Introduction

The large landslide complex on the east side of Black Mesa (also known as Mesa Prieta), New Mexico, formed due to a combination of mid-to-late Pleistocene Rio Grande incision and millennial-to-decadal variations in regional precipitation. In northern New Mexico, the Rio Grande has carved several prominent gorges. To the north, and the location of this study, is the Rio Grande Gorge. To the south, and adjacent to the Jemez Mountains volcanic complex, is White Rock Canyon. Landslide-dammed lakes formed at least four times between 18 and 12.4 ka at the upstream end of White Rock Canyon, and three of these lakes existed between 13.7 and 12.4 ka1. Removal of lateral support at the base of steep slopes by Rio Grande erosion, particularly during high river discharge or rapid incision, is a key trigger for these river-damming landslides1,2. Furthermore, late Pleistocene climate change is considered a major factor in triggering the White Rock Canyon and other landslides along the upper Rio Grande, including the eastern flank of Black Mesa2. Although morphological signatures of past landslide activity are present in the deep-seated, slow-moving landslides of Black Mesa, there is very limited information on their movement history, and their absolute ages have yet to be determined. The Black Mesa is capped by 3–20 m thick erosionally resistant 3.3–3.8 Ma year old Servilleta Basalt formed during the late Pliocene epoch that is underlain by upper Miocene terrestrial sedimentary strata of the Chamita Formation of the Santa Fe Group, creating a predisposition for instability3,4. Differential weathering and erosion of these layers, exacerbated by the permeable nature of Mesa’s basalt caprock, enable significant water infiltration during periods of increased precipitation. This infiltration raises pore water pressure within the underlying weakly lithified sedimentary rocks, reducing their shear strength and promoting gradual downslope movement. Additionally, the Rio Grande’s erosional activity at the base of the Mesa removes lateral support, further destabilizing the toe slope region and contributing to the movements of these landslides2. There are no recorded catastrophic failures of the Black Mesa landslide complex, and little is known about the movement history of the extensive slide complex. The presence of tens of thousands of Native American and colonial Spanish pictographs on gravity-transported basalt blocks within the complex5,6 suggests that movements have been subdued in the last hundreds to thousands of years. However, the presence of landslide-dammed lake deposits above White Rock Canyon, 40 km downstream, begs the question, could the same happen along the Black Mesa section of the Rio Grande Gorge?

Topographic surface roughness is a geomorphic metric successfully used to calibrate the ages of known landslides to determine the ages of unknown landslides7,8,9,10. Younger landslides typically exhibit higher surface roughness due to fresh debris, sharp scarps, and irregular surfaces7,11. In contrast, older landslides become smoother over time because of weathering and in-filling of topographic closed depressions (TCDs), erosion of micro-topographic highs, vegetation growth, bioturbation, and the reestablishment of drainage networks. The downslope movement of unconsolidated soils is a gradient-driven (diffusion-like) phenomenon, which gradually smooths out the depositional morphology, and can act on the topography for hundreds to thousands of years in undisturbed terrain. However, reactivating older landslide deposits through rapid mass wasting processes, such as shallow soil failure or debris flows, can disrupt this process, altering the previously preserved topographic signatures. Consequently, the surface roughness measurements reflect the most recent phase of landslide activity. This loss of surface roughness from the earlier landslide events can lead to inaccuracies or weaken the correlation when establishing the empirical relationship between landslide age and surface roughness.

