Assessing the effect of lithological setting, block characteristic and slope topography on the runout length of rockfalls in the Alps and on the La Réunion island

. In high mountain regions, rockfalls are common processes, which transport different volumes of material and therefore endanger populated areas and infrastructure facilities. In four study areas within different lithological settings, LiDAR (light detection and ranging) data were acquired for a morphometric analysis of block sizes, block shapes and talus 15 cone characteristics. Based on these high-resolution terrestrial laser scanning (TLS) data, the three axes of every block larger than 0.5 m in the referenced point cloud were measured. Block sizes and shapes are used to investigate them in the context of runout distances and to analyse the spatial distribution of blocks on the talus cone. We also investigate the influence of terrain parameters such as slope inclination, roughness and profile curvature (longitudinal profiles). Our study shows that the relation of block size within different lithological settings on runout length is complex, because we can neither confirm nor reject the 20 theory of gravitational sorting. We also found that the block shape (axial ratio) does not have a simple influence on runout length, as it plays the role of a moderating parameter in two study sites (Gampenalm: GA, Dreitorspitze: DTS) while we could not confirm this for Piton de la Fournaise (PF) and Zwieselbach valley (ZBT). The derived roughness values show a clear difference between the four study sites. This also applies for the parameter of slope inclination and longitudinal profiles. properties, among other influencing factors (Haas et al., 2012, Fityus et al., 2013). Some studies described the deposition of boulders on the talus cone as gravitational sorting, where larger blocks are deposited in the proximal part of the talus slope and smaller blocks at the upper part (e.g. Statham, 1973, Whitehouse and McSaveney, 1983, Kotarba and Strömquist, 1984, White, 1981, Jomelli and Francou, 2000, Sanders et al., 2009, Messenzehl and Dikau, 2017, Popescu et al., 2017, Kenner, 2019). The sorting of the blocks on the slope is a key factor for the roughness component of talus slopes and 65 3 thus on the runout length of following rockfall events (Hungr and Evans, 1988), indicating a potential feedback loop in the formation of talus landforms. The aim of this study is to carry out a comparative investigation of the morphometric properties and runout distances of rockfall fragments in mountain regions within different lithological settings. We selected four sites with different lithological conditions and different rockfall activity. For the two study sites PF and ZBT the blocks cannot be assigned to one single rockfall. Whereas 70 the blocks of the other two study sites GA and DTS can be assigned to a rockfall event. Due to lithological differences of the cliffs, we expected different statistical distributions of block sizes and block shapes on the talus slopes. The study is conducted using high-resolution digital terrain models (DTMs) created from TLS surveys. Based on the research of Haas et al. (2012), we determined different block properties (size and shape) and analysed them in the context of runout distances and talus morphology. 80 the roughness the same direction Especially the PF, where the highest roughness the lower end of the The largest spatial dispersion of the roughness at GA (median = 6.92), where large blocks and thus high roughness seems to be distributed over the whole talus cone. To a slightly lesser extent this also applies to DTS (median = 4.86). These high roughness values can influence the deposition of rock fragments as they act like a natural obstacle 250 and can influence the trajectories of future rockfall events. This deposition and accumulation of material can be explained by the straight and steep slope with a minor basal concavity. At the same time, the highest roughness values can be found at GA and DTS, the lowest at ZBT (median = 4.75) and PF (median = 4.67), which can be associated with the different block volume distribution. The analyses show that surface complexity and roughness are not the only parameters which are important for gravitational sorting to take effect.

