Brief communication: A ~50 Mm ice-rock avalanche on 22 March 2021 in the Sedongpu valley, southeastern Tibetan Plateau

On 22 March 2021, a ~50 M m ice-rock avalanche occurred from 6500 m asl in the Sedongpu basin, southeastern Tibet. The avalanche transformed into a highly mobile flow which temporarily blocked the Yarlung Tsangpo river. The 15 avalanche flow lasted ~5 minutes and produced substantial geomorphological reworking. This event, and previous ones from the basin, occurred concurrently with, or shortly after recorded positive air temperature anomalies. The occurrence of future large mass flows from the basin cannot be ruled out, and their impacts must be carefully considered given implications for sustainable hydropower and associated socioeconomic development in the region.

average (Yao et al., 2019). The region has also experienced a series of high-magnitude ice-rock avalanches, glacier detachments, and glacial lake outburst floods (GLOF) in recent decades (Tong et al., 2019;Veh et al., 2020;Kä ä b et al., 2021;Zheng et al., 2021). Furthermore, the region is the focus of considerable investment by the People's Republic of China (PRC), 35 including the construction of the high-speed Sichuan-Tibet Railway (anticipated completion 2030), with development of its attendant economic corridor, and the planned construction of new, large-scale hydropower projects to serve an increasing regional and international demand for electricity. Indeed, trans-border hazards of cryospheric origin are of significant concern across High Mountain Asia, particularly when combined with the often rapid changes in both exposure and vulnerability to these geohazards (Shugar et al., 2021). 40 The Sedongpu basin (29.80° N, 94.92° E) in Nyingchi prefecture, PRC, has a history of large avalanches and low-angle glacier detachments which have transformed into powerful debris flows in the southeastern Tibetan Plateau (Tong et al., 2019;Chen et al., 2020;Wang et al., 2020;Kä ä b et al., 2021). The basin has a total area of 67 km 2 and ranges from a maximum elevation of 7294 m asl at Gyala Peri peak to a minimum of ~2750 asl m at its confluence with the Yarlung Tsangpo, a major tributary of the Brahmaputra River (Fig. 1a,b). The Randolph Glacier Inventory (V6.0) identifies 18 glaciers covering a total 45 of ~17 km2 in the basin (RGI Consortium, 2017), which are nourished by snow, ice, and debris from very steep mountain flanks. Analysis of historic satellite imagery has revealed evidence of mass flow activity in the basin (Kä ä b et al., 2021) and at least 11 mass flows originating in the basin have partially or entirely blocked the Yarlung Tsangpo in past decade (Tong et al., 2019;Chen et al., 2020). The basin has recently experienced large ice-rock avalanches with a total of ~50 Mm 3 on October 2017 and into 2018, and the detachment of the tongue of Sedongpu Glacier in two separate events with a total of ~130 Mm 3 50 on 17/18 October and 29 October 2018 (Kä ä b et al., 2021). Both detachments in 2018 transformed into debris flows; the earlier detachment blocked the Yarlung Tsangpo for ~60 hours, and the rapid rise in upstream water level damaged or seriously threatened roads, power lines, hydropower stations, and other riverside infrastructure and prompted relocation of more than 6,000 local residents (Chen et al., 2020).
In this brief communication we reflect on the large ice-rock avalanche that originated in the Sedongpu basin on 22 March 55 2021 which, similar to past events, also temporarily blocked the Yarlung Tsangpo. We utilise in-situ field investigations, highresolution satellite imagery and digital elevation models (DEMs) of difference, seismic records, and meteorological data to analyse the evolution of the event and its impact, discuss potential drivers, and briefly reflect on implications for the sustainable development of the region.

