An important first step in the geotechnical asset management of Great Smoky Mountains National Park (GRSM) is the creation of an unstable slope inventory along major transportation corridors. Slope-stability problems are frequent in GRSM, often initiated in highly weathered and fractured metasedimentary rocks. In this study, an unstable slope inventory was created using the Unstable Slope Management Program for Federal Land Management Agencies protocols. Hazards and risks were evaluated for 285 unstable slopes along 243.67 km of roadway. Kernel density estimation was used to identify unstable slope hotspots and establish 14 sites for site-specific investigations to evaluate potential impacts of discrete unstable slopes along major roadways. Two-dimensional probabilistic rockfall simulations and acid-base accounting tests were used to predict rockfall pathways and evaluate the acid-producing potential of rocks. Simulations indicated that rock material would likely enter the roadway at all 14 sites. Acid-base accounting test results indicated that slaty rocks of the Anakeesta Formation and graphitic schist of the Wehutty Formation are primary acid-producing rocks in rockfall-prone areas. This research illustrates an approach for prioritizing areas for site-specific investigations towards the goal of improving safety in GRSM, including developing mitigation strategies for rockfall by widening ditches, installing barriers, and encapsulating acidic rockfall material.

The terms slope failure and landslide are often used interchangeably to describe a wide variety of natural geomorphic processes that result in downward movement of earth materials, including rock, soil, artificial fill, or a combination of these (Varnes, 1978; Turner and Schuster, 1996). The different types of slope failures can be distinguished based on the nature of materials involved and their movement. Failures occur frequently in the mountainous terrains of the United States (e.g., Colorado Plateau, Appalachian Mountains, Coastal Ranges of California, Southern Rocky Mountains, Pacific Northwest Coast Range of Oregon and Washington, Olympic Mountains, and Cascade Range) and range from small rockfalls or slope creep to complex landslides, rock avalanches, and debris flows. The behaviors and outcomes of slope-failure events will vary based on their location and many underlying factors. For example, in the Pacific Northwest region, failure events are mostly triggered by rainfall, earthquakes, or volcanic activities (Wieczorek and Leahy, 2008). In the Appalachian Mountains, including the Great Smoky Mountains National Park (GRSM), slope failures result from complex interactions among various rock and soil types, joint geometries, precipitation duration and intensity, topographic profiles, and hydrological conditions (Wieczorek et al., 2000; Moore, 2004; and Nandi and Shakoor, 2017).

Unstable Slope Failures in GRSM

Great Smoky Mountains National Park straddles the border of North Carolina (NC) and Tennessee (TN), covering an area of more than 500,000 acres (Figure 1 ) (NPS, 2019). The park is the most visited park of the 62 national parks in the United States, accommodating more than 14 million visitors in 2021 (NPS, 2021). The park generates more than $1.05 billion in visitor spending and provides employment for more than 15,000 people in local communities (Cullinane and Koontz, 2020). Each year, unanticipated road closures due to slope-failure events occur within the park. These events interfere with park objectives and have a significant negative economic impact on the regional economy (Anderson and Cuelho, 2017). For example, in 2013, heavy rainfall in the GRSM resulted in a landslide needing about $4 million for repair. Smaller-scale slope failures related to maintenance costs are frequent and range from $25,000 to $200,000, excluding added vehicle and emissions costs, travel time, and maintenance of detour routes (according to correspondence with GRSM maintenance personnel).
Figure 1.

Major transportation corridors in Great Smoky Mountains National Park, TN and NC.

Figure 1.

Major transportation corridors in Great Smoky Mountains National Park, TN and NC.

Large-scale landslides are not common along the GRSM transportation corridors, but when they do occur, they can completely or partially close the road network, causing economic loss as well as social costs. When the ground conditions are favorable, rainfall from cloudbursts, hurricanes, and storms can trigger fast-moving flows (Wieczorek et al., 2000). Bogucki (1976) identified numerous rockslides and debris flows in GRSM during a September 1951 rainstorm. About 50 percent of the debris flows from those slides occurred in the Mount Le Conte–Sugarland Mountain area and Alum Cave Creek watershed, significantly damaging the roads and hiking trails. More than 60 percent of the debris flows happened on slate and phyllite of the Anakeesta Formation, and the rest occurred on metasandstone of the Thunderhead Formation (Bogucki, 1976). In 2010, three rockfall events occurred on roads that serve GRSM park visitors. The largest and most disruptive failure event occurred on January 25, 2010, along a southbound section of Route 0011S (Gatlinburg Spur), an arterial access route within the park. As a result, both southbound lanes of the spur were closed for more than 30 days (TDOT, 2010a). Though the Tennessee Department of Transportation (TDOT) was responsible for $700,000 in emergency expenditure and cleanup for the January 25, 2010, rockslide, each event also posed a risk to GRSM park visitors who frequently travel along this route (TDOT, 2010b). A recent slide, close to the Trout Branch tributary of Little Pigeon River, transformed into a debris flow in August 2012 and damaged the Alum Cave trail (Nandi and Shakoor, 2017). A heavy rainfall event in January 2013 triggered a large cut-slope embankment slope failure and created a large landslide that destroyed about 200 m (∼600 ft) of Route 0010S (Newfound Gap Road or U.S. Route 441) in the GRSM towards NC, a major economic commerce corridor for communities on either side of the park (USGS, 2013). Regrettably, slope-failure events in the park have led to fatalities. On August 1, 2019, a man was killed by a fallen tree on the Gatlinburg Spur where multiple rockslides occurred following heavy rainfall. According to a local news station, more than 10 cm (4 four in.) of rain fell in just over 1 hour, which triggered the event (Cherokee One Feather, 2019).

