Frozen debris lobes (FDLs) are slow-moving landslides in permafrost. FDL-A, the largest monitored FDL in the Brooks Range of Alaska, has steadily progressed downslope toward the Dalton Highway, which is the only road to the oil and gas fields of the North Slope. To avoid this encroaching landslide, the Dalton Highway was realigned farther downslope in 2018. The abandoned portion of the highway was left in place, providing a unique opportunity for a full-scale field experiment to monitor the impact of a landslide on an engineered structure. In 2020, we conducted a subsurface investigation, drilling and sampling the subsurface soils and installing geotechnical instrumentation within the abandoned highway embankment. Here, we present the integration of multiple datasets to provide a detailed description of the landslide–embankment collision. FDL-A is shearing within weathered bedrock ∼7.6 m below the embankment surface. It horizontally displaced a portion of the embankment ∼0.3 m as of November 1, 2023. Using the infinite slope approach and strength properties determined from laboratory testing, we estimate that FDL-A is impacting the highway embankment and underlying soils with at least 77.8 kN/m width shear force. As this force occurs approximately 4 m below the bottom of the embankment, we postulate that the presence of the abandoned Dalton Highway embankment does little to stop FDL-A’s downslope progression.

We dedicate this paper to one of our co-authors—our long-time colleague in FDL research and dear friend—Dr. Ronald (“Ronnie”) Daanen, who died in a helicopter crash while conducting field work on the North Slope of Alaska on July 20, 2023.

Landslides are destructive events, causing an estimated $3.5 billion worth of damage annually in the United States (USGS, 2017). Linear infrastructure is especially susceptible to damage because, by its nature, it must sometimes transverse difficult and landslide-susceptible terrain. Alaska’s Dalton Highway, “one of the northernmost roads in the world” (ADOT&PF, n.d.), is an excellent example of linear infrastructure crossing difficult terrain. The Dalton Highway is 666 km (414 mi) long, stretching from Livengood in Interior Alaska to Deadhorse on the Arctic Coastal Plain (ADOT&PF, n.d.; Figure 1a). Originally built in the 1970s to support the construction of the Trans Alaska Pipeline System, currently the Dalton Highway facilitates the transportation of goods and services from cities like Anchorage and Fairbanks to the oil and gas fields of the North Slope. Underlain by continuous and discontinuous permafrost and crossing the Brooks Range in northern Alaska, the Dalton Highway has experienced its share of geohazards, including thaw settlement, avalanches, slush flows, and landslides. In this paper, we focus on a specific geohazard impacting the Dalton Highway in the south-central Brooks Range, namely, frozen debris lobes (FDLs; Daanen et al., 2012; Darrow et al., 2016).

FDLs are slow-moving (Cruden and Varnes, 1996) landslides in permafrost, typically composed of silty sand with gravel based on the Unified Soil Classification System (USCS) following ASTM D2487-17e1 (ASTM, 2017a). In addition to the soil component, FDLs contain some quartz-rich cobbles and boulders that likely are remnants of resistant veins in the heavily fractured metasedimentary rocks that form the catchments from which these features originate (Darrow et al., 2016). Although the soil matrix is ice-poor, FDLs contain areas of massive infiltration ice, which forms in cracks that are open at the surface due to downslope movement throughout the year (Darrow et al., 2016). FDLs also contain organic material and woody debris. Because FDLs predominantly consist of soil, mature spruce forests grow on their surfaces. As FDLs move downslope, their movement knocks over these trees and those on the slope in front of them, and they entrain that woody debris, as well as the original organic mat into their moving mass.

Our FDL research began in 2008 with the initial investigation of four FDLs (Daanen et al., 2012). We now regularly monitor nine of these landslides (Figure 1b). Our previous investigations included mapping and testing of rocks in the surrounding catchments, analysis of historic and current movement rates, change detection analysis using lidar and interferometric synthetic aperture radar, and thermal modeling (Darrow et al., 2016, 2017, 2019; Gyswyt et al., 2017; Byrd et al., 2019; and Gong et al., 2019).