Due to the economic and societal importance, especially in the context of global change, many studies exist about rockfall processes, focusing on the modelling of runout trajectories and the prediction of rockfall events (e.g. Kirkby and Statham, 40 1975, Meißl, 1998, Agliardi and Crosta, 2003, Dorren 2003, Copons et al., 2009, Jaboyedoff and Labiouse, 2011, Frattini et al., 2012, Nappi et al., 2013, Wichmann, 2017, Volkwein et al., 2018, Caviezel et al., 2019, as well as on the measurement of rockfall activity by seismic monitoring (e.g. Vilajosana et al., 2008, Hibert et al., 2011, Farin et al., 2015, Dietze et al., 2017a, Dietze et al., 2017b, Durand et al., 2018, Feng et al., 2019. Most of these studies use LiDAR (light detection and ranging), techniques such as airborne laser scanning (ALS) as well as terrestrial laser scanning (TLS) to improve the understanding of 45 this geomorphic process (e.g. Jaboyedoff et al., 2007, Abellán et al., 2011, Haas et al., 2012, Heckmann et al., 2012, Royán et al., 2014, Strunden et al., 2015, Sala et al., 2019. In recent times Structure-from-Motion (SfM) is also increasingly used for such kind of studies (e.g. Kromer et al. 2019, Vanneschi et al., 2019, Guerin et al., 2020. In the context of hazard assessment, but also for geomorphological models, not only the transported volumes, but also the analysis of the maximum runout distance of blocks plays an important role especially in populated mountain regions (e.g. 50 Jaboyedoff and Labiouse, 2011, Volkwein et al., 2011, Lambert et al., 2013, Caviezel et al., 2019. Factors influencing the runout distance and the track of blocks include properties of the blocks itself (size and shape) and the characteristics of the topographical conditions of the talus cone, including e.g. roughness (e.g. Meißl, 1998, Frattini et al., 2012. The size and shape of the blocks are considered to be a controlling factor for the travel distance (e.g. Pfeiffer andBowen, 1989, Leine et al., 2014), as they influence the inertia moment of a block and as a consequence its runout trajectory (e.g. Frattini et al., 2012). 55 The influence of block shape, block size and slope topography on the runout distance was investigated only by a rare number of studies (Azzoni and de Freitas, 1995, Haas et al., 2012, Fityus et al., 2013 and mainly in case studies or tests under laboratory conditions (Okura et al., 2000, Glover et al., 2015, Wang et al., 2018. Haas et al. (2012) already stated in their case study that the influence of the lithology on block shape and block size must be investigated with a broader view. This could be done e.g. in areas with different lithological settings, as the size and shape of rock fragments are determined by their 60 lithological properties, among other influencing factors (Haas et al., 2012, Fityus et al., 2013. Some studies described the deposition of boulders on the talus cone as gravitational sorting, where larger blocks are deposited in the proximal part of the talus slope and smaller blocks at the upper part (e.g. Statham, 1973, Whitehouse and McSaveney, 1983, Kotarba and Strömquist, 1984, White, 1981, Jomelli and Francou, 2000, Sanders et al., 2009, Messenzehl and Dikau, 2017, Popescu et al., 2017, Kenner, 2019. The sorting of the blocks on the slope is a key factor for the roughness component of talus slopes and 65 https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. thus on the runout length of following rockfall events (Hungr and Evans, 1988), indicating a potential feedback loop in the formation of talus landforms.
The aim of this study is to carry out a comparative investigation of the morphometric properties and runout distances of rockfall fragments in mountain regions within different lithological settings. We selected four sites with different lithological conditions and different rockfall activity. For the two study sites PF and ZBT the blocks cannot be assigned to one single rockfall. Whereas 70 the blocks of the other two study sites GA and DTS can be assigned to a rockfall event. Due to lithological differences of the cliffs, we expected different statistical distributions of block sizes and block shapes on the talus slopes. The study is conducted using high-resolution digital terrain models (DTMs) created from TLS surveys. Based on the research of Haas et al. (2012), we determined different block properties (size and shape) and analysed them in the context of runout distances and talus morphology. 75

