PART-IV
Long-term
trends in air temperature, precipitation, and modeled mass balances
No
long-term climate data are available near South Lhonak Glacier; therefore, we
analyzed ERA 5-Land reanalysis data using a 0.1° grid cell over the lake (91).
Following well-established approaches in climatology, the Mann–Kendall
non-parametric test was used in combination with Sen’s slope estimator to
calculate the trend and magnitude of any change over time. Over the 71-year
(1951 – 2023), the annual mean temperature has warmed by around 0.56°C (0.08°C
decade?1) (fig. S10A). If only the monsoon/summer period [June, July, August,
and September (JJAS)] is considered, the warming has been about half this rate,
with a total warming of 0.28°C (0.04°C per decade?1) (fig. S10B). Despite a
moderate warming rate, the past four years have been characterized by
anomalously warm summers – with 2020, 2022, and 2023 being the three warmest
summers on record (40). While regional attribution studies are lacking, there
is a strong anthropogenic signal seen in general warming over Asia since around
the 1950’s (92).
The mean
annual precipitation is ?1150 mm w.e. over 1950-2023 with a maximum
contribution of 48% during monsoon months (JJAS), followed by 28% from winter
months (DJFM) and almost equal contributions of ?12% from pre- [May and June
(MJ)] and post-monsoon [October and November (ON)] months. The maximum
precipitation in summer months (MJJAS) suggests that South Lhonak Glacier is a
summer-accumulation type glacier where accumulation and ablation occur
concurrently. The mean long-term trend over mean annual precipitation sums
suggested an increasing precipitation trend of ?21 mm w.e. over 1950-2013 (fig.
S11). The mean total precipitation for JJAS is around 560 mm w.e. a?1 for the
reference period of 1991 - 2020, with a statistically significant increasing
trend of 8.6 mm decade?1 over 1951 - 2023. The 2023 monsoon was not remarkable,
bringing near-average conditions.
The
annual glacier-wide mass balances of South Lhonak Glacier were estimated by
applying a temperature-index model (93) using the ERA5-Land data over 1950-2023
(fig. S12). This model is specially tailored for the data-scarce Himalayan
region and has been successfully applied on several glaciers. For the South
Lhonak Glacier, the model was calibrated using the threshold temperature and
precipitation gradient against the available geodetic mass of -0.49 ± 0.05 m w.e.
a?1 over 2000-2019 (37). The other model parameters (melt factors for snow and
ice, threshold temperature for snow/rain, and temperature lapse rates) were
adopted from DokrianiBamak Glacier, which is a monsoon-dominated glacier
similar to South Lhonak. The uncertainty in annual mass balances is estimated
following the procedure in (94).
The mean
annual glacier-wide mass balance was estimated to be -0.45 ± 0.33 m w.e. a?1,
corresponding to a cumulative mass loss of -33.16 ± 2.82 m w.e. over 1950-2023
(fig. S12). This mass wastage is similar to the observed wastage at the
regional scale (36, 95). An increased mass wastage of -0.52 ± 0.33 m w.e. a?1
was observed post-2000 compared to -0.42 ± 0.33 m w.e. a?1 over the pre-2000
period, which is in line with continued warming (fig. S12), the continued lake
expansion (methods section “Expansion of South Lhonak Lake” of section
“Climatological drivers”) and the previous studies (95, 96). The wastage (-0.58
± 0.33 m w.e. a?1) has increased significantly over the past four years, marked
by unusually warm summers, with 2020, 2022, and 2023 being the three warmest
summers on record (fig. S12).
Expansion
of South Lhonak Lake
SLL was
first noted as a small supraglacial lake in the 1960s (12, 14), expanding
dramatically from an area of around 0.15 km2 in 1975, to 1.68 km2, in September
2023, just before the outburst event on 3 October 2023 (fig. S19A and table
S8). This equates to an average rate of areal expansion of 0.032 km2a?1 from
1975 – 2023. There has, however, been a notable doubling in the rate of
expansion over the past 2 decades, from a rate of 0.023 km2a?1 over the period
1975 - 2004, compared to a rate of 0.046 km2 a?1 since 2004. While the initial
formation of glacial lakes is directly linked with the thinning and retreat of
the parent glaciers from their little ice age moraines in response to
20th-century climate warming (90), calving processes and feedbacks decouple
lake expansion from the climate signal over time (97, 98). Hence, a continuous
expansion of SLL is observed despite fluctuations in the overall long-term
warming trend.
