Rock glaciers in Daxue Shan, southeastern Tibetan Plateau: an inventory, their distribution, and their environmental controls

Rock glaciers are typical periglacial landforms. They can indicate the existence of permafrost, and can also shed light on the regional geomorphological and climatic conditions under which they may have developed. This article provides the first rock glacier inventory of Daxue Shan, southeastern Tibetan Plateau. The inventory is based on analyses of Google Earth imagery. In total, 295 rock glaciers were identified in Daxue Shan, covering a total area of 55.70 km, between the 10 altitudes of 4,300 and 4,600 m above sea level. Supported by ArcGIS and SPSS software programs, we extracted and calculated morphometric parameters of these rock glaciers, and analyzed the characteristics of their spatial distribution within Daxue Shan. Our inventory suggests that the lower altitudinal boundary for permafrost across the eight aspects of slopes observed differs significantly, and that the lower altitudinal permafrost boundary is ~ 104 m higher on western rather than eastern facing slopes. Moraine-type and talus-derived rock glaciers exhibit mean gradients that are all concentrated within the 22°–35° range. 15 However, lobate rock glaciers (27°–45°) have a higher mean gradient than tongue-shaped (22°–35°). Shady (i.e., N, NE and E) slopes appear related to the presence of moraine-type rock glaciers, whereas sunny (i.e., W, SW and S) slopes appear related to the presence of talus-derived rock glaciers. Rock glaciers in Daxue Shan are more concentrated within tertiary monzonitic granite, which is more sensitive than other lithological components to the freeze-thaw process. Continuous weathering of this substrate provides the ideal raw material for the rock glacier development. These results show that environmental controls (i.e., 20 topographical, climatic, lithological factors) greatly affect the formation and development of rock glaciers. This study provides important data for exploring the relation between maritime periglacial environments and the development of rock glaciers on the southeastern Tibetan Plateau (TP). It may also highlight the characteristics typical of rock glaciers found in a maritime setting. 2


Introduction
The term 'rock glacier' was first proposed by the American scholar Capps (1910) when the investigating Kennicott Glacier in Alaska. By definition, rock glaciers consist of perennially frozen masses of ice and debris that creep downslope under the weight of gravity (Haeberli, 1985;Barsch, 1996;Haeberli et al., 2006). As many Himalayan rock glaciers develop out of 5 moraines, it is hard to distinguish where the moraine ends and the rock glacier begins. The bodies of rock glaciers are similar to moraines, in that as their ice mass moves over a pore ice surface, they do not sort materials in relation to the thickness of the debris they contain. Statistically, rock glaciers occupy extensive areas above the tree line in the mountainous regions of the world (Haeberli, 1985). Indeed, there are ~ 73,000 rock glaciers globally (Jones et al., 2018a), with ~ 1,000 active rock glaciers in the Swiss Alps alone. The freeze-thaw process experienced by the ice masses within rock glaciers can exert a major impact 10 on the hydrological cycle (Azócar and Brenning, 2010;Jones et al., 2018a;Jones et al., 2018b), and creep of rock glaciers can significantly negatively influence any infrastructure built on top. Rock glacier research may therefore aid a more detailed and accurate understanding of the genesis of periglacial geomorphology, and of the relations between rock glaciers and their local environments.
Over the last twenty years, with the rapid development of more advanced Geographical Information System (GIS), remote 15 sensing (RS), and statistical techniques, rock glacier research has entered a new, accelerated phase. This phase has included compilations of rock glacier inventories (e.g., Bolch and Marchenko, 2009;Cremonese et al., 2011;Bolch and Gorbunov, 2014;Falaschi et al., 2014;Colucci et al., 2016;Janke et al., 2017;Jones et al., 2018a), mapping of their spatial distributions and their relations with environmental controls such as topography and climate (e.g., Chueca, 1992;Brazier et al., 1998;Brenning, 2005;Janke, 2007;Johnson et al., 2007;Kenner and Magnusson, 2017;Onaca et al., 2017;20 Jones et al., 2018b), estimations of the distribution of permafrost based on rock glaciers (e.g., Allen et al., 2008;Boeckli et al., 2012;Schmid et al., 2015;Sattler et al., 2016), and investigations of rock glacier dynamics (e.g., Haeberli et al., 2006;Liu et al., 2013;Muller et al., 2016;. However, compared with ice glaciers, rock glaciers remain poorly described 3 and infrequently studied. One reason is because they are mixtures of rock fragments of different sizes and are spectrally similar to their surroundings, and therefore cannot easily be automatically mapped from optical RS data (Brenning, 2009). In addition, both supraglacial-debris (upon the glacier) and debris along the glacier margins originate from surrounding valley rock (Jones et al., 2018b), thus their debris surfaces do not produce distinct spectral signals. As a result, it is often difficult to distinguish relict rock glaciers from inactive rock glaciers that still contain ice using optical RS imagery (Millar and Westfall, 2008;5 Kenner and Magnusson, 2017).