In many cases, large, deep-seated, slow-moving landslides are the relict, or paleo-landslides that may have initially failed catastrophically, often inferred to occur under different climatic conditions. Slow-moving active landslides generally show distinct features such as hummocky topography, active primary and internal scarps, and TCDs, which provide a qualitative understanding of their activity12,13,14. In many cases, slow-moving landslides have failure planes extending up to tens of meters below the surface, making the surface manifestations of such deformations not easily identifiable. Slow-moving deep-seated landslides can remain active for hundreds of years, exhibiting episodic activity based on climatic and geological conditions15,16,17. These movements can range from millimeters to a few meters per year15,18,19,20,21,22,23,24,25,26. Identifying these movements provides valuable insights into the current rate of slope deformation and the associated hazards. Recently, the Interferometric Synthetic Aperture Radar (InSAR) technique, a form of microwave satellite remote sensing, has revolutionized the detection and monitoring of slow-moving landslides15,27,28,29,30. Microwave remote sensing enables monitoring of the Earth’s surface, day and night, under all weather conditions. It is ideal for performing closely-spaced time-series analysis to understand surface deformation even on a millimeter scale. InSAR time-series analysis enables the rapid-repeat (weekly or bi-monthly) monitoring of surface deformation over decadal time scales, effectively representing present-day surface deformational activity31. This technique is widely used for monitoring and detecting slow-moving landslides across the world25,27,32,33,34.

The deep-seated landslides of Black Mesa are reported to be creeping gradually, as evidenced by the formation of Toreva blocks, extension fractures, and the presence of multiple cracks at the head scarp visible from high-resolution LiDAR data4,35,36 (Figs. 1 and 2a–c). The exact formation time of these landslides is unknown; however, it is assumed that they were initiated sometime during the middle Pleistocene to Holocene epoch3. The objectives of this study are to (i) identify the present-day instability and degree of activity in the landslide complexes of Black Mesa near Espanola; (ii) evaluate the long-term morphological signatures of surface deformation and its interaction pattern with present-day slope instability to understand the dynamics and geomorphological evolution of this most impressive landslide complex.

Study area

Black Mesa extends approximately 21 km in a northeast direction, located north of the Española Basin, which is a west-tilted half-graben within the Rio Grande Rift in northern New Mexico37 (Fig. 1). The mesa is capped by 3 to 20 m of Servilleta Basalt from the late Pliocene epoch (Fig. 1B,C). The La Mesita Fault, a left-oblique, northeast-striking fault, is part of the Embudo Fault system and intersects the face of the landslide complex in Black Mesa. This fault has caused a vertical offset of the Servilleta Basalt by 60–70 m, lowering the western side, with a potential left-lateral displacement of 460 m38,39,40. At the southern end of the study area lies the Santa Clara Fault, which extends from the southern tip of Black Mesa to the southern end of Santa Clara Canyon. Along the western edge of Black Mesa is the Black Mesa Fault, which has caused a vertical displacement of the capping Servilleta Basalt by 18–40 m, lowering it to the southeast38. Beneath the Servilleta Basalt lie the fluvial deposits of the Chamita Formation of the Santa Fe Group, characterized by an abundance of quartz-rich gravel and sand. The Chamita Formation is unconformably overlain by basin-fill Quaternary alluvium and eolian sediments from the middle Pleistocene to the Holocene epoch. Landslides disrupt the Servilleta Basalt, the Chamita Formation sediments, and Quaternary deposits except in some locations in the southern part of the study area where floodplain deposits remain intact (Fig. 1c). The study area has a semi-arid climate with hot summers and mild winters. The average annual rainfall is about 340 mm (13.4 inches)41.

Fig. 1
figure 1

(A) Map showing the location of the Rio Grande Gorge (RGC) study area, White Rock Canyon, and some major cities in New Mexico. (B) Slopeshade map of the RGC. (C) Simplified geological map of the study area and vicinity. Latitude and longitude values represent the coordinates of the corresponding corners of maps B and C. Map generated using ArcGIS Pro version 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Full size image
Fig. 2
figure 2

Field photographs from different locations of the study area.

Full size image

Results and discussion

Landslides, scarps, and closed topographic depressions

A total of 12 landslides were identified and demarcated by recognizing the head scarps, lateral scarps, and toe regions using high-resolution LiDAR digital elevation model (DEM) (Fig. 3A). The mean size of the landslides is \(2\,\text {km}^2\), with the largest measuring \(7.35\,\text {km}^2\) and the smallest covering only \(0.55\,\text {km}^2\). Secondary scarps are prominent in landslides no.1, 7, and 9 and are located mainly at the head of the landslides. Landslide no. 7 has the highest number of secondary scraps, with fracture planes extending several tens of meters deep, as observed during the field survey (Fig. 2a–c). The presence of numerous such scarps and dislocated basalt layers at various locations down the slope within the landslide bodies indicates paleo-landslide activity, with the present head scarps being the youngest zone of detachment. These landslide blocks possibly have gradually failed over multiple periods, exhibiting rotational movement along concave upward surfaces, resulting in the formation of Toreva blocks (Fig. 2a,b).