Study Sites
Four areas in high mountainous regions were selected for this investigation. Three of these areas are situated in the Alps ( Fig.   1), one area is located on the island of La Réunion (Fig. 2). The areas differ mainly with regard to the lithological conditions. All areas are characterized by a recent rockfall activity and a clearly distinguishable rock face with an associated scree slope.
A further criterion for the selection of the area was that both the rock faces and the talus cones were clearly and completely 80 visible to ensure a complete and dense LiDAR acquisition.
https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.    (Table 1). The GA area is located in the Dolomites and is dominated by the thick banked/untreated Rosengarten Dolomite (c.f. Haas et al., 2012), the DTS is dominated by the thick banked/untreated Wetterstein limestone. In both areas, major rockfall events occurred in recent years.
The Zwieselbach valley (ZBT) in the Stubaier Alps is located in the area of the crystalline Central Alps and is characterized by slated gneiss and metamorphic granites. Major rockfall events during the last years are not known, but the deposits on the 100 talus cone show indicate rockfall activity. The test site Piton de la Fournaise (PF, Dolomieu crater) on La Réunion is the only area outside the Alps and is located in the 105 Indian Ocean, east of Africa, but politically it belongs to France as an overseas department. PF is one of the most active volcanoes in the world (e.g. Peltier et al., 2009a) with an average of one eruption every 5.3 months since 2014 (e.g. Derrien et al., 2018). Due to a summit collapse during an eruption in 2007 (e.g. Peltier et al., 2009b) a 340 m deep caldera was formed (e.g. Staudacher et al., 2016) on an area of 1100 x 800 m (e.g. Urai et al., 2007). On this volcano, the composition of the lava is mainly bimodal with a combination of aphyric basalts and olivine rich basalts (e.g. Peltier et al., 2009a, Lénat et al., 2012. 110 Due to the high tectonic stress (e.g. Merle et al., 2010, Staudacher et al., 2016, the high volcanic activity that generate https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. deformation and seismic activity (e.g. Sens-Schönfelder et al., 2014 and the layering of different lava flows, the rim is very unstable and thus prone to high rockfall dynamics. This area certainly differs most clearly from all other studied areas. Besides the volcanic rocks, both the cliff and the talus cones are very young landforms, as the geomorphic forming started right after the emergence of the caldera in 2007. Further differences are the high deformation and seismic 115 activity and the extremely high precipitation. Both factors certainly play an important role for the rockfall activity (daily rockfall activity), but should not have any influence on the runout lengths of single blocks, so that for the present investigations primarily the differences in lithology and in the case of PF the age have to be considered.

Data acquisition and processing (TLS) 120
The data of all study sites have been acquired with a terrestrial 3D long range laser scanner. Two systems were used: the Riegl LMS-Z420i and the Riegl VZ-4000. Both scanner work on the same principle (time of flight), but due to laser configurations, the scanning distance of the VZ-4000 is four times longer (4000 m compared to 1000 m). Both systems provide colour information due to integrated camera systems in order to colorize the point clouds. All important technical information of both devices is listed in Table 2. 125 Due to the special conditions of the Dolomieu crater (longer distance between scanner and target, poor reflectance of the volcanic material) we used the VZ-4000 for this study site, all other test sites were surveyed using the LMS-Z420i. To minimize shadowing effects, several scan positions were necessary at each site, which had to be referenced using manual adjustment and ICP algorithms (Table 3), which are implemented in the software RiScan Pro (v2.2.1 www.riegl.com). After the referencing, the data were exported as ASCII files containing x, y, and z coordinates as well as RGB values for 135 further analysis in SAGA GIS/LIS (Conrad et al., 2015; Laser Information System LIS: www.laserdata.at). Based on the point https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. clouds we created digital terrain models (DTMs) for all test sites with a raster resolution of 0.75 m using the lowest z value. Table 3 provides information about each study site including the vertical and horizontal scan resolution, the referencing precision, number of points in the raw data set and the point density (points/m²).

Determination of block size, block shape and runout length
According to the workflow of Haas et al. (2012) we measured the dimensions of the three axes (a, b and c) of single boulders for every study area (PF: n=255, GA: n=618, ZBT: n=65, DTS: n=182) from the upper to the lower parts of the talus cones. 145 Based on the coloured LiDAR point clouds every block larger than c. 0.5 m (longest axis) was manually measured in the software RiScan Pro. Figure 3 shows an idealized sketch of a boulder with the three measured axes.

