Field Testing of Geomorphic Landform Design Features in Central Appalachia Open Access

Restoration of abandoned mine lands from bond-forfeited mining permits is ongoing in Appalachia and across the United States. The diversity of mining-impacted sites necessitates adaptive engineering approaches for sustainable land restoration. Challenges include mine refuse characteristics, such as coarse and fine coal gradations, as well as blended refuse that all exhibit unique erosion and strength properties. Reclamation sites, such as those located on mountain ridges, add complexity to land restoration challenges, because of the topography setting and the precipitation runoff effects to headwater streams, especially when acid mine drainage conditions exist. Advancing mining reclamation techniques from the approximate original contour method is needed to address inherent problems, such as high soil erosion, lack of vegetation, planar landforms, and landslides (Bell et al. 1989). Advancing landform restoration beyond flat slopes toward the natural geomorphology can minimize further disturbances in a watershed, compared with the original undisturbed land (Ferrari et al. 2009).

One potential technique to reclaim coal mining–affected areas is geomorphic landforming also termed geomorphic landform design (GLD). The geomorphic approach attempts to approximate the long-term, steady-state landform condition, leading to reduced erosional adjustment as compared with standard engineered designs (Toy & Chuse 2005). Recent research has considered the use of geomorphic landform principles for surface-mined land reclamation in central Appalachia. For example, application of the GLD approach included the characterization of mature landforms in southern West Virginia to inform the design processes, the development of conceptual geomorphic landform designs, geotechnical slope stability evaluations, groundwater seepage evaluations, hydrologic response comparisons, and erosion estimations (Buckley et al. 2013; DePriest et al. 2015, 2020; Hopkinson et al. 2015, 2017; Michael et al. 2015; Quaranta et al. 2013; Russell & Quaranta 2013; Santos et al. 2020; Sears et al. 2013, 2014, 2018).

The GLD approach is now moving to field implementation in West Virginia. Currently, we are working to establish a field demonstration site located in Greenbrier County, West Virginia (38°00′41″N, 80°36′08″W). The field site (19 ha) is a bond forfeiture coal refuse pile and is being reclaimed by the West Virginia Department of Environmental Protection (WVDEP). The refuse site was originally planned to be reclaimed using a conventional land cover approach consisting of benched and flat slope profiles covered with a vegetative growth layer. This was a low-cost option to site closure; however, it did not address the long-term acid mine drainage treatment costs associated with groundwater seepage infiltration and discharge compliance requirements. The WVDEP collaborated with the West Virginia University, Department of Civil and Environmental Engineering, and the West Virginia Water Research Institute to investigate a robust reclamation alternative for the site to address site closure and seek to mitigate acid mine drainage generation. The outcome of the collaboration produced a design strategy using a hydraulic barrier cover structure, having compound surface slope profiles, and incorporating a surface vegetative layer composed of paper mill sludge by-products. The design integrated an outwardly draining network of reclaimed surface drainage channels. This approach followed the general principles of GLD, customized to a mountaintop application. The full design is described by Hopkinson et al. (2017).

Vegetation cover is critical for the long-term stability of reclaimed coal refuse piles (Daniels et al. 2018). Paper mill sludge was included in the design as part of the growth layer in the cap and cover system to support the establishment of vegetation. Paper mill sludge, also called short paper fiber, is a by-product from the water treatment process from paper production. Daniels et al. (2013) evaluated the nutrients in the paper mill sludge to determine its suitability for use as a soil amendment. The C to N (carbon-to-nitrogen) ratio was 22.2:1, P (phosphorous) concentration was 79 ppm, the pH value was 8.1, and the calcium carbonate equivalence was 33.3%, aiding acidic neutralization. Camberato et al. (2006) identified that paper mill sludge can increase organic matter, nutrients, and the water-holding capacity of soil. The low unit weight of paper mill sludge decreased bulk soil density, aiding slope stability. Laubenstein (2004) reported that the use of paper mill sludge in coal refuse reclamation aided seepage water quality by increasing pH and reducing concentrations of Al (aluminum), manganese, and acidity.

The design includes features related to geomorphic landforming and the cover system that have not yet been field tested. The objectives of the research described herein were to design, implement, and monitor a pilot test design of a geomorphic landform with a cap and cover system at a coal refuse site in Greenbrier County, West Virginia.