While this expansion has dramatically increased the water volume of SLL and thereby increased the potential volume and intensity of the outburst event, the lake expansion and associated glacial retreat have also played a major role in destabilizing the lateral moraine and eventual collapse of the northern (orographic left) moraine on 3 October 2023. Along this moraine, the lake has expanded, and the glacier retreated at an average rate of 47 m a?1 (fig. S19B). However, there was a period of enhanced calving and retreat at a rate of 130 m a?1 between 2010 and 2013, when the subsequent failure zone began to lose the buttressing support of the glacier and became exposed to lake water. It cannot be excluded that a subaqueous toe of the glacier extended further out into the lake and remained in contact with the base of the moraine within this zone for some years thereafter. Again between 2019 - 2020 and 2021 - 2022, glacial retreat/lake expansion along the 3 October 2023 failure zone exceeded 100 m a?1.
***
Meteorological conditions during the October 2023 Sikkim flood
Analysis
of the geopotential height at 700 hPa isobaric surface, IMERG daily rainfall
data, and specific humidity at 700 hPa reveals that the October 2023 Sikkim
flood was significantly influenced by the proximity of a low-pressure system or
cyclonic circulation located south and southeast of Sikkim from 28 September to
6 October. As this system moved over West Bengal and Bangladesh on 3 and 4
October, it triggered substantial increases in heavy rainfall along its path,
impacting most parts of northeast India, including West Bengal, Sikkim, and
Bangladesh.
A
detailed daily comparison of moisture, circulation, rainfall, and pressure
fields indicates that when the weather system was near the Myanmar coast on 28
September 2023, the weather over northeast India was mostly clear. As the
system moved northwest into the northern Bay of Bengal and reached the state of
Odisha in India during 29-30 September, a significant increase in moisture
content and rainfall was observed over the states of Odisha, Bihar, Jharkhand,
and Bengal. Notably, after 1 October 2023, the system ceased its westward
movement and began drifting northeast toward Bengal. During this eastward
movement, it triggered heavy rainfall over its eastern sector, where northward
winds dominated. The interaction of these northerly winds with the Himalayan
topography likely enhanced orographic rainfall over the Sikkim region.
As the
system advanced eastward, heavy rainfall was recorded in several places in
Bangladesh from 5-7 October. Once the system moved past Bengal, rainfall over
Sikkim significantly decreased, indicating that the flood event was closely
modulated by this low-pressure system. Additionally, persistent rainfall
occurred in southern Sikkim before the lake burst event on the night of 3
October. As the rainfall system moved from south to north due to the northward
background winds of the low-pressure system, conditions deteriorated further.
The lake water flowed through regions already affected by previous rainfall,
exacerbating the situation.
In
essence, the heavy rainfall and subsequent floods in Sikkim, Bengal, and
Bangladesh were directly influenced by the low-pressure monsoon system that
originated in the Bay of Bengal. This system recurved toward Sikkim during 2-4
October 2023, modulating the cascade of flood and landslide events.
Mapping
exposed elements
We mapped
elements exposed to the 2023 SLLGLOF and areas impacted by triggered landslides
along the Teesta River in India and Bangladesh (tables S1, S2, and S4 to S7).
We focused on quantifying the change of exposure to buildings in the past
decade and the totals of buildings were surveyed for two points in time (2013
and 2023). The pre-event building footprints for May 2023 (v3) are sourced from
(117, 118). For data accuracy of buildings refer to (117, 118). Taking the 2023
building dataset as baseline, we employed 2011-2014 high-resolution Maxar
imagery (Google Earth) to identify buildings that existed at that time.