In 2008, the National Park Service (NPS) published its most recent Great Smoky Mountains National Park Geologic Resource Evaluation Report. The report compiled information related to geologic issues (e.g., erosion and slope processes, abandoned mines, air and water quality) as well as geologic features and processes (e.g., major faults, views, tectonic windows). The report was designed to be used by park officials, scientific researchers, conservation and environmental constituencies, and the public. A section related to geohazards can be found in the report. However, it does not provide a usable database for tracking potential geohazard sites along park routes (Thornberry-Ehrlich, 2008).

The NPS is responsible for operating and maintaining 510 km (315 mi) of roadway within GRSM boundaries, 243.67 km (151.41 miles) of which are paved (NPS-GRSM, 2014, Figure 1 and Table 1 . Significant roads within the park include: Route 0008 A, E, F, G, and H (Foothills Parkway), Route 0010 N and S (Newfound Gap Road or U.S. Route 441), Route 0011 N and S (Gatlinburg Spur), Route 0014 (Little River Gorge Road), Route 0015 (Laurel Creek Road), Route 0017 (Clingmans Dome Access Road), Route 0019 (Lakeview Drive East), Route 0026 (Cades Cove Loop Road), Route 0105 (Cherokee Orchard Road), and Route 0107 (Heintooga Ridge Road). Figure 2 shows the current conditions of some representative slopes in the park. Several paved roads that traverse mountainous terrain serve not only park visitors, but also local and regional traffic. After nearly 80 years of use on some roads, GRSM's transportation corridors require effective long-term management (Anderson, 2016). NPS has recognized the need to implement a proactive, risk-based strategic unstable slope management approach for GRSM transportation routes in the face of fluctuating annual budgets and aging geotechnical assets that become more unstable as they are continually exposed to the environment.
Table 1.

Paved roads at GRSM, within the study area, where route ID corresponds to the road designations.

Figure 2.

Current slope condition photos in (a) Newfound Gap (0010N), (b) Little River Gorge (0014), (c) Gatlinburg Spur (0011S), and (d) Clingmans Dome (0017).

Figure 2.

Current slope condition photos in (a) Newfound Gap (0010N), (b) Little River Gorge (0014), (c) Gatlinburg Spur (0011S), and (d) Clingmans Dome (0017).

Transportation Corridor Risk Assessment

Roadway and trail slopes are transportation or geotechnical assets, and their reliable performance helps the transportation system to operate safely. These assets have a life cycle; if the slopes fail, the cost of repair can be much greater than periodically intervening with risk-reduction improvements. Unfortunately, the slope assets are often overlooked until they directly damage and impact the transportation system. Risk-based geotechnical asset management (GAM) is fundamental for slope maintenance; it reduces risk, improves system performance, and, if actively managed, can reduce slope life-cycle costs and improve safety (Stanley, 2011; Anderson, 2016). To help foster GAM, federal and various state departments of transportation have developed roadway unstable slope, landslide hazard, and/or rockfall rating systems to rate high- and low-hazard areas. This information allows departments to prioritize areas of concern for slope failure. At the federal level, a platform introduced in 2019 known as the Unstable Slope Management Program for Federal Land Management Agencies (USMP for FLMA) has gained much recognition (Beckstrand et al., 2019). USMP for FLMA is designed to guide efforts by federal land management agencies (FLMAs) and lower-traffic-volume transportation departments to assess slope hazards and risks along transportation corridors in order to achieve their own transportation maintenance goals and objectives (Anderson and Cuelho, 2017; Stanley and Anderson, 2017; and Beckstrand et al., 2019). USMP includes management tools that are important components of any GAM program, such as: condition assessments, examples of performance measures, and quantitative risk assessment (QRA) prioritization techniques (Beckstrand et al., 2019). The program was formulated by adopting and adapting methods from accepted transportation asset management practices used for bridges, pavement, etc., as well as existing GAM programs such as Oregon's rockfall hazard rating system (RHRS) and Alaska's USMP (Thompson, 2017). Alaska's USMP, which built upon progress made by programs like Oregon's RHRS, was completed in 2009 and provided a model for stakeholders (NPS, U.S. Forest Service [USFS], Bureau of Land Management [BLM], Bureau of Indian Affairs [BIA], and Western Federal Lands Highway Division [WFL]) to develop the USMP for FLMA (Beckstrand et al., 2019).

