Paleoglacial footprint and fluvial terraces of the Shaluli Shan, SE Tibetan Plateau

ABSTRACT This study provides mapping of glacial and fluvial geomorphology in the Shaluli Shan region on the eastern margin of the south-eastern Tibetan Plateau. Based on TanDEM-X 12 m elevation data and GoogleEarth imagery, glacial valleys, ice-marginal moraines, glacial lineations, scoured terrain and fluvial terraces were mapped. Covering around 11,000 km2, this map is the first for this region to display geomorphology at a spatial resolution of 0.4 arcsec (= c. 11 m) and to include fluvial terraces. Its glacial landform distribution is largely consistent with previous mapping. The substantially higher level of detail in this study is reflected in an approximately tenfold number and smaller median sizes of individual landforms such as moraines and glacial lineations. These results underscore the importance of high-resolution DEM data such as TanDEM-X for the identification of glacial and fluvial geomorphology. The map presented here will be used for detailed paleoglacial reconstructions and landscape evolution studies combining both glacial and fluvial landforms.


Introduction
Global climate change induces an increasing need to understand spatial, temporal, and ultimately causal relationships between climate and landscape development. Moreover, landscapes are sensitive to changes in exogenic (climatic) and endogenic (tectonic) forcing and their responses can be complex (Cordier et al., 2017;Molnar & England, 1990;Raymo & Ruddiman, 1992). This complexity typically results in palimpsest landform assemblages formed by one process coexisting with, or overprinted by, a landform assemblage of another process (Glasser & Bennett, 2004;Schmidt & Andrew, 2005).
Due to its dimension (Figure 1), the Tibetan Plateau exerts a regional dominance on atmospheric circulation patterns (Yao et al., 2013;Zhang et al., 2018), reflected in the extents of glaciers and their trajectories of change (Yao et al., 2012a). This is because glacier mass balance, and ultimately glacier extent, is highly sensitive to temperature and precipitation (Oerlemans & Fortuin, 1992). Reconstructions of glacier extent can therefore serve as evaluation datasets for glacier and climate modelling Yan et al., 2018). Evaluating climate proxy data (e.g. Liu et al., 2020;Yao et al., 2007) against paleoglacial reconstructions can reveal if glaciation on the Tibetan Plateau is most sensitive to changes in precipitation or temperature (e.g. Dortch et al., 2013). Nonetheless, these relationships remain challenging to disentangle, in part because of tectonic activity. Uplift affects climate and drives or modifies responses in glaciation on geological timescales and adds an additional layer of complexity to landscape evolution reconstructions .
To add further complexity, strong regional gradients in the response of glaciers to increases in temperature and concomitant changes in precipitation have been observed on the Tibetan Plateau (Farinotti et al., 2020;Hewitt, 2005;Yao et al., 2012a). Controversies about the timing and extent of past glaciations remain. Spatial gradients in the timing and extent of paleoglaciation have been inferred between the southern and eastern margins of the Tibetan Plateau, which receive snowfall from monsoon circulation, and its western and northern margins that are dominated by snowfall supplied by westerly circulation (Lehmkuhl & Owen, 2005;Yan et al., 2020). This diverse evidence indicates that glaciers on the Tibetan Plateau respond, and have responded, to climate change in complex ways. This may partly reflect tectonic modification of the glacier response to climate change (Barr & Lovell, 2014;Shulmeister et al., 2009). Consequently, the question to