150
By using the measured three axes, the volume (Eq. 1) and the block shape (axial ratio) (Eq. 2) can be approximated. We followed the work of Haas et al. (2012), which used the formula of Valeton (1955) as an indicator of the block shape, where values of 1 means more or less a round shape and values of >>1 more or less an elongated or platy shape.
For the calculation, the parameter of the axis b must be set to 1 (cf. Valeton, 1955).
The Euclidean distance of each measured boulder to the detachment zone was determined to obtain the runout length (DTS, 160 GA). Since the exact detachment zone could not be determined at PF and ZBT, we measured the Euclidean distance from the beginning of the transition between cliff and talus cone to each boulder instead. To compare the runout distances between the study areas, we normalized the runout lengths for each talus cone to the interval [0,1].

Morphometric slope properties 165
Based on the high-resolution DTMs of the study sites we performed a spatial analysis of the talus cones including all morphometric properties with a presumed influence on the deposition (e.g. Wang and Lee, 2010, Frattini et al., 2012 and runout distances of rock fall boulders (e.g. Glover et al., 2015): slope inclination, surface roughness (debris texture) and profile curvature. The slope inclinations were derived based on the DTMs according to Zevenbergen and Thorne (1987). The surface roughness was derived after the approach of Frankel and Dolan (2007). They defined the surface roughness 170 as the standard deviation of slope (SDS), where is the width of the moving windows (number of cells), is the slope of the -th cell and the mean slope within 175 the moving window (Eq. 3; Frankel andDolan, 2007, Berti et al., 2013). We set N to 9 (a 3x3 cell neighbourhood) corresponding to an area of 5.1 m². This results in a spatial distributed roughness map for all talus cones (Grohmann et al., 2010) where the roughness of the slope can also serve as a proxy for the particle size distribution on the talus cone.
Additionally, we created three longitudinal swath profiles (width=10 m) from the highest part of the cone to the distal boundary in order to characterize slope morphology of all four test sites (e.g. segmentation indicated by changes in slope inclination, 180 profile concavity etc, cf. Hergarten et al., 2014) and we applied a kernel density estimation (cf. Cox, 2007) to compare the distributions of slope inclination between rock faces and talus cones, and between the study sites. In contrast to the block sizes, the block shapes (Fig. 4) in all four areas show a high dispersion between round and rather elongated blocks. The most elongated blocks can be found in the ZBT (median = 2.63) area, which again can be explained by the slated structure of the gneisses. The limestones of the areas GA (median = 2.33) and DTS (median = 2.40) are very similar in both the median block shape and the dispersion of the data. The lowest values and therefore the most round shaped blocks 200 can be found in the area PF (median = 1.73) with its volcanic rocks. Whether this is due to lithology or to tectonic stresses caused by the high seismic activity and the resulting eruptive fissures cannot be finally clarified with the data of this study. With regard to the block shapes and block size distributions in the areas, however, it can be concluded that the investigated 210 areas show sufficient differences in both block sizes and shapes, and that this is a consequence of the lithological setting. This allows for a more detailed investigation of the relationships between block shape, block size and runout distance. https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.