The largest and closest FDL to the Dalton Highway is termed FDL-A (Figure 1c). Our initial studies of FDL-A included subsurface investigations and analysis of geotechnical instrumentation data (Darrow et al., 2013; Simpson et al., 2016). We informed the Alaska Department of Transportation and Public Facilities (ADOT&PF) about FDL-A’s hazardous increasing rate of movement, which prompted ADOT&PF to realign the Dalton Highway, moving it ∼125 m downslope in 2018. A portion of the abandoned highway alignment was left in place in front of FDL-A, providing a unique opportunity to observe the impact of a landslide with a compacted embankment. In 2020, we—along with personnel from the ADOT&PF Northern Region Material Section (NRMS)—conducted a subsurface investigation, drilling and sampling the subsurface soils and installing geotechnical instrumentation within the abandoned part of the highway embankment. The objectives of this paper are to 1) summarize results of this full-scale field experiment to monitor the impact of a landslide with an instrumented embankment, 2) evaluate the movement mechanisms at FDL-A’s toe, and 3) estimate the force FDL-A is imparting to the embankment and subsurface soils. We integrate data from above-ground and subsurface instrumentation (including local climate, ground temperatures, groundwater pressure, and slope movement), laboratory testing, lidar, and repeat photography to build a conceptual model of the interaction between FDL-A and the abandoned Dalton Highway embankment. This paper includes data collected from a previous 2018 subsurface investigation and from the 2020 subsurface investigation until November 1, 2023.

Field Investigations

FDL Surface Movement Measurements

In 2012, we installed an array of surface marker pins on FDL-A (Figure 1c), whose changing positions we have documented since that time using a real-time kinematic global positioning system (RTK-GPS). For this project, we continued those measurements, using a Leica Viva system (including GS14 receivers and a CS15 controller) as the RTK-GPS device. With post-processing, this system is accurate to within 5 cm in both horizontal and vertical directions.

2018 Subsurface Investigation

Our joint University of Alaska Fairbanks (UAF)–Alaska Division of Geological and Geophysical Surveys (DGGS) research team completed two boreholes in 2018, each to 3.05 m (10 ft), using a handheld power head and auger assembly (Figures 1d and 2a). The purpose of this investigation was to measure changes in the subsurface temperature and porewater pressure as FDL-A approached and covered the instrument locations. Thus, we located each boring close to both the toe of FDL-A and a stream draining from it. Into each borehole, we installed a temperature sensor cable and a vibrating wire piezometer (VWP)—which we previously tested and calibrated in the laboratory—and backfilled each boring with cement–bentonite grout. Finally, we hand-dug a ∼15-cm-deep trench into which we buried the instrumentation cables, routing them to an automated data acquisition system (ADAS). The ADAS consisted of a data logger, multiplexer, solar regulator, solar panel, enclosure, battery bank with enclosure, and temperature sensors to measure air and surface temperatures at the ADAS location.

2020 Subsurface Investigation

Our UAF-DGGS research team, along with personnel from ADOT&PF NRMS, performed the subsurface investigation and geomechanical equipment installation in the abandoned section of the Dalton Highway in mid-August 2020. Prior to going to the field, we calibrated the temperature sensor cables and VWPs and tested the components wired into the ADAS. Once in the field, using a small track excavator, ADOT&PF NRMS personnel dug a shallow trench (∼0.4 m deep) into the embankment surface for burial of instrumentation cabling. All boreholes were advanced using a truck-mounted CME 55 drill rig, which was centered over the trench (Figure 2b). We completed four 12.19-m-deep boreholes (40 ft deep as measured from the embankment surface) using a 16.5-cm (6.5 in.) outside diameter (OD) hollow-stem auger, which allowed for split-spoon sampling during drilling and served as a casing to keep each borehole open during instrument installation (locations are provided in Figure 1d). We collected samples using a 1.51-kN (340 lb) hammer to pound a 7.6-cm (3 in.) OD split-spoon sampler at 0.76-m (2.5 ft) intervals, providing near-continuous visualization of the subsurface materials. We collected representative samples from each borehole, which were sealed and transported to the Frozen Soil Testing (FROST) Laboratory at UAF. In each boring, we installed one micro-electro-mechanical system–based in-place inclinometer (M-IPI), a temperature sensor cable, and a VWP and then backfilled with cement–bentonite grout. All cabling was routed to an ADAS through the trench and buried. Simultaneous with the drilling, the UAF-DGGS research team laid out tilt meters and pressure plates for installation on the east or uphill side of the embankment. For each borehole location, a 2-MPa pressure plate and tilt meter (both affixed to a single steel plate) were installed vertically into a slot dug into the edge of the embankment. The exception was at TH20-02, where an additional 5-MPa pressure plate on its own steel plate was installed immediately adjacent to the 2-MPa pressure plate and tilt meter (Figure 2c). It was during this effort that we discovered that the tilt meter cables were provided to the specified length in feet instead of meters and thus were too short to be installed where desired. To accommodate the short cables, we moved the ADAS associated with the 2018 subsurface investigation into the middle of the embankment and dug an additional trench to route the remaining 2018 instrumentation cables, plus the tilt meter and pressure plate cabling, to this central ADAS. Later, in summer 2022, we exposed the buried cables, extended them through splicing, routed the extended cables to the south, and moved this ADAS out of the way of FDL-A’s imminent impact with the embankment (Figure 2d).