which degree temperature decreases or precipitation increases drive Tibetan Plateau glacial advances remains unresolved (Chevalier & Replumaz, 2019;Murari et al., 2014;. It is conceivable that some of the spatial and temporal complexity derives from a lack of detailed geomorphological mapping and associated dating of glacial landforms (Fu et al., 2013a(Fu et al., , 2013bHeyman et al., 2008;. While such mapping is essential for accurate, detailed, and meaningful reconstruction of paleoenvironments, in practice it is often limited by the availability of high-resolution data and imagery. Previous geomorphological mapping of the Tibetan Plateau and surrounding mountain ranges has primarily been based on digital elevation models with spatial resolutions of 30-90 m (e.g. SRTM and ASTER), Landsat ETM+ satellite imagery (30 m horizontal resolution, 15 m for its panchromatic band), and, in some studies, Google Earth imagery ( Figure 1). This paper advances previous mapping of the Shaluli Shan region on the south-eastern margin of the Tibetan Plateau in two critical aspects and ultimately contributes to a reconstruction of its paleoglacial history and landscape evolution. First, we utilise highresolution TanDEM-X 12 m data to map glacial landforms (glacial valleys, ice-scoured terrain, ice-marginal moraines, and glacial lineations) in unprecedented detail. There are individual studies that utilise TanDEM-X data for geomorphological purposes (e.g. Boulton & Stokes, 2018;Grohmann, 2018;Pasquetti et al., 2019;Pánek et al., 2020;Pipaud et al., 2015;Viveen et al., 2020), but to our knowledge, this is the first study employing such high-resolution data for the purpose of extensive geomorphological mapping on the Tibetan Plateau. A geomorphological map of glacial landforms in the Shaluli Shan region was presented by Fu et al. (2012). However, because they used SRTM and Landsat ETM+, the availability of 12 m TanDEM-X in this study enables a considerable increase in the level of detail at which landforms are mapped.
Geomorphological mapping for paleoglacial research commonly focusses on glacial landforms (e.g. Glasser & Bennett, 2004). To add a systematic mapping of fluvial terraces to glacial features is novel for this region. It also is the prime reason for employing TanDEM-X data because it provides the required spatial resolution, as shown by the mapping of fluvial terraces elsewhere (Charrier & Li, 2012;Viveen et al., 2020). We here jointly map glacial and fluvial landforms as part of one study to distinguish between climatic and tectonic drivers of landscape evolution. This is because river terrace sequences form through cycles of sediment deposition and fluvial incision and their formation has been mainly attributed to climatic cyclicity, even though tectonic uplift also favours their formation (e.g. Bridgland & Westaway, 2008;Maddy et al., 2001;Olszak, 2017;Starkel, 2003).  (Fu et al., 2012); (b) Bayan Har Shan (Heyman et al., 2008); (c) Dalijia Shan (Kassab et al., 2013); (d) Parlung Zangbo Valley, (Chen et al., 2016); (e) Maidika region (Lindholm & Heyman, 2016); (f) Tangula Shan (Morén et al., 2011). Overview map of the Tibetan Plateau, indicating the mapping area of this and of previous mapping studies.
The purpose of this study is to test and illustrate the potential of 12 m TanDEM-X for high-resolution mapping of glacial and fluvial landforms. As an added advantage, it also provides a comparison dataset to the mapping of glacial landforms conducted by, and the reconstruction of the glacial footprint presented in Fu et al. (2012). The resulting main map has guided the identification of targets for geochronological research and field studies in support of paleoglacial and landscape evolution studies on a regional scale.