Talus cone characteristics
In addition to the block shape and block size, the topography of the talus cones also plays an important role and has to be 215 considered for a runout analysis. Thus, we analysed the following form parameters: slope inclination, curvature/slope profiles and roughness (Fig. 5).
The average slope inclination of the talus cones lies between 28° and 36° and thus within the range for such landforms (e.g. Pérez, 1989, Pérez, 1998, Francou and Manté, 1990, Jomelli and Francou, 2000, Sanders et al., 2009, Luckman, 2013b, Popescu et al., 2017, Volkwein et al., 2018. This can provide an indication of the dependence between rock material and slope 220 inclination of gravel or blocky material and thus supports the findings of other studies. According to Gerber (1974), the slope inclination is defined by the shape and roughness of the deposited blocks and lies 225 between 18°-43°. In his study he measured slope inclination of 33° for Gneisses and metamorphic granites (ZBT) and 32° for limestone formations (GA, DTS) (Gerber, 1974). Serrano et al., (2019) described in their study slope inclination for limestone formations between 32-36°. Knoblich (1975) gives values of 28°-43°. Yamamoto et al., (2005) conclude that basaltic material cannot be deposited on slopes of more than 33°. The relatively higher average slope inclination at the PF is also interesting.
Here, rock surface conditions could also play an important role, since volcanic rocks in particular are characterized by a high 230 micro-roughness of the rock surface and surface friction can govern the natural slope angle of debris material. This is supported by the distributions of the slope inclinations (c.f. Fig. 5D), where the data scatter is lowest in the PF area. Here the distribution is very peaked and differs clearly from the distributions of the other three areas. The shape of the slope profiles (Fig. 5A) also points in a similar direction. The talus cone of PF follows a straight line over the entire length of the slope, whereas the other cones show a slight convexity in the upper and middle slope and a basal concavity at the end of the slope, which is in good 235 agreement with the works of Kotarba and Strömquist (1984), Luckman (2013b), and Popescu et al. (2017). Nevertheless, it is very likely that the talus cone of the PF represents a pure talus cone and is reshaped since then by persistent rockfalls, whereas the talus cones of the other areas represent much older forms where different types of geomorphic processes occur.
The derived roughness shows a clear difference between the study sites (Fig. 5C, 6A). The roughness in PF seems to be continuously increasing in the downslope direction (with the lowest roughness at the upper slope and the highest roughness at 240 the end). On the other talus cones it is noticeable that the highest roughness can be found at the upper slope (somewhat less pronounced in ZBT), then fall off and rise again towards the end of the slope. Roughness can be seen as a factor of the runout length. As seen in Fig. 6A and 7B it can be assumed that for smaller particles roughness on the slope is more decisive than for larger particles that do not get stopped by coarser material. The very straight line of the slope with a high mean slope inclination (PF) compared to the other areas play an important role. Here large blocks seem to overcome almost the full length of the 245 slope, while in the other areas large blocks seem to stop in the upper parts as well as in the lowest parts. The spatial distribution https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. of the roughness points in the same direction (Fig. 6A). Especially at the PF, where the highest roughness can be found at the lower end of the slope. The largest spatial dispersion of the roughness can be found at GA (median = 6.92), where large blocks and thus high roughness seems to be distributed over the whole talus cone. To a slightly lesser extent this also applies to DTS (median = 4.86). These high roughness values can influence the deposition of rock fragments as they act like a natural obstacle 250 and can influence the trajectories of future rockfall events. This deposition and accumulation of material can be explained by the straight and steep slope with a minor basal concavity. At the same time, the highest roughness values can be found at GA and DTS, the lowest at ZBT (median = 4.75) and PF (median = 4.67), which can be associated with the different block volume distribution. The analyses show that surface complexity and roughness are not the only parameters which are important for gravitational sorting to take effect. 255 https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.

Relationship between block size, block shape and runout length 275
In order to analyse the relationship of block volume, block shape and runout length, we calculated a Spearman rank correlation (Table 4). The results show only weak correlations between block volume and runout distance as well as weak correlations between runout length and block shape for the four test sites, which indicate no monocausal relationship.
These results are in contrast to other studies. Jomelli and Francou (2000) state that longitudinal sorting of talus cones shows an increase of block sizes downslope. In their study, Popescu et al. (2017) conclude that there is a gradual increase of boulder 280 https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. size towards the slope base. Serrano et al. (2019) showed that the distal part of a slope is defined by the accumulation of large blocks. In contrast to these studies Caine (1967) found a decrease of blocks sizes with distance downslope, but his results are statistically insignificant. Messenzehl and Dikau (2017) found, that there is a distinct downslope increase in block size and sphericity, which indicates a combination of these block characteristics govern the runout length. Thus and in accordance to Haas et al. (2012) we plotted the relative distance against log10 block volume to show the relationship for every single talus cone in more detail (Fig. 7). We combined this analysis with the axial ratio of the boulders. 290 Figure 7A shows boxplots with six different quantile classes (<q10, q10-q25, q25-q50, q50-q75, q75-q90, >q90), which correspond to the 10, 25, 50, 75 und 90 % quantiles of the log10 block volume and log10 axial ratio. To analyse the relationship of the parameters in more detail, we highlighted the 10 % (low axial ratio, most spheroidal blocks, small volume) and the 90 % (high axial ratio, least spheroidal blocks, large volume) quantile classes, which are visualized with different symbols and colours (Fig. 7B). 295 Figure 7 shows that there are some significant differences between the individual areas and that the areas can be divided into two groups with regard to runout length, block size and block shape. The first group consists of the areas GA and DTS, the second group consists of PF and ZBT. Furthermore, GA and DTS primarily include blocks of one rockfall event, while the blocks of PF and ZBT cannot be assigned to one rockfall event.

