculations. Based on this calculation, the flux from the Pennsylvania mines ranged up to 0.016 Mt-CO2/year (Fig. 7; Table 5) with a total flux of approximately 0.076 Mt-CO2/year. Both the largest and smallest fluxes of CO2 were obtained from low-pH samples (Fig. 7).

A regional-scale perspective was provided by Raymond and Oh (2009) who estimated the long-term flux of CO2 due to mining for the entire Susquehanna River basin. By modeling pH and alkalinity data, they estimated that 3.1 Tg of carbon were released to the river due to mining over the past century. The furthest downstream location used in this study was the USGS station on the Susquehanna River at Danville, PA. The drainage area of the river at this location (USGS National Water Information System 2014) covers a land area approximately 25 % of the size of the state of Pennsylvania. On an annual basis, this flux of carbon in this system averages 0.11 Mt-CO2/year (Table 5). Raymond and Oh (2009) concluded that (1) their flux may be overestimated because they did not include CO2 drawdown from silicate weathering, and (2) that current fluxes are far less than what was likely during peak periods of mining in the basin.

The fluxes from individual drainage sites cannot be directly compared to the estimates from the Susquehanna River basin, but it is notable that the average flux for the entire basin is only one order of magnitude greater than the C discharged from the single mine with the largest flux (Table 5). Although these two datasets have different sources of data, temporal range, spatial range, and estimation methods, the comparison suggests that carbon release due to mining has a measurable impact but is difficult to accurately estimate.

An additional comparison can be made between the fluxes from the mine waters and regional power plants. The 2014 CO2 emission data were obtained from the USEPA Air Markets Program Data (USEPA 2014) for power plants located in West Virginia (WV), USA. The John E. Amos plant had the highest annual emissions (12.2 Mt-CO2/year) which are approximately two orders of magnitude greater than the total emissions estimated from the mine drainage data set from Cravotta (2008). The three smallest WV plants generated between 0.04 and 0.22 Mt-CO2/year; the estimated annual flux from the Susquehanna River Basin and the total flux from the Cravotta data set fall within this range (Fig. 7).

Limitations in estimating CO2 evasion from mine waters

In their review of CO2 evasion from terrestrial waters, Regnier et al. (2013) stated that the lack of direct CO2 and DIC data limit the accuracy of flux estimates. In lieu of direct data, alkalinity has commonly been used to estimate the aqueous CO2 by assuming that the total alkalinity equals the carbonate alkalinity (Cravotta and Trahan 1999; Kharaka et al. 2010; Keating et al. 2010; Raymond and Oh 2009). This approach works well in carbonate-dominated waters but in mine waters it can result in a bias in the estimated CO2 concentration. The presence of colloidal floccs of hydroxides in mine waters can contribute to the total alkalinity (Adams 2012; Vesper and Edenborn 2012) resulting in an overestimate of the CO2 concentration. Filtering prior to an alkalinity titration may not remove all hydroxide-alkalinity because of the small size of the particles (Waychunas et al. 2005).

An additional limitation in quantifying CO2 is the chemical disequilibrium of the mine waters after they are discharged from a portal. Once the mine water is exposed to the atmosphere, oxidation occurs concurrent with CO2 degassing. The significance of these changes is illustrated by the change in field and laboratory measurements reported in the dataset from Pennsylvania (Cravotta 2008). Of the 140 mine-water samples, the pH increased from the field-measured pH to a laboratory-measured pH in 54 % of the samples and decreased in 44 %. The maximum changes in pH were 2.78 and –2.17. If everything else remains constant, a pH change of two units corresponds to a change in the estimated dissolved CO2 by a factor of 100. The change in alkalinity between field and laboratory measurements has a lesser effect. Alkalinity decreased in 90 % of the samples with detectable concentrations. The greatest change in alkalinity was a sample with 6.2 meq/L in the field that decreased to 1.9 meq/L in the laboratory. The changes in pH and alkalinity were attributed to chemical instability, owing to atmospheric disequilibrium, at the time of collection (Kirby and Cravotta, 2005). Given that the laboratory measurements were made on aged samples, the impacts reported above constitute a worst-case comparison of field and laboratory values but it illustrates the challenges inherent in collecting mine water samples.

Fonyuy and Atekwana (2008b) evaluated post-sampling changes by tracking mine water chemistry in open reactors to determine the rates at which change occurs. They measured pH, alkalinity, DIC, iron, and stable carbon isotopes for 15–88 days after sample collection and found that the mine waters changed significantly more and lost more DIC than did the non-mine control samples. Both of these examples illustrate that even when careful sampling and processing protocols are followed, sample instability can contribute to inaccuracies in CO2 evasion estimates in mine waters.