The TCDs were found to occur near the secondary scarps, insinuating a possible spatial correlation between the formation of the depressions and the scarps (Fig. 3A). The highest intensity of occurrences of the TCDs is ≈ 12 m distance from the scarps, again indicating a correlation between these two features (Fig. 3B). To check whether the occurrences of TCDs near the scarps are random or not, a Spatial Kolmogorov–Smirnov (SKS) test was performed, which yielded a statistically significant value indicating a spatial dependency of the TCDs on the proximity to the scarps. The cumulative distribution of the distances of present TCDs from the scarps differs from the cumulative distribution of the distances of randomly occurring TCDs from the scarps (Fig. 3C). This statistically significant association implies that these scarps must have some geological influence on the formation of the TCDs. The secondary scarps within the landslide bodies define Toreva blocks formed due to rotational movements of the landslide blocks where the block tilted backward and moved down along a concave slip surface36. Such rotational movements along the failure plane create a scarp at the head of the sliding mass, and the downward movement of the block forms a depression behind it. Additionally, the settling process of the dislocated block can lead to further subsidence of the landslide surface, forming new depressions or deepening the existing ones. Therefore, TCDs are a strong indicator of the timeline of past landslide activity, as they are zones that trap sediments and water—useful for reconstructing sedimentary infilling and potentially providing insights into local paleoclimate conditions. The TCDs are ideal locations for collecting samples for landslide age dating since they are formed almost concurrently with landslide activity. Two carbon samples were collected from a TCD located in landslide no. 5 from 1 m and 2 m below the ground surface, which yielded ages of \(3910 \pm 24 \ ^{14}\text {C} \ \text {ybp}\) and \(6874 \pm 14 \ ^{14}\text {C} \ \text {ybp}\) years, respectively (Fig. 2d). The age of the sample collected from the greater (2m) depth indicates the minimum age of this TCD. In a landslide complex like Black Mesa, where landslide movements have occurred multiple times across different geological periods, determining the precise age of landslides can be challenging. The age of a TCD is only a minimum estimate for the initial timing of movement and the formation of the TCD from which the sample was collected. This uncertainty arises because it is unknown whether the block experienced one or multiple movement events before forming the TCD. To better understand the approximate ages of these failures, a series of samples from different TCDs need to be collected and dated to estimate the landslide ages. In such cases, Optically Stimulated Luminescence (OSL) dating may be more useful for estimating the minimum age of the landslide with greater reliability. The OSL dating technique calculates the time when quartz grains in the soils were last exposed to sunlight, effectively determining the burial periods of these grains. A forthcoming paper will explore the possibility of using the OSL technique to determine the ages of landslides in the Black Mesa complex.

Fig. 3
figure 3

Map A showing landslide boundaries, TCD, and radiocarbon sample location. Graph B is the intensity plot of the occurrence of TCD as a function of distance from scarps; Graph C is the result of the SKS test indicating a non-random distribution of the TCDs against distance from the scarps. Latitude and longitude values represent the coordinates of the corresponding corners of the map. Map generated using ArcGIS Pro version 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Full size image