National parks, including Acadia, Crater Lake, Denali, GRSM, Hawaii Volcanoes, Olympic, Yellowstone, Yosemite, and Zion, have begun utilizing USMP for FMLA by performing slope condition hazard and risk assessments. Additional NPS units such as Vicksburg National Military Park, Delaware Water Gap National Recreation Area, and the Heritage Partnerships Program of the NPS Intermountain Region have also begun using the USMP for FMLA. Positive outcomes from proactively managing geotechnical assets are becoming clearer as growing numbers of organizations utilize the program. Recently, Bauer et al. (2021) and Banks et al. (2021) utilized the USMP to rate the unstable slopes along the Blue Ridge Parkway and suggested best practices based on the experiences gained from the extensive mapping. Researchers in Zion National Park concluded that reactive management can be four to five times more expensive for rockfall events than proactive management (FHWA, 2020). Additionally, Capps et al. (2017) concluded that QRAs are critical to understanding where funds should be allocated to avoid the common mistake of fixing the “worst first” reconstruction-only policy, which occurs when funding is spent without careful consideration of the exposure to associated risks. This conclusion was supported by the findings of Beckstrand et al. (2017), which estimated a value of $19.7 billion for the state of Alaska's geotechnical assets, i.e., more than three times greater than the value of their bridge inventory based on current reconstruction costs. The technical report estimated that managing these assets using a preservation model would reduce overall life-cycle costs by 5 percent (Beckstrand et al., 2017).

Objective of Study

With the goal of managing geotechnical assets along roadways, this research evaluated slope-failure risk along the primary GRSM transportation corridors to determine how to prioritize limited financial resources for risk-reduction maintenance or full-mitigation-level work. As such, the specific research objectives were to (1) create an inventory of unstable slopes and associated transportation-related hazards and risk ratings using USMP organized in a geospatial database, (2) delineate unstable slope hotspot areas that have high likelihood of roadway interruption using geospatial analysis, (3) perform site-specific investigations that predict roadways susceptible to unstable slope impact using probabilistic simulations, and (4) perform site-specific acid-base accounting (ABA) tests to evaluate the acid-producing potential (APP) of waste rock from slope failures.

The inventory of unstable slopes along with the hazards and risk rating digital geodatabase and maps will enable GRSM officials to take steps towards prioritizing maintenance and mitigation efforts using cost-benefit analyses based on short- and long-term budgets. The research provides an example of high-risk unstable slope prioritization using data-driven hotspot analysis, and application of USMP to provide a geo-logic and environmental framework for site-specific slope remediation to maintain the integrity of roadways in GRSM.

Study Area

Most of GRSM is in the Western Blue Ridge Physiographic Province, with a limited area in the Tennessee Valley and Ridge Physiographic Province to the northwest (Southworth et al., 2012). Bounded to the south by series of en echelon zones collectively called the Swannanoa Lineament, the park is home to some of the highest peaks in the eastern United States, some of which reach more than 2,025 m (6,644 ft) above adjacent valley floors. In some areas of the park, mountain slopes may be as great as 44° (Southworth et al., 2012; Hill, 2018). Much of GRSM is within the highlands of the Western Blue Ridge Physiographic Province, which is composed primarily of Neoproterozoic metasedimentary rocks of the Snowbird Group and Great Smoky Group (Southworth et al., 2012) (Figure 3 ). The northwestern portion of the park is within the foothills of the Western Blue Ridge Physiographic Province, which is characterized by rolling hills. The foothills are primarily low-grade greenschist facies or have not been metamorphosed and range from Neoproterozoic to Early Ordovician in age (Southworth et al., 2012). Quaternary deposits of alluvium and colluvium occur in low-lying areas of the park, along drainage features, or along the base of cliffs and slopes.
Figure 3.

Geology and major faults in Great Smoky Mountains National Park.

Figure 3.

Geology and major faults in Great Smoky Mountains National Park.

Along Route 0010, the primary rock formations encountered included metasandstone of the Thunderhead Formation and slate and metasiltstone of the Anakeesta Formation. Metasandstone of the Thunderhead Formation, Metcalf Phyllite, and Pigeon Siltstone and metasandstone of the Elkmont Formation were encountered along Routes 0013, 0014, and 0015. Route 0011 is mainly composed of Roaring Fork Formation metasandstone and the Pigeon Siltstone. Route 0008 (H, G, F, and E) crosses Hesse Quartzite, Wilhite Formation phyllite, and conglomerate, sandstone, and slate of the Shields Formation. Route 0019 traverses mostly through Wehutty Formation, consisting of slate graphitic and sulfidic schist. The Anakeesta, Wehutty, and part of the Copperhill formations are prime examples of acid-producing rock because they contain sulfide minerals such as pyrite and little or no carbonate minerals. GRSM is dominated by four major structural systems: (1) The Greenbrier and Dunn Creek faults in the highlands and foothills, (2) the Miller Cove and Great Smoky thrust faults in the foothills, (3) the Gatlinburg and Pigeon Forge faults in the foothills, and (4) the thrust sheets of the Tennessee Valley, which are bounded by the Pine Mountain Thrust Fault and the Great Smoky Fault (Thornberry-Ehrlich, 2008; Southworth et al., 2012). Most of the major faults are part of a connected fault system and can be a source of rockslides (Southworth et al., 2012) (Figure 3).

