The Sikkim flood of October 2023: Drivers, causes and impacts of a multihazard cascade

RESEARCH ARTICLE

PART-I

Abstract

On 3 October 2023, a multihazard cascade in the Sikkim Himalaya, India, was triggered by 14.7 million m3 of frozen lateral moraine collapsing into South Lhonak Lake, generating an ~20 m tsunami-like impact wave, breaching the moraine, and draining ~50 million m3 of water. The ensuing Glacial Lake Outburst Flood (GLOF) eroded ~270 million m3 of sediment, which overwhelmed infrastructure, including hydropower installations along the Teesta River. The physical scale and human and economic impact of this event prompts urgent reflection on the role of climate change and human activities in exacerbating such disasters. Insights into multihazard evolution are pivotal for informing policy development, enhancing Early Warning Systems (EWS), and spurring paradigm shifts in GLOF risk management strategies in the Himalaya and other mountain environments.

Catastrophic water release from glacial lakes can cause far-reaching Glacial Lake Outburst Floods (GLOF), with impact up to hundreds of kilometers downstream (1–5). GLOFs often involve complex, cascading multihazard processes [c.f. (6)] and are particularly evident in steep mountainous regions like the Himalaya. These impacts and reach may be further extended by interaction with entrained and relocated deposits, leading to debris flows and debris floods (7–11).

South Lhonak Lake (SLL) in Sikkim, India is located at 5200 m above sea level (asl) in the Upper Teesta basin. It is one of the largest, fastest-growing, and most hazardous lakes in Sikkim with potential to cause significant downstream damage in the event of a GLOF (12–17). On 3 October 2023, SLL experienced an outburst, triggering a devastating flood cascade that killed 55 people, left 74 missing (18), and destroyed the 1200-megawatt (MW) Teesta-III hydropower dam. The flood cascade (3-4 October) impacted Sikkim, West Bengal, and had transboundary implications in Bangladesh (Fig. 1A).

(Fig. 1A. Overview of the 3-4 October 2023 GLOF cascade from South Lhonak Lake.

(A) Flood-impacted stretch of the Teesta River showing the location of SLL (27°54?20??N and 88°10?20??E) and the dominant flood processes along the channel (lake outburst, GLOF cascade), as well as the impacted hydropower plants, flood-triggered landslides, major population centers, discharge, and gauging stations. (B) Schematic showing the SLL outburst including the lateral moraine that collapsed into the lake and the breaching of the frontal moraine. (C) Pre- and post-GLOF high-resolution SkySat imagery (imagery © Planet Labs, 2023) showing the lake, failure zone of the northern lateral moraine, and the breached frontal moraine. (D to F) Field photographs show (D) the breached frontal moraine; pre- and post-surface along AA? and the eroded cross-sectional area (E), and (F) the northern lateral moraine failure zone, and the post-GLOF lake level drop. (G) Pre- and post-GLOF surface along cross-section BB?. Cross-section locations of AA? and BB? are shown in panel C. Photo credits: KSK and ITBP)

This paper presents a collaborative effort involving scientists, non-governmental organizations, and diverse stakeholders to investigate the SLL GLOF and the subsequent multihazard cascade. Our motivation is not only to understand this event but also to identify major findings of wider relevance given rapid climate warming in mountain regions worldwide. We analyze the drivers, causes, and downstream impacts of the hazard cascade using high-resolution satellite imagery, seismic data, meteorological data, field observations, and numerical modeling. We explore the triggers of the GLOF, prevailing meteorological conditions, long-term climatological influences, glacier mass balance, permafrost conditions, and reconstruct the hydraulic dynamics of the GLOF. Downstream implications are evaluated from the GLOF source to the confluence of Teesta and Brahmaputra rivers in Bangladesh at 385 km downstream, including (i) mapping of damaged infrastructure (buildings, roads, bridges, and hydropower plants) and agricultural land; (ii) erosion and deposition by the flood; (iii) impact of secondary triggered landslides; and (iv) transboundary impacts. Finally, we evaluate the long-term impact of the event on future hazards within the Teesta River system. The assessment indicates that the high hazard level arises not only from the flood itself but also from the series of subsequent processes it triggers. The hazard increases the vulnerability of Teesta valley to future events. Understanding these cascading and enduring effects is essential for developing effective strategies to manage GLOF risks in the Himalaya and similar mountainous regions worldwide.