Data and Methods 60
The basin's recent event history led to the installation of time-lapse optical/thermal cameras and an automatic weather station (AWS) at the exit of the basin in September-October 2019 (3 hours interval transmitted by the Inmarsat maritime satellite system) for the purpose of obtaining photographic and meteorological evidence of future mass flows and their timing, and a Campbell CS477 radar-based water level monitoring system at Gyala, which is located on the Yarlung Tsangpo ~6 km upstream of the exit of the Sedongpu basin ( Fig. 1b) to provide early warning and a real-time (10 min interval) record of water 65 level and, by implication, flow impedance in the case of future blockages. We used a DJI Phantom 4 drone to obtain aerial photographs of the basin in October 2019 (i.e. pre-event) and on 25 March 2021 (post-event), and used these imagery to document geomorphological modification of the landscape around the basin outlet ( Fig. 1c-g).
Vertical seismic waveforms (10 Hz) from long-period seismographs recorded at Nyingchi station, ~60 km away from the Sedongpu basin, were used to determine the start time of the March 2021 event and track its evolution. In addition, we used 70 United States Geological Survey (USGS) seismic and meteorological records from monitoring stations at Nyingchi, Bomi (~80 km distance) and Milin (~80 km distance) to analyse the regional tectonic and climatic context of the event.
We used pre-and post-event 0.5 m-resolution tri-stereo optical Plé iades images (Plé iades-1A on December 30 2018 and Plé iades-1B on April 30 2021, respectively) to establish the source location of the avalanche, and used high-resolution (1 m) DEMs to establish the volume and size distribution of the initial detachment and quantify net elevation change along the 75 immediate flow path. The DEMs were derived from tri-stereo images by using PCI Geomatica software (Banff sp4) with the OrthoEngine module based on the photogrammetry principle. A sufficient number of Ground Control Points and Tie Points were also automatically collected to improve the computed math model in OrthoEngine module. In the module, the Semi-Global Matching (SGM) algorithm was used to match pixels. And the implementation of SGM is done by image-matching along epipolar lines. We used the demcoreg Python package to refine horizontal and vertical alignment of the DEM data (Shean 80 et al., 2020); the final products had a relative vertical accuracy of 0.51 ± 4.1 m over stable ground (Fig. S1).

Field observations of March 2021 avalanche and river blockage
The last automatic data transfer from our monitoring system at the outlet of Sedongpu valley (Fig. 1b) (Fig. 1h). The water level continued to increase at a rate of 0.6 -0.8 m/hour and rose by a total of 11 m before stabilising at around 18:00 PM on 23 March (Fig. S2). Both data streams (or sudden lack of) implied the blockage of the Yarlung Tsangpo, likely by a large mass flow event, and were useful for constraining its timing.
We undertook a field visit to area on 25 March 2021 and found evidence that a mass flow from Sedongpu basin had overtopped the 200 m-high hill at the basin outlet, destroying the combined AWS-time lapse camera monitoring station and 100 stripping the surrounding slope of vegetation ( Fig. 1d-f). This included 200 year-old pine forest, thereby providing insight into the minimum return period of an event of such magnitude from this basin. By applying a simple frictionless point mass model of = √2ℎ (Iverson et al., 2016) and assuming a runup height (h) of 200 m, and g = 9.81 m/s 2 , we estimate that a flow velocity (u) of around 60 m s -1 (~225 km/h) is required to achieve run up / superelevation to the height required for moraine overtopping. Post-event field observations showed the valley bottom to be covered by fresh, water-rich debris which was in 105 the process of dewatering, and widespread destabilisation of valley flanks along the flow path (Fig. 1g).
Examination of the seismic record at Nyingchi station reveals the onset of a clear ground-shaking event at 23:41 PM on 22 March (Fig. 2a). The seismic waveform has the typical characteristics of a landslide because it lacks the clear P-and S-wave arrival times typical for earthquakes (Ekströ m and Stark, 2013). The waveform suggests that the avalanche-mass flow lasted ~300 seconds and consisted of an initial phase (lasting 100 s) exhibiting a high-amplitude signal which we infer as representing 110 the detachment and downslope passage of the initial avalanche, followed by a ~200-second waveform with a weaker amplitude, which most likely reflects the passage of the continuous mobile avalanched mass flow through the Sedongpu basin and its outlet valley. Based on the duration of the seismic waveform and known distances (~11 km), we estimate that the mean velocity of the whole avalanche-mass flow reached ~37 m s -1 (~132 km/h). If the avalanche material runout the Sedongbu basin within the first 100 second, the mean avalanche velocity can reach as high as ~110 m s -1 (~396 km/h). These observations and 115 inferences were corroborated by local people at Gyala village, who report hearing a continuous, 'loud' sound originating from the direction of the Sedongpu basin (pers. comm.).