Present day landslide activity

To understand the present-day landslide activity, an InSAR time-series analysis was performed over seven years (2017–2023). Given the area’s abundant rock exposures, good reflectors for SAR signals, and sparse vegetation, the Persistent Scatterer (PS) technique was implemented to analyze surficial deformation. The extent of the InSAR investigation encompasses a larger area than the main study area, which is marked by the dashed black line in Fig. 4A. Several landslide complexes are located upstream of the main study area, exhibiting cumulative displacements of up to 50 mm over the past seven years. These areas are moving away from the satellite, suggesting possible subsidence associated with landslide activity (Fig. 4B). In contrast, some locations are moving toward the satellite, indicating potential uplift (Fig. 4C). However, the present study focuses on landslide activity downstream from La Mesita along the eastern flank of Black Mesa, as these landslides may pose a significant hazard to the local population (Fig. 4A). This region is sparsely populated, which includes unincorporated communities such as Canova and Lyden in Rio Arriba County, New Mexico. According to the 2020 Census, these communities are home to approximately 102 and 250 residents, respectively.

Fig. 4
figure 4

Kriging map (A) shows the average velocity of PS points computed from InSAR time-series analysis for the study area. Plots on the top right show slope movements outside the main landslides. The color and number in the plot (B and C) match the location’s color and number of rectangular boxes in the krigging map (A). Relative to the satellite’s line of sight (LOS), lines 1–3 show subsidence (B), and lines 4–5 show uplift (C). The two graphs below figure A show the average velocity distribution of PS points in ascending and descending directions. Latitude and longitude values represent the coordinates of the corresponding corners of the map. Map generated using ArcGIS Pro version 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Full size image

To measure the average movements of the PS points with greater certainty, active PS points (average velocity > \(\pm\,1 \ \text {standard deviation}\)) were filtered from the results of both ascending and descending image analyses and merged for further examination. Kriging was performed on the active PS points to visualize general deformation areas, superimposed by individual PS points showing significant uplift (PS moving towards the satellite along the LOS) and subsidence (PS moving away from the satellite along the LOS) (Fig. 5).

Fig. 5
figure 5

A kriging map showing the average velocity of the landslides developed using only active PS points. Some of the active PS points are superimposed on the map. Small graphs in the box illustrate the movement magnitude and direction of various locations within the landslides. Latitude and longitude values represent the coordinates of the corresponding corners of the map. Map generated using ArcGIS Pro version 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Full size image

Subsidence is the most prominent feature in the present landslides, particularly in areas where cumulative subsidence exceeds 40 mm, indicating active deformation zones. Significant subsidence movements have been identified near the toes of the landslides, likely resulting from undercutting by stream channels and the Rio Grande. This indicates that these areas are significantly active in the present day. Additionally, the rotational nature of failures of the landslides along concave planes causes subsidence in many parts of the landslides, allowing the PS points to move away from the satellite. At the same time some landslides would experience the building or uplift of the surface where the PS points would move towards the satellite. In a larger picture, this has been captured by the InSAR time-series analysis. Gradual erosion of exposed soil and sediment deposition in certain areas of the landslides are also observable in the PS movement data. Although these processes are not directly related to landslide movements, they indicate zones of active geomorphic surface processes. Interestingly, the crown or upper portions of the landslides, particularly those numbered 1 through 7, exhibit very limited surface movement. However, this region contains most TCDs and secondary scarps, which indicate past landslide activity but show very limited present-day movements. To better understand the magnitude of landslide activity, the landslides were divided into 250 × 250 m grid cells, and the absolute velocity of the active PS points, regardless of movement direction, was averaged within these grids (Fig. 6). Most parts of the landslide crowns have an average absolute velocity of less than 1 mm/year (marked in green). In contrast, grids near the toe and middle of the landslide bodies exhibit the highest average absolute velocities, ranging from > 1.12 to 3.72 mm/year. Landslide number 10 shows significant average absolute velocity, with most grid cells falling within the > 1.12 to 3.72 mm/year velocity range. Analyzing the average magnitude or absolute velocity of landslides provides information on the overall intensity of the deformation, regardless of the direction of the movement. Since both active uplift and subsidence indicate deformation of the landslide surface, a higher average absolute velocity of the 250 × 250 m grid cells suggests increased intensity in landslide activity.

