In order to identify and map long-runout landslides with longitudinal ridges in Iceland, we used available high-resolution satellite/aerial imagery from Google Earth and the Icelandic Map website (https://www.map.is, last access: 31 January 2023). In ArcGIS, we used digital elevation models (DEMs) provided by ArcticDEM (Porter et al., 2018; 2 m px-1 resolution) and their hillshade models to extrapolate landslide morphometric data (elevation drop and horizontal length; H and L, respectively, in Table S1 in the Supplement). In combination with the image data, the hillshade was used to identify the locations of the extreme ends of the headscarp and toe of the deposit so their elevation values could be extracted from the DEM.

We acquired drone imagery of one Icelandic landslide near the town of Dalvík (I-landslide ID 47 in Table S1). We used 10 ground control points (GCPs) in order to correctly georeference the drone images (Table S2) acquired by differential GPS and marked by 0.8×0.8 m square orange targets easily visible in the images. We obtained the latitude and longitude coordinates (WGS84) and the height above the ellipsoid of 10 locations, of which 8 were on the landslide deposit and 2 were on the slope adjacent the distal frontal edge of the deposit (Fig. 4), using a Leica dGPS GS09. We occupied the points for >5 min and post-processed the data using the permanent GPS station in Akureyri (AKUR run by LMI Landmælingar Íslands – National Land Survey of Iceland) 38 km away from the site. The resulting positions had errors much smaller than the error produced by inherent uncertainty in identifying the control points in the drone images (the worst position had a standard deviation of 2.4 cm in vertical position, yet this was typically 3 mm). We calculated the elevation above the geoid (i.e. elevation above sea level) using an online geoid calculator (GeoidEval). We built a digital elevation model (4.65 cm px-1 resolution) and an orthoimage (2.38 cm px-1 resolution) of the Dalvík landslide using Agisoft Metashape (dataset provided in Magnarini et al., 2023).

We made use of the landslide inventory provided by Crosta et al. (2018) to select Martian landslides with longitudinal ridges that have a comparable length to Icelandic long-runout landslides with longitudinal ridges. We selected Martian landslides that are shorter than 5 km long (Table S3), which is double the length of the longest Icelandic landslide considered in this study (I-landslide ID 103; L=2629 m). We added additional landslides with longitudinal ridges that are not found in the Crosta et al. (2018) inventory, which were found using high-resolution satellite imagery acquired by the ConTeXt (CTX) camera (6 m px-1 resolution; Malin et al., 2007) and, where available, the High-Resolution Imaging Science Experiment (HiRISE) camera (nominal resolution 25 cm px-1; McEwen et al., 2007) on board the NASA Mars Reconnaissance Orbiter (MRO). In total, we selected 112 landslides. As Martian landslides can be up to 70 km long (Lucchitta, 1979), those selected for this work will be also referred to as “small-scale”, as done in the literature (Guimpier et al., 2021).

In ArcGIS, we calculated the elevation drop (H) and measured the horizontal length (L) using the HRSC MOLA Blended global DEM at 200 m px-1 resolution (Fergason et al., 2018). CTX images were used to identify the top of the headscarp and toe of the deposit so that their elevation values could be extracted from the DEM. We generated digital elevation models and orthoimages of four landslides (M-landslide ID 31; M-landslide ID 42; M-landslide ID 46; M-landslide ID 47) using CTX (Malin, 2007; Malin et al., 2007) image pairs (Table S4). CTX-derived DEMs were generated using the United States Geological Survey (USGS) Integrated Software for Imagers and Spectrometers (ISIS) software and commercial image analysis software SOCET SET (Kirk et al., 2008). DEMs were vertically and horizontally controlled to Mars Orbital Laser Altimeter topographic data (Smith et al., 1999; Zuber et al., 1992). The resolution of CTX DEMs and orthoimages is 20 and 6 m px-1, respectively. We estimated the vertical precision of the CTX-derived DEMs using the established method of Okubo (2010) (Table S4).

In this study, we investigated a landslide in the field located 1 km northwest of the town of Dalvík in the Tröllaskagi peninsula (Fig. 4). The Dalvík landslide (I-landslide ID 47) is a good example of a well-preserved long-runout landslide with longitudinal ridges in Iceland. The landslide has a horizontal runout (L) of 1.210 km and a vertical drop (H) of 0.480 km (Fig. 1), resulting in a H/L ratio of 0.396.