The field test site represents a small-scale geomorphic landform integrated with three subtest plots draining into a center, geomorphic channel (Fig. 1). The three test plots were used to test functions of the composite hydraulic barrier and growth layers. The three test plots were defined with three different cap and cover systems as follows: the refuse control plot without the growth layer; 60:40 (60% refuse with 40% paper mill sludge) growth layer over a hydraulic barrier; and 80:20 (80% refuse with 20% paper mill sludge) growth layer over a low hydraulic barrier. The entire site was designed to be seeded according to WVDEP specifications for permanent seeding (WVDEP 2006).

The design of the cover system was guided by using elements of the US Environmental Protection Agency pertaining to municipal solid waste landfills (Closure Criteria 1991). The municipal solid waste landfill requirements stipulate a low-permeability bottom layer, an infiltration layer at least 45.7 cm thick of earthen material, and an erosion layer at least 15.2 cm thick that can sustain native plant growth. The cover system used in this pilot test included the growth layer and low-permeability zone (hydraulic barrier) over the in situ coal refuse (Fig. 2).

The growth layer is the uppermost portion of the cover system and consisted of 0.3 m of on-site coarse coal refuse amended with a paper mill by-product material MGro (WestRock, Covington, VA). This paper mill by-product material is primarily composed of water (50%–60%), pulp mill waste solids (15%–30%), and cellulose wood fiber (10%–30%; G. Cox, personal correspondence, 8/24/2015). According to the material safety data sheet, the MGro calcium carbonate equivalency was 29%. Nitrogen, P, K (potassium), and Ca (calcium) compositions were 1.3%, 1.3%, 0.13%, and 11.6%, respectively (MeadWestvaco 2009). Two ratios (60:40 and 80:20) were selected based on the potential improvement to ground cover compared with the refuse (Cyphers et al. 2018). The 80:20 blend was considered for cost savings.

The low-permeability layer of the cover system included 0.6 m of compacted coal refuse as a hydraulic barrier to reduce precipitation infiltration into the pile. It was also used to maintain strength for slope stability. The low-permeability layer was designed for material placement held at >95% of maximum dry density at standard Proctor energy, defined by the standard method D-698-12 (ASTM International 2012). The third layer consisted of in situ refuse material that was not otherwise modified (Fig. 2).

The center draining channel was designed using SEDCAD (Schwab 2010), assuming a 10%, “steep” channel following procedures discussed by Hopkinson et al. (2017). The bottom width, top width, and depth with freeboard of the channel were 1.22, 1.98, and 0.20 m, respectively. The channel was lined with sandstone gravel with D₅₀ equal to 22.9 cm. The hydraulic barrier was designed beneath the filter layer and riprap liner.

The pilot test area was regraded according to the design criteria (Fig. 1). The borrowed refuse was scraped with a bulldozer and loaded by an excavator for transport by truck to the fill placement locations. Slopes and final elevations were rough graded and finished with light regrading. The backslope of the plots to the existing ground was constructed at a 1H:1V slope (45°). According to the final survey, the total fill volume needed for construction was 305 m³.

The channel was cut into the compacted refuse below the final grade elevation. Sandstone gravel (D₅₀ = 22.9 cm) lined the channel; the recommended filter material was not added but should be considered in future designs. The channel slope was constructed as 7.1%, rather than the proposed value of 10%, due to site conditions.

The outermost layer after regrading was used as the hydraulic barrier. The coal refuse hydraulic barrier was compacted by tracking the bulldozer over the refuse without blading as one compaction layer, in contrast to the design that included the compaction of multiple layers with a vibratory roller to achieve the required 95% of maximum dry density. To simulate local terrain characteristics, the designed slopes in geomorphic reclamation may exceed the steepness for the operational safety of traditional compaction equipment. Alternatives for compaction in these areas may include options, such as pulling the compaction equipment with a heavier piece of equipment at the slope crest or the use of compaction attachments to an excavator.

Soil compaction density testing was conducted in five locations using a Troxler 3440 moisture and density gauge (Research Triangle Park, NC). Field moisture content did not meet the specified range, exceeding optimum moisture content up to 7%. The coarse coal refuse maximum dry density yielded 59%–74% of the targeted value (15 kN/m³). With compaction not reaching 95% of maximum dry density, the infiltration rate would be expected to be higher than designed (Figure 3).

Multiple techniques were explored for mixing and placing the MGro and coal refuse blends. The first method attempted on-site was to spread a layer of MGro onto the hydraulic barrier with a bulldozer, followed by a proportional layer of refuse (Fig. 4a). For the 60:40 subplot, the MGro was spread 20.3 cm thick, followed by a 30.5-cm layer of refuse. The 80:20 subplot growth layer was composed of a 7.6-cm layer of MGro with a 30.5-cm refuse layer on top. This method was expected to maintain the appropriate MGro and refuse ratio and create a 30.5-cm growth layer after mixing, because the MGro takes the volume of the refuse voids (Park 2017).