Fig. 6
figure 6

Map showing the average absolute velocity of the landslide grids for active PS points averaged within 250 × 250 m grid cells. Map generated using ArcGIS Pro version 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Full size image

The Index of Activity (IA) was calculated for individual grids to assess landslide activity. A higher IA value indicates more active PS points within the grid, representing enhanced landslide activity. In other words, a higher IA value signifies a higher density of active PS points, which are highly precise and suggest increased surface movement, interpreted as landslide activity with greater confidence. The IA values ranged from 0.01 to 1, with a mean of 0.28. The IA values were classified into three groups: high (> 0.33–1), medium (> 0.22–0.33), and low (0.1–0.22) (Fig. 7). Approximately 37% of the landslide grids were classified as high IA zones, 33% as medium IA zones, and 30% as low IA zones.

Fig. 7
figure 7

Map showing the Index of Activity (IA) of each 250 × 250 m landslide grid for active PS points. Map generated using ArcGIS Pro version 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Full size image

High IA zones are predominantly located near the lower portions and toes of the landslides, whereas the crowns of the landslides largely coincide with low IA zones. Findings from mean velocity, average absolute velocity, and IA are consistent, indicating limited present-day landslide activity in the head scarp regions and enhanced movement activity in the lower portions and toes of the landslides. Specifically, landslides numbered 3, 4, 6, and 11 show signatures of low to moderate activity, while landslide number 10 is identified as the most active. These results highlight the importance of monitoring landslides’ lower regions and toes, as they may be more prone to activity due to various geological and hydrological factors. The increased activity observed in the lower regions of the landslides can be attributed to several factors. The geologically recent undercutting of the landslide toe by the Rio Grande causes small collapses and subsidence of the landslide surfaces, as detected by the InSAR analysis. Additionally, increased groundwater seepage and pore water pressure in the lower parts of the landslides reduce effective stress, leading to higher activity, especially in areas with poor drainage. Overburden stress can also contribute to increased activity in the lower regions. The composition of the landslide material likely also plays a significant role, with the lower portions consisting of landslide-disturbed components of the terrestrial fluvial muds and sands of the Chamita Formation of the Santa Fe Group that are more prone to weathering and erosion. In contrast, the crown is composed of more stable basalt lithology. These factors combined make the lower portions of the landslides more vulnerable to frequent movement, resulting in the observed enhanced activity relative to the head regions of the slides.

Analysis of surface roughness

Studies have found that topographic surface roughness can be an external geomorphic manifestation of relative landslide age and activity7,9,10. This geomorphic metric can be successfully utilized to empirically calibrate the age of dated landslides to nearby, undated slides. Therefore, surface roughness can be a proxy of landslide age in thousands of years time-scale7,10. The surface roughness of the landslides was estimated by calculating the standard deviation of the topographic slope in a 15 × 15 m window, which ranges from \(0^\circ\) to \(34.22^\circ\) (Fig. 8). The highest surface roughness is observed in the crown of the landslides near the primary and secondary scraps. The lowest surface roughness is associated with TCDs and portions of the landslides capped by blocks of basalt, which were transported to their present down-slope locations by previous landslide activity. Interestingly, surface roughness is high at the crown near the scarps, but present-day movement is limited in those regions. Many grids near the toes of landslides exhibit high present-day movements but low surface roughness. Figure 9A shows that grids with high velocity magnitude largely have low surface roughness. Only a limited number of grids have high present-day movement and surface roughness. Figure 9B shows the average absolute velocity or velocity magnitude of the PS points plotted against the average surface roughness of each grid, while Fig. 9C presents the absolute velocity (velocity magnitude) plotted against the surface roughness of the individual PS points. These plots reveal that points with low present-day movement have high and low roughness values, but points having higher velocity magnitude, say velocity >3mm/year, are mostly associated with low surface roughness. Therefore, there is no straightforward relationship between the present-day instability captured by InSAR and the long-term deformation signature presented by the surface roughness. Landslide activity is conspicuous along the head scarp regions as evidenced by block displacement and TCD; however, such displacement signatures were not significantly captured by the InSAR time series. It is possible that the displacements along the scarps occurred over timescales ranging from 100 to 10,000 years or more. These failures were most likely triggered when the Rio Grande was actively eroding the toes of the landslides at a much greater rate than observed today. This significant removal of lateral support caused the blocks to fail due to gravitational pull. A similar landslide-triggering mechanism is invoked for White Rock Canyon, when the Rio Grande was blocked several times by Toreva-style landslides during the late Pleistocene epoch1. Climatic conditions are interpreted to have played a crucial role in activating these landslides in the late Pleistocene. A wetter regional climate favored initiating and reactivating many landslides along the Rio Grande1,2. A drier and warmer climate would significantly retard the movements of deep-seated slow-moving landslides, while a wetter climate would accelerate them. The present-day slower movements of these landslides can be attributed to drier climatic conditions compared to the late Pleistocene in the Española region. However, most parts of the landslides remain slowly active. Changes in climatic conditions and erosion activity by the Rio Grande could significantly alter the behavior of these landslides in the coming decades. Also, estimating the ages of these previously failed landslides will provide insights into the approximate timing of the failures and climatic conditions under which these landslides were initiated. It’s also important to note that the northern Rio Grande Rift is seismically active42,43,44. Seismic activity is believed to have triggered landslides in the past, especially when combined with wet antecedent conditions1,2,45,46. Therefore, to minimize modern-day landslide hazards it is important to implement comprehensive monitoring of these slow-moving landslides, especially in light of the Embudo, Black Mesa, and Santa Clara faults that run through or adjacent to Black Mesa.