The Himalaya contains more than 2,400 lakes larger than 0.1 km², and many of these are growing rapidly (19). The rivers draining High Mountain Asia also have a hydropower potential of 500 gigawatts (20), 80% of which remains untapped (21). Increasing demands for stable and renewable energy have driven a surge in hydropower development with >650 projects planned or under construction in High Mountain Asia (20). However, these hydropower installations are susceptible to a wide range of hazards, including GLOFs and their associated multihazard chain [e.g., (20, 22–26)]. Recent disasters include the 2016 Gongbatongsha/Upper Bhotekoshi GLOF in Nepal (9), and the 2021 Chamoli rock-ice avalanche in India (27), both of which destroyed hydropower plants. Through our analysis of this flood disaster, we aim to identify key insights to reduce risks and enhance multihazard management strategies for GLOFs across High Mountain Asia and similar regions around the world.

Drivers and causes of the 3 October 2023 GLOF from South Lhonak Lake

On 3 October 2023, the hazard cascade began with the collapse of ~14.7 × 10⁶ m³ of lateral moraine into SLL lake (Fig. 1, B and C ) at 22:12:20 Indian Standard Time (IST) [~16:42:20 Universal Time Coordinated (UTC)] which caused a tsunami (Figs. 2B) and 3A). The wave overtopped and eroded the frozen frontal moraine with a maximum wave run-up ~15 m over the frontal moraine. This resulted in a breach 165 m wide (top width) and 55 m deep (Fig. 1, D and fig. S1C) that released ~50 × 10⁶ m³ of water, approximately half of the total SLL volume (see methods sections “DEM of difference and uncertainty” and “Reconstruction of the GLOF cascade” for GLOF volume calculations and uncertainty). We calculated the observed GLOF volume from the lake level drop (~28 m) and the volume of collapsed moraine material deposited in the lake (Fig. 1, F and G), and figs. S1B and S8). The breach exposed massive buried dead ice embedded within the permafrost of the frontal moraine (fig. S1A).

(Fig. 2. Details of the moraine collapse.

(A) Displacement of the perennially frozen northern lateral moraine from 2016-17 to 2022-23; the moraine failure zone is marked on the displacement map of 2022-23. (B) Seismic waveforms and inverted force history of the lateral moraine collapse into the lake. (C) Trajectory of the mass movement (collapsed northern lateral moraine) for the first 200 s, from the force history inversion of the seismic signals.)

(Fig. 3. Summary of the reconstructed GLOF process chain.

(A) (left) Depth distribution of the collapsed moraine and the lake bathymetry immediately before the initial collapse of the northern lateral moraine (at 22:12:20 IST, reconstructed from the seismic data); maximum flow height and the reconstructed timing of the GLOF process chain; the reconstructed GLOF is compared to the observed inundation limits and arrival flood time at the ITBP camp located 10 km downstream of the lake. (B) Observed maximum erosion of the frontal moraine derived using DoD is compared to the (C) reconstructed timeline of the frontal moraine erosion. (D) Reconstructed GLOF process inferred from the modeled discharge (Q) vs. time plot of the two phases: lake water (P_L) and eroded sediments of the frontal moraine (P_S) at a cross-section CS-1 located immediately below the lake (see panel A for location). (E) Discharge (Q) and flow depth/flow velocity vs. time of P_L and P_S at cross-section CS-2. (F) Routed P_L from SLL to Teesta-III hydropower at Chungthang; subplots show discharge at four cross-sections along the flow channel (red hydrographs); at CS-6 the time vs. accumulated GLOF volume is shown (blue). Reconstructed average flow depths, velocity, and time of GLOF arrival every 5 km along the flow path are shown at the bottom and matched with the reported GLOF arrival time at the Teesta-III hydropower)