Similar to other long-runout landslides, the edges of the deposit are steep (on average about 30° slope) and about 15–25 m high (Fig. 5d). The bedrock in the headscarp exposure shows typical alternation of Tertiary lava layers and red palaeosols (Fig. 5a and b). The landslide deposit sits on a sedimentary cover that has an undulating surface appearance (Fig. 5c) and could be due to previous mass movements or glacial deposits. The zone of depletion is characterised by a Toreva block (about 200 m long, 450 m wide, and 80 m high), while the zone of accumulation is composed of a debris apron about 750 m long. Boulders decimetres to metres in size are found on the surface of the deposit. However, size sorting of boulders from proximal to distal area of the landslide seems not to occur.

The debris apron is characterised by longitudinal ridges which can be identified from both aerial/satellite images and ground photos. The longitudinal ridges occur across the entire length of the debris apron (Fig. 6). At some locations, longitudinal ridges bifurcate (e.g. locations marked with a star in Fig. 6; red arrowheads in Fig. S2a and b), forming two ridges commonly narrower than the parent ridge. Consequently, the number of longitudinal ridges increases from the proximal to the distal area of the apron. Transverse ridges appear only on the SW side of the deposit.

In addition to longitudinal ridges, the Dalvík landslide exhibits other linear structures superposed on, and broadly transverse to, the longitudinal ridges (Fig. 7). These linear structures appear to have en echelon configuration. The en echelon structures appear to be oblique to the longitudinal direction of the ridges. Towards the frontal edge of the deposit, the orientation of these linear structures changes and gradually becomes perpendicular to the extension of the ridges, which corresponds to the direction of movement of the landslide.

We compiled the first catalogue of Icelandic long-runout landslides with longitudinal ridges, which comprises of 129 of these landforms (Table S1). We included only landslides that have distinct longitudinal ridges, and we omit those landslides for which the presence of longitudinal ridges was uncertain. Ridges were identified by their horizontal continuity in the downslope direction, and their presence is evaluated using a top-view observation of the landslide deposit.

The horizontal length of the Icelandic landslides spans from 0.287 km (I-landslide ID 127) to 2.629 km (I-landslide ID 103). The median and mean values of length are 1037 and 1150 m, respectively. The median and mean values of elevation drops are 317 and 337 m, respectively. The H/L ratio of this set of Icelandic long-runout landslides ranges from 0.149 (I-landslide ID 18) to 0.58 (I-landslide ID 42). The statistics of the morphometry of Icelandic landslides are summarised in Figs. 12 and 13.

Some of the landslides have their headscarps at the top of cliffs and mountains (e.g. I-landslide ID 8, ID 9 in Fig. 3), whereas other cases develop from the mid-/lower part of slopes (e.g. I-landslide ID 74 and ID 88 in Fig. 3). All the landslide deposits are characterised by steep edges, typically about 20 m high. About 25 % of the landslides have a Toreva block. Another characteristic found in some landslides is the presence of lateral levees that develop continuously from the depletion zone. In about the 80 % of the cases, the deposit does not entirely reach the slope break at the base of the slope. The landslides often rest on the inclined surface that connects the slope to the bottom of the valley (e.g. I-landslide IDs 45, 69, and 74, ID 115 in Fig. 3, and the Dalvík landslide). Such connecting surfaces have slopes of about 8–15°. They correspond to sedimentary units, whose origin cannot be established from remote sensing only (in Fig. 5, such connecting surface is labelled “undulating pre-landslide surface”). A river gorge near the Dalvík landslide exposes outcrops showing the sedimentary origin of the inclined surface over which the landslide deposited (Fig. 5c). In the case of absence of the connecting inclined surface, landslides extend beyond the slope break at the base of the slope (e.g. I-landslide ID 8 and ID 9 in Fig. 3).

In this work, we compiled a catalogue of Martian long-runout landslides shorter than 5 km long that exhibit longitudinal ridges (Table S3; Figs. 8–10) and are thus comparable to the length of Icelandic long-runout landslides with longitudinal ridges. We only included long-runout landslides that have distinct longitudinal ridges, and we did not consider landslides where the presence of longitudinal ridges is uncertain. Moreover, we have excluded deposits with longitudinal ridges that do not have matching headscarps. The catalogue that we compiled comprises of 112 of these landforms, which represent the 3 % of the latest Martian landslide database (3118 entries) provided by Crosta et al. (2018).

