, Volume 713, Issue 1, pp 87–95

Exploring the legacy effects of surface coal mining on stream chemistry


    • Biology DepartmentUniversity of Rio Grande
  • Bradley M. Altier
    • Chemistry DepartmentUniversity of Rio Grande
  • Derek Haselman
    • Biology DepartmentUniversity of Rio Grande
  • Andrea D. Merry
    • Chemistry DepartmentUniversity of Rio Grande
  • Jacob J. White
    • Chemistry DepartmentUniversity of Rio Grande
Primary Research Paper

DOI: 10.1007/s10750-013-1494-9

Cite this article as:
Hopkins, R.L., Altier, B.M., Haselman, D. et al. Hydrobiologia (2013) 713: 87. doi:10.1007/s10750-013-1494-9


Surface coal mining results in dramatic alterations of the landscape in central Appalachia, leading to a myriad of environmental problems. In this study, we explore the long-term effects of surface coal mining on stream chemistry and endeavor to gain a better understanding of the efficacy of reclamation. We examined 30 sites in the Raccoon Creek watershed in southeastern Ohio, where the majority of surface mine sites are in various stages of reclamation. Our results show that conductivity (r = 0.862; P = 0.000), sulfate (r = 0.619; P = 0.000), and aluminum (r = 0.469; P = 0.009) levels increase linearly as a function of the areal extent of reclaimed mines in each subwatershed, suggesting limited success of reclamation to restore natural stream chemistry. In contrast, pH was not significantly linearly correlated with the areal extent of surface mines. This suggests that local acid mine drainage remediation projects are able to regulate acidity levels in the watershed but not conductivity and certain heavy metal concentrations. Many sites had conductivity levels high enough to impair aquatic biota via ionic and osmoregulatory stress. In sum, surface coal mining appears to have a strong legacy effect on stream chemistry in the Raccoon Creek watershed.


Water qualityAppalachiaReclamationStream ecosystemsSurface coal mining


In central Appalachia, surface coal mining is the chief driver of landscape change (Palmer et al., 2010; Bernhardt & Palmer, 2011; Lindberg et al., 2011). During the surface mining process, overburden layers of soil and rock are removed to expose the underlying coal seam. There is little debate that during the mining process stream habitat and water quality can be severely degraded (e.g., Merricks et al., 2007; Petty et al., 2010). Surface mining greatly affects stream systems by modifying topography, removing vegetation, exposing previously buried geologic materials, and even directly burying streams (Negley & Eshleman, 2006; Northington et al., 2011). As a result, streams in surface mined regions show much more variation in flow than expected and also show increased sediment loading (Ferrari et al., 2009; Reynolds & Reddy, 2012). Groundwater and surface waters from mined areas are often laden with a myriad of potentially toxic solutes including sulfate, iron, aluminum (Al), and selenium (Agouridis et al., 2012). Shifts in pH, higher electrical conductivity, and increased dissolved ions in streams are all hallmarks of surface mining (Palmer et al., 2010). Not surprisingly, the increasing level of surface coal mining has been accompanied by an extirpation of species in affected streams in the central Appalachian region (Pond, 2010; Bernhardt & Palmer, 2011; Northington et al., 2011), a biodiversity hotspot. Moreover, there is increasing evidence of threats to human health. Levels of hazardous dust, prevalence of chronic health problems, and drinking water pollution have all been shown to increase as a function of surface coal mining production (Hendryx & Ahem, 2008; Palmer et al., 2010).

Once the coal is mined, reclamation efforts take place and are aimed at mitigating environmental effects; the overburden is used to reshape the landscape to restore pre-mining topography, and vegetation (mostly grasses and herbs) is re-planted (Holl, 2002; USEPA, 2011). It is the efficacy of reclamation efforts that remains uncertain, specifically in terms of restoring natural stream function. While there is some evidence that reclamation efforts offer partial alleviation of sediment loading (Northington et al., 2011), the majority of studies show continued elevation of dissolved ions, increased conductivity, and modified flow regimes in streams downstream of reclaimed mine areas (Merricks et al., 2007; Pond et al., 2008; Johnson et al., 2010; Palmer et al., 2010; Lindberg et al., 2011; Northington et al., 2011). Thus, in spite of extensive reclamation efforts, surface coal mining may have a significant legacy effect on important parameters of stream chemistry and ecology.