Study area and previous research
The Tibetan Plateau is a unique topographic feature with an average elevation of more than 4000 metres above sea level and an extent of 5 million km 2 (Fielding et al., 1994;Yao et al., 2012b). It started uplifting during the early Cenozoic as the Central Asian and Indian plates collided, with profound consequences for landscape evolution and global and regional climate change (Clark et al., 2005;Raymo & Ruddiman, 1992;Wang et al., 2008Wang et al., , 2012Yao et al., 2012b). Because of its importance for climate dynamics, considerable efforts have been made to reconstruct climate change from the history of glaciation on the Tibetan Plateau and surrounding mountain ranges through glacial geomorphological mapping and geochronological studies (Blomdin et al., 2016a(Blomdin et al., , 2016bChen et al., 2016;Chevalier & Replumaz, 2019;Gribenski et al., 2018;Heyman et al., 2008Heyman et al., , 2009Heyman et al., , 2011aHeyman et al., , 2011bLehmkuhl & Owen, 2005). Most of these recent studies conclude that alpine-style glaciation has dominated on the Tibetan Plateau, with strong regional differences in the timing of maximum glaciation across the region (Blomdin et al., 2016b;. The Shaluli Shan mountain range extends for 450 km in a north-south direction and is part of the Hengduan Mountains forming the southeastern margin of the Tibetan Plateau (Figures 1 and 2). The study area is located in the province of Sichuan, China, and encompasses around 11,000 km 2 of the central part of the Shaluli Shan ( Figure 1). The regional topography is characterised by high, deeply-incised mountain ranges and low-relief granitic plateaus such as the Haizishan Plateau, which are interpreted to be remnants of an older, uplifted paleosurface ( Figure 2; Clark et al., 2005). Past and present-day tectonic activity results from active faults intersecting the Shaluli Shan, for example the Litang fault system (Figure 2), and is expressed in the presence of pull-apart basins, landform displacements, and rockfalls triggered by earthquakes (Chevalier et al., 2016;Zeng et al., 2019).
An abundance of preserved glacial landforms in the Shaluli Shan have attracted research on its regional paleoglacial history (Chevalier & Replumaz, 2019;Fu et al., 2012Fu et al., , 2013aFu et al., , 2013bFu et al., , 2019Xu & Zhou, 2009Zhang et al., 2015). Extensive geomorphological mapping of the Haizishan Plateau using SRTM, Landsat ETM+ satellite imagery, and field-based paleoglaciological observations, revealed zonal patterns of glacial erosion and deposition indicative of a former ice cap (Fu et al., 2012(Fu et al., , 2013a with outlet glaciers descending the plateau flank and, where present, into the pull-apart basins. The presence of a polythermal paleo ice cap on the Haizishan Plateau was further corroborated by cosmogenic nuclide exposure dating and numerical glacier modelling (Fu et al., 2013b(Fu et al., , 2019. It is expressed in the landscape as relict and scoured terrains on top of the Haizishan Plateau and linear erosion of glacial valleys on its flanks, defining zones of differing erosion strength.

Data and map production
Geomorphological mapping of glacial landforms and fluvial terraces was performed using TanDEM-X 12 m digital elevation data. To map glacial landforms, including ice-marginal moraines marking the maximum extent of glaciation (end moraines), in close proximity to fluvial terraces, we chose a map area that covers the Ge'Nyen massif and the northern part of the Haizishan Plateau (the main source areas of the maximum glaciation) and three tectonic basins, which serve as storage areas for fluvial sediments.
Geomorphological mapping relies on the premise that landforms have characteristic appearances which allow them to be delineated and classified (Hubbard & Glasser, 2005;Minár & Evans, 2008). Factors such as the size of the investigation area, age and degradation of targeted landforms, applied mapping methodology, and the purpose of the final map product will determine applied classification criteria. Transparency regarding the mapping criteria and techniques is therefore a fundamental part of mapping ethics (Chandler et al., 2018).
Here, a pre-defined set of criteria guided the mapping from remotely sensed imagery (Table 1) and cross-checking of the mapping was achieved for several key areas during field observations in June 2019 (field verification route shown in black in Figure 2). We used four tiles of 12 m TanDEM-X with a local spatial resolution of 11.11 m (Wessel, 2018) and combined this with Google Earth imagery for perspective viewing where additional clarification was required. Mapping was conducted in ArcMap through manual delineation of landforms and interpretation of Tan-DEM-X hillshade-derivatives, using different azimuth (315°, 280°, 0°, 45°, 80°) and zenith (10°, 20°) values, which defines the angle of incoming insolation and the pattern of shading. The use of a range of illumination angles is important because shadowing affects the level of detail at which landforms can be mapped.