To evaluate the error in the calculations, the flux of CO2 from Iron Falls was recalculated using the same method as used for the Pennsylvania data set: (1) alkalinity was used to estimate the dissolved CO2 in the water, and (2) the CO2 degassed until atmospheric equilibrium was obtained. When the flux is estimated using these assumptions, the resulting values are 11–82 % greater than the flux calculated from measured data (Table 6). This comparison suggests that the estimated fluxes from the Pennsylvania data (Fig. 7), which assume the water degasses to atmospheric equilibrium, are likely to be overestimated. However, without measured CO2 concentrations at these sites, the exact extent of the error is unknown.

Table 6 Estimate of flux at Iron Falls based on measured and estimated data
Full size table

In contrast, CO2 flux from mine waters may be underestimated when calculations exclude water samples with no detectable alkalinity. For example, the sample taken from the portal at LRM on 8/6/12 had a pH of 3.13 yet contained 3.7 mM DIC in the CO2 form (Table 2). Fonyuy and Atekwana (2008b) also reported samples that had low pHs (3.72, 3.62) with high concentrations of DIC (17.7, 14.3 mM-C/L).

Assuming an average CO2 concentration for ground water samples, as we did for the low-pH sample data, adds error to flux estimates. If the same CO2 concentration assumption is applied to the high-pH sites, the estimated CO2 flux obtained is approximately 40 % higher than the flux based on the results from PhreeqC. This suggests that assuming a constant CO2 concentration may cause an overestimate in the total flux values. In contrast, the exclusion of any low-pH high-DIC samples would result in an underestimation of the total CO2 flux. If the low-pH samples are excluded from the total estimate, the annual flux for the Cravotta sites decreases from 0.076 Mt-CO2/year to 0.043 Mt-CO2/year, a 44 % change.

In addition to needing a more accurate means of determining CO2flux from mine waters, more data are needed to estimate the percent of the carbon flux coming from geologic storage and evaluating how it changes over time. Sharma et al. (2013) parsed the carbon sources for three waters samples collected from the portal from Iron Falls using a two-endmember carbon-isotope mixing model. The model found that 65–82 % of the carbon being discharged at Iron Falls was from carbonate dissolution while the rest was of atmospheric origin. To obtain a good estimate of the release of geologically-bound carbon due to coal mining, these percentages require refinement. Additional data need to be obtained regarding the distribution of carbon sources vary (1) between sites, (2) in different geochemical settings (e.g., acid vs. alkaline mine waters), (3) with hydraulic conditions (e.g., storm vs. baseflow), (4) seasonally, and (5) over long-term time scales. With the limited amount of directly-measured CO2 data reported for mine waters, it is not possible to meaningfully assess bias or error at this time.

Conclusions

The Iron Falls and LRM sites discharge mine drainage waters that contain measurable concentrations of CO2 and DIC. The sites differ in physical layout and availability of in-stream carbon sources, but the evasion of CO2 to the atmosphere exceeded the downstream export of dissolved DIC at both sites. The magnitude of the CO2 evasive flux from these sites was comparable to values estimated from literature data for similar sites. These examples illustrate that coal mining can and does contribute to the release of geologically-bound carbon to the surface environment.

Limitations exist in obtaining accurate site-specific or regional estimates of carbon flux due to coal mining. Assumptions regarding the nature of alkalinity and the end-point of degassing may result in overestimation of the flux while conversely a lack of direct CO2 measurement in low-pH waters may result in underestimated fluxes. Regnier et al. (2013) concluded that more directly-measured data are necessary to understand CO2 evasion from surface waters; the data from Iron Falls and LRM suggest this is also true, and perhaps even more so, for mine drainage waters. Furthermore, it is critical that the distribution of carbon sources to the discharge be quantified to a greater spatial and temporal extent. Without additional research on these topics, it may not be possible to determine if current flux measurements overestimate or underestimate the impact of coal mining on terrestrial carbon export.

The preliminary estimates of CO2 flux from coal mine drainage suggest that these waters may be important to carbon cycling on a regional scale. Although considerable research has been conducted regarding the other environmental legacy of abandoned mine lands, their role in carbon cycling is poorly constrained and a better estimate of the CO2 flux is necessary to determine the significance of the estimates reported herein.