Rock glacier research in China has, up to this point, focused principally on the Tianshan Mountains (Cui and Zhu, 1989;Qiu, 1993;Zhu et al., 1996;, rather than the rock glaciers of Daxue Shan on the southeastern margins of the TP. This region has been, and continues to be, strongly uplifted and deformed due to the extrusion and collision of the Indian and Eurasian continental plates since the beginning of the Quaternary, and is therefore characterized by an extremely complex 10 matrix of relations between different environmental factors such as climate and geomorphology. It is therefore of particular importance to study the environmental controls on the rock glaciers here to better understand the complex geographical environment, identify potential natural hazards, and aid environmental planning and management. The purpose of this study was twofold: first, to describe and complete a systematic inventory of the previously undocumented rock glaciers in Daxue Shan; and second, to characterize their distribution and environmental controls (i.e., climatic, topographical and lithological 15 factors). Stemming from these goals, there is an analysis and discussion of the mechanisms driving the formation, development, and spatial distribution of the rock glaciers in relation to different environmental controls in this maritime setting.

Study Area
The study area is situated in China's Sichuan Province between 29.956-30.573° N and 101.477-101.974° E (Fig. 1). To the west is the uplifted eastern sector of the TP, and to the east are mountain gorges, both of which are important geographical 20 boundaries . The topography of Daxue Shan is characterized by the strong downcutting of high energy water courses, resulting in a great altitudinal range (1,349-7,321 m asl). The region's climate is relatively warm and humid, and is strongly influenced by a southwesterly monsoonal atmospheric circulation . East of Daxue Shan is 4 a subtropical monsoon climatic zone which is principally affected by the aforementioned southwesterly monsoonal atmospheric circulation, but also by a southeasterly monsoonal atmospheric circulation, all of which transport abundant precipitation to this region. West of Daxue Shan the subtropical monsoon and continental plateau climatic zones intersect, producing a cold-temperate climate, as well as abundant precipitation. Geologically, Daxue Shan is located where the Songpan, Chuandian and South China tectonic blocks intersect, with the Xianshuihe (Ganzi-Yushu) Fault passing to the northwest of 5 the region (Zhang, 2013).

Rock glacier inventory, classification and database
The availability of more powerful RS tools such as Google Earth has transformed geomorphological fieldwork and has, on the whole, made the recognition of landforms in remote and poorly accessible areas both fast and easy (Slaymaker, 2001;Bolch, 10 2004;Kaab et al., 2005). This is beneficial to the present study as Daxue Shan is remote and difficult to access, therefore we compiled an inventory of the rock glaciers in this region using high-resolution Google Earth satellite imagery (for the period October 2014-January 2017). Google Earth contains the best freely available imagery for detecting rock glaciers across large spatial areas, and it has been previously used for rock glacier identification in the Bolivian Andes (Rangecroft et al., 2014) and the Hindu Kush-Himalayan region (Schmid et al., 2015). 15 Rock glaciers are characterized by distinct flow features and structural patterns. Transversal or longitudinal flow features (ridges and furrows) are common on rock glaciers due to the deformation of their internal ice structures (Clark et al., 1998;Humlum, 2000;Haeberli et al., 2006;Berthling, 2011). Many rock glaciers also exhibit structural patterns such as steep frontal slopes and side slopes with swollen bodies. Due to the constant supply of talus or debris, the surface textures of rock glaciers are usually different from those of the surrounding slopes. Depending on their mobility and permafrost presence, rock glaciers 20 are usually divided into active, inactive, and relict types (Sattler et al., 2016). In general, the presence of ice within active/inactive rock glaciers is indicated by a steep (> 35°) frontal slope (Ikeda and Matsuoka, 2002), a well-developed flowlike morphology defined by sets of parallel and curved ridges separated by long V-shaped furrows (Barsch, 1996;Roer and 5 Nyenhuis, 2007), and an absence or sparse occurrence of vegetation (Onaca et al., 2013). Inactive rock glaciers also contain ice, but are immobile. In contrast, relict rock glaciers, characterised by surface collapse features as a result of permafrost degradation, have gentler frontal and marginal slopes and are often vegetated (Wahrhaftig and Cox, 1959;Haeberli, 1985;Scotti et al., 2013). Based on these criteria, we visually examined the Google Earth images and identified any potential rock glaciers. We then mapped the rock glaciers in the study region, using the Advanced Spaceborne Thermal Emission and 5 Reflection Radiometer (ASTER) Global Digital Elevation Model Version 2 (GDEM V2) dataset (to within a horizontal accuracy of 30 m) and the Google Earth imagery as guides, marking the geographical location of each identified rock glacier and delineating its outline.