This technique assumed that the bulldozer tracks helped adequately mix the materials during spreading. This was not the case, and the materials remained separated in layers. The MGro became compacted and did not mix with the refuse. The blade of the bulldozer was then used to scrape the MGro and refuse of the growth layer and mechanically work the materials together on the plots in a second attempt (Fig. 4b).

The additional mixing created a slightly better blend (evaluated by visual observation) than the first trial, but large accumulations of paper sludge remained. The compaction of the paper fiber from the first attempt left the paper fiber in a more difficult condition for mixing than when the material arrived on-site, as seen by later mixing trials.

The third technique required the materials to be mixed before spreading. Both the refuse and paper sludge were in large piles. For the 60:40 blend mixture, three full excavator bucket scoops (2.3 m³) of refuse were mixed with every two buckets (1.5 m³) of paper fiber. The excavator operator mixed the smaller pile with the teeth of the bucket (Fig. 4c).

A small tiller was used to mix the materials that were in layers from the previous attempts, as the growth layer did not appear to be an adequate mixed and uniform growth layer (Fig. 4d). The tiller was smaller than the equipment that will be used for the full site reclamation but was studied conceptually. The tiller worked well within the top few inches; however, a tiller would need to be able to reach the entire growth layer for even mixing (= 30.5 cm). In addition, large shale particles may also present a hazard and damage the tiller.

The growth layer for the 80:20 blend subplot was placed by using material mixed with the excavator bucket. However, the 60:40 subplot was placed before mixing. Instead of replacing the material to be mixed in the same method as the 80:20 subplot, the unsuccessfully mixed layers of the 60:40 subplot were then remixed in place with the excavator bucket. This method did not mix as well as mixing before placing but was more uniform than the prior methods of placing before mixing. The different mixing techniques can be a source of error in the results.

Seeding was completed 4 September 2017 and reseeded on 9 May 2018, due to poor initial growth. The entire site was designed to be seeded according to WVDEP specifications for permanent seeding (WVDEP 2006). The permanent seed mixture included orchard grass (Dactylis glomerata), birdsfoot trefoil (Lotus corniculatus), red clover (Trifolium pratense), annual ryegrass (Lolium multiflorum), bicolor lespedeza (Lespedeza bicolor), and foxtail millet (Setaria italica) (spring) or winter wheat (Triticum aestivum) (fall). Wood cellulose fiber mulch was mixed with the seeding slurry and applied via hydroseeding.

Performance of the pilot test site was monitored for infiltration, ground cover, surface water runoff quality, and surface temperature. Monitoring began at the completion of the site on 4 September 2017 (Fig. 5).

Infiltration of both the hydraulic barrier and growth layer were measured by using Turf-Tec International 15.2-cm infiltrometer (Tallahassee, FL) following standard methods. Infiltration readings were taken before site disturbance, upon completion of the hydraulic barrier, and through the growth layer.

Vegetation was monitored monthly in the fall, suspended during winter, and resumed in spring. Ground cover was measured at five random sample locations within each test plot using a portable point frame (1.0 m × 1.0 m) following the procedures of Coulloudon et al. (1996) and Elzinga et al. (1998). The grid has 100 equally spaced intersections. Each intersection was classified as vegetated or not vegetated, resulting in a local value for ground cover. The mean of the five sample locations was recorded for the ground cover of each test plot.

Water runoff from the cover system was monitored on each subplot, before reaching the channel. Fifteen collectors were placed throughout the test subplots (five on each subplot; Fig. 6). The collectors were constructed from 15.2-cm-diameter polyvinyl chloride pipe cut to 20.3 cm tall and capped at the bottom. A hook was added and attached to a rod driven into the ground below the collector, approximately 30.5 cm. This helped to minimize vertical displacement when pore pressure exerted an upward force that caused the top to move above ground level, making them ineffective. A screen (apparent opening size = 0.25 cm) was added to filter larger solid particles.

A YSI Pro-Plus multiparameter water quality meter (Yellow Springs, OH) measured total dissolved solids, dissolved oxygen, conductivity, pH, and temperature in the field. Water samples were collected and analyzed by REI Consultants (Beaver, WV) for: Al, Ca, Fe, Mg, Mn, and Na (EPA 200.7 Rev. 4.4); chloride (Cl⁻) and sulfate (SO₄²⁻) (EPA 300.0, Rev. 2.1); and organic N (EPA 351.2, Rev. 2.0).