The methodology for mapping of glacial landforms follows the framework for best practice (Chandler et al., 2018) and is comparable with previous mapping strategies on the Tibetan Plateau. Individual landforms were identified and mapped by iterative passes over the mapping area at different scales (for exceptionally large or small landforms, the applied onscreen mapping scales may have varied): . 1:100,000large-scale identification of glacial valleys, large ice-marginal moraines, scoured terrain, and glacial lineations . 1:50,000identification of smaller ice-marginal moraines, mapping of glacial valleys, and glacial lineations . 1:25,000mapping of ice-marginal moraines and glacial valley heads (cirques) . 1:5000 -1:10,000validation and refinement of map details The distribution of contemporary glaciers comes from the Global Land Ice Measurements from Space glacier database (GLIMS).

Mapped landforms and identification criteria
Glacial valleys, ice-marginal moraines, scoured terrain, and glacial lineations are mapped because of their paleoglaciological significance (Glasser & Bennett, 2004). In addition to these glacial landforms, fluvial terraces were identified as part of a larger project to understand whether their formation by fluvial deposition and incision is intrinsically coupled to the advance and retreat of glaciers. Identification and detailed mapping of these landforms largely depend on the spatial resolution of available data and hillshade illumination angles. Consistent mapping results require detailed landform identification criteria and recognition of potential error sources. These are adapted for mapping from TanDEM-X from previous studies (Blomdin et al., 2016a;Fu et al., 2012;Heyman et al., 2008;Stroeven et al., 2013) and outlined in Table 1.
Glacial valleys are smooth valleys, sometimes with truncated spurs or lake-filled overdeepenings, and with U-shaped or parabolic cross profiles that, because of their commonly km-scale size, often stand out from remote sensing and topographic data. They form when glaciers deepen and widen pre-existing valleys (   glacial valleys have segments that start with steep, arcuate-shaped headwalls and gently sloping, glaciallyeroded valley floors (cirques), and these are included in this category. Because fluvial valleys are commonly V-shaped and narrow, they are quite distinct from glacial valleys even though uncertainty in either identification can arise for valleys with a tectonic origin or where glacial valleys have a postglacial fluvial overprint ). Mapping of glacial valleys was based on identification criteria in Table 1, and features were outlined as polygons. Headwater valley outlines were identified from TanDEM-X-derived slope rasters and hillshades. The ArcGIS-profiling tool was employed to visually determine transitions in valley profiles, from upstream parabolic to downstream V-shaped.
Ice-marginal moraines are ridges that are mostly composed of unsorted debris deposited along the terminus of a glacier (Figure 4(A)). No subdivision into different moraine types has been pursued for the purpose of this study because all mapped ice-marginal features delineate former maximum ice extents (cf. Benn & Evans, 2013). Most moraines are indicative of warm-based glaciation (erosion of its substrate by sliding ice) because they typically contain subglacially transported sediment (Bierman & Montgomery, 2014). The outline and slope of ice-marginal moraines can be used to reconstruct the spatial extent and thickness of the depositing glaciers. The timing of deposition can often be construed from concentrations of cosmogenic nuclides in boulders residing on moraine crests (e.g. Chevalier & Replumaz, 2019;Fu et al., 2013b;Heyman et al., 2011b;. Uncertainty in the delineation of a moraine can arise if its size is small compared to the spatial resolution of the TanDEM-X satellite data, if the debris cover is thin or fragmented, or if the moraine is highly degraded (Blomdin et al., 2016a).
Ice-marginal moraines were mapped by outlining the ridges as polygons. The break-in-slope between the moraine ridge and the surrounding terrain, which is consistently visible in the hillshaded DEMs, was employed as a demarcation criterion.
Glacial lineations are elongated, streamlined ridges formed parallel to ice flow (Figure 4(B)). They form through subglacial deposition of till or erosion of bedrock and can typically be found in glacial valleys (Benn & Evans, 2013). Erosional glacial lineations consist of streamlined bedrock, while their depositional equivalents consist either solely of sediment, or have a bedrock core and sediment tail (e.g. crag-and-tails). No subdivision into different lineation types has been pursued for the purpose of this study because all lineation types indicate the direction of warm-based ice flow. Confusion with non-glacial bedrock lineaments or tectonic structures may lead to identification errors (Blomdin et al., 2016a).