The topographical characteristics of the rock glaciers identified in the inventory were extracted from the GDEM V2 dataset and attributed to each rock glacier in a GIS environment. These characteristics were both qualitative and quantitative and 10 included each rock glacier's geographical location (i.e., the coordinates of its center), type as determined using dynamic, genetic, and geometric criteria (moraine-talus; tongue-lobate), aspect, mean gradient of slope (°), area (km 2 ), centerline length (m), average width (m), and average altitude (m asl). A geological layer (Li et al., 1999) was added to the GIS so that the relevant class of bedrock could be attributed to each rock glacier.
Based on the main source of the mass input of debris into each rock glacier and its subsequent transport downslope, we 15 subdivided rock glaciers into two distinct categories: talus-derived rock glaciers developing below talus slopes; and morainetype rock glaciers evolving mainly from glaciogenic materials (Lilleøren and Etzelmüller, 2016;Onaca et al., 2017) (Fig. 2).
In terms of their planar geometry, the length/width ratio was used to distinguish between lobate (length/width ratio < 1) and tongue-shaped (length/width ratio > 1) rock glaciers ( Fig. 2) (Giardino and Vick, 1987;Martin, 1987;Barsch, 1996;Guglielmin and Smiraglia, 1998;Onaca et al., 2017). The overall aspect of each rock glacier was manually derived according 20 to the main direction of flow. For spatial analysis, these aspects were then recoded into 8 orientation classes.
However, due to the lack of data regarding the flow behavior of rock glaciers, it remains to be determined whether these landforms are currently active, or whether they represent inactive rock glaciers. In addition, some aspects of digitisation were challenging based on visual interpretation of remotely sensed imagery alone and thus the mapped rock glaciers are inherently associated with some spatial uncertainty (Sattler et al., 2016;Jones et al., 2018b). Consequently, some rock glaciers may not 25 6 be correctly delineated as delimitation of the upper boundary of rock glaciers through geomorphic mapping is arbitrary (Krainer and Ribis, 2012), and delineation of individual polygons where multiple rock glaciers coalesce into a single body is inherently subjective (Scotti et al., 2013;Schmid et al., 2015). Moreover, several complex landforms delineated as rock glaciers may be landslide deposits or relict rock glaciers. Therefore, future research may benefit from integration of additional data sources and further in situ observations that could be used to constrain methods for rock glacier identification using remote sensing 5 and digital elevation data. Further, use of a higher resolution DEM paired with in situ climate datasets would likely produce a more accurate representation of the distribution of the rock glaciers in Daxue Shan. Due to the limitations imposed by the 30 m spatial resolution and the uncertainties inherent in any visual identification, we may have failed to identify all the rock glaciers in the study area. As a result of these uncertainties, we chose to examine ranges of values during our statistical analyses.

Spatial and statistical analyses 10
When there is collinearity between the variables, principal components analysis (PCA) can used to determine the relationships between them (White and Copland, 2015;Ran, 2017). However, in this study, we performed the Kaiser-Mayer-Olkin (KMO) and Bartlett's tests to examine the suitability of the data for factor analysis and we found a KMO value of 0.387 (Table 1), which indicates that the original variables are not suitable for PCA because there is weak collinearity (KMO < 0.5) between them. Therefore, we retained the original variable information, which allows for convenient interpretation and calculation (not 15 too many dimensions), without dimensionality reduction.