Avalanche source, magnitude and landscape change
The comparison of 0.5 m-resolution Plé aides satellite orthophotos (dated 30 December 2018 and 30 April 2021) and field investigation revealed that the avalanche originated from the western flank of a ridge that extends north from Gyala Peri ( Fig.   2b and Fig. 3a). The elevation difference between Plé aides -derived DEMs reveal that about ~50.0 ± 1.5 Mm 3 of ice and rock 125 detached from the mountain ridge (Fig. 3b,c), and we assume that the majority of this was released in the March 2021 event; we have not detected any large avalanching events, including those capable of transforming into a debris flow, since the October 2018 glacier detachment. The mean and maximum depth of the ice-rock avalanche was ~140 m and ~300 m, respectively, and the detachment scar spans an altitudinal range of ~6000 -~6500 m asl. with a total 2D area of ~0.36 km 2 .
The avalanche caused headward erosion of the ridgeline of up to ~160 m, and ~0.8 km laterally (Fig. 2b).
We posit that the avalanche contained a mixture of rock and glacier ice; a perched ice mass was on the ridgeline pre-event ( Fig. S3), whilst the 'fresh' appearance of the rock face immediately beneath the ridge implies the incorporation of at least some rock debris, but we remain uncertain on the exact ice:rock ratio. Analysis of previous events in the basin imply the potential for a high proportion of avalanche rock debris (Kä ä b et al., 2021)

Possible drivers for the 2021 ice-rock avalanche
The Gyala Peri region, and the southeastern Tibetan Plateau more broadly, are tectonically active. Seismic activity, a known mass movement avalanche trigger has been forwarded as a possible trigger mechanism for avalanche-driven debris flows originating from the Sedongpu valley in 2017 and into 2018 (Zhao et al., 2019). Indeed, the Gyala Peri region has experienced considerably more earthquakes compared to other nearby glacierized centres (e.g. Namjagbarwa, ~25 km distant; 155 Fig. S6). Whilst we do not detect heightened seismic activity around the time of the massive March 2021 avalanche, it is possible that seismic activity over preceding years and decades may have been a conditioning (rather than triggering) factor.
The frequent earthquakes, particularly the M6.9 earthquake on November 18 2017 and the following small magnitude aftershock (Zhao et al., 2019) near the Gyala Peri mountain, are likely to have enhanced any instabilities of both glaciers and rock masses in the Sedongpu valley, especially in the high mountain ridges where topographic amplification can occur. The 160 eventual 2021 failure at the ridge crest is more commonly associated with earthquake triggering in historical inventories (Densmore and Hovius, 2000) although the 2021 Chamoli event, which was also aseismic, was also sourced close to a ridge crest (Shugar et al., 2021).
Meteorological records from nearby monitoring stations show a significant increase in mean air temperature and a decrease in precipitation since the early 21 st century (Fig. S7a,b); the latter affects the southeast Tibetan Plateau more broadly. 165 We found that the 2021 ice-rock avalanche, and previous ones from the basin, occurred concurrently with, or shortly after the record positive air temperature anomalies (Fig.3d-f for mean minimum air temperature, Fig.S7c-e for mean air temperature).
The period January to March 2021 saw positive temperature anomalies in the range +1.6 -+1.8 °C, and which exceeded (Nyingchi, Bomi) or came close to exceeding (Milin) historical records (Fig. 3d). Similarly, in 2017 and 2020 we observe unprecedented positive temperature anomalies in the months of September and October at all three stations (+2.0 -3.3 o C); the 170 former coincides with the occurrence of the October 2017 ice-rock avalanche in the Sedongpu basin, whilst the record positive temperature anomalies in Autumn 2020 and late Winter-early Spring 2021 occur immediately before, and during, the period when the March 2021 avalanche occurred (Fig. 3f). In addition, the summer in 2018 was the warmest season during the past four decades. The mean air temperature during the June-August was about 0.95-1.