It is important to note that the vast majority of recent studies on the effects of surface coal mining have been carried out in regions of central Appalachia in which mountaintop mining with valley filling (MTM/VF) is performed. This mode of surface mining is extremely controversial, involving the burial of extensive lengths of headwater streams and dramatic transformation of regional topography (Lindberg et al., 2011; USEPA, 2011). Comparable practices occur in southeastern Ohio, which is situated on the western edge of the central Appalachian coalfield on the Allegheny Plateau. However, the topography is not as deeply dissected in this region and influences of surface mining on topography and hydrology are more subtle. Admittedly, compared to MTM/VF operations in more mountainous areas, the effects of surface mining on stream systems in this region may not be as profound. Nonetheless, studies of the influence of surface mining and reclamation practices on stream chemistry in the transitional physiographic region of southeastern Ohio are much needed and may provide insight into the broader debate on MTM/VF practices.

In this study, we conduct a synoptic exploration of the legacy effects of surface coal mining on stream chemistry of the Raccoon Creek watershed in southeastern Ohio. While there is some active surface mining in the watershed, the majority of mined areas are in various stages of reclamation—providing an excellent opportunity to investigate the general effectiveness of reclamation processes and potential legacy effects. Our objective was to examine the relationship between stream chemistry and the areal extent of reclaimed and unreclaimed surface mines in subwatersheds with various histories of mining activity.


Study area

Raccoon Creek, situated in southeastern Ohio, is a direct tributary of the Ohio River and drains a total land area of about 1,770 km2 (Fig. 1). Lying on the western edge of the Appalachian Plateau, the watershed has a moderately dissected topography and is underlain by multiple coal seams of the Pennsylvanian Allegheny Formation (OGS, 2008). The coal in the area is heavily pyritic (FeS2) and acid mine drainage is a persistent environmental concern (OEPA, 1996; OGS, 2008). Our study focused on the upper region of the watershed, which is where the majority of the coal mining activity has occurred. Surface mining has been the predominant mining technique since the 1960s (OGS, 2008). Prior to that, deep mining was the choice method of coal extraction. Admittedly, the occurrence of deep mining presents a potentially confounding effect because the practice is known to contribute to acid mine drainage and heavy metal pollution in the region. However, based on the relative area impacted by surface mining and the high degree of spatial autocorrelation with deep mining sites, we believe our analyses remain pertinent.
Fig. 1

Map showing major streams in Raccoon Creek watershed, extent of surface mining, and location of sample sites. Inset map shows relative position in southeastern Ohio

Subwatershed analysis

We randomly selected 30 nested sample sites along the length of the upper Raccoon Creek watershed. Therefore, each site is not completely independent and our setup represents a longitudinal gradient analysis—subwatersheds contain variable levels of mining disturbance and reclamation. Based on permit dates, the age of the reclamation sites range from less than two to more than 25 years, with the majority of reclaimed sites being more than 10 years old. Subwatersheds contained anywhere from one to multiple different mining operations in various stages of reclamation. Besides mining activities, the area consists predominantly of forest and herbaceous land covers with only limited developed land and agriculture. Sites located at confluences were moved 100 meters upstream of the largest tributary. Areas affected by reclaimed and unreclaimed mining areas were identified and classified based on surface mining permit map layers provided by the Ohio Division of Mineral Resources, aerial photography, a digital elevation model, and the 2001 National Land Cover Dataset (NLCD). Reclaimed mines were zones within permitted areas with contrasting slopes compared to the surrounding landscape and some form of vegetation present (usually herbaceous). In a few cases, reclaimed mines have been converted to residential areas; however, these areas were still treated as reclaimed mines due to presumed fundamental shifts in surface geology and hydrology. Unreclaimed mines were zones within permitted areas with contrasting slopes compared to the surrounding landscape and no vegetation present. We used aerial photography and Google Earth images to complete a cursory accuracy assessment of our mining classifications. Based on 100 points confined to zones classified as some form of mining, our classification accuracy was ~77%. We did not assess omission errors and relied on the NLCD 2001 for all other land use/land cover (LULC) types. Non-mining land use/cover types were condensed into the following categories: developed land, barren land, forest, herbaceous, and agriculture. For each site’s subwatershed, we calculated the subwatershed area and percent areal coverage of each land cover type using ArcGIS 10.1.