In 2020, we installed two ADASs at the south end of the embankment section, one for the time-lapse camera system and one to serve as the main ADAS for the majority of the instrumentation. The time-lapse camera system (Nupoint Systems Inc.) was set up to take one photograph per day of the toe of FDL-A, which was transmitted to UAF via a satellite uplink. We mounted two additional stand-alone field cameras (Reconyx) to acquire images of the landslide collision from different angles. The main ADAS consisted of a data logger, solar regulator, solar panel, battery bank, and the necessary instrumentation to serve as a weather station for the project area, including measurements of air temperature, precipitation, and snow depth. ADAS components were mounted inside an enclosure or directly onto a rugged tripod that was anchored into the embankment surface. Table 1 is a summary of the geomechanical instrumentation installed during both drilling programs.

Laboratory Testing

We transported all samples to the UAF FROST Laboratory for standard tests, including moisture content (ASTM D2216-19, ASTM, 2019), Atterberg limits (ASTM D4318-17e1, ASTM, 2017b), unit weight (ASTM D7263-21, ASTM, 2021a), and particle-size distribution consisting of hydrometer tests (ASTM D7928-21, ASTM, 2021b), specific gravity tests (ASTM D854-14, ASTM, 2014), and sieve analyses (ASTM D6913M-17, ASTM, 2017d). We classified samples using the USCS (ASTM D2487-17e1, ASTM, 2017a). We were able to collect three split-spoon samples of the bedrock material in TH20-04. The high moisture content and weathered condition of the bedrock samples allowed us to prepare test specimens for direct shear tests (ASTM D6528-17, ASTM, 2017c).

Lidar Data Acquisition and Change Detection Analysis

We conducted lidar scanning of the FDL-A toe area in May 2021 and June 2022, using a Riegle VUX-LR integrated into a fully dynamic lidar system by Phoenix LiDAR. We used a backpack to transport the lidar system, which allowed high-resolution scanning of the small toe area (Figure 2e). The system is setup with RTK-GPS for accurate location information at 10 points per second. Between RTK-GPS readings, the system relies on a Northrop Grumman inertial momentum unit at a rate of 100 times per second. The combination of these two devices makes it possible to know the location and orientation of the device with an accuracy of about 1 cm. We produced digital elevation models (DEMs) at a resolution of 10 cm from each lidar scan. We used these datasets, as well as a previously collected 2020 dataset, to conduct the change detection analysis. We also used the DEMs to produce longitudinal profiles across the Dalton Highway embankment and up the toe of FDL-A, which were used to produce scaled conceptual models of movement.

Local Climate

Figure 3a contains air temperature data from the toe of FDL-A and from Coldfoot, Alaska (approximately 70.8 km south of the research site; Figure 1a). For the 4 years (2019–2022) of complete temperature data from the FDL-A research site, the mean annual air temperature (MAAT) was −5.0°C, whereas for the Coldfoot location, the MAAT was −4.0°C. The warmer Coldfoot air temperatures also are apparent in the air thawing index (ATI) and air freezing index (AFI) summarized in Table 2. Despite the slight difference in air temperature between locations, we use the Coldfoot data as a proxy for FDL-A, as we do not have a complete record of precipitation or snow depth from the research site due to intermittent equipment malfunctions. Figure 3b and Table 3 contain snow depth and precipitation data for Coldfoot by water year (a water year begins on October 1 and ends on September 30; USGS, 2016). The winter of 2020/21 (or water year 2021) is the most notable during this period of observation, as it experienced nearly half the typical snowfall, which also started accumulating about a month later than normal (see bold values in Table 3). This lack of snowfall is manifested in other data at the FDL-A research site, which are discussed later.