We assigned the eight geographical and topographical variables (i.e., latitude, longitude, rock glacier (RG) area, length, width, altitude, mean gradient, and aspect) for each of the rock glaciers to an eight-dimensional random variable (i.e., X1, X2, X3 ... X8). A correlation coefficient ρij (i, j = 1, 2 ... 8) of Xi and Xj was introduced into the correlation matrix of the random dimensional vector as an eight order matrix for each element, and was denoted by R, thus: 20 The diagonal element of the correlation matrix was 1, and the correlation matrix itself was symmetrical. We performed the statistical analysis using SPSS20® software. Correlations between the topographical variables were then evaluated using Pearson correlation coefficients at a corresponding 0.05 level of significance.

Results
In total, 295 rock glaciers were identified in Daxue Shan (Fig. 3), covering an area of 55.70 km 2 (Table 2). Of these, 50.51 % 5 were talus-derived rock glaciers, the other 49.49 % were moraine-type rock glaciers. Most (94.58 %) of the rock glaciers were tongue-shaped, and remaining were lobate-shaped.
The rock glaciers are concentrated at altitudes between 4,300 and 4,600 m asl, with a mean altitude of 4,471 m asl (Fig. 4a).
Moraine-type rock glaciers are mainly concentrated between 4,400-4,600 m asl, and talus-derived rock glaciers between 4,300-4,550 m asl. In terms of general morphology, both tongue-shaped and lobate-shaped rock glaciers are mainly located 10 between 4,350-4,600 m asl (Fig. 4a). The upper boundaries for the vast majority of rock glacier types were ~ 4,600 m asl, because at higher altitudes there are often ice glaciers. Figure 4b shows the ranges in areas covered by different types of rock glaciers. Apart from a few outliers, the area of most rock glacier types area is < 0.3 km 2 , and in this regard there is no clear difference between rock glacier types. Figure 4c shows the ranges in the mean gradients of the slopes of different types of rock glaciers. Moraine-type and talus-derived rock glaciers exhibit mean gradients that are concentrated within the 22°-35° range. 15 However, tongue-shaped and lobate rock glaciers display a difference in mean gradient. Tongue-shaped rock glaciers have slopes with mean gradients which are concentrated in the 22°-35° range, whereas those of lobate rock glaciers fall within the 27°-45° range. This means that the upper (~ 10°) and lower (~ 5°) slopes of tongue-shaped rock glaciers are both shallower than for lobate rock glaciers. Figure 4d displays the ranges in the lengths of different types of rock glaciers. Moraine-type and talus-type rock glaciers have similar lengths (500-1,000 m), but when the rock glaciers are categorized by shape, tongue-20 shaped ones (500-1,000 m) are generally longer than lobate-shaped (200-400 m). Figure 5 shows rock glacier abundance versus aspect. Our dataset revealed that, apart from south-facing (5.44 %), southeastfacing (3.06 %) and northeast-facing (20.75 %) slopes, the rock glaciers are fairly evenly distributed on slopes with the 8 remaining five aspects that each account for ~ 15 % of the total. Moraine-type rock glaciers are most often northeast-facing (30.34 %) or north-facing (20 %), whereas talus-derived rock glaciers are most often southwest-facing (22.82 %) or westfacing (17.45 %). Lobate rock glaciers tend not to be found less on south-facing (6.25 %) and southeast-facing (0 %) slopes, but more commonly on north-facing, northwest-facing and east-facing, which each aspect accounting for ~ 18.75 % of the total. We compared all our results and discovered that shady (i.e., N, NE and E) slopes have more moraine-type rock glaciers, 5 and sunny (i.e., W, SW and S) slopes have more derived from talus. North-facing (i.e., N, NW and NE) slopes also seem to be more favorable for the formation of lobate rock glaciers than do south-facing (i.e., SW, S and SE) ones (Fig. 5), possibly in relation to the abundance of debris-producing steep rock walls on north faces. In contrast to other regions (Lilleøren and Etzelmüller, 2016;Onaca et al., 2017), we found that in Daxue Shan both moraine-type and talus-derived rock glaciers have developed in the monzogranitic areas, and that rock glaciers and monzonitic granite exhibit a high spatial correlation and 10 interdependence (Fig. 6). Table 3 shows the results of the correlation analysis. Positive correlations between rock glacier latitude, altitude, and length, and a negative correlation between altitude and mean slope were significant (p-value ≤ 0.05). Negative correlations between latitude and longitude, longitude and altitude, length and mean slope, and area and mean slope, and positive correlations of area with length and width were highly significant (p-value ≤ 0.01). 15

Discussion
The spatial distribution and dynamics of rock glaciers are especially dependent upon the local topography and climate (Springman et al., 2012;Delaloye et al., 2013). Lithology also exerts considerable control on the rock glaciers (Onaca et al., 2017). Analyzing local environmental factors (i.e., climatic, topographical and lithological factors) is therefore crucial to obtaining an understanding of the formation, development and spatial distribution of rock glaciers. 20

Topographical controls on rock glaciers