Fig. 8
figure 8

Map showing surface roughness of the landslides. Map generated using ArcGIS Pro version 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Full size image
Fig. 9
figure 9

Map A shows the qualitative interaction between the average surface roughness and the average absolute velocity (velocity magnitude) derived from InSAR analysis on a 250 × 250 m grid space. Graph B is a scatter plot between average surface roughness value and average absolute velocity (velocity magnitude) of PS points in each grid space. Graph C is a scatter plot between surface roughness and absolute velocity (velocity magnitude) of all the study area’s PS points. Map generated using ArcGIS Pro version 3.3.0 (https://www.esri.com/en-us/arcgis/products/arcgis-pro/overview).

Full size image

Conclusion

Understanding slope instability and the dynamics of slow-moving landslides is critical for fundamental research and practical applications in landslide hazard assessment and mitigation. Landslides repeatedly dammed the Rio Grande at White Rock Canyon in the late Pleistocene.  A similar landslide complex exists along the eastern slope of Black Mesa, 40 km upstream, though its age remains undetermined. This study represents the first quantitative assessment of landslide movements in this region, employing InSAR time-series analysis from 2017 to 2023 to examine twelve landslides along the Rio Grande. The results reveal significant slope deformation, with some areas exhibiting over 40 mm of cumulative subsidence during the study period and annual movement rates exceeding − 6 mm/year (subsidence) and 4 mm/year (uplift). The lower portions of these landslides demonstrate greater activity, whereas the upper sections exhibit comparatively slower deformation.

The Rio Grande Gorge between Española and Taos is characterized by Toreva landslides resulting from rotational slope failures within the sedimentary rocks of the Chamita Formation, which underlies the Servilleta basalt caprock. Active erosion by the Rio Grande from the late Pleistocene to the Holocene likely contributed to the initiation of these failures by removing slope material at the base, inducing gravitational pull and tensile fracturing at the head scarps. These Toreva landslides are likely initiated from the bottom of the landslide bodies due to the removal of slope mass at their toes, which generates a gravitational pull that mobilizes both the overlying basalt and underlying sediment along curved failure planes. However, contemporary landslide activity cannot be solely attributed to fluvial erosion, as some active landslides are located at considerable distances from the modern river course. Instead, increased pore pressure from precipitation and snowmelt likely significantly influence the velocity of these slow-moving landslides.