Annual rainfall throughout the park ranges from 1.14 m (45 in.) to 2.41 m (95 in.). Most of the primary roads are in the 1.50 m (59 in.) to 2.06 m (81 in.) range, and in higher sections of the park, over 2.16 m (85 in.) of precipitation falls annually (NPS-GRSM, 2017). More slope movements are expected to occur during early spring and late fall, when frost wedging conditions and large storm events create ideal slide conditions (Matsuoka, 2001; Sass, 2005; and Nandi and Shakoor, 2017). Over 3,379 km (2,100 mi) of streams and rivers are contained within GRSM, of which 1,175 km (730 mi) are fish-bearing and 2,092 km (1,300 mi) are tributaries (NPS-GRSM, 2017). Tributaries, springs, and precipitation replenish waterfalls and surface streams (McKenna, 2007). GRSM streams are vulnerable to acid rain because of nearby power plants, factories, and volume of traffic (McKenna, 2007). Water in GRSM can be acidic from pollutants in rain, and from rock formations that have acid-producing potential (e.g., Anakeesta Formation, Copperhill Formation, Wehutty Formation). Schaeffer and Clawson (1996) conducted geologic mapping, petrographic analysis, and ABA tests as part of a road and transmission line construction project in southwestern NC, where the acid-producing rocks of interest included Anakeesta Formation graphitic schist and thin layers of sulfidic rock within the Ammons Formation, both of which are present in GRSM. The construction project required the use of an encapsulating embankment design similar to several large highway projects in the Blue Ridge Physiographic Province in TN and NC to prevent acid drainage (Byerly, 1996; Schaeffer and Clawson, 1996). Their study exemplifies the special handling required for acid-producing material to minimize acid rock drainage (ARD) and avoid costly mitigation of adverse environmental impacts (Byerly, 1996). The potential negative impacts on physical infrastructure and surface waters illustrate how evaluation of the acid-producing potential at rockfall prone cut slopes can help to inform waste rock management strategies and why ARD represents an important consideration for the GAM strategy in GRSM. A study by Mathews and Morgan (1982) showed the adverse effect of ARD on aquatic life: The salamander (Leurognathus marmoratus) population was almost destroyed downstream from the highway cut-and-fill areas due to the presence of sulfide minerals in the Anakeesta Formation. These rock types are more prone to rockfalls and landslides, and they also have the potential to negatively impact flora and fauna via acidification of waters (Schaeffer and Clawson, 1996; Latham et al., 2009).

This study utilized the USMP for FLMA protocols to develop a digital database of unstable slopes and their current conditions along 243.67 km (151.41 mi) of road in GRSM. Site investigation field data were added to a geodatabase in ArcGIS Pro 2.7 and analyzed to better understand the spatial distribution of unstable slopes. Kernel density estimation (KDE) was used to identify clusters of unstable slopes with high likelihood of roadway disruption and establish study areas for site selection. Two-dimensional probabilistic slope stability simulations and ABA tests were used to predict unstable slope pathways and evaluate the acid-producing potential of rock fragments. The study methods are displayed in a flowchart (Figure 4 ) and described in the following sections.
Figure 4.

Methodology flowchart, where the gray color code corresponds to the study objectives.

Figure 4.

Methodology flowchart, where the gray color code corresponds to the study objectives.

Data Collection and Preparation of the Geodatabase

Primary data were collected using the USMP for FLMA standardized field form that organizes hazard and risk data into discrete attributes and quantifies the observations (Capps et al., 2017; Beckstrand et al., 2019). The protocols can be used to assess several types of unstable slopes, such as soil and rock landslides, rockfalls, debris flows, and thaw-unstable slopes (Capps et al., 2017). Site assessments ranged from July 2019 to July 2020. A field rating was conducted for each unstable slope using the USMP for FLMA rating form, which included parameters listed in Table 2 . Photographs of each slope and global positioning system (GPS) coordinates were also collected, and site data were uploaded to the USMP.info web portal. Preliminary and total USMP ratings were calculated based on the hazard and risk parameters as indicated by FHWA (2020) observed in the field or reported by park officials.

Table 2.

Parameters used to calculate USMP ratings.

Secondary data were acquired as spatial data layers from state and federal data download websites. The NPS Integrated Resource Management Applications (IRMA) web portal (IRMA.NPS.gov) was used for road centerlines, the park boundary shapefile, and the 2016 geologic map of GRSM. Sub-meter-resolution light detection and ranging (LiDAR) digital elevation models (DEMs) were downloaded from the Tennessee GIS Clearinghouse (TNGIS.org/LiDAR) and North Carolina's Spatial Data Download website (SDD.NC.gov). Primary and secondary data were compiled and organized to create a geodatabase of unstable slopes along primary transportation corridors in GRSM.

Establish Priority Areas: KDE

The KDE method is an interpolation routine used to identify hotspots or high-risk areas based on a set of point or line data. For this study, the Kernel Density tool from ArcGIS Pro 2.7 was used to identify clusters of poorly rated unstable slopes. Line data were used that represent the length of the affected roadway associated with known unstable slopes. Each line was associated with a symmetrical surface centered on the line called a kernel. A quartic kernel with a fixed-interval bandwidth (search area) was used in this study (Silverman, 1986; ESRI, 2021). The following formula was used to calculate the density value at each output raster cell or (x, y) location (ESRI, 2021).
formula
(1)
This equation was used for disti < radius, where i = 1, …, n are the input line segments within the radius distance of a (x, y) location; the population field popi is the total USMP score; and disti is the distance between line segment i and the (x, y) location. The default search radius was used in the study and was determined using an algorithm that (1) calculated the weighted mean center of input unstable slopes; (2) calculated the distance from the weighted mean center for all sites; (3) determined the weighted median of these distances, Dm; and (4) calculated the weighted standard distance, SD. Once these values were established, they were applied to the following formula:
formula
(2)

where n is the sum of the population field values and either SD or graphic , whichever value is smaller. The output KDE raster was used to establish priority study areas for site-specific analysis.