Composite soil samples were collected from each subplot from the top 15.2 cm of the growth layer. The composite samples were analyzed by AgSource Harris Laboratories (Lincoln, NE) for organic matter, pH, N, P, K, Fe, Mg (magnesium), Ca, and Na (sodium). Analytical procedures followed the methods reported by AgSource (2022).

Surface temperatures were measured by using a General IRT207 infrared thermometer (Secaucus, NJ) in five random locations per plot. Surface temperature was introduced to monitor the difference in surface temperatures of each subplot due to differences in aspect.

Infiltration rates were expected to be high because compaction requirements were not met for the hydraulic barrier; however, infiltration rates in the hydraulic barrier were less than the rates prior to disturbance. The infiltration rates of the hydraulic barrier were 15.5%–20.6% and 9.7%–16.5% less than the refuse before disturbance for the 60:40 and 80:20 plots, respectively (Fig. 7). Although the target compaction density was not reached, the barrier still effectively reduced infiltration with compaction during the construction of the hydraulic barrier.

The 60:40 and 80:20 growth layers were placed as uncompacted layers. The infiltration rate of the 60:40 plot was twice that of the 80:20 treatment. Differences in placement technique likely contributed to this difference. The 60:40 growth layer was handled nearly 3 times more than the 80:20 layer, allowing more opportunity for introducing air into the voids. This difference is supported by soil testing results reported by Stevens (2016) and Park (2017) that suggested the hydraulic conductivity of the 60:40 blend is almost double the 80:20 blend.

In a small-scale growth study, Cyphers et al. (2018) reported that 60:40 and 80:20 blends of MGro and refuse resulted in significantly higher ground cover than refuse alone; however, no significant differences were determined between blend ratios. In the present study, the 80:20 plot had greater ground cover than the 60:40 plot (Fig. 8). This difference may be explained, in part, due to the aspect of each test plot. The 60:40 plot was directed toward the southeast, and the 80:20 plot was directed toward the northwest. A previous study in coal mining spoil heaps showed a positive correlation between slope aspect and vegetation cover. Northwestern slopes retain more moisture in the soil from less evaporation, only receiving evening sun, and benefiting vegetation establishment efforts (Nyssen & Vermeersch 2010).

An increase in sunlight exposure can increase water stress and ground temperatures (Sandholt et al. 2002). Water stress can constrain plant growth and decrease total nutrient content because of decreased uptake of water and nutrients and dieback of roots and shoots (Lenhart et al. 2015). These factors may explain, in part, the differences in success of establishing vegetation, rather than only the intended difference of mixing proportion in the growth layer.

Ground cover increased above 75% following a second planting in May 2018. Although the initial planting date was within the recommended fall planting period (August 8 to October 17), the permanent seed mixture contained several species that is preferred for spring planting, including annual ryegrass and bicolor lespedeza (US Department of Agriculture 2022). In addition, limited precipitation occurred following the fall planting date. Daily precipitation totals from the Rupert 4 N weather station (38° 01′ 05″ N, 80° 40′ 56″ W) located 7 km from the field site indicated that significant precipitation occurred on only 4 days within 3 weeks of the fall planting, but significant precipitation occurred 13 days within 3 weeks of the second planting in the spring (Fig. 9).

The decrease of vegetation over the winter months was likely due to the biological inactivity of the grasses from low temperatures (Beard 1973). The temporary fencing around the site become damaged in the winter months, possibly allowing animals onto the test subplots to consume some vegetation.

With the low instance of ground cover in the test subplots after recent construction, the water surface runoff quality was expected to be low (Tables 1 and 2). In general, the 60:40 plot resulted in better pH, conductivity, total dissolved solids, and dissolved oxygen than the other plots (Table 1). Concentrations of Al, Ca, Fe, Mg, Mn, and SO₄²⁻ in the runoff from blended plots were less than in the refuse plot. The conductivity values for all plots exceeded the benchmark value for Appalachia streams (= 300 μs/cm; US EPA 2011); however, surface runoff quality is expected to increase once vegetation is established and erodibility decreases (Nyman et al. 2013).