285
GA and DTS clearly show that the size of the blocks obviously does not play a major role for the runout length on the talus 300 cones (the median values do not show a significant trend), but the dispersion of the runout distances clearly increases with increasing block size. This is visible in both areas, but is more pronounced at DTS. But it is also visible that blocks with larger volumes are also having smaller runout distances (Fig. 7). The scatterplots show, in accordance with Fig. 7A, that the block size scatters strongly over the entire talus cone. The block size can therefore not be used as an explanation for the runout distance alone, which is in agreement with the Spearman rank correlation. 305 The situation is different regarding the block shapes in the two areas. Here it is clearly visible that with increase in axial ratio, the runout distance decreases. This is also quite consistent with the results of the Spearman-Rank correlation, because in these two areas a slight correlation between axial ratio and range can be stated (GA: r = -0.35, DTS: r = -0.42). However, this is https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License. also visible in Fig. 7B, which shows the roundest and longest blocks combined with the ranges: round blocks (<q10) reach larger distances than elongated blocks (>q90) (e.g. Pérez 1998). This can be seen in both areas from a relative distance of about 310 0.7 on the talus cone. We observed that the parameter axial ratio acts like a moderating parameter with regard to the deposition of blocks. Pérez (1998) and Messenzehl and Dikau (2017) also conclude in their study that rounder blocks are deposited predominantly at the talus toe. As already shown by Glover et al. (2015) in their rockfall simulation, uniformly shaped blocks followed by elongated blocks achieve the largest ranges. This is due to the fact, that boulders with a low axial ratio roll over any axis and do not lose momentum. Platy blocks on the other hand reach the shortest runout distances (Glover et al., 2015). 315 It is particularly noticeable, however, that in the DTS area all very round blocks (<q10) achieve very high ranges and are not found at all in the shorter distances. This is different in the GA area. Here the majority of the round blocks can be found at the end of the slope, but also at the upper slope. This could be due to the fact that during the rockfall event the blocks collided with each other (maybe split into smaller blocks), resulting in the different block shapes being deposited in both the lower and the upper area of the talus cone (Ruiz-Carulla and Corominas, 2020). However, this could not be fully clarified in the context 320 of this analysis.
In the areas ZBT and PF the situation is clearly different, here the block sizes seem to have a recognizable influence on the runout length. Although the dispersion of the data and the Spearman-Rank correlation show that this correlation is not significant ( Table 5, PF: r = 0.15, p-value = 0.02, ZBT: r = 0.33, p-value = 0.007111), large blocks above a certain size are no longer found in the short runout lengths. However, Fig. 7 also shows that for PF, smaller and larger blocks are deposited in all 325 areas of the slope. At the same time, unlike GA and DTS, the block shape does not seem to play a role for the runout distance, which is also indicated by the Spearman rank correlation, PF: r = 0.06, ZBT: r = 0.27. Round and elongated blocks are actually found over the entire talus cone without any visible clustering (Fig. 7).