The horizontal length of the Martian landslides included in the catalogue spans from 0.508 km (M-landslide ID 3) to 4.902 km (M-landslide ID 43). The median and mean values of length are 1930 and 1803 m, respectively. The median and mean values of elevation drops are 527 and 536 m, respectively. The H/L ratio of this set of Martian long-runout landslides ranges from 0.076 (M-landslide ID 14) to 0.468 (M-landslide ID 12). The statistics of the morphometry of small Martian landslides is summarised in Figs. 12 and 13.

Small-scale Martian long-runout landslides with longitudinal ridges are found between 38° N and 34° S. The largest concentration is found in the chaotic terrains east of Valles Marineris and in the channels and valleys that flow into Chryse Planitia. Other major clusters are found in Noctis Labyrinthus and north of Terra Cimmeria. Some of the landslides have their headscarps at the top of cliffs and mountains (e.g. in Figs. 9a, b and 10a, g), whereas other cases develop from the lower part of slopes (e.g. in Figs. 9c and 10b, c, e, f, h). All the landslide deposits are characterised by steep edges, typically about 20–50 m high. Only about 20 % of these landslides have a Toreva block. Another typical characteristic is the presence of lateral levees that develop continuously from the depletion zone (e.g. Fig. 11a). The debris aprons rest on sub-horizontal surfaces, which are either valley or crater floors. In only five cases (M-landslide ID 67, 71, 79, 102,and  112) does the deposit not entirely reach the slope break at the base of the slope (e.g. in Fig. 10b, e and f).

Longitudinal ridges are visible in both CTX and HiRISE images across the entire length of the debris aprons. Similar to what we observed at the Dalvík landslide, longitudinal ridges split, generating two ridges from a parent ridge (Fig. 11). Unfortunately, the resolution of the CTX-derived DEMs (20 m px-1) is not able to resolve the topography of longitudinal ridges; and no HiRISE image pairs are available to generate higher-resolution DEMs of the landslides discussed in this work. Therefore, a topographic comparison between longitudinal ridges of the Martian and the Dalvík landslide could not be conducted.

The gigantic, long-runout landslides on Mars have attracted much interest, giving rise to a commonly accepted belief that Martian landslides are an order of magnitude larger than terrestrial landslides (e.g. Brunetti et al., 2014; Legros, 2002; Lucchitta, 1978, 1979; McEwen, 1989). However, the availability of higher-resolution orbital imagery has shown the existence of Martian long-runout landslides that are smaller and similar in size to terrestrial long-runout landslides (Crosta et al., 2018; Guimpier et al., 2021, 2022).

The mobility of long-runout landslides, expressed as H/L ratio, increases with increasing volume (Heim, 1932; McEwen, 1989; Legros, 2002). When the largest landslides are taken into account, terrestrial and Martian long-runout landslides seem to show distinctive trends, suggesting that Martian long-runout landslides are less mobile than terrestrial long-runout landslides; that is, for long-runout landslides with equal volume, the terrestrial H/L ratio is lower than the Martian one. However, such a trend becomes less obvious when larger datasets that include smaller landslides are considered (Crosta et al., 2018), and the large scatter of the data points in the H/L vs. L plot suggests that other parameters affect the mobility of long-runout landslides.

When volume values are not available, length (L) is used as a proxy for the volume of landslides (Legros, 2002; Singer et al., 2012; Lucas et al., 2014; Johnson and Campbell, 2017; Schmidt et al., 2017; Beddingfield et al., 2020; Johnson and Sori, 2020). In this work, we used L as a proxy for landslide volume (Fig. 13b).

Figure 13a is a visual representation of the H/L ratio that characterises the populations of the landslides that we have analysed in this study. The plot shows that the H/L ratio range that characterises the two landslide populations overlaps and that there is no distinct trend in the H/L ratio of the Icelandic and small-scale Martian landslides. When instead plotted against elevation drop (Fig. 13c), the H/L ratio shows a correlation that is increasing with increasing fall height. This trend was shown to also exist for large landslides on Earth, Mars, and Iapetus by Johnson and Campbell (2017) and confirmed by their numerical simulations. The plot in Fig. 13a also seems to show that there are two distinct lower boundaries for the Martian and Icelandic landslides.

