The maximum elevation of the landslide dam deposit is at 845 m. This exposure is revealed in a series of hillslope gullies above an active sand quarry near the confluence of the Jinsha and Xiaojiang River (26° 30’ 22” N; 102° 03’ 56” E). A distinctive bench at the same elevation extends along the eastern side of the Xiaojiang River and can also be traced in pockets of the steep-sided Jinsha valley upstream of the confluence. The downstream extent of the deposit was then traced to identify the most likely landslide location. The most northerly extent of the deposit was observed in the sidewall of a gully in agricultural land in the village of Yanpengzi (26° 50’ 13” N; 106° 52’ 56” E) at an elevation of 790 m on the western side of the river. Downstream of this location the material exposed in gully walls is indicative of debris flow deposits with poorly sorted angular clasts of mixed lithology in a reddish matrix. On the opposite (eastern) side of the river an extensive rampart of material with a similar reddish matrix extends to approximately 1200 m, while an elongated ridge with a crest elevation of 900 m extends some 6 km downstream, parallel to the present river course (Figure 2). The source area for the landslide is on the western (Sichuan) side of the river. Triassic sandstones form the ridge line above the river and occur in a syncline which is dipping northwards at 10-20°. The Geological Survey of China has mapped the above-mentioned rampart on the eastern side of the river as an outcrop of similar Triassic rocks. However, these Triassic rocks are not in situ east of the fault line and must originate from detachment from the Sichuan side of the river, where they form a prominent escarpment at 2000 to 2500 m elevation at a distance from 8–10 km from the river (Figure 3).
At ground level it is difficult to delimit the landslide complex but using satellite imagery and digital elevation models the landslide location and volume can be estimated. The topography suggests two principal headscars and run-out tracks a short distance below the Triassic escarpment (Figure 4). The length of the track from the headscars to the river is at least 7.5 km with a width of 5 km. Given the antiquity of the slide estimating the depth to the shear surface from the topography is difficult and hence any estimate of the volume of the disturbed mass is provisional. Analysis of some of the largest terrestrial landlsides [7] suggested an assumed average depth of 100 m is reasonable. The volume of the landslide displaced from the Sichuan side of the river would then be at least 3.75 km3. Though the site requires a more detailed field investigation, the estimate suggests it is among the largest terrestrial landslides previously identified (within the top 100 sites identified) and may be considerably larger. A back-tilted linear feature extends along a fault line running north from the dam site suggesting a block detachment process continuing for 8 km downstream. A rampart at an elevation of 900 m extends for the same distance along the eastern side of the river as far as Qiaojia.
At the landslide location, the present day elevation of the river is 650 m. Based on a maximum elevation of sand deposits and associated benches in the inundated area of 845 ± 10 m, it is estimated that the minimum height of the landslide dam was approximately 200 m. Backfilling a digital elevation model derived from 3-arc-second Shuttle Radar Topography Mission (STRM) data, the dimensions of the lake can be estimated (Figure 5). The lake extended a length of 105 ± 4 km up the Jinsha Valley and was 1–2 km wide between the Xiaojiang confluence and the dam site. The estimated lake volume was 11.4 ± 1.3 km3. Contemporary annual discharge measured at Huantan (1962–1987), the nearest gauging station downstream is 122 ± 21 km3, suggesting that the lake would fill rapidly (between 15 to 100 days depending on whether the event occurred during summer monsoon or winter dry season). However, this also raises the question of how such a large river could be blocked by landslide debris. Ongoing work is seeking to date the lake sediments to establish both the occurrence and duration of the landslide dam. As climate conditions during the Last Glacial Maximum (c. 22 kBP) were colder and drier than present, the discharge of the Jinsha would have been significantly lower and the probability of landslide blocking the valley would have been higher [15].
The lake volume of the Jinsha landslide dam is similar to the largest known event in historic times – the 1911 Usoi Dam in Tajikistan [16]. However, its significance is the recognition that landsliding is capable of disrupting one of the world’s largest rivers at a considerable distance downstream. This merits reappraisal of the potential hazard posed by earthquake triggered landslide dams. In comparison, the 2008 Wenchuan Earthquake generated more than 250 landslide dams, but the largest of these at Tangjiashan, which causes considerable hazard management challenges, impounds a lake volume three orders of magnitude smaller than the Jinsha landslide dam [17, 18]. An empirical relationship to predict peak discharge from catastrophic failure of natural dams [19], Qp = 0.3 [VD]0.49 where Qp = peak discharge (m3 s−1); V = lake volume (m3); D = lake depth (m), suggests a peak discharge > 300,000 m3 s−1 if the dam failed catastrophically with a resultant flood wave impacting the channel downstream for several hundred kilometers. Furthermore, as the river enters a steep gorge downstream of Qiaojia where the active channel width is only 120–200 m, the unit stream power of a dam failure flood would be extremely high.