Conclusion
In this study a comparative analysis of morphometric properties and run out lengths of rockfall fragments in mountainous regions within different lithological conditions was conducted.
Our study shows a recognizable correlation between block size and block shape depending on the lithology, which results in larger blocks with higher sphericity in areas with thick banked/untreated limestones (GA, DTS), smaller and platy/elongated 345 blocks in areas with slated gneiss and metamorphic granites (ZBT) and smaller and more spherical blocks in basaltic lava material (PF).
Our analyses reveal a complex relation between block size and block shape with respect to the travel distance of deposited blocks with divergences between different lithological settings. Compared to other studies (Whitehouse and McSaveney, 1983, Jomelli and Francou, 2000, Sanders et al., 2009, Luckman, 2013a, Messenzehl and Dikau, 2017, Popescu et al., 2017 and 350 with respect to our results we can neither confirm nor reject the theory of gravitational sorting (Fig. 7), as we cannot confirm, that blocks with smaller volumes have decreasing runout distances. In fact, some of the results provide a negative relationship of block volume and runout distance, which results in increasing block volume with decreasing runout distances (Fig. 7).
As the block size does not seem to be the only control variable, the axial ratio of boulders seems to influence the deposition of rock fragments on the talus cone such that blocks with increasing sphericity have increasing runout distances (Pérez, 1998, 355 Glover et al., 2015, Messenzehl and Dikau, 2017, Volkwein et al., 2018. This is due to the fact that boulders with a low axial ratio (high sphericity) roll over any axis and do not lose momentum. In our study, we could show that spherical shaped boulders are deposited not only in the distal part of the talus cone but also in the proximal area (Fig. 7). For the study area GA and DTS we observed that the parameter axial ratio acts like a moderating parameter with regard to the deposition of blocks.
Accordingly, most spheroidal blocks have increasing runout distances and least spheroidal blocks have decreasing runout 360 distances. In case of PF and ZBT we cannot confirm the hypothesis of a moderating role of the axial ratio. Whether this is due to the different lithological setting or the fact that GA as well as in DTS primarily included blocks of one rockfall event, while the blocks in PF and ZBT represent several individual and presumably temporally unrelated block falls, cannot finally be clarified and has to be tested in the future.
Regarding the talus cone characteristics, PF has the highest slope inclinations, followed by GA, DTS and ZBT with the lowest 365 values (Fig. 5D, 6B). When considering the talus cone characteristics, it must be taken into account that PF represents a pure slope, which is reshaped by persistent rockfalls. The other areas are older slopes that are influenced by different types of geomorphic processes besides rockfalls. Apart from the slope inclination, differences between the longitudinal profiles (straight slope of the PF and more slight convexity in the upper and middle slope and a basal concavity at the end of the slope of the other areas) should be mentioned (Fig. 5A). The roughness values can be associated with the different block volume 370 distribution and can be seen as a factor of the runout length. The highest roughness values as well as block volumes are found https://doi.org/10.5194/nhess-2020-322 Preprint. Discussion started: 6 October 2020 c Author(s) 2020. CC BY 4.0 License.
in the GA area and in DTS (Fig. 5C, 6A). In contrast, the smallest values have PF and ZBT (Fig. 5C, 6A). Roughness influences the deposition of blocks and their trajectories as it acts like a natural obstacle.
As the sample of different source rocks is quite small (four test sites), this must be seen as an indication of a relationship between lithological conditions and block size/block shape, but should be verified by further studies with larger samples. We 375 are quite sure, that such a study can be done on the base of ALS data or photogrammetric elevation models based on aerial photographs, where point densities are high enough to resolve blocks of a certain size, which are nowadays available for the whole Alps and mountain ranges worldwide.

Data availability.
The data used in this study are accessible upon request by contacting Kerstin Wegner (KWegner@ku.de). 380

Author contributions.
KW, FH and TH designed the conceptual idea of the manuscript. KW and FH collected the terrestrial laser scanning data on La Réunion. FH collected the TLS data of DTS, ZBT and GA. TH contributed to statistical analysis of data. VD, NV and PK acquired the dGPS data on La Réunion. NV, PK and AP contributed to fieldwork on La Réunion. AM contributed to the financial support of the travel costs to La Réunion. MB contributed to the acquisition of the terrestrial laser scanner. KW wrote 385 the manuscript with discussions and improvements from all co-authors.

Competing interests.
The authors declare that they have no conflict of interest.