In order to isolate and test the effects of the length and elevation drop on the mobility, we locked the landslide populations to equivalent maximum length. Therefore, we restricted the selected Martian landslides to a smaller subset (Nsubgroup=51), whose maximum length matches that of the Icelandic landslides (L<2800 m). It is interesting that these two populations with equivalent L (that is, with assumed equivalent volume) show similar mobility (Icelandic average H/L ratio = 0.303; Martian average H/L ratio = 0.292), thus supporting the absence of distinctive trends of mobility in small-scale terrestrial and Martian long-runout landslides. Additionally, the Martian landslide subgroup shows cases that are characterised by higher-elevation drops (Icelandic max elevation drop = 810 m; Martian max elevation drop = 1071 m). As explained in Melosh (2011), the lower surface gravity of Mars means that for the same material strength Mars can support proportionally higher topography than Earth. These results are again in agreement with Johnson and Campbell (2017).

Lateral spreading of landslide deposits has been considered an important factor controlling the behaviour of longitudinal ridges and the thickness of a landslide deposit (Magnarini et al., 2021b). Unconfined long-runout landslides tend to spread laterally, causing the thinning of the deposit with distance and hence divergence of the longitudinal ridges. However, the existence of a scaling relationship between the thickness of the landslide and the spacing between the ridges (i.e. the distance between ridges ranges between 2 and 3 times the thickness of the deposit; Magnarini et al., 2019, 2021a, b) implies that as the landslide thins while laterally spreading, and then the distance between the ridges has to decrease accordingly. Therefore, the scaling relationship is maintained by the splitting of parent ridges and also by the appearance of new smaller ridges between spreading longitudinal ridges (Magnarini et al., 2019).

We observed longitudinal ridges bifurcating, thus splitting and developing two ridges from the parent ridge, in both the Dalvík landslide (Figs. 6 and S2a, b) and small-scale Martian landslides (Fig. 11). The resulting new ridges are less wide. This process leads to an increase in the number of longitudinal ridges with distance. Similar behaviour has been described for the 63 km long Martian Coprates Labes landslide, in Valles Marineris, by Magnarini et al. (2019) (Fig. S2c and d). Therefore, the occurrence of ridge splitting seems to be scale-independent. However, in the Dalvík landslide, we do not observe the appearance of new, smaller ridges between two existing ridges, which is observed in the small-scale Martian landslide ID 109 (Fig. 11b). Similar ridge behaviour is observed in two other Martian landslides that are not included in the current selection, namely the landslide shown in Fig. 11c and the Coprates Labes landslide (Magnarini et al., 2019).

The fact that the Dalvík landslide shows only the splitting of parent longitudinal ridges into two smaller ridges and does not show the appearance of new ridges between two existing ridges may be due to limited lateral spreading at Dalvík compared to the Martian examples. Limited lateral spreading in turn limits the thickness reduction in the deposit and limits the space for new ridges to develop. Lateral spreading may be limited in Dalvík by the basal surface over which the landslide moved and thus not providing sufficiently low friction, the rheology of the landslide prevented it from spreading further laterally, or the landslide was not large enough and thus did not have enough kinetic energy to expand further. The latter proposed explanation may have support in the fact that the landslides in which we report the appearance of new ridges (in addition to ridge splitting) are all much larger than the Dalvík landslide.

Similar to the Dalvík landslide, linear en echelon structures are seen in an 8 km long Martian landslide (e.g. Fig. 14a and b) and have also been described in the 63 km long Coprates Labes landslide by Magnarini et al. (2019) (Fig. 14c and d). Similar linear structures superposed on longitudinal ridges were also described in the Sherman Glacier landslide deposit, Alaska, as “transverse fissures” by Shreve (1966) and McSaveney (1978); in the Lamplugh landslide, Alaska, by Dufresne et al. (2019); and in the Iymek landslide, China, by Shi et al. (2022). Such linear structures have been interpreted as kinematic indicators, as they indicate the differential velocity of adjacent sections of the landslides during emplacement (e.g. Dufresne et al., 2019; Magnarini et al., 2019; McSaveney, 1978; Shreve, 1966).

In the Dalvík landslide, towards the frontal edge of the deposit, these linear structures change orientation and gradually become perpendicular to the direction of the ridges (Fig. 7), which corresponds to the direction of movement of the landslide. The observed change in the orientation of the en echelon structures may represent the expression of changing velocity during emplacement; the oblique configuration shows that different parts of the deposit were characterised by different velocities, and the transition to transverse configuration, i.e. perpendicular to the landslide movement direction, shows that different parts of the landslide progressively attained similar velocity. Given that the oblique–transverse transition occurs towards the terminus of the deposit, it may represent evidence of slowing down of the landslide. However, such a transition does not seem to take place towards the terminal part of the Martian landslide in Fig. 14a and b; therefore, either a different interpretation is required or the Martian landslide came to a halt in a different manner than the Dalvík landslide.