Glacial lineations were identified from hillshades and slope DEMs and mapped as lines along their centrelines. They were also delineated as polygons where they showed distinctive asymmetric shapes (Supplementary Material 1).
Scoured terrain is an undulating, low-relief bedrock surface of a knock-and-lochan type, typically with lakes in depressions and streamlined bedrock highs (glacial lineations), with a deranged river drainage pattern (Figure 4(B)). The scoured terrain is a result of areal erosion through glacial plucking and abrasion; processes which also require basal ice at the pressure melting point (Sugden & John, 1976).
Scoured terrain was mapped based on hillshades and slope DEMs to identify streamlined bedrock features and the absence of consistent river drainage patterns. In addition, the use of Google Earth imagery facilitated the identification of lakes as features typical of scoured terrain.
Fluvial terraces are characterised by even surfaces that gently slope towards the valley centre and in a downstream direction ( Figure 5). These terraces terminate in steep slopes roughly parallel to the valley centre and may form staircase sequences (Stokes et al., 2012). Even though terraces can often be confidently identified from topographic data, it is equally often impossible to distinguish between strath terraces (which are eroded into bedrock) and alluvial terraces (consisting of incised floodplain deposits) without additional field verification. During field inspection ( Figure 2) we sometimes identified bedrock outcrops, but, in all cases, those were found near the base of the lowest terrace slope. The terraces in our study area were therefore classified as alluvial terraces. Furthermore, they are defined as 'fluvially incised', because we cannot differentiate between different types of alluvial deposits at the scale of this map, which is largely based on remote sensing imagery.
In contrast to the straightforward strategy for the mapping of glacial landforms, the identification of terrace surfaces from hillshades and Google Earth imagery can be difficult. Especially in large tectonic basins, the differentiation between terrace surfaces and gently sloping alluvial or colluvial deposits poses a challenge. To reliably distinguish these landform types, the average slope of all terrace surfaces identified and mapped based on field observations were calculated and taken as the threshold value to create a classified slope model. Supported by field verification observations, surface inclinations below 4°were deemed indicative of fluvial terraces. The same threshold value was applied to areas outside of the field verification area unless distinguishable lobate patterns indicated the presence of alluvial or mass movement deposits.

Delineation of past ice extents
Based on the distribution of glacial landforms, we reconstruct a minimum extent of maximum glaciation (Figure 6; cf. Heyman et al., 2009;Blomdin et al., 2016a). It cannot be assumed that ice caps and glaciers within this area advanced to their maximum positions synchronously. Indeed, Fu et al. (2013a) show that the region experienced multiple cycles of glaciation. The chronological record for the wider region remains sparse (Chevalier et al., 2016;Chevalier & Replumaz, 2019;Fu et al., 2013b;Schäfer et al., 2002).

Results and discussion
Due to the high resolution of 12 m TanDEM-X, mapping was performed at an unprecedented level of detail.
Differences in resolution between TanDEM-X and digital elevation models used for previous mapping on the Tibetan Plateau (SRTM-90, SRTM-30) are clear from both a visual comparison (Figure 7), and from quantitative comparisons between our mapping results and a previous study by Fu et al. (2012) (Figure 8).
Evidence for paleoglaciation occurs primarily on plateaus and in mountains, with most extensive ice having built ice-marginal moraine complexes beyond valley confines. To document the extent of former glaciation, we mapped its four most common landforms; glacial valleys, glacial lineations, ice-marginal moraines, and scoured terrain. Glacial valleys cover 4080 km 2 , or 37% of the mapped area, and showcase the importance of deep glacial erosion during repeated glaciation. These valleys are common landscape elements of most mountains (as, for example, the Ge'Nyen massif in our study area) and ornament the margins of the Haizishan Plateau. Indeed, in the latter case, these glacial valleys are trunk valleys that lack alpine-style glaciation accumulation areas, and they signify the imprint of outlet glaciers from an ice cap that repeatedly covered the Haizishan Plateau (Fu et al., 2013a). In our study area, most glacial valleys span several kilometres in length and width. In contrast, valleys draining the lowest-elevation mountains (< 4000 m a.s.l.) show no signs of glaciation.