Stream chemistry analysis

We sampled each of the 30 sites in early spring: our sampling was timed to occur prior to any row-crop agricultural activities and during baseflow conditions. Due to the timing of our sampling, our estimates of the effects of surface mining on water chemistry are likely conservative as relative effects tend to worsen during low flow conditions in summer (USEPA, 2011). At each site, we analyzed pH and conductivity using a YSI Quatro multimeter probe (Yellow Springs Instruments, Inc.). Water samples were collected for laboratory analysis of sulfate, phosphate, lead (Pb), and Al in March 2011. Sulfate was analyzed following the methods of ASTM D516-02 (ASTM Standard, 2002). Al was analyzed following the Standard Method 3120 B (APHA et al., 2005). Phosphate was analyzed using the Standard Method 4500 D (APHA et al., 2005). Lead was analyzed using ASTM D3559-08 (ASTM Standard, 2008).

Sulfate, conductivity, and pH were selected for analysis because they are classic physiochemical markers of coal mining (USEPA, 2011). In Raccoon Creek, mining activities are associated with increased conductivities and sulfate concentrations and decreases in pH (OEPA, 1996). The inclusion of Al and Pb was more exploratory and was based on anecdotal reports of high concentrations of these metals in some regions of the watershed (OEPA, 1996). Al is often associated with coal mining but its solubility is highly dependent on pH. Conversely, Pb is normally associated with the combustion process of coal but can be found released from abandoned coal refuse piles, slurry ponds, and areas where coal is typically cleaned. Lastly, phosphate was analyzed to help ascertain a limited influence of agriculture and developed land on water chemistry.

Statistical analysis

Considering the relatively homogenous bedrock geology, we postulate that land use/cover differences are primarily responsible for any variation observed in stream chemistry between subwatersheds. We used a forward-selection multiple regression analysis to model each water chemistry parameter. An additional threshold of α = 0.05 was applied. Because these variables are compositional metrics, there exists naturally some level of dependence. Therefore, prior to our multiple regression analyses, we arcsine-square root transformed the LULC variables. For presentation clarity, the results presented are shown in back-transformed format. We also computed partial regression analyses and graphed partial dependence plots to further explore the relationship between each LULC variable and water chemistry measurements. A lack of relationship between the areal composition of mining land uses and stream chemistry parameters may indicate successful reclamation and/or pollution control practices.


Twenty-seven of the 30 subwatersheds assessed showed some level of mining disturbance, with the majority of mining disturbance in the form of reclaimed mine lands (\( \bar{x} = 6.40\% \)) and very little unreclaimed mine lands (\( \bar{x} = 0.53\% \)) (Table 1). The subwatersheds also varied considerably in size, ranging from 0.8 to 765.6 km2 with a mean of 127.7 km2. Many stream chemistry parameters vary in a predictable longitudinal fashion in a stream system; however, subwatershed area was not sub-selected as a significant predictor of stream chemistry in any of the models. In addition, our phosphate analyses showed all samples were below the detectable level, indicating limited influence of agriculture and developed land in the watersheds (Cormier et al., 2013).
Table 1

Summary of areal extent and landscape composition for the 30 subwatersheds analyzed



SE of mean



WS area (km2)





% Water





% Developed





% Barren





% Forest





% Herbaceous





% Cropland





% Reclaimed





% Unreclaimed





Stream chemistry varied considerably between subwatersheds (Table 2). The exception was pH, which showed limited variation overall. Resultantly, no significant multiple regression model was identified. In contrast, Pb concentrations varied considerably with a range of 0–81 ppb; however, there was still no detectable relationship between Pb concentrations and subwatershed LULC. Significant models were produced only for conductivity, Al, and sulfate (Table 3). Notably, the % reclaimed mines was identified as a significant predictive factor in each of these models.
Table 2

Summary of stream chemistry for the 30 subwatersheds analyzed



SE of mean








Lead (ppb)





Conductivity (μS/cm)





Al (mg/l)





Sulfate (mg/l)





Table 3

Forward-selection, multiple linear regression models for conductivity, Al, and sulfate


Regression equation

P value

Adjusted r2


\( 3 5 7 + 3 4.0(\% {\text{ reclaimed}})- 3 5. 7 (\% {\text{ cropland)}} \)




\( 0. 7 7 6 + 0.0 4 1 (\% {\text{ reclaimed)}}-0.0 4 6 (\% {\text{ herbaceous)}} \)




\( 4 2. 7 + 1 4. 9 (\% {\text{ reclaimed)}} + 2 9.0(\% {\text{ unreclaimed}}) \)