The significantly positive correlations between rock glacier latitude, altitude, and length (Table 3) are locally determined by the topographical characteristics. With the increase of latitude from the south to the north, the high altitude rock glaciers 9 increase in count, and flow further downvalley than those at low altitude. The altitudes of the mountains and rock glacier lengths increase with latitude, whereas air temperatures decrease, which implies that the northern sector of Daxue Shan has an environment that is more conducive to the formation of rock glaciers and other periglacial landforms. Likewise, the significantly negative correlation between latitude and longitude (Table 3) indicates local environmental factors dominantly control rock glacier distribution despite a general increase in relief in the northeast of the Daxue Shan region compared to the 5 southwest (Fig. 3a). There is also a significantly negative correlation between longitude and altitude (Table 3), as lower altitude areas to the east are less conducive to the development of rock glaciers where warmer and more humid conditions are common.
The negative correlation that exists between rock glacier length and mean gradient of slope is likely because the shortest rock glaciers are the talus-derived variety, and these have usually developed in steep topographical environments. Rock glacier area and mean gradient of slope, and altitude and mean slope are significantly negatively correlated (Table 3), likely because larger 10 and high altitude rock glaciers are mostly concentrated on gentle slopes that may be more conducive to their development. In summary, the topography of Daxue Shan is an important environmental control on the formation, development, and spatial distribution of the region's rock glaciers.
The mean altitude of a rock glacier's front (MAF) has often been taken to be a good approximation of the lower boundary of the discontinuous permafrost zone (i.e., Scotti et al., 2013). We found a substantial altitudinal difference between the lower 15 permafrost boundaries identified on the abovementioned eight aspects. For example, permafrost was assumed to be probable above 4,300 m asl on east-facing slopes, and above 4,403 m asl on west-facing slopes. The mean lower permafrost boundary was calculated as occurring at 4,352 m asl (derived from a mean value of 4,315 m asl for east-facing slopes, and 4,419 m asl for west-facing slopes). The mean lower permafrost boundary on east-facing (shady) slopes would therefore be ~ 104 m lower than that of west-facing (sunny) slopes (Fig. 7). 20 In addition, the formation and development of the Daxue Shan rock glaciers are also strongly influenced by the landforms created by glacial erosion and deposition. The southeastern margins of the TP are in a region of Quaternary glaciation which has been, and continues to be, strongly affected by monsoonal atmospheric circulations (Owen et al., 2005). This region possesses numerous ancient glacial relics and abundant landforms created by glacial erosion and deposition (Li and Yao, 1987).
We found that the distribution of rock glaciers is in close association with ice glaciers, as the upper boundaries for rock glaciers 25 were ~ 4,600 m asl, and at higher altitude there are often ice glaciers. In the context of global warming, it is widely accepted that the majority of glaciers on the Tibetan Plateau (TP) and its surroundings have experienced accelerated reduction (Bolch et al., 2012;Yao et al., 2012). The rate of glacier decline in Daxue Shan was -0.25 ± 0.20 % a -1 during 1990-2014 , with some ice glaciers transforming to rock glaciers. Glacial depositional landforms (e.g., moraine ridges) are highly conducive to the formation and development of moraine-type rock glaciers. Moraine ridges or moraines left after the retreat 5 of the ancient glaciers can provide significant quantities of boulders, erratic blocks, debris, sand, and ground ice. In the process of down-slope movement, rock glaciers can incorporate old moraine material as well as the debris from both sides of the moraine ridge. Glacial erosional landforms in particular have a close relation with the formation and development of talusderived rock glaciers. Ice structures, snow layers and moraines within glaciers collapse from time to time, supplying talus to the feet of mountains. As a result of the freeze-thaw process and the effect of gravity, talus creep then forms rock glaciers. 10

Climatic controls on rock glaciers
The west-facing slopes of Daxue Shan lie in the intersection between a sub-frigid monsoonal and a continental plateau climatic zone, and therefore experience a cold-temperate climate. At the Daofu meteorological station (2,957.2 m asl), mean annual precipitation (MAP) is ~ 613.5 mm, and mean annual temperature (MAT) is ~ 8.14 C (Fig. 8b). Based on an adiabatic rate of 0.65 C/100 m, we estimated the MAT at 4,311 m asl (i.e., the lower permafrost boundary) to be ~ -0.66 C. The east-facing 15 slopes of Daxue Shan are affected by a subtropical monsoonal climatic environment, and are affected principally by a southwesterly monsoonal atmospheric circulation, but also by a southwesterly monsoonal atmospheric circulation. East-facing slopes therefore experience high levels of precipitation (snowfall). MAP at the Kangding meteorological station (2,615.7 m asl) reaches 858.3 mm, and MAT is ~ 7.29 °C (Fig. 8a). We calculated the MAT at 4,352 m asl (i.e., the lower permafrost boundary) to be ~ -4.00 C. Here, the freeze-thaw process would be frequent (Fig. 8), meaning that the climatic environment 20 would provide temperature and precipitation conditions highly favorable to the formation and development of rock glaciers.