History of FDL-A’s Movement Rate

Figure 4 is a summary of FDL-A’s rate of motion, including rates obtained from analysis of historic imagery and recent RTK-GPS measurements (refer to Darrow et al., 2016, for more information on the historic imagery analysis). A linear trend is the best fit for the historic rate of motion, with an R2 of 0.95 (blue series, Figure 4). When the recent RTK-GPS rates are added, the best fit is no longer linear (orange series with R2 of 0.62) but becomes exponential (red series with R2 of 0.82). The RTK-GPS measurements demonstrate a peak in FDL-A’s movement rate in 2020, followed by 3 years of lower annual movement rates. We attribute this slowing to the below-average snowfall during the winter of 2020/21. The lack of insulation provided by the snow had the effect of cooling the subsurface, which increased the shear strength of the frozen soil and slowed the downslope movement in subsequent years. We observed a similar below-average snowfall during the winter of 2012/13, which resulted in colder ground temperatures within the lobe and lower strain rates (Darrow et al., 2017).

Embankment and Subsurface Analysis

Subsurface Soils and Laboratory Testing Results

Figure 5 contains schematics of the four boreholes, including a summary of the laboratory testing results. Where drilled, the embankment ranged from 3.05 to 3.66 m (10–12 ft) thick, consisting of moist, brown, non-plastic silty gravel with sand to silty sand with gravel. The foundation soils typically consisted of brown, non-plastic slightly organic silty sand with gravel to silty gravel with sand. These soils were moist to wet where unfrozen. We intercepted gray, white mica schist bedrock in all boreholes, varying between 4.27 and 9.75 m (14.0 and 32.0 ft) below the embankment surface. The highly weathered bedrock was moist to wet where unfrozen. We intercepted the permafrost table 6.04–7.62 m (19.8–25.0 ft) below the ground surface (bgs) in TH20-01 through TH20-03; TH20-04 was unfrozen. We attribute this depth to permafrost to the lack of winter maintenance and the insulating effects of snow accumulation on the abandoned embankment’s surface. A water table was perched on the permafrost table in TH20-01 and TH20-02. We also intercepted a water table in TH20-04.

We conducted direct shear tests on two sets of three unfrozen samples of the weathered white mica schist collected 4.9–5.2 m bgs in TH20-04 (sample locations not shown on Figure 5). These samples were collected from the unfrozen layer above the permafrost table and tested at above-freezing temperatures. Although the samples were not oriented in a cardinal direction when taken in the field, the foliation within the schist was approximately perpendicular to the long axis of the sample, or approximately parallel with the ground surface. The following is a summary of the average results obtained from the tests: dry unit weight of 16.76 kN/m3; gravimetric moisture content of 6.4 percent, friction angle of 30.8 degrees, and cohesion of 8.79 kPa.

Ground Temperature

Due to hardware conflicts and programming issues, we were unable to obtain long-term data from the temperature cables installed in the borings, despite multiple attempts to troubleshoot the system with the aid of the vendor. Because of interference with recording other data from the site, we disconnected the temperature cables in 2022. Fortunately, the M-IPIs contain temperature sensors in each of their instrumented segments. Although uncalibrated, these sensors can serve as a backup for temperature measurements (Darrow and Jensen, 2012). As an example, Figure 6 contains temperature profiles for TH20-01 for months for which we have complete data. At 1 m bgs in TH20-01, the embankment temperature ranged between −5°C and 15°C. The depth of zero annual amplitude (ZAA) was at 6.6 m bgs in 2020 and 2021, deepening to 7.1 m bgs in 2022 (the interpretation of ZAA depth is limited to the spacing of the sensors within the M-IPI). The deepest thaw at TH20-01 occurred in September or October, depending on the year.

Porewater Pressure

Figure 3c contains the porewater pressure measurements from both 2018 VWPs (each at 3.0 m bgs) and from the southernmost VWP (TH20-04 at 8.0 m bgs) as examples of the 2020 installations. The VWP instruments also contain thermistor beads, allowing temperature measurements at the installation depth (Figure 3d). Typical data indicated two peaks in water pressure, the first occurring with the end of snowmelt in mid- to late May and the second occurring with the increase in late summer precipitation in late July. The TH18-01 installation failed on February 24, 2020, after it was overrun by the toe of FDL-A. Following the below-normal snowfall during the winter of 2020/21, the VWP in TH18-02 froze in early April 2021, causing a spike and subsequent drop in water pressure. FDL-A covered the TH18-02 location in July 2021, after which the VWP thawed; however, as the freezing may have damaged the diaphragm within the VWP, the data may no longer be accurate. The other VWPs did not experience freezing. Peak temperatures occurred in early to mid-September at 3.0 m bgs in TH18-01 and in late September to late October at 8.0 m bgs in TH20-04.