After selecting priority sites based on the results from the KDE, rockfall simulations and ABA tests were conducted at each slope coinciding with hotspots to develop a geologic and environmental prioritization framework for slope remediation (Figure 4). Rockfall simulation was conducted because field investigation revealed that common unstable slopes along the roadways were mostly categorized as rockfalls. Field assessments were conducted at each rock slope to record bedrock lithology, block dimension, slope material properties, seeder or starting location, and potential rockfall pathway data. The topographic profiles were extracted from the 1 m DEM and revised in the field using a laser range finder. Rockfall simulations were completed using RocFall software utilizing the rigid body analysis method with tangential Colorado Rockfall Simulation Program (CRSP) damping (RocScience, 2002). The slope material properties of the topography are listed in Table 3 . One thousand (1,000) rocks, distributed evenly between seeders, i.e., starting locations, were thrown for each simulation, and a recommended default initial horizontal velocity of 1.5 m/s (4.9 ft/s) was used for every seeder, while initial vertical and rotational velocities were set to 0 m/s (0 ft/s), as recommended by the RocFall User Guide (RocScience, 2002). Point seeders were added to each slope based on field observations, and line seeders were added along slopes where point sources were not obvious, for example, where rock debris and fragments were observed along the length of a slope and within the ditch. Each seeder required block shape, dimensions (0.3 to 1.2 m [1 to 4 ft] in the elongated direction), and a density value that was specified using the rock type library. An appropriate rock type for the site was selected to determine density, while block shape(s) and dimensions were taken from field notes. The block size ranged from 0.3 to 1.2 m (1 ft to 4 ft) in the long direction, and the block shapes were various shapes of polygons selected from the library, as closely represented in the field. Additional input data such as roadway width, ditch properties, and presence of mitigation measures were collected during the field visits. Validation of model results was performed by comparing the rock pathways and end points to photographs taken during field visits and notes recorded in the field. Photographs and field notes provided an account of rock block locations along the slope, contained within the ditch, and occasionally within the roadway. For several sites, traces of scars associated with impacts of blocks on the roadways were also observed and recorded.

Table 3.

Earth materials parameter properties used for simulated rockfall pathways for the 14 investigated sites.

Rock samples were collected during field assessments and were sent to a commercial laboratory for ABA tests. Rock samples were collected as loose material along the toe of slopes, in compliance with the scientific research and collecting permit granted by the NPS to minimize impact to park resources (Figure 5 ). Three samples were collected at roughly equal distance along the base of each slope and placed in labeled plastic bags for storage. A composite sample was prepared for each site using approximately 333 g of material from each sample point for a total weight of 1 kg. An ABA test using the modified Sobek method described by Sobek et al. (1978) and Lawrence and Marchant (1991) was used in this study. ABA test results are reported in units of kg CaCO3 per tonne of material. Samples with net neutralization potential (NNP) values <−5 kg CaCO3/t are considered to have a significant acid-producing potential. In practical terms, an NNP value of −5 means that 5 kg of CaCO3 are required to neutralize 1 t (1 metric ton) of sample material.
Figure 5.

Composite samples were prepared for each unstable slope. This representative outcrop (GRSM-155) (35.4574186°N, 83.4956021°W) was in Wehutty Formation composed of dark metagraywacke and metasiltstone, with black graphite schist; outcrop is covered with Fe oxides, secondary sulfur minerals, and gypsum. The hand sample shown in the figure is a graphite schist. Net neutralization potential (NNP) for the composite sample was −26.4 (kg CaCO3/t).

Figure 5.

Composite samples were prepared for each unstable slope. This representative outcrop (GRSM-155) (35.4574186°N, 83.4956021°W) was in Wehutty Formation composed of dark metagraywacke and metasiltstone, with black graphite schist; outcrop is covered with Fe oxides, secondary sulfur minerals, and gypsum. The hand sample shown in the figure is a graphite schist. Net neutralization potential (NNP) for the composite sample was −26.4 (kg CaCO3/t).

USMP Inventory

In total, 285 discrete unstable slopes assessed along 243.67 km (151.41 mi) of roadway in GRSM were added to the USMP database. Of these, 280 slopes were designated as localized rockfall, dominated by wedge and planar failure mechanisms. The five (5) remaining sites were designated as small-scale landslides in soil-fill embankments along stream banks. The USMP for FLMA classification system defines slope conditions as “good” when the total USMP score is <200, “fair” when it is ≥200 and ≤399, and “poor” when it is ≥400. This classification system is based on experience and was designed for federal land management agencies with low to very low traffic volumes (Beckstrand et al., 2019). In the assessment, 133 slopes ranked as “poor” (45 percent), 147 ranked as “fair” (53 percent), and five ranked as “good” (<2 percent) based on the USMP for FLMA classification system. Figure 6 shows the distribution of 285 slopes classified by quartile range to better compare local sites. Because 280 out of 285 slopes were rockfalls, the five landslide sites on soil slopes were discarded from further site-specific analysis.
Figure 6.

Inventory map of unstable slopes classified by USMP total score quartile range.

Figure 6.

Inventory map of unstable slopes classified by USMP total score quartile range.