Swarms of glacial lineations and scoured terrain are prominent on the Haizishan Plateau. Glacial lineations are also present in some of the glacial valleys. The asymmetric shapes of some can be used to determine ice flow direction. The presence of scoured terrain, covering 354 km 2 of the mapped area, indicates the presence of ice-cap-style glaciation and supports the presence of warm-based ice as inferred by Fu et al. (2012Fu et al. ( , 2013a. On the very northern Haizishan Plateau, glacial lineations are difficult to distinguish from bedrock structures and fracturing, and so appear absent on the map (we were unable to access this area during fieldwork, see Figure 2 for the field verification route). The bulk of the 1310 glacial lineations mapped, ranging in length from 29 to 1335 m (Figure 8(A)), are from the Haizishan Plateau. Where they have distinct asymmetric (streamlined) shapes, they are indicative of ice flow direction (Supplementary Material 1). For example, they indicate ice flow outwards from a central ice dome towards the margins of the plateau, which is in accordance with paleoglaciological reconstructions by Fu et al. (2012Fu et al. ( , 2013aFu et al. ( , 2013b. Furthermore, a funnelling of ice flow into glacial trunk valleys incised in the margins of the plateau can be reconstructed based on a spatial pattern of convergence of glacial lineations. In total, 1237 individual ice-marginal moraine ridges were mapped, ranging from 2254 m 2 to over 2 km 2 (Figure 8(B)). They are primarily located along glacial valleys (where they appear as straight or arcuate ridges, or complexes of ridges) and at or beyond their mouths in tectonic basins, which are otherwise largely free of glacial landforms. Fu et al. (2012) found no evidence for ice-marginal moraines on the northern sector of the Haizishan Plateau, in the area considered in this study. We can largely confirm that absence (and thus its contrast with the southern half of the plateau) even though the increased resolution offered by TanDEM-X has allowed us to outline a small number of recessional moraines.
In addition to glacial landforms, 910 individual terrace surfaces were identified which are primarily located in four tectonic basins (i.e. Maoyaba, Litang, Kangga, Cuopo; Figure 2). Terraces appear as flat (< 4°inclination) surfaces in tectonic basins or river valleys with steep edges towards rivers. Their extents range from 1328 m 2 to 24.4 km 2 and amount to a surface area of approximately 265 km 2 . Hence, terraces comprise around 2.4% of the mapped area. The Figure 6. Minimum extent of maximum glaciation in the study area (65% of mapped area), compared with present-day glaciation (1% of mapped area). Light-blue arrows indicate reconstructed ice-flow directions from lineations. TanDEM-X topographic data © DLR 2019. A map showing the inferred minimum extent of maximum glaciation which comprises 65% of the mapping area of this study. Contemporary glaciation is also shown; it comprises an area of 1% of the mapping area. This map also shows the locations of the landforms presented in Figures 2-5. presence of extensive terraces in tectonic basins indicates that these basins act as local sinks in the sediment budget, storing sediment emanating from upstream catchments.
The Haizishan Ice Cap and surrounding mountain glaciers have at times advanced across 65% of the mapped area judged from the distribution of glacial landforms (7101 km 2 , Figure 6). We consider this footprint to represent the minimum extent of maximum glaciation because it is plausible that landforms associated with even further ice extents are either too small (e.g. degraded), buried underneath fluvial sediments, or only represented in the sedimentary record (till, glaciofluvial deposits) or by scattered erratics . To put these numbers in perspective, the paleoglaciological footprint of the Haizishan Ice Cap has been estimated to 3600 and 4000 km 2 (Fu et al., 2013a;Li et al., 1996), which is just over half the area of the contemporary Vatnajökull Ice Cap in Iceland. Ice expansion towards these limits of maximum glaciation was likely asynchronous across the mapped area, as evidenced by cosmogenic nuclide dating (Chevalier & Replumaz, 2019;Fu et al., 2013b).