Pressure Plate and Tilt Meter Measurements

Figure 7 contains graphs of the temperatures recorded by the tilt meters and pressure plates (Figure 7a) and the pressure plate (Figure 7b) and tilt meter (Figure 7c) readings. As all of these devices were buried within 0.6 m (2 ft) of the surface, the measured temperatures demonstrated seasonal fluctuations. The temperatures from pressure plate 1 (PP 1) and tilt meter 1 (Tilt 1; both installed near TH20-01) were about 2.4°C warmer during the winter of 2022/23 and about 4.2°C cooler in the summer of 2023 than the other readings. We attribute this to the toe of FDL-A covering this part of the embankment in the late summer of 2022, which we observed in the field in July 2022.

The pressure plates are composed of stainless steel plates with de-aired hydraulic oil as a working fluid (Geokon, 2020). The pressure readings are corrected for temperature; however, we attached the pressure plates to steel plates to make them more robust against the impact of FDL-A. The steel plates may be adding additional temperature sensitivity to the pressure plate readings (Figure 7b), as all readings increased with sub-freezing temperatures and then decreased as the embankment thawed in the spring. The winter increase in pressure also may be attributed to pressure exerted by the embankment soils as they underwent the seasonal freeze–thaw cycle. The PP 1 readings increased more than the others during the winters of 2021/22 and 2022/23, and although the readings also dropped with the spring thaw, they continued to increase through the summer. Although the PP 4 readings also increased during winter of 2022/23, we know from visual observations that FDL-A is not impacting the TH20-04 location. One of the tilt sensors at the TH20-01 location (Tilt 1A; Figure 7c) also started to deviate from the other readings on August 2, 2022. These changes are most likely due to the impact of FDL-A. Further observations during the landslide–embankment collision are required for a full explanation of the pressure plate and tilt meter data.

M-IPI Measurements

Figure 8a contains plots of the M-IPI measurements for the four 2020 boreholes (locations provided in Figure 1d) in the A0 direction (i.e., the direction of greatest movement—downslope). We detected movement in TH20-01, TH20-02, and TH20-03 immediately upon the first reading on September 6, 2020, while the toe of FDL-A was still more than 10 m away. Figure 8b is a graph of the rate of horizontal displacement for selected depths along each of the four M-IPIs. In this analysis, for TH20-01 through TH20-03, we selected the sensor at the depth experiencing the most change; for TH20-04, we chose a median depth within the embankment. Through the period of observation, the rate of displacement in TH20-04 has been steady, at an average rate of 0.013 mm/d. After their installation, the northern three M-IPIs demonstrated similar rates of displacement, peaking at an average rate of 0.223 mm/d. Unfortunately, there are missing data from the winters of 2020/21 and 2021/22 due to hardware conflicts and programming issues; however, the periods of existing data suggest that TH20-01 started to diverge with a greater horizontal displacement rate in June 2021 and has continued this trend throughout the observation period. The northern three M-IPIs all demonstrated peaks in movement rate sometime between mid-October and mid-November and the slowest movement rates in mid-May to early June. These data are in good agreement with previous analysis of strain rate data from a M-IPI installed within FDL-A in 2012, from which a predictive formula indicated a maximum velocity in the fall and a minimum velocity in early spring (Darrow et al., 2017).

Figure 9 is schematic illustrating the simplified subsurface and includes the cumulative displacement of the M-IPIs as they are positioned along the embankment; in this schematic, all cumulative displacements are drawn to the same horizontal scale. These data indicate that the base of FDL-A is shearing into the weathered bedrock (TH20-01) and below the permafrost table (TH20-01 through TH20-03). These data also indicate that FDL-A has displaced the area around TH20-01 approximately 0.32 m horizontally, with this M-IPI demonstrating about three times more horizontal displacement than TH20-02 or TH20-03. The M-IPI in TH20-04 demonstrates a slow and steady downslope movement of the embankment material, indicating that FDL-A has not yet started to impact this portion of the embankment. The subsurface interpretation in Figure 9 also indicates that there are some variations in the bedrock surface beneath the embankment. For example, we interpret the TH20-03 area as a paleo-drainage, as this area was also occupied by the modern drainage before FDL-A modified its path. Results from other subsurface investigations in the area (e.g., Simpson et al., 2016) indicated that the bedrock surface is typically 3.7 m bgs, similar to the TH20-01 location.

