PART-II
Moraine failure
conditioning factors
South
Lhonak glacier has undergone rapid mass loss in recent decades (figs. S12 and
S19), in common with many Himalayan glaciers (36, 37). This mass loss is driven by both long-term climate
warming and local topographic and glaciological forcings such as glacier-lake
interactions [e.g., (38, 39)]. As there are no local meteorological
observations, we use ERA5 Land to determine climatic trends and reconstruct the
mass balance of South Lhonak Glacier since 1951 (methods section
“Climatological drivers”). Annual mean temperature has warmed 0.08°C per decade
(fig. S10A), with monsoon/summer temperatures in June, July, August, and
September (JJAS) increasing at a slower rate of 0.04°C per decade (fig. S10B)
whereas JJAS total precipitation increased by 8.26 mm per decade since 1950
(fig. S11). The modeled mean mass balance of South Lhonak Glacier was -0.45 ±
0.33 m w.e. a?1 from 1950 to 2023 (fig. S12 and methods section
“Climatological drivers”). Over the past four years, mass loss increased to
-0.58 ± 0.33 m w.e. a?1, coinciding with lake expansion up to 100 m
a?1. The three warmest summers on record occurred in 2020, 2022 and
2023 (fig. S10B) (40).
Large-scale
deformation of the northern lateral moraine is consistent with our
understanding of other steep, frozen, and warming mountain permafrost slopes (41–43). Models incorporating climate data, incoming
solar radiation, and ground truth from viscous creep features (rock glaciers)
all predict permafrost occurrence within the SLL moraines (44–46),
of which we assess the properties and implications (figs. S13 to S18 and
methods section “South Lhonak Lake: Permafrost and related aspects”). Estimated
near-surface temperatures and permafrost depths are -1 to -3°C and around 100 m
for the northern lateral moraine (failure zone) and -3 to -6°C and >200 m
for the shaded southern moraine. We estimate permafrost warming reaching about
100 m below the surface, close to the slide detachment depth of 85 m (likely
near the local permafrost base) (methods section “South Lhonak Lake: Permafrost
and related aspects”). Field investigations reveal exposed dead ice and ductile
deformation on the exposed scarp of the breached frontal moraine, indicating
ice-supersaturated morainic material (fig. S1). The final collapse of this
northern moraine was connected to glacier retreat and lake growth (fig. S19),
as well as water input from a stream draining adjacent glacierized basins.
Pre-GLOF mapping of the collapsed moraine surface (1 January to 28 September
2023) exhibited mass movement scars depicting small-scale slope failures (fig.
S20). In these circumstances, slopes can progressively evolve toward a critical
threshold, or an external event can trigger slope failure (11).
Velocity mapping shows that the slope exhibited extensive, rapid deformation
for years preceding the collapse (Fig. 2A and figs. S2 and S3).
A
low-pressure cyclonic system, bringing heavy rainfall to Bangladesh, West
Bengal, and Sikkim on 3 and 4 October, was proposed as one potential trigger
for this collapse (Fig. 4, figs. S21 and S26, and methods section
“Short-term meteorological drivers”). ERA5 reanalysis shows 29-39 mm of
rainfall was recorded over SLL from 28 September - 5 October, with modest
amounts of ~5 mm and ~14 mm on 3 October and 4 October, respectively (fig.
S23). Also, ERA5, CPC, and IMERG rainfall pattern and amount all showed good
agreement over the rainfall gauging stations, over SLL, and for the region
10°N-30°N, 70°E-100°E, with the available data providing no evidence of a
triggering cloudburst event in the vicinity of SLL (figs. S22 to S27, table S9,
and methods section “Short-term meteorological drivers”). The rainfall
intensities observed are typical for this region and season, which suggests
that the impact of the event has been conditioned by processes that have
increased sensitivity of the hazard cascade to landscape and extreme rainfall
events.
(Fig. 4.
Meteorological conditions before, during, and after the GLOF.
(A)
Spatial distribution of daily geopotential height with winds at 700
hectopascals (hPa) isobaric surface over eastern India and Bangladesh from 28
September to 6 October 2023. (B) Spatial Distribution of daily ERA5
rainfall with winds at 700 hPa isobaric surface over eastern India and
Bangladesh from 28 September to 6 October 2023. The ERA5 rainfall was compared
to two station datasets: Lachen (in Sikkim) and Dalia (in Bangladesh) (figs.
S22, S24, and S25). Spatial distribution of ERA5 daily specific humidity is
shown in fig. S21.)