The distribution of landforms presented in this study and the reconstructed glacial footprint quite  consistently match the mapping by Fu et al. (2012). This congruence between the two independent mapping studies strengthens the presented reconstruction of the maximum paleo ice extent and indicates that a mapping approach based on SRTM and Landsat data remains suitable for reconstructing regional Tibetan paleoglaciology. This new study adds more and smaller landforms and therefore important detail to established patterns (Figures 7 and 8). For example, because Fu et al. (2012) were limited to landforms larger than 70 m due to the spatial resolution of their data, the minimum length among the 136 glacial lineations in their study is 109 m, compared to 29 m in this study. The possibility to identify a larger range of lineation lengths and more lineations makes reconstructions of ice flow directions more robust, especially in places where streamlined shapes can be distinguished ( Figure  4(B)).
When only the area of overlap between the two studies is considered, Fu et al. (2012) found that the smallest out of their 174 mapped individual moraines had a size of 13,011 m 2 , which is more than five times larger than the smallest moraine mapped here (2254 m 2 ). On the contrary, their largest moraine of almost 5.2 km 2 (Figure 8(B)) is more than twice the size of the largest moraine in this study (almost 2.1 km 2 ). The 12 m TanDEM-X largest moraine is significantly smaller because individual ridges were differentiated in instances where the previous study aggregated these into complexes of several moraines. This differentiation attests again to the level of detail afforded by 12 m TanDEM-X.
The fluvial terraces mapped in this study provide an addition to the glacial landform record by adding information about the distribution of a prominent class of primarily fluvial features. Their correct identification has been confirmed through field verification in the Kangga and Maoyaba Basins ( Figure 2). Moreover, the dating of these terraces allows for a bracketing of moraine ages with independent ages of upstream and downstream terraces in geochronological studies.
The reconstructed glaciation footprint supports a prevalence for ice cap-and alpine styles of glaciation on the SE Tibetan Plateau. In comparison with previous studies, the level of detail achieved using 12 m TanDEM-X illustrates the potential for increasingly detailed reconstructions of ice dynamics and ice margin locations. The presence of fluvial terraces in direct conjunction with traces of glaciation (end moraines, glacial valleys), may imply a modification of the glacial landscape by fluvial processes (removal or burial of moraines). The novel approach of this study, to map fluvial terraces in addition to glacial landforms, allows for testing the temporal relationship between glaciation and fluvial deposition and, if coupled, to increase the precision in the timing of glaciations.

Summary
We present a map of the distribution of glacial valleys, ice-marginal moraines, glacial lineations, and scoured terrain, using 12 m TanDEM-X digital elevation data, to provide an alternative estimate on the former extents of ice caps and valley glaciers in the Shaluli Shan, south-eastern Tibetan Plateau, an area previously mapped by us using lower-resolution satellite data. As an additional approach, novel for this region, we mapped fluvial terraces. The overall distribution of landforms is objectively highly similar to the previous product and also the paleoglaciological footprint of glaciation (i.e. the minimum extent of maximum glaciation) is consistent with previous mapping. Although we have mapped many more and smaller landforms, this similarity indicates that a mapping approach based on SRTM and Landsat data remains suitable for reconstructing regional Tibetan paleoglaciology. Landform distribution indicates a prevalence of ice cap-and alpine styles of glaciation. The higher mapping resolution of our study and the addition of fluvial terraces provides a basis for detailed geomorphological and geochronological studies of these glaciation patterns and highlights the potential of highresolution 12 m TanDEM-X for the identification of glacial and fluvial geomorphology and the benefits of such data for landscape evolution studies.

Software
Data processing, analysis and landform mapping was conducted in ArcMap 10.7.1. The final map layout was created in Inkscape version 0.92.3.

Data
ESRI shapefiles used in the production of the map are provided as supplementary material.