Lidar Data and Change Detection

We took the difference between the 2021 dataset and our dataset collected in 2020 (not part of this project) and the difference between the 2022 and 2021 datasets, producing DEMs of difference (DoDs); see Figure 10. We assumed spatially uniform errors and applied the square root of the sum of squares to calculate the error of the DoDs (Wheaton et al., 2010), which was 0.43 m. Both DoDs were masked to exclude differences of less than 0.5 m to be conservative. Some errors still remain in the DoDs; for example, the 2021–2020 differencing indicated movement where there should be little (e.g., green areas west of the embankment in Figure 10a). This may be due to acquiring the 2020 lidar after leaf-out in the spring, as this part of the embankment is covered by dense shrubs. The 2021 dataset contained pixelated areas that produced artifacts and an apparent elevation gain (green area indicated by purple arrow in Figure 10b) along the eastern part of the embankment, which is erroneous based on visual observations of the site.

Despite these errors and artifacts, the DoDs also illustrate dramatic changes in the position of FDL-A, with vertical changes up to 8.2 m as the toe advanced toward the embankment. The 2021–2020 DoD also captured sediment filling a small basin at the inlet of a culvert that used to provide cross-drainage for the stream draining from FDL-A (indicated by the red arrow in Figure 10a). As another means of change detection, the locations of FDL-A’s toe, as mapped using the RTK-GPS device from 2020 until 2023, are provided in Figure 10c. These curves illustrate the proximity of the toe to TH20-01 by July 2023.

Based on the M-IPI data, FDL-A was shearing approximately 4 m below the bottom of the Dalton Highway embankment at the TH20-01 location before this project started in 2020 and while the lobe was still meters away. Figure 11 illustrates the movement processes evident in March 2021. The daily photograph (Figure 11a) indicated an overall steepening of the toe. During a trip to the site that month, we noted a shear plane that daylighted in FDL-A’s toe, about 3 m above the ground surface in front of the lobe (Figure 11b). Figure 11c is a scaled conceptual model of the toe of FDL-A; the ground surface was derived from the 2020 lidar DEM. We determined the thickness of FDL-A near its toe using TH22-02; although this borehole was drilled for a different project and is not discussed in this paper, we include its position here to constrain the conceptual model. Based on field observations and the M-IPI data, we hypothesize that a décollement forms the base of FDL-A, where its mass shears along the bedrock surface (Figure 11c). A series of minor thrust faults originate at the décollement and daylight within the toe slope; they also form the interface between FDL-A soil and the original ground downhill of it. Only the décollement was evident in the TH20-01 M-IPI data in March 2021, approximately 7.6 m below the embankment surface and about 1.3 m below the bedrock surface (Figure 11d).

Beginning September 19, 2021, in the daily photographs we observed a mound of soil rising from the ground surface (red oval in Figure 12a). This mound, located between FDL-A and TH20-01, was obvious during a trip to the site in October 2021 (Figure 12b). Our camera installed at the north end of the FDL-A site recorded the formation of the soil mound (red oval in Figure 12c), as well as FDL-A pushing a near-surface layer of soil in front of it toward the embankment (northern extent of sheared soils indicated by the orange arrow and curve in Figure 12c). We hypothesize that the mound formed as soil was forced upward due to rotational movement and buckling as FDL-A approached the embankment (Figure 12d). During the fall of 2021, the M-IPI at TH20-01 continued to demonstrate steady horizontal displacement at the décollement located within the upper bedrock (Figure 12e).

Based on field observations, and RTK-GPS measurements of FDL-A’s toe and the toe of the embankment, FDL-A started to overlap the side slope of the highway embankment in July 2022. The next noteworthy change in the westward progression of FDL-A was the back-tilting of trees that became visible in the daily images on March 1, 2023 (Figure 13a). This suggests a shallow rotational failure of the toe, possibly along one of the thrust faults developed during the fall of 2021. During a field visit in June 2023, we noted the presence of scarps and down-dropped blocks in the vicinity of the back-tilted trees, which is additional evidence of rotational failure (Figure 13b). Figure 13c illustrates the location of the shallow rotational failure relative to the thrust faults and décollement, the latter of which continued to be demonstrated by the M-IPI in TH20-01 (Figure 13d).