The majority (72 percent) of unstable slopes were identified along three main roads in the park: Routes 0014 (Little River Gorge Road), 0010 N (Newfound Gap Road), and 0011 N, S (Gatlinburg Spur Road). Of these, 32 percent were located along Route 0014 in the Metcalf Phyllite, Cades Sandstone, and Thunderhead Sandstone geologic units, including four of the 10 highest-rated slopes; 18 percent were located along Route 0010 N, which crosses the NC-TN state border in the Anakeesta Formation, Copperhill Formation, and Thunderhead Formation; and 22 percent were identified along Route 0011 N and S, primarily in the Pigeon Siltstone and to a lesser extent in the Rich Butt Sandstone.

The remaining 28 percent of unstable slopes were distributed along the other primary transportation corridors. Notably, 12 percent of slopes were identified along Route 0008 E, F, G, and H from Chilhowee at the southwest to Wears Valley at the north near Sevierville, TN. Additionally, 5 percent of unstable slopes were assessed along Route 0019 near Bryson City, NC, within the Wehutty and Copperhill formations. Most of the primary roads and unstable slopes in GRSM were located on the TN side of the park in the foothills of the Western Blue Ridge Physiographic Province.

Kernel Density Estimation

The output density surface created using KDE had a spatial resolution of 10 m and was presented using equal interval classification. Dark purple patched areas in Figure 7 have the greatest density of poorly rated unstable slopes, as labeled in Figure 6 and subset in Figure 7. The geologic formations at the greatest density of poorly rated areas included the Anakeesta Formation, Thunderhead Sandstone, Cades Sandstone, Metcalf Phyllite, Wehutty Formation, Shields Formation, and Pigeon Siltstone. Six noticeable clusters of unstable slopes with a high likelihood of roadway disruption were identified along the Gatlinburg Spur (0011), Newfound Gap Road (0010) near the TN-NC border, Little River Gorge Road (0014), and Laurel Creek Road (0015) (Figure 7). Lakeview Drive East Road (0019) showed a medium- to low-density cluster. Foothills Parkway West (0008) did not show any leading clusters; however, the route was included as an additional area of interest for further site-specific studies based on its documented history of sporadic rockfall and environmental hazards, such as acid rock drainage. Using the KDE output, 14 sites were selected within the clusters and along Route 0008 for site-specific analysis, including probabilistic rockfall simulation and ABA (Figure 7).
Figure 7.

High-density clusters of poorly rated slopes were identified using KDE. Fourteen sites were selected within the clusters for site investigation. The subset map includes the locations of unstable slopes color-coded by risk rating.

Figure 7.

High-density clusters of poorly rated slopes were identified using KDE. Fourteen sites were selected within the clusters for site investigation. The subset map includes the locations of unstable slopes color-coded by risk rating.

Probabilistic Rockfall Simulations

The RocFall output includes end-point analysis, kinetic energy (total, translational, and rotational), velocity (translational and rotational), and bounce height. End-point analysis is a significant factor concerning safety on the roadway. Therefore, this study primarily focused on the distribution of rockfall end points as the percentage of rocks running out of the ditch and passing the edge of the roadway closest to the slope, passing the centerline, and exiting the roadway away from the slope. Validation of rockfall simulations was performed by comparing model results to Google Maps street view, site photographs, and field notes.

Results from the simulations showed rock material entering the roadway at all 14 sites (Figure 8 ). The distributions of end-point locations for each unstable slope are presented in Table 4 . Across all sites, most rocks (63.4 percent) were contained by ditches and did not enter the roadway. End points for rocks that did enter the roadway were generally confined to one lane of traffic closest to the slope. Only 3.4 percent of rocks reached the centerline, and only 0.2 percent of rocks crossed both lanes of traffic. The predicted percentage of rocks contained within ditches varied widely among slope models. For example, GRSM-168 on Foothills Parkway Section 8E had the most effective containment of material, with only one rock out of 1,000 (0.1 percent) entering the roadway. In contrast, GRSM-088 on Little River Gorge Road had the least effective containment, with 99.5 percent of rock-path end points within roadway, 3.3 percent of which reached the centerline. An inverse relationship between ditch width and the percentage of rocks entering the roadway (Table 4) was noted; however, the statistical relationship was not analyzed due to the small sample size.
Figure 8.

Simulated rockfall pathways for the 14 investigated sites.

Figure 8.

Simulated rockfall pathways for the 14 investigated sites.

Table 4.

Distribution of end-point locations for each unstable slope.

Environmental Impact (ABA)

Total sulfur concentration was reported as weight percent and ranged from below the detection limit (0.02 weight percent) to 1.5 weight percent. A full account of ABA test results is included in Table 5 . Samples from five sites contained significant concentrations of total sulfur (>0.5 weight percent). These values directly correlated with the sulfide concentration and therefore the acid-generation potential of the samples. Test results indicated a wide range of NNP values, from −31.1 to +69.2 kg CaCO3/t (Figure 9 ). Notably, samples from GRSM-013 and GRSM-168 had significant sulfide concentrations and acid-generation potentials that did not result in NNP values <−5 kg CaCO3/t due to relatively high neutralization potentials. Four rock samples collected from three discrete slopes had NNP values <−5 kg CaCO3/t. The most negative values, −31.1 and −27.6 kg CaCO3/t, were from the Anakeesta Formation and were duplicate samples collected at GRSM-010 along Newfound Gap Road North. Two samples collected from two discrete slopes in the Wehutty Formation along Lakeview Drive East also indicated significant acid-producing potential with NNP values of −26.4 and −20.7 kg CaCO3/t.
Table 5.

Complete ABA test results for the 14 investigated discrete slopes.