Compared with Gruber's (2012) global Permafrost Zonation Index (PZI) map, the distribution of rock glaciers in Daxue Shan is in general agreement, but many rock glaciers are situated within the PZI fringe of uncertainty (Fig. 3b). However, the PZI is strictly controlled by temperature that decreases with increasing altitude, thus our results indicate the importance of the local climatic controls on development of rock glaciers, and thus permafrost. In addition, compared with the distribution of ice glaciers in Daxue Shan, the distribution of rock glaciers also has characteristic small differences between the south and north, owing to a north-south corridor effect for water and heat transport and diffusion through the longitudinal gorges .

Lithological controls on rock glaciers 5
Lithology is a critical control for the supply of talus to ice-and rock-glacier surfaces (Haeberli et al., 2006). Figure 6 shows that the major exposed strata in the Daxue Shan region are composed of Tertiary monzonitic granite, consistent with the NW-SE trending Xianshuihe Fault. The surrounding mountains in this area generally consist of biotite-muscovite granite that intruded 16-13 Ma ago (Roger et al., 1995). Also located in this region is the tectonically important Zheduotang Fault, which runs through the Zheduo Valley, and is one of the most active fault systems on the TP's margins (Allen et al., 1991). It can be 10 seen from Figure 6 that the distribution of rock masses along the Xianshuihe Fault in the Daxue Shan region is clearly controlled by this NW-SE left-lateral strike-slip fault.
The Tertiary monzogranites are clearly highly conducive to the formation and development of rock glaciers. This is consistent with the findings of Onaca et al. (2017) in the southern Carpathian Mountains. According to Popescu et al. (2015), rock glaciers located in granitic and granodioritic massifs are composed of larger clasts compared with those found in metamorphic massifs. 15 Thus, the higher porosity of the substrata in granitic and granodioritic massifs allows for a significant cooling beneath the bouldery mantle because the denser cold air is trapped between the large boulders (Balch, 1900). The lithological and mineralogical characteristics which accompany the high porosity of tertiary monzogranites are therefore more favorable to the formation and development of local rock glaciers than are other lithologies. In addition, rock glacier formation is also controlled by slope and sedimentation rates contributing debris to the landforms (Müller et al., 2016). There are large sources 20 of sediment and sediment storages in Daxue Shan, which are controlled by the processes occurring within this setting (Müller et al., 2014). An abundance of steep rock walls and deepened valley sides provides catchment areas for rock glacier development and, when combined with intense monsoonal precipitation and tectonic activity, drives sediment transport processes and rock glacier development in Daxue Shan. Several researchers (e.g., Cui and Zhu, 1989;Zhu, 1992;Zhu et al., 1992;Liu et al., 1995;Bolch and Gorbunov, 2014) have previously identified hundreds of rock glaciers in the northern Tianshan Mountains. They found that most of the identified rock glaciers were tongue-shaped, and were located at altitudes between 3,300 and 3,900 m asl, on north-facing slopes. Most rock glaciers in Daxue Shan are also tongue-shaped. However, the altitudes at, and the aspects on, which these rock glaciers are found differ between the Daxue and the Tianshan mountain ranges. First, in terms of altitude, the rock glaciers of Daxue 5 Shan are located between 4,300 and 4,600 m asl, higher than the Tianshan rock glaciers by approximately 700-1,000 m. It would be reasonable to assume, therefore, that the rock glaciers located in lower latitudes are more likely to be found at higher altitudes. Second, in terms of aspect, the rock glaciers of Daxue Shan are more evenly distributed across all eight abovementioned aspects than are the rock glaciers of the Tianshan Mountains. This could be explained by several factors, including the differences in overall altitude, as well as in the orientation of the main massif of each mountain range. Daxue 10 Shan lies along an approximately NW-SE axis, whereas the Tianshan Mountains are roughly W-E in