By September 2023, there were several pulses of soil slumping down the toe, causing the trees to rotate forward toward the embankment (Figure 14a). The removal of this wet and loose surface layer again exposed the minor thrust faults daylighting in FDL-A’s toe (indicated by orange lines in Figure 14b). Figure 14c illustrates the multiple types of movement now observed within the toe of FDL-A. The M-IPI readings in TH20-01 (Figure 14d) indicated downslope leaning of the device beginning about 2 m below the embankment surface. We interpret this as FDL-A’s mass finally impacting the embankment at and above the original ground surface, which is corroborated by the pressure plate and tilt meter readings near TH20-01 (Figure 7) and through visual observations.

Certain deformation features demonstrated during the collision of FDL-A with the abandoned part of the Dalton Highway embankment are analogous to pro-glacial push moraines. For example, the area of near-surface soil shearing captured by the north camera in October 2021 (Figure 12c) is similar to a foreland wedge, and the soil mound formed during the same time may represent distal folding ahead of the thrust faults within the toe (Bennett, 2001). Likewise, the major shear zone at the base of FDL-A can be interpreted as a décollement, above which FDL-A shears downslope and below which no significant deformation occurs (Aber and Ber, 2011). The results of the direct shear tests of the unfrozen weathered white mica schist indicate a moderate friction angle of 30.8 degrees and low cohesion of 8.79 kPa along foliation. For comparison, González de Vallejo and Ferrer (2011) list values of 20–30 degrees and 2–15 MPa for friction angle and cohesion, respectively, for fresh schist along foliation. Recognizing that these samples were tested under thawed conditions and do not necessarily reflect the strength of the frozen bedrock, this low cohesion supports the hypothesis that the weathered bedrock represents a weak layer with an unfavorable foliation orientation along which FDL-A shears downslope.

Based on the limited subsurface data used to produce the conceptual models, we hypothesize that the bedrock surface is parallel with the ground surface at the research site. This geometry, as well as FDL-A’s shearing within a weak zone in the upper bedrock, lends itself to infinite slope analysis (Cornforth, 2005; Duncan et al., 2014). Using the geometry defined in the schematic in Figure 15a and assuming no groundwater, the resisting shear force (S) can be calculated using:
where c is cohesion, σ is geostatic stress, and ϕ is the friction angle (Cornforth, 2005). For this calculation, we assume a 1 by 1 m soil column (b equals 1). Eq. 1 represents the total shear force. To determine the portion of the shear force resisting the movement of FDL-A (SFDL), we must resolve the geostatic stress into its principle components and subtract the component oriented in the downslope direction (i.e., the last term in Eq. 2):

For the TH20-01 location, there is 3.66 m of embankment with a moist unit weight of 21.95 kN/m3, based on typical maximum unit weights, optimum moisture content of soil from local material sites (Maxwell, 2016), and 95 percent compaction. Recognizing that the embankment is not continuous across the entire slope, we treated it as a strip load and reduced this vertical stress using an influence factor and the principle of superposition (Osterberg, 1957). Figure 15b is a schematic of the position of TH20-01 within the embankment; we assumed a uniform embankment thickness for these calculations. The values A1, B1, A2, and B2 were determined from lidar data. We reduced the vertical stress caused by the embankment at the depth of shear (indicated by a red dot in Figure 15b) by dividing the embankment into two portions, and calculating the terms forumla, forumla, forumla, and forumla. Using the two pairs of terms (i.e., either forumla and forumla or forumla and forumla), we determined influence factors for each portion of the embankment using the Osterberg (1957) chart, added the influence factors together, and then multiplied the result to the vertical stress from the embankment, resulting in 71.5 kPa. The embankment is underlain by 2.59 m of silty sand with an assumed moist unit weight of 19.75 kN/m3 (Coduto et al., 2011), in turn underlain by 1.35 m of weathered white mica schist bedrock above the shear surface with a unit weight of 17.83 kN/m3. This results in a total geostatic stress of 146.7 kPa at the depth of shear (7.6 m). For the β angle, we used the lidar data to calculate a slope angle of 7.0 degrees adjacent to FDL-A. For c and ϕ, we used the results from the direct shear tests on the white mica schist samples (c of 8.79 kPa and ϕ of 30.8 degrees). These assumptions yielded an SFDL of 77.8 kN/m width. Since FDL-A is moving, we assume a factor of safety of 1, meaning that 77.8 kN/m width is the minimum value of the driving force as well. This simple approach, with all of its assumptions, provides an order of magnitude for the impact from FDL-A.