The Spatial Kolmogorov–Smirnov (SKS) test revealed a statistically significant correlation between topographic closed depressions (TCDs) and Toreva scarps, with the highest concentration of TCDs occurring approximately 12 m from the scarps. Surface roughness analysis, representing landslide activity over timescales of \(10^2\) to \(10^4\) years indicates a distinct contrast between the upper and lower portions of the landslides. The upper sections of the landslides exhibit greater surface roughness than the lower sections due to the presence of secondary scarps formed by subsidence and the rotational displacement of semi-coherent Servilleta basalt slabs. In contrast, InSAR analysis indicates that the lower portions of the landslides are more dynamically responsive to contemporary geomechanical processes than the upper sections. This discrepancy in movement velocities between the upper and lower segments suggests that headward propagation of these landslides is driven by cumulative deformation occurring over centuries to millennia. These findings enhance our understanding of landslide processes in the region and provide critical insights for land-use planning and hazard mitigation efforts.

Material and methods

Landslide, topographic scarps, and TCD mapping

The landslides were manually delineated using a hillshade map derived from a high-resolution LiDAR DEM (\(\approx 1\,\text {m}\)). Identification was based on hummocky topography, clearly visible head scarps, and lateral scarps that helped define the boundaries. The landslide head scarps were mapped through visual interpretation of the hillshade map and previously published topographic and geologic maps. The upper portions of the landslides are characterized as Toreva block failures, with sharp head scarps, tensional fractures, and back-rotated blocks of the capping Pliocene basalt flows. By about the midpoint of the landslides, the Toreva blocks have disintegrated sufficiently and are difficult to discern. The lower portions are best characterized as rotational debris slides. Some landslide boundaries were slightly modified from previously mapped versions38, where boundaries were not distinctly demarcated. TCDs, characterized by a downward-concave surface surrounded by higher terrain, were mapped using a LiDAR DEM and the closed depression detection toolbox in ArcGIS.

Spatial association of TCDs with head scarps

The randomness in the spatial distribution between TCDs and their proximity to topographic scarps was evaluated by applying a Spatial Kolmogorov–Smirnov (SKS) test47. This non-parametric test evaluates whether the spatial distribution of the observed TCD points is independent of a covariate (proximity to head scarps) by comparing the cumulative distribution of the covariate values at the observed TCD points with the cumulative distribution expected under the assumption of complete spatial randomness (CSR). In this case, the null hypothesis states that there is no statistically significant difference between the distribution of proximity values of the observed TCDs and those occurring randomly under CSR, indicating that proximity to the head scarp does not influence the spatial distribution of the TCDs. Subsequently, the intensity of the TCDs as a function of proximity to head scarps (spatial covariate) was estimated to understand the intensity distribution of the TCDs across proximity values using the following equation:

$$\begin{aligned} \lambda (u) = \rho (Z(u)) \quad \ \end{aligned}$$
(1)

Where, \(\lambda\) is the intensity of the TCD point process (TCDs per unit area) and \(Z(u)\) is the value of covariate at the spatial location \(u\), \(\rho\) is the non-parametric function that describes the relationship between TCD intensity and proximity to head scarps (e.g.,48). To perform the SKS and intensity analyses, the central point of each TCD was derived by converting the polygons into points in ArcGIS. The spatial analysis was performed using the Spatstat package in R48.

Surface roughness as a proxy for landslide activity

Surface roughness estimated from high-resolution LiDAR DEM (\(\approx 1\,\text {m}\)) is useful for assessing various properties of a landslide, such as size, inferred age, estimated activity state, failure type, etc. It can facilitate automated mapping of individual landslides within larger complexes49,50. Active landslide areas should exhibit a higher surface roughness due to ongoing deformation of the slope material than the adjacent less active or stable terrain11,51,52,53. Likewise, in a landslide complex, dormant or inactive landslides should exhibit lower surface roughness and become smoother over time through the diffusion of surficial materials10. Studies have shown that topographic surface roughness an be used as a proxy for landslide age and activity over thousand-year scales8. Still, such a relationship may not stand valid for shorter periods. Various methods are available in the literature for computing surface roughness. This study computed surface roughness using the average standard deviation of slope (SDS) within a 15 × 15 m window in ArcGIS, using a 1m LiDAR DEM7. Features that could artificially increase surface roughness, such as gullies, stream erosion, roads, and other infrastructures, were manually removed from the landslide areas before calculating surface roughness. Subsequently, the landslides were divided into 250 × 250 m grids to understand better the spatial variation of surface roughness across different parts of the landslide complex. The average surface roughness for each grid was computed as a proxy of landslide activity, assuming that surface roughness reflects underlying geological processes over hundreds to thousands of years. This qualitative observation was then compared to the present-day topographic signatures of landslide activity, which operate on time scales from years to decades on deep-seated, slow-moving landslides of the Rio Grande Gorge.