Soil samples were collected on 11 October 2017 after seeding. The 60:40 and 80:20 plots had 4% and 12% ground cover, respectively. The pH of each of the samples was similar, ranging from 7.1 to 7.5, a value higher than expected for the refuse plot. Compared with previous results, the refuse sample displayed uncharacteristic values (i.e., pH, P, Mg, Ca, and Na); therefore, data from sampling before site disturbance and construction were also reported. Coal refuse piles are composed of fragments of shale, mudstone, and coal residue (Quaranta et al. 2022). The organic matter variation is likely due to sampling, mixing techniques, and heterogeneity of the material (Table 3).

It is possible that some of the paper fiber became mixed with the refuse while constructing the growth layer 60:40 and 80:20 plots. The refuse plot was between the mixing area and the paper sludge growth layer. For the first mixing technique, the paper fiber was transported and spread with the bulldozer blade. Inevitably, paper fiber was mixed into the refuse subplot and did not leave bare refuse to sample.

The MGro material is of varying composition. The material safety data sheet (MeadWestvaco 2009) shows that the residual solids are composed primarily of pulp mill waste solids (15%–30%) and cellulose wood fiber (10%–30%). Up to 5% of the residual solids may be composed of fly or wood or both that may have the following elements present as oxides or hydroxides: Al, Ca, Fe (iron), Mg, K, titanium, and silicon. Additional components can include quartz (0.1%–10%), titanium dioxide (<3%), kaolin clay, calcium carbonate, and calcium phosphate (<5%). This composition may contribute to the increase in Fe, Ca, and K in the plots containing MGro.

With the dark nature of the refuse material, ground surface temperatures became a concern for vegetation germination and persistence. With the differing slope aspect of the subplots, surface temperatures were expected to vary among the subplots. On average, the 60:40 subplot was 5.1°C warmer than the 80:20 subplot (Fig. 10).

Depending on the time of the year, ground temperatures will affect the subplots differently. In early spring, the higher surface temperatures are beneficial to growth. In the peak of summer, the subplots are more prone to surface burnout. The 60:40 subplot stands a greater chance because of the slope aspect causing a longer period of sun exposure. Figure 11 shows that snowmelt is more rapid on the 60:40 subplot.

Compaction of the refuse was completed by the weight of the dozer used to place the material because a vibratory roller could not be safely operated. Due to the compaction technique and excessive moisture during soil compaction, the low-permeability layer did not attain 95% of the maximum dry density required (15 kN/m³); therefore, the rate of infiltration was expected to be higher than designed. Infiltration was expected to be four orders of magnitude less than what was produced in the field. If unable to operate a vibratory roller safely on all areas of the full reclamation, alternative compaction techniques should be implemented, such as pulling the compaction equipment with a heavier piece of equipment at the slope crest or using compaction attachments to an excavator.

The growth layer materials must be mixed thoroughly to have success in establishing and maintaining vegetation. This layer should be mixed before placement, or it may be spread first and then mixed with a large tiller that is able to reach a depth of 30.5 cm from the ground surface to mix the full layer thickness.

To aid in the success of the reclamation, construction should begin in the dry season of the summer. The low compaction experienced in the test subplots was a result of low compaction energy and high-moisture contents. By regrading in the driest time of the year, compaction may be increased with a small increase in energy combined with a lower in situ moisture content. The grass mixture is to be seeded in the spring or fall. Shortly after the completion of the regrading and growth layer placement during the summer, seeding could take place in the fall, if construction is finished. If that does not happen, seeding could occur in the following spring months but should not take place in winter.

Vegetation ground cover above 80% was measured in both plots that included paper mill sludge. Infiltration of the refuse was reduced by 9.7%–20.6% in the hydraulic barrier of the plots containing paper fiber from the in situ material, despite not meeting the 95% required compaction density. The surface temperature varied across the plots. The findings suggest that the mixture ratio containing 20% paper mill sludge may be sufficient for supporting vegetation. Reducing paper mill sludge for the growth layer will result in cost savings for the final reclamation.

The study demonstrated that reclamation using the GLD aspects with a cap and cover could be implemented for coal refuse reclamation; however, this article reports the results of a single, nonreplicated study. The conclusions stated herein will need further research and replications to fully understand the design implications and coarse coal refuse and MGro blend soil characteristics.

The work described in this publication was supported by the Office of Surface Mining, Reclamation, and Enforcement (OSMRE; grant S15AC20020). The contents in this manuscript are solely the responsibility of the authors and do not necessarily represent the official views of the OSMRE. This work was also in collaboration with the West Virginia Department of Environmental Protection. The authors thank Mike Sheehan, Nathan Parks, Rick Pino, Dave McCoy, and Mike Richardson for continued support.