Figure 9.

Acid-base accounting test data for GRSM.

Figure 9.

Acid-base accounting test data for GRSM.

The geodatabase and inventory maps created in this study represent an important step towards implementing long-term GAM protocols in GRSM. The cluster map created using KDE highlights sections of road where slopes with a high likelihood of roadway disruption are most concentrated and can be used to communicate risk to park visitors and commuters. Further site-specific kinematic investigation of the structural analysis of bedrock discontinuities along with the rock friction and cohesion within these clusters will provide insights into whether some geo-logic or geometric condition influences slope stability. Once study areas were established based on results from KDE and input from park officials, site-specific rockfall simulations and ABA tests were conducted at 14 selected as high-risk sites. These investigations provided a better understanding of the potential impacts of rockfalls on roadway infrastructure and the environment.

USMP Inventory and KDEs

Most unstable slopes identified in this study are located on the north side of the park in TN (88 percent) with only one KDE cluster identified in NC. Many of these slopes are within the foothills of the Western Blue Ridge Physiographic Province. This province is bounded to the north by the Great Smoky Fault and to the south by the Gatlinburg Fault and is characterized by rolling hills with predominately sedimentary bedrock (Neoproterozoic, Cambrian, Lower Ordovician), which is either low-grade greenschist facies or has not been metamorphosed. About a quarter of all slopes were in the higher-grade metamorphic rocks of the highlands of the Blue Ridge, and less than 7 percent of sites were in the Tennessee Valley and Ridge Physiographic Province. Geologic units with the greatest number of unstable slopes along major transportation corridors are Neoproterozoic in age and include the Pigeon Siltstone (n = 45) and Metcalf Phyllite (n = 45) of the Snowbird Group and the Thunderhead Sandstone (n = 30), Cades Sandstone (n = 27), Anakeesta Formation (n = 28), and the Copperhill Formation (n = 25) of the Great Smoky Group. The remaining 85 slopes were distributed among 11 other rock formations. The Great Smoky Mountains National Park Geologic Resource Evaluation Report by Thornberry-Ehrlich (2008) and previous slopes stability studies at GRSM hinted upon the same susceptible rock units (Moore, 2004; Wieczorek et al., 2000; and Nandi and Shakoor, 2017).

The cluster analysis created using the USMP total score helped to highlight areas where unstable slopes pose significant risk to park visitors and commuters along GRSM primary routes. The cluster map was also helpful in establishing priority areas within the park where site-specific studies were concentrated. More clusters and more unstable slopes in general occur on the north side of the park because this is where the majority of roadways within the study area exist. This represents a limitation of the study because the presence of clusters is controlled by the roadways and data collection sites. However, it may also be true that rock units within the foothills of the Western Blue Ridge Physiographic Province are more susceptible to rockfalls and rockslides where road cuts exist than rock units within the highlands. Future studies could evaluate whether a relationship exists between the metamorphic grade of geologic units and instability.

Ultimately, the aim of this study was to assess unstable slopes along major transportation corridors in GRSM, so data collection was constrained to accomplish that goal. The research provided examples of site-specific investigations like probabilistic rockfall simulation and ABA for selected sites that could be prioritized from cluster analysis using the USMP inventory database. This type of application could be adapted by a state department of transportation, FLMA, or future researcher to suit their specific needs. In addition to ongoing condition assessments and performance monitoring, future effort should be directed to develop forecasting models, such as topographical change detection using GPS combined with real-time kinematic (RTK) capabilities, unmanned aircraft system (UAS) structure from motion (SfM) analysis to generate three-dimensional slope models that can detect the temporal change of a surface, and terrestrial laser scanner (TLS) and aerial laser scanner (ALS) data from UAS to detect slope change and displacement. These forecasting models can provide estimates of future changes in the performance of discrete slopes, which can help GRSM park officials to anticipate changes to management costs and evaluate program alternatives.

Probabilistic Rockfall Simulations

Accurately predicting rockfalls is difficult due to variability in slope geometry, uncertain material properties, and the sensitivity of analysis methods (Stevens, 1998). However, results from probabilistic simulations provide an effective and acceptable method for evaluating the potential impact of rockfall on transportation corridors. Results from this study showed rock material entering the roadway at all 14 sites, which confirms the premise that GRSM's major transportation corridors are vulnerable to localized slope failures. Model results also indicated that some sections of roadway are more vulnerable than others, mainly where ditch effectiveness is limited. These predictions were validated using a combination of Google Street View, field notes and photographs, and comments in the USMP for FLMA geodatabase. GRSM-136 on Foothills Parkway Section 8E stands out as somewhat unique from the other sites due to its long and consistent slope, wide ditch, and vegetation near the slope's toe. Also, a feature of interest is that vegetation has a significant damping effect on simulated rockfalls; however, this relationship is complicated by the fact that vegetation can contribute to biological weathering, especially in fractured rocks. Sites like GRSM-087 and GRSM-105 along Little River Gorge Road stand out because they feature blocks that slightly overhang the roadway.