In this work, we showed that small-scale Icelandic and Martian long-runout landslides share similar mobility and morphological characteristics. It is reasonable to consider that the high mobility that these landslides have in common and their morphological similarities are a result of a similar mechanism of emplacement. However, the geomorphological issue of “equifinality” (i.e. similar landforms originating from different processes) is a major challenge in the interpretation of Martian landforms, as in situ data are limited to robotic exploration at few localised sites, and must be taken into consideration when linking morphology to processes (e.g. Balme et al., 2011). It follows also that interpretations based on morphological analysis may be biased due to morphological similarity (Johnson and Campbell, 2017; Magnarini et al., 2019; Johnson and Sori, 2020). With this in mind, we can, nevertheless, attempt to draw considerations about the implications of such similarities in terms of the emplacement mechanism, emplacement environment, and different gravity.

The effect of gravity on the formation of long-runout landslides on different planetary bodies is highly debated, and the matter is not settled yet. In this work, we have attempted to remove biases that populations of landslides with non-negligible size differences could carry with them. Our results do not suggest that gravity affects the mobility, as we do not see distinct trends between Martian and Icelandic landslides in our plots. Simulating terrestrial and Martian landslides, Johnson and Campbell (2017) reached similar conclusions that the lower surface gravity does not lead to significantly reduced mobility.

To our knowledge, no other region on Earth shows the same high spatial density of long-runout landslides with longitudinal ridges as Iceland does. Such unicity suggests that since the LGM (which broadly corresponds to the age of Icelandic landslides; Decaulne et al., 2016; Mercier et al., 2013, 2017), no other terrestrial setting has had conditions favourable for their development, and also preservation, to such a scale. On Mars, small-scale long-runout landslides with longitudinal ridges are not characterised by a high spatial density distribution, as in Iceland, although they tend to occur in steep-slope settings provided by chasmata, outflow channels, and chaotic terrains. However, there is evidence that the Martian record of small-scale landslides could have been largely removed. In fact, we identified several deposits almost fully obliterated by sedimentary burial (Fig. S1a and b in the Supplement). One of the best examples is found in Coprates Chasma, Valles Marineris, on Mars, where we found evidence of landslides with longitudinal ridges that have been partially buried and subsequently exhumed (Fig. S1c). The incomplete record of these landforms hampers our reconstruction of regions where conditions were favourable for their formation. In addition to the matter of where, the incomplete record prevents us from establishing when such conditions were optimal. Based on the existing age estimation of the largest long-runout landslides on Mars, long-runout landslides in Valles Marineris span from 3.5 Gya to 50 Ma (Quantin et al., 2004; Hager and Schedl, 2017); near Hydraotes Chaos, they span from 835 to 252 Ma (Molaro et al., 2024); and younger, small-scale landslides (<2 km long) have been found to be younger than 20 Myr old (Guimpier et al., 2021). According to these results, it appears that long-runout landslides on Mars have been recurring throughout much of Martian history, suggesting that the formation mechanism of long-runout landslides and their longitudinal ridges is persistent or recurrent in time.

The age difference between Icelandic and Martian landslides simply reflects different landslide populations of which we have accessible records. Therefore, it is not possible to establish whether the lack of records of Martian landslides more recent than 20 Ma is evidence that the mechanism favourable to their formation may have become unavailable on Mars.

The availability of an extensive, well-preserved, geomorphological record of long-runout landslides with longitudinal ridges in Iceland represents a unique opportunity to study their formation mechanisms and the influence of the environment on their development (both the geological and climatic contexts) at the regional scale. Reconstructing the palaeoenvironmental conditions at the time of their emplacement will help to better constrain the relationship between the occurrence of landslides with longitudinal ridges and the potential presence of ice or ice-cemented material (permafrost) during their emplacement. Not only will this represent important advancement in understanding the formation of these catastrophic landslides on Earth, but it will also provide insights into possible past Martian scenarios. However, formation mechanisms of long-runout landslides with longitudinal ridges that solely rely on climatic and environmental conditions should be carefully considered, especially on Mars, as it is well established that the planet has undergone dramatic climatic changes throughout its geological history (e.g. Laskar et al., 2004; Head et al., 2003; Bibring et al., 2006; Kite and Conway, 2024).