For each model, the adjusted r2 and P value are shown

Conductivity levels in our study ranged from 126 to 1,631 μS/cm and were best predicted by the % reclaimed mines and % cropland in the subwatershed (P = 0.000; adjusted r2 = 0.79) (Table 3). Partial dependence plots illustrate a positive correlation of % reclaimed mines with conductivity and a negative correlation of cropland with conductivity (Fig. 2). Al levels varied from 0 to 2.30 mg/l across the subwatersheds and were best explained by % herbaceous and % reclaimed mines (P = 0.004; adjusted r2 = 0.283) (Table 3). As shown in partial dependence plots, the % herbaceous land cover was negatively correlated with Al concentrations while the % reclaimed mines was positively correlated with Al concentrations (Fig. 3). Sulfate levels ranged from 28 to 1,000 mg/l, and was most highly correlated with measures % reclaimed mines and % unreclaimed mines (P = 0.000; adjusted r2 = 0.800) (Table 3). Partial dependence plots show that both of these LULC types were strongly positively correlated with sulfate concentrations (Fig. 4).
Fig. 2

Partial regression analysis for conductivity in response to A % reclaimed and B % cropland. Partial correlation coefficients and P values are shown for each plot

Fig. 3

Partial regression analysis for Al in response to A % herbaceous and B % reclaimed. Partial correlation coefficients and P values are shown for each plot

Fig. 4

Partial regression analysis for sulfate in response to A % reclaimed and B % unreclaimed. Partial correlation coefficients and P values are shown for each plot


Our results suggest surface mining exerts long-term impacts on stream chemistry with sustained high levels conductivity, Al, and sulfate concentrations. Similar long-term effects of coal mining on stream chemistry have been documented worldwide. Wood et al. (1999) showed that water discharges from coal-mined areas in Scotland could show high levels of sulfate and iron for up to 40 years, while low pHs stemming from acid mine drainage usually returned to normal in less than 10 years due to rapid weathering of carbonate rock. In regard to pH, our study area has a history of acid mine drainage but is naturally lacking in limestone bedrock, suggesting the effects of acid mine drainage may persist longer. However, the pH levels were well within tolerance limits of aquatic life (Dodds & Whiles, 2010), and pH was not correlated with any measure of surface mining. These data suggest that current attempts by various government and non-profit agencies (e.g., Ohio Environmental Protection Agency, Raccoon Creek Partnership) to alleviate acidity problems stemming from abandoned mine lands and coal refuse piles are generally successful. Numerous acid mine drainage restoration techniques including lime dosing, steel slag leach beds, wetland construction, and open limestone channeling (Johnson & Hallberg, 2005) have been employed across the watershed and seem effective at controlling stream acidity.

Regarding the heavy metals analyzed, the extent of past mining in a watershed is positively associated with Al concentrations, suggestive of long-term pollution impacts. Al is a common trace metal in coal and can pose serious environmental and human health risks if introduced into water systems. The OEPA drinking water limit for Al is 0.20 mg/l, which was exceeded in 20 of 30 sample sites. Moreover, the acute toxicity of fishes has been observed at concentrations as low as 0.1 mg/l (Baker & Schofield, 1982). However, the highest Al concentrations were largely restricted to areas with pH values <6. This pattern is expected as the solubility (and toxicity) of Al increases in acidic waters (Nordstrom & Ball, 1986). In light of this, we suspect the stream acidity restoration projects are having a secondary influence in reducing heavy metal pollution with the decreased solubility of these elements at neutral and basic pHs. For example, in areas of central Appalachia with extensive limestone formations discharges from mining valley fills are typically alkaline and contain low dissolved Al concentrations (Griffith et al., 2012).

In contrast, Pb is less common in coal and is typically associated with the combustion process. However, storage piles of coal can leach significant quantities of Pb and other heavy metals if exposed to precipitation and surface waters (USEPA, 2011). Given the known occurrence of several abandoned mine lands, refuse piles, and slurry ponds in the watershed, we felt a Pb analysis was appropriate. These Pb levels detected are extraordinarily high given the US EPA maximum contaminant level of 15 ppb for drinking water (http://water.epa.gov/drink/contaminants/index.cfm). However, Pb concentrations were not significantly correlated with any of the landscape factors analyzed. Past sampling by Ohio EPA also indicate high levels of Pb in different portions of the Raccoon Creek watershed, particularly in sediments, but no source could be identified (OEPA, 1996). In any case, these levels of Pb are enigmatic and alarming—warranting further investigation.