Environmental Impact (ABA)

Insights from ABA tests can be used by GRSM park officials to help develop solid waste management protocols at cut slopes. Additional costs associated with encapsulating or transporting acid-producing rock debris are important to consider for budget allocation, which is an essential part of the GAM process. As part of ongoing condition assessment and performance monitoring, park officials should take note of the acid-producing potential of rock units. Weathering of Precambrian metasedimentary rocks in the Southern Appalachian Mountains is well recognized, and the Federal Highway Administration developed guidelines on evaluation and handling of acid-producing materials (Byerly, 1996). ABA test results indicated significant acid-producing potential at three discrete rock slopes of the 14 sites sampled. The study confirmed that the sulfide minerals contribute to the acid-generating potential, and the Anakeesta Formation and the Wehutty Formation present the greatest hazard regarding ARD. At these sites, it is reasonable to take special precautions when handling rockfall materials. Field investigation revealed that slaty metasiltstone members of the Copperhill Formation may also require special handling due to ARD; however, no samples were analyzed in this study. Significant ARD seems to be limited to a short length of roadway, about 21.2 km (13.2 mi) out of 243.7 km (151.41 mi), almost exclusively between mile markers 10 and 20 of Newfound Gap Road (GRSM-0010N, S), the first 1.6 km (1 mi) of Clingman's Dome Access Road (GRSM-0017), and the first 8 km (5 mi) of Lakeview Drive East (GRSM-0019), where units of the Anakeesta Formation, slaty metasiltstone member of the Copperhill Formation, or Wehutty Formation are exposed.

Schaeffer and Clawson (1996) concluded that the Anakeesta Formation is a potential acid-producing graphite schist unit, with NNP for the graphite schist units ranging from −19.27 to 1.81 CaCO3/t. Hammarstrom et al. (2003) conducted a thorough investigation of metal cycling in GRSM and identified soils at the Hazel Creek Mine with an NNP value of −61 kg CaCO3/t. That study presented important considerations for sulfide minerals at historic mine sites within the park; however, the study did not discuss how sulfide minerals and ARD could impact transportation infrastructure or how solid waste management practices should be incorporated into GAM protocols. Latham et al. (2009) found an association of sulfide minerals with unstable slopes in metagraywackes and graphitic muscovite schists along the Blue Ridge Parkway. Further, sulfide-induced heave was not observed during field observations; however, Bryant (2003) documented the same in the Sevier Shale near the study area and presented chemical tests procedures and various ARD mitigation options.

Implementing long-term, risk-based strategic GAM is imperative for public lands, like GRSM, where maintenance officials are responsible for achieving performance objectives with a fluctuating annual budget. The goal of the work described here was to provide data to guide GAM efforts by prioritizing sites and informing the selection of site-specific interventions. The study succeeded in creating the first exhaustive inventory of unstable slopes along major transportation corridors in GRSM and provides an example of high-risk rock slope prioritization using cluster analysis. Additionally, 14 site-specific investigations were completed that predicted rockfall pathways using probabilistic simulations, and acid-base accounting tests were performed to evaluate the acid-producing potential of unstable rocks. The study provides a geologic and environmental framework for slope remediation to maintain the integrity of roadways in GRSM. The study will assist park officials in their efforts and foster a better understanding of life cycles of discrete unstable slopes.

This study utilized the USMP for FLMA protocol to (1) create a detailed inventory of 285 unstable slopes, of which five slopes were ranked as being in good condition, 147 slopes were ranked as fair, and 133 slopes were ranked as poor according to the USMP for FLMA classification system. (2) Five noticeable clusters of unstable slopes with high likelihood of roadway disruption were identified along three major transportation corridors using KDE. As state departments of transportation and FLMAs across the country adopt and implement GAM programs, cluster analysis can be used to target remediation and mitigation efforts. This is significant because, once an inventory has been created, the decision of where to target proactive management or mitigation can be daunting.

The site-specific analysis of the 14 high-risk slopes indicated that (3) rock fragments entered the roadway at all 14 sites, (4) sections of roadway where ditch effectiveness is limited are more vulnerable to rockfall, such as along Little River Gorge Road (0014), and (5) significant APP is limited to a short length of roadway overall, because only about 21.2 km (13 mi) of roadway exist where Anakeesta Formation, the slaty metasiltstone member of the Copperhill Formation, or Wehutty Formation are exposed. Probabilistic rockfall simulations can provide valuable information for park officials who are responsible for GAM protocols. Because rockfall events interfere with transportation corridors during most years, which can have a negative impact on the local economy, rockfall modeling has a role in future management and mitigation efforts. To a lesser extent, the same is true for analysis of acid-producing waste rocks at GRSM. Future studies can evaluate the correlation between acid-producing rocks and slope instability in the park.

Finally, results from this study affirm that GRSM's major transportation corridors are vulnerable to localized slope failures. Insights from the study can be used by GRSM park officials to help develop short- and long-range management and mitigation plans, such as widening ditches, installing barriers, and encapsulating acidic rockfall material. These strategies can inform park officials’ efforts to monitor the performance of geotechnical assets and make periodic updates to the GAM in GRSM. In addition to ongoing condition assessments and performance monitoring, future effort should be directed to develop forecasting models that estimate future changes in performance of discrete slopes. These forecasting models can facilitate efforts by GRSM park officials to anticipate changes to management costs and evaluate program alternatives.

The study was financially supported by The United States Department of the Interior – The National Park Service/Great Smoky Mountain National Park. The Appalachian Highlands Science Learning Center of Great Smoky Mountains National Park provided the research permit at the park. The reviewers and the editor of the Environmental & Engineering Geoscience (E&EG) journal provided detailed and helpful comments, for which we are grateful.