GLOF-induced erosion,
channel aggradation, and landslides
A total
of 45 secondary landslides (noted L1-L45) were observed, triggered by the GLOF
cascade (fig. S28). We mapped the landslides using 0.7 m resolution post-event
Pléiades imagery (acquired on 24, 29, and 31 October and 5 November 2023) in
the first 67.5 km downstream of SLL (figs. S28A, S30, and S31 and table S1) and
used 3 m PlanetScope imagery (acquired between 9 to 19 October 2023) further
downstream. Landslides mapped from satellite imagery were cross-referenced with
field evidence (fig. S28, B to J). No co- or post-GLOF landslides were detected
beyond 108 km downstream of the lake (520 m a.s.l). Erosion and deposition
volumes and their uncertainty along the Teesta Valley were calculated using DoD
of pre- (1 and 8 December 2018) and post-GLOF (24, 29, and 31 October and 5
November 2023) DEMs (methods section “DEM of difference and uncertainty”). At
35 km downstream, the Teesta River was dammed by a series of landslides (L6-L8)
(fig. S28M). We mapped the area of this landslide-dammed lake at ~8.185 × 103 m2 from
24 October 2023 Pléiades imagery. Results showed that the L6 deposits created a
dam with a maximum height of 19 m and volume of ~5.8 x104 m3.
The lake persisted as of 24 May 2024. Partial drainage of the lake occurred
through a channel cutting through the landslide deposits (fig. S29). Most
landslides resulted from lateral erosion of the valley walls by the GLOF,
destabilizing slopes and leading to their failures.
Between
35 km downstream of SLL and Teesta III, the flood wave eroded both vertically
and laterally (Fig. 5). Elevation differences measured before and after
the flood indicate lateral channel shifts of up to 100 m. Many of the
landslides were deep-seated, with depths up to 150 m (Fig. 5A). Lateral
erosion of the valley caused slumping in L43, where roads and concrete walls
were offset by several meters (fig. S34). The sustained geomorphic impact of
the flood along this part of the river can be attributed to channel steepness,
which generally increases shear stresses at the channel bottom, and prevents
attenuation of the flood peak (hydrographs in Fig. 3for CS-4 and
CS-5) (47). Also, downstream of 35 km, valley side walls (within
500 m around the river) have average slopes between 30-40°. These values are
consistent with the effective angle of internal friction controlling hillslope
stability (48), suggesting that this part of the river runs
through a valley prone to mass wasting. Valley cross-sections (Fig. 5F)
moreover show that landslides mostly led to a slope-parallel retreat of the
topographic surface, rather than decline in hillslope angle, suggesting that
hillslopes instantaneously adjusted to undercutting by lowering to a threshold
hillslope angle.
(Fig. 5. Summary
of the GLOF-induced erosion (observed).
(A)
(left) Distance from the lake (in km) vs. elevation (along GLOF flow channel),
maximum erosion depth, and maximum erosion volume; (right) the GLOF valley
showing the erosion zones in the 67.5 km stretch from the lake to Chungthang;
marked are the co- or post-GLOF landslides and field observations (fig. S37). (B to E)
Spatially distributed erosion depth of different sections from upstream to
downstream along the Upper Teesta Valley; the co- or post-GLOF landslides are
marked. (F) Pre- and post-GLOF elevation vs. distance plots across
cross-sections showing undercutting and lateral erosion at different locations
along the Teesta valley; the locations of the cross-sections (a-a? to d-d?) are
shown in (B) to (E).)
The total
eroded volume is estimated at ~270 × 106 m3, of
which combining GLOF erosion and triggered landslides occurred upstream of
Chungthang is ~233 × 106 m3 (in the first 67.5
km stretch) (Fig. 5Aand table S3). In terms of volume of material, this
would be equivalent to basin-wide erosion of 9 cm across the entire catchment
(area = 3021 km2) upstream of the Lachung-Teesta confluence. Only 7%
of the total eroded volume is observed in the first 30 km of the channel where
GLOF-triggered landslides are absent (Fig. 5A). The erosion volume
increases in the landslide-dominated stretch from 30 km downstream of the lake
to Chungthang (Fig. 5A). Maximum erosion of 66.5 × 106 m3 occurred
40-45 km downstream (table S3), where reconstructed GLOF flow velocity is
maximum (fig. S36). The triggered landslides can be attributed to river erosion
induced by high flow velocities with substantial lateral and vertical erosion
observed in the field and remotely at various locations downstream along this
stretch (fig. S37). Field observations suggest that the transition from erosion
to aggradation occurred downstream of the Lachung-Teesta confluence (near
Mangan) (Fig. 6). The town of Rangpo, ~135 km downstream of the lake,
was severely impacted by the debris, burying buildings and automobiles (Fig.
6, C and D). Other severely impacted areas include Geli Khola, Teesta
Bazaar, and Bardang (Fig. 6, B, E, and F).
(Fig. 6. Field
evidence of sediment aggradation.
(A to F)
Photographs taken along the Teesta River show the aggradation of the sediments
transported by the flood cascade and its impact. Latitude, longitude, and
elevation (in m a.s.l) are at top right; locality name and distance from SLL
are at bottom right. Photo credits: Praful Rao (co-author).)
Impacts on population,
infrastructure, agricultural land, and transboundary implications
The flood
cascade damaged ~25,900 buildings, 59% built in the last decade (Fig. 7A,
table S4, and methods section “Mapping exposed elements”). Most affected
buildings are located below Chungthang, within 200-385 km of the lake, with the
most heavily inundated zone between 290 and 385 km downstream in Bangladesh.
Similarly, ~276 km2 of agricultural land was flooded (Fig.