Bridges
and roadways impacted by GLOF cascade and triggered landslides were mapped
using 0.7 m resolution post-event Pléiades imagery (acquired on 24, 29, and 31
October and 5 November 2023) for the first 67.5 km downstream of the lake. The
PlanetScope imagery (3 m resolution) from 9 October to 24 October 2023 were
utilized further downstream (fig. S39). The existing bridges (pre-GLOF) were mapped
using high-resolution Maxar imagery from Google Earth. Road networks for the
states of Sikkim, West Bengal, and Bangladesh obtained from OpenStreetMaps
(119) were overlaid on the post-event imagery to extract the impacted roads
both by flood and triggered landslides. We used pre-flood PlanetScope imagery
(acquired on 31 July, 3 September, 16 September, and 28 September 2023) for
mapping the areal extent of impacted agricultural land.
The
limitations/uncertainties resulted from the size of the study area (hundreds of
km along Teesta River), the number of exposed elements (tens of thousands), and
the data used. Our analysis of exposure change considers only two points in
time (2013 and 2023 before the GLOF) and cannot capture any changes with finer
temporal resolution, e.g., exposure changes associated with damages caused by
seasonal floods. Rather than going into details, we aim to provide elementary
statistics to document the changing number/area of exposed elements between a
decade before and shortly before the 2023 GLOF. Further, it is important to
mention that the location of elements within the 2023 GLOF impact area does not
necessarily imply that these elements were damaged or destroyed. Therefore, we
primarily report changing exposure of buildings rather than conclusively
stating the extent of damages associated with the 2023 GLOF, unless explicitly
known (e.g., as for bridges).
Transboundary
implications and sediment transport
To evaluate the transboundary implication of the flood cascade we collected gauged data from the BWDB for water level in the Teesta River, rainfall, and sediment discharge. The data was collected for the Dalia station (26.1758 N, 89.0505 E) located in DimlaUpazila (an administrative region or sub-district) of the Nilphamari District, the first station to encounter the floods along the Teesta’s path in Bangladesh (see Fig. 1Opens in image viewer for location). We analyzed sediment discharge available at 7-day temporal intervals from 17 September 2023 to 29 October 2023. Daily water level and rainfall data were analyzed for this period. BWDB employs a Binckley Silt Sampler, a cylindrical device with a uniform opening, for collecting suspended sediment samples (120). This instantaneous sampler is lowered into the water column and triggered at specific depths (0.2 and 0.8 of the total depth) by pulling a wire. At each depth, 1000 ml of water is collected. The samples undergo a two-step process to determine the total suspended sediment concentration:
Coarse
sediment analysis
Samples
are allowed to settle for 100 s. The settled portion (coarse sediment) is then
collected and analyzed to determine its concentration per liter volume using a
dispersion method.
Fine
sediment analysis
A separate sample, collected from the top of the water column for all verticals, is sent for analysis. The analysis includes filtration techniques to obtain the average concentration of finer sediment particles. The total suspended sediment concentration for each vertical is calculated by adding the concentration of coarse sediment (obtained from on-site analysis) to the average concentration of fine sediment. The sediment transport load is calculated and expressed in kilograms per second (kg s?1).
Flood
impacts
The collapse of a ~90 m high embankment in Gangacharaupazila (administrative division) of Rangpur district was reported to have destroyed 11 houses in the PaschimIsli village along the banks of the Teesta River (121). Immediately downstream of this region, in the Char Isli area of the Gangacharaupazila (Rangpur district), the flood inundation and erosion destroyed several houses (fig. S41E). The flood washed away 73 houses while temporarily displacing 33,000 people in the five districts including Rangpur, Lalmonirhat, Kurigram, Gaibandha, and Nilphamari. At Rajarhat in Kurigram district, the collapse of another embankment was reported (121). Significant asset losses were recorded in Lalmonirhat, Kurigram, and Rangpur due to the submersion of residential and agricultural properties (122). In these districts, 21 unions faced inundation, with houses submerged under up to about 1 m of water and extensive agricultural lands affected. High rainfall was recorded in Bangladesh on 5 October 2023, a day after the GLOF cascade entered Bangladesh. The impacts in Bangladesh were due to a combined effect of the GLOF cascade on 4 October and the intense rainfall that followed immediately on 5 October 2023.
CONCLUDED