Evaluation of the present day landslide activity

Synthetic Aperture Radar (SAR) interferometry, also known as InSAR, is a powerful active remote sensing technique capable of detecting and measuring ground deformation with high precision. InSAR can map surface displacements over large areas by analyzing the phase differences between successive SAR images taken from slightly different positions with millimeter-level accuracy. Estimating surface deformation from a single SAR interferogram developed from two SAR images can be affected by temporal decorrelation effects, local topography, and atmospheric noise54. Various techniques for time-series analysis of SAR data have been developed to overcome these limitations, allowing surface deformation to be measured over a long temporal series of SAR acquisitions. One widely used technique is Persistent Scatterer (PS) InSAR, which focuses on identifying and analyzing stable, radar-reflective points, known as persistent scatterers, that remain coherent over multiple SAR acquisitions. In this study, an InSAR time-series assessment was performed using the PS technique to quantify the present-day instability across the Black Mesa landslide complex.

101 descending (Path 158) and 114 ascending (Path 151) Sentinel-1A images, captured between February 2017 and December 2023 with VV polarization, were obtained and processed using ENVI SARScape software. This processing measured the displacement of these landslides along the satellite line of sight (LOS) and, when projected along the maximum slope, aids in illuminating the short-term landslide complexes’ dynamics. The assessment of the degree of landslide activity was conducted using the following metrics and utilizing PS data from both ascending and descending image analyses with a coherence threshold of 0.76:

  1. (i)

    Index of Activity Zoning: The Index of Activity (IA) is the ratio of active PS to the total number of PS within a given area55. A PS is classified as active if its average velocity exceeds \(\pm\,1 \, \text {standard deviation}\) of the velocity of all PS within the landslides. This classification approach is adapted and modified from Frattini et al.23, who defined a PS as active if its displacement rate was less than \(-1\, \text {standard deviation}\) of all PS within a landslide. In this study, \(\pm\,1\, \text {standard deviation}\) was used as a velocity threshold to identify the active PS points exhibiting both positive (uplift) and negative (subsidence) displacement movements. The IA value ranges from 0 to 1, where 0 indicates very low activity, and 1 indicates very high activity. IA zoning was performed by averaging the IA values of PS within \(250 \times 250 \ \text {m}\) grids, assigning a unique IA value to each grid.

  2. (ii)

    Absolute Velocity of Landslides: The average displacement values of the scatterers can be either positive or negative, indicating movements towards or away from the satellite LOS, respectively. Movement towards the satellite signifies uplift, while movement away from the satellite indicates subsidence. The absolute velocity considers the magnitude of movements irrespective of direction and is a robust measure of landslide activity. Higher absolute velocities indicate greater movement activity. The average absolute velocities were calculated within 250 × 250 m grids for active PS to identify zones within the landslides that either undergo deformation or are relatively stable.

  3. (iii)

    Correlation between Absolute Velocity of Landslides and Surface Roughness: Finally, a qualitative assessment of the spatial correlation between the absolute velocity of all PS points and the surface roughness values of the landslides was conducted within 250 × 250 m grids. This analysis aimed to identify zones where high and low values of velocity magnitude and surface roughness interact, highlighting the pattern of interaction between geological processes acting on the landslides over thousand-year scales (surface roughness) versus decade scales (InSAR time series).