presentation. Rock glaciers are therefore less commonly found on the east-and west-facing slopes of the Tianshan. The effect of solar radiation is stronger on the south-facing slopes of the Tianshan Mountains than on its north-facing ones, meaning that conditions on these south-facing slopes are less conducive to the development of rock glaciers; most of the range's rock glaciers are therefore found on its north-facing slopes. Furthermore, when higher altitudes are reached, all aspects experience lower air temperatures, 15 resulting in a lessening of the impact caused by the difference between air temperature and solar radiation exposure; this phenomenon is similar to that found in Daxue Shan, and explains why rock glaciers there are fairly evenly distributed on all eight aspects. However, when altitudes are lower, the impact of solar radiation, combined with warmer air temperatures, is greater, particularly on south-facing slopes; both temperature and solar radiation are lesser on shady north-facing slopes, however, explaining the predominance of north-facing rock glaciers in the Tianshan Mountains. 20

Conclusions
Rock glaciers are widespread in Daxue Shan, and of these, tongue-shaped rock glaciers cover the largest area. The occurrence and characteristics of these rock glaciers can mostly be explained by local environmental controls (i.e., climatic, topographical and lithological factors).
In total, 295 rock glaciers were identified in Daxue Shan, covering a total area of 55.70 km 2 . The altitudes at which moraine-5 type rock glaciers are found (i.e., 4,400-4,600 m asl) are at least 50-100 m higher than for talus-derived rock glaciers (i.e., 4,300-4,550 m asl), although the upper altitudinal limit for both these types of rock glacier is ~ 4,600 m asl. At higher altitudes there are often ice glaciers. Except for a few outliers, the area of each type of rock glacier is no greater than 0.3 km 2 . There is no significant difference between moraine-type and talus-derived rock glaciers in terms of the mean slope gradients (i.e., they are all clustered within the 22°-35° range), but the upper and lower mean slope gradients of tongue-shaped rock glaciers are 10 ~ 10° and ~ 5° lower than for lobate rock glaciers, respectively. In terms of length, moraine-and talus-derived rock glaciers have similar lengths (~ 500-1,000 m), but according to shape, lobate-shaped rock glaciers are distinctly shorter than tongueshaped by ~ 300-600 m. We found shady (i.e., N, NE and E) slopes more conducive to the presence of moraine-type rock glaciers than sunny (i.e., W, SW and S) ones that appear more conducive to the presence of talus-derived rock glaciers. In addition, north-facing (i.e., N, NW and NE) slopes appeared more favorable to the formation of lobate rock glaciers than south-15 facing (i.e., SW, S and SE) ones. The mean regional lowest altitudinal limit of rock glaciers is 4,352 m asl, an altitude which was taken to indicate the local permafrost's mean lower boundary. On east-facing slopes, the permafrost's lower boundary can therefore reasonably be assumed to be ~ 104 m lower than on west-facing slopes.
The correlation matrix of rock glacier variables indicates that the formation of rock glaciers is closely related to local environmental conditions. The local climatic environment leads to a frequent freeze-thaw process within these rock glaciers, 20 a process which is also beneficial to their formation and development. Tertiary monzonitic granite, with its large clastic and highly porous characteristics, is more sensitive than other lithological components to the freeze-thaw process, and continuous weathering of this monzogranitic substratum thus provides the ideal raw material for the rock glaciers of Daxue Shan.
14 Data availability. The data associated with this article can be found in the Supplement. These data include maps of the most important areas described in this article, as well as a tabulation of the parameters of the rock glaciers found in Daxue Shan.
Competing interests. The authors declare that they have no conflict of interest.
Acknowledgements. This work was funded by the National Natural Science Foundation of China (Grant Nos. 41230743 and 41371082). We thank the editor and three reviewers for detailed and constructive suggestions for revisions to the manuscript. 5