As previously indicated, the M-IPI in TH20-01 recorded ∼0.3 m of horizontal displacement between September 6, 2020, and November 1, 2023. The displacement is not yet discernible through close visual inspection of the daily photographs or through the DoDs, but this horizontal displacement is manifested as cracks in the embankment (Figure 10c and d). In 2022, we noticed persistent cracks in the embankment surface and mapped their positions using the RTK-GPS device in May 2023. It is possible that these cracks formed through thaw settlement, as this portion of the embankment is no longer maintained; however, their locations and orientations also suggest that they could be due to horizontal shearing from FDL-A. By July 2023, the north camera was in danger of being overrun by FDL-A’s toe. We moved this camera to the west of the embankment in a location where we hope to capture evidence of the embankment’s horizontal displacement.

FDL-A’s annualized movement rate for July 2022 to July 2023 was 7.96 m/yr. The data presented in Figure 4 indicated a reduction in the movement rate since 2020, which we attribute mainly to cooling of the lobe due to low and late snowfall during 2020/21. This hypothesis is supported by a similar slowing in nearly all monitored FDLs after that winter. This reinforces previous work indicating that FDL movement is closely tied to its internal temperature, which is controlled by air temperature and precipitation, especially snow (Darrow et al., 2017). Unlike the other FDLs, FDL-A also has been interacting with the highway embankment, which may serve as an additional impediment to slow—although not stop—its downslope movement. Continued monitoring will resolve this question.

Results from our initial research indicated that FDL-A was an impending threat to the Dalton Highway. Based on these results, ADOT&PF realigned the highway to avoid the geohazard, leaving in place a section of the abandoned embankment, which we instrumented in 2020. The integration of multiple datasets obtained through this full-scale field experiment has allowed us to produce a detailed description of the collision of a landslide in permafrost with an engineered embankment. Although we have made field observations of this site for over a decade, this project made it possible to understand the nuances of movement. For example, while still meters away on the surface, FDL-A was already impacting the Dalton Highway embankment by shearing within the upper bedrock at 7.6 m below the embankment surface in one of the borehole locations. FDL-A made “official” contact with the highway embankment in July 2022 and has horizontally displaced a portion of the embankment by ∼0.3 m as of November 1, 2023. Using the infinite slope approach and strength properties determined from laboratory testing, we estimate that FDL-A is impacting the highway embankment and underlying soils with at least 77.8 kN/m width shear force. As this force, manifested by the measured horizontal displacement, occurs at depth below the bottom of the embankment, we postulate that the presence of the embankment will do little to stop FDL-A’s downslope progression.

Analysis of the data presented here supports the hypothesis that FDL-A’s annualized movement rate is tied to ground temperature and thus sensitive to changes in air temperature and precipitation (namely, snow). We also reconfirm that FDL-A’s peak movement rate occurs in late fall—the period of deepest thaw, as measured in the embankment. Many features observed at the toe of FDL-A between 2020 and 2023 are similar to push moraines produced in a pro-glacial environment. All of these observations can be incorporated into modeling of the landslide to determine appropriate and effective mitigation measures. With continued monitoring, we hope to 1) obtain proof of concept for the tilt meter and pressure plate instrumentation, 2) observe horizontal displacement within the embankment to verify the cause of cracking in the surface, and 3) determine how the presence of the abandoned Dalton Highway embankment affects the movement characteristics and long-term movement rates of FDL-A.

In several places in this article, we referred to changes observed in daily images from the cameras at the research site. As animated sequences provide a better means for visualization than static images, the FDL website (UAF, n.d.) provides time-lapse movies from the north and south cameras that show the progression of FDL-A’s approach and collision with the abandoned Dalton Highway embankment.

This research was supported by a Pacific Northwest Transportation Consortium grant (UWSC10217), an ADOT&PF grant (ADN 45-2-1065), and ongoing support from the Alyeska Pipeline Service Company (APSC). We thank J. Schwarber for her true grit during the 2018 field work, and P. Presler, J. Simpson, P. Lanigan, and S. Parker for their expertise and enthusiasm in the field during the 2020 geotechnical investigation. We thank D. Cronmiller for guiding us to recent work on push moraines and M. Bray, Y. Shur, B. Leshchinsky, and the anonymous reviewers for improving this paper through their thorough reviews and suggestions.

The views, opinions, findings, and conclusions reflected in this paper are the responsibility of the authors only and do not represent the official policy or position of the ADOT&PF, APSC, U.S. Army Corps of Engineers, or other entity.