7Band table S5). A total of 31 major bridges made up of Reinforced Cement
Concrete (RCC) or steel (Bridges: B1-B31) (18) along the Teesta
River were damaged, including 14 upstream of Chungthang (Fig. 7C, figs.
S39 and S40, and table S6). Moreover, ~20 small pedestrian bridges in Sikkim
were also affected (18). A road length of ~18.5 km was damaged,
~6.4 km of which was due to secondary landslides (fig. S32 and table S7).
Approximately 200 buildings were impacted by these triggered landslides, 90% of
which were caused by the two largest adjacent landslides, L33 and L35, located
60 km downstream of SLL (figs. S28, K and L, and S33 and table S1). A total of
10 landslides damaged the road network (fig. S32 and table S2), L43 (known as the
Naga landslide) causing maximum damage in terms of road length (fig. S34).
(Fig. 7. Summary
of the damage assessment.
(A)
Flood inundated buildings along the entire stretch of the Teesta valley;
Bar-plots show the number of inundated buildings in every 10 km stretch along
the flood path; Pie-charts show the percentage of damaged buildings existing in
2013-15 and the damaged buildings constructed in the last decade in every 40 km
stretch along the flood path. (B) Flood inundated agricultural land; bar
plot shows inundated agricultural land for every 10 km stretch along the flood
path. (C) Flood-damaged major bridges (B1 - B31) and hydropower plants;
bar plot shows the number of bridges damaged in every 10 km stretch along the
flood path.)
The GLOF
and associated erosion volumes destroyed the 1200 MW Teesta-III hydropower dam
at Chungthang. The cascading flood continued downstream, affecting another four
dams: Teesta V, Teesta VI, Teesta Low Dam III, and Teesta Low Dam VI (Fig.
7Cand fig. S38). Field visits to assess the impact of the flood were
undertaken along the Teesta Valley (figs. S38 to S40). Post-disaster surveys by
a multi-stakeholder team constituted by the Sikkim State Disaster Management
Authority (SSDMA), including sector experts, government representatives,
international organizations, and others (18) revealed that the GLOF impacted 100 villages in
Mangan, Pakyong, Gangtok, and Namchi districts, causing 55 deaths, 74 missing
persons, over 7025 displaced individuals, and significant livestock losses,
including 547 cattle, 62 sheep, 664 goats, 586 pigs, 7252 poultry, 51 calves,
and 200 rabbits (18). Transboundary flood impacts included
infrastructure damage in Bangladesh, particularly in Rangpur district (fig. S41
and methods section “Transboundary implications and sediment transport”). Other
affected districts were Lalmonirhat, Kurigram, Gaibandha, and Nilphamari before
the floodwaters discharged into the Brahmaputra River. Water levels in the
Teesta River in Bangladesh rose around noon on 4 October, ~16 hours after the
initiation of the GLOF. Rainfall, water level, and sediment discharge data from
17 September 2023 to 29 October 2023, collected by the Bangladesh Water
Development Board (BWDB) at the Dalia station (26.1758°N, 89.0505°E, in Dimla
Upazila Nilphamari District) (methods section “Transboundary implications and
sediment transport”), which is the first station to encounter the flood along
the path of Teesta in Bangladesh (see Fig. 1for location),
indicated that water levels on 4 October 2023 reached ~52 m, perilously close
to the dangerous threshold of 53.15 m. Despite minimal rainfall on 4 October,
the water levels mirrored those of 24 September, when Dalia Station recorded
substantial rainfall (~150 mm), suggesting that the elevated discharge on 4
October was primarily due to the upstream flood cascade.
Post-flood,
weekly suspended sediment discharge at the Dalia station between 8 October and
15 October 2023, reached 6587.5 kg s?1 (on 15 October 2023),
which is respectively 5 times and 2.8 times higher than the average and maximum
discharge in the preceding month (September 2023) (fig. S41D). This spike in sediment
discharge was 17 times higher than in the week preceding the flood event. An
increase in the river turbidity also occurred upstream of Dalia and at the
confluence of Teesta and Brahmaputra rivers (figs. S42 and S43). The coarse
sediment discharge peaked on 8 October 2023 and was respectively 8 times and
6.5 times higher than the average and maximum discharge in the preceding month.
The analysis shows that more than ~17,000 buildings in Bangladesh were impacted by the flood, with ~50% built in the last decade (Fig. 7A). The total agricultural land inundated in Bangladesh was 168 km2 (Fig. 7B). The easterly movement of the low-pressure system caused heavy rainfall, exceeding 300 mm per day in several places, in Bangladesh, and 75 mm at Dalia station (Fig. 4B and figs. S25 and S26), from 5 to 7 October (Fig. 4B), contributing to the flooding impact. Thus, the effects in Bangladesh were due to both the GLOF cascade on 4 October and the intense rainfall that followed immediately on 5 October 2023.
TO BE CONTINUED
(https://www.science.org/doi/10.1126/science.ads2659?fbclid=IwZXh0bgNhZW0CMTEAAR2Xx8WLq9ZqVg_UhG3M2O2UZCFz5kBGOtSJ0TLxsZhNeYJytkyNaH7bH8M_aem_XDjyoEsPLbbclUL2C8-G3w)