Conductivity typically increases in coal-mined watersheds when precipitation, surface water, and groundwater become loaded with dissolved minerals as they flow through and over the exposed and broken up bedrock (Bernhardt et al., 2012; Griffith et al., 2012; Cormier et al., 2013). Our study shows that conductivity levels are best predicted by the % reclaimed mines in a subwatershed. These data suggest a continued negative influence in spite of reclamation efforts. Eight of the 30 sites had conductivity levels >500 μS/cm: streams at this level have been shown to display ecological impairment and decreased biological diversity compared to undisturbed streams, which often have a specific conductance <150 μS/cm2 (Pond et al., 2008; Agouridis et al., 2012). It is likely that many sites within the watershed had conductivity levels high enough to induce ionic and osmoregulatory stress in aquatic organisms (Pond et al., 2008; Bernhardt & Palmer, 2011; Bernhardt et al., 2012). These observations suggest that efforts at controlling stream acidity may be futile if conductivity levels remain elevated. As aforementioned, in some areas of central Appalachia, acid mine drainage is not an issue due to high incidences of carbonate minerals. However, declines in aquatic macroinvertebrate taxa have been documented in this region when conductivity levels exceeded ~300 μS/cm in spite of near neutral pH levels (Griffith et al., 2012). Interestingly, Agouridis et al. (2012) observed that reclaimed areas consisting predominantly of gray sandstone spoil showed more rapid decreases in conductivity over time compared to those having more brown sandstone spoil. Thus, long-term impacts of conductivity could potentially be reduced by selective use of spoil when completing the reclamation process.

The pattern of sulfate concentrations closely parallel that of conductivity. In addition to the % reclaimed mines, the % unreclaimed mines was also an important predictor for sulfate concentrations indicating that current mining operations are exacerbating sulfate loading; thus, strengthening our postulation that surface coal mining is the dominant driver of this aspect of stream chemistry in Appalachia, including the upper Raccoon Creek watershed (OEPA, 1996; Palmer et al., 2010; Petty et al., 2010; Lindberg et al., 2011; USEPA, 2011). In the sulfur rich coal seams of the Allegheny Plateau, sulfate (SO42−) is often the dominant dissolved ion. In fact, sulfate concentrations are often used to quantify mining disturbance impacts (Pond et al., 2008; Bernhardt & Palmer, 2011). Unmined streams in this region typically have 40–60 mg/l of sulfate (OEPA, 1996), yet 22 of 30 sites in our study exceeded this range. In fact, 17 sites exceeded 100 mg/l and three sites had over 500 mg/l sulfate. Singleton (2000) suggested that streams should contain <100 mg/l sulfate to protect freshwater biota, yet many streams in Raccoon Creek have shown sulfate concentrations in excess of 250 mg/l for the last 20+ years (OEPA, 1996). Comparable long-term stresses have been observed in heavily coal-mined areas of Germany, where intensive water quality management strategies have been created to address impacts of sulfate loading on drinking water sources (Maassen et al., 2012). Moreover, in a study of Australian coal-mined areas, Carroll et al. (2000) found that spoil type dramatically effects levels of sulfate loading—providing additional impetus to consider selective use of spoil materials during reclamation stages.

In sum, comparable to MTM/VF practices (see Bernhardt and Palmer, 2011), surface coal mining appears to have a strong legacy effect on stream chemistry in the Raccoon Creek watershed. Aquatic systems are highly sensitive to surface mining disturbances, and the negative effects on stream chemistry appear to persist over time, in spite of reclamation efforts. On a positive note, localized restoration projects aimed at remediation of acidity problems seem to effectively manage pH and reduce Al concentrations. However, this positive effect may be overwhelmed by the continued high conductivities observed. In many cases, subwatersheds with suitable pH had conductivity levels high enough to impair aquatic biota via ionic stress. For future studies, we aim to classify reclaimed mine sites according to dominant land cover and time since restoration and correlate those data with more proximal measures of stream chemistry to gain better insight into the long-term effectiveness of reclamation efforts.


This project was funded by a Provost Academic Excellence Initiative Grant from the University of Rio Grande and utilized instrumentation purchased through a grant from the Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy. Lastly, the comments of two anonymous reviewers also substantially improved the quality of this manuscript.

Copyright information

© Springer Science+Business Media Dordrecht 2013