The moraine slide, measuring ~900 m in width and ~88 m in thickness, occurred on the North flank of the lateral moraine, close to the South Lhonak Glacier terminus (Fig. 1B). The dimensions and volume of this moraine collapse, as well as the frontal moraine erosion, were calculated from DEMs of Difference (DoD) created by differencing high-resolution (4 m resolution) pre- and post- Digital Elevation Models (DEMs), (methods section “DEM of difference and uncertainty”). SPOT-6 (1.5 m) and Pléiades (0.7 m) stereo-pairs were used to create the pre-GLOF (1 December 2018) and post-GLOF (29 October 2023) DEMs respectively. To quantify the pre-and post-GLOF lake level and moraine changes, we computed DEMs at 1 m resolution for 18 October 2022 and 29 October 2023, both from Pléiades stereoscopic images (28). We also obtained a 16 July 2017 DEM from the High Mountain Asia 8 m DEM Mosaics (29) (Fig. 1G).

We computed the displacement of the SLL northern lateral moraine using optical feature tracking (30, 31), applied to 257 satellite image pairs between January 2016 and September 2023 (methods section “Pre- and post-GLOF dynamics of the lateral moraine”). The moraine had a maximum coherent displacement >15 m per annum (m a^?1) between 2016 and 2023 (Fig. 2A and fig. S2) (median velocities are shown in fig. S3). We distinguish two primary displacement zones, one up-glacier from the 2023 glacier terminus (Zone 1; see Fig. 2A) and a second to the east of the terminus (Zone 2). The fastest slope velocities are found in Zone 1 (>10 m a^?1 since 2016) while Zone 2 accelerates from ~1 m a^?1 to ~10 m a^?1 from 2016 to 2023, with the most rapid speed immediately preceding the failure on 3 October (fig. S2). The two zones coalesce in 2022 to form a continuous complex of fast moraine displacement centering on the failure zone.

Seismic waveforms and spectrograms, from broadband stations near Mount Everest (EVN; 135 km away), Kathmandu (KKN; 286 km away), and Lhasa (LSA; 349 km away), indicated a potential landslide signal (methods section “Seismic records and GLOF signals”) (fig. S4). Seismic data force inversion pinpointed the moraine failure timing (32, 33) (Fig. 2B and fig. S5). We used the inverted force history and a mass of 2.875 × 10¹⁰ kg, based on a volume of 12.5 × 10⁶ m³ (failure mass above the lake surface) and an estimated density of 2300 kg m^?3 (considering it to be a mixture of ice and rock mainly comprising of Phyllite and Biotite-Gneiss) (34), to estimate the slide trajectory (Fig. 2C), which suggests a runout distance of 690 m, and movement to the southeast, consistent with the moraine collapse into the lake. The total maximum force was 2.8 × 10¹⁰ N, oriented largely N-S.

We employed a multiphase numerical model to reconstruct the SLL GLOF process chain (methods section “Reconstruction of the GLOF cascade”) that propagates as debris flood based on the mixture of the lake’s water and eroded moraine debris (35). We ran a simulation ensemble, varying the erosion coefficient and basal friction angle, and comparing this to reference datasets including flood arrival time, seismic records, observed GLOF inundation, and moraine erosion to identify the most suitable parameter combinations (fig. S7) (methods section “Reconstruction of the GLOF cascade”). The modeled collapse of the north lateral moraine, starting at 22:12:20 IST (as per seismic data in Fig. 2B), generated a tsunami ~20 m high at the impact site. The resulting overtopping wave initiated erosion of the frontal moraine until the maximum breach depth of 55 m (observed erosion, Fig. 3B) was reached at ~22:24:00 IST (Fig. 3C). Our model predicts up to ~16 m of moraine sediment accumulation at the bottom of SLL following the collapse (fig. S8). The reconstructed outflow water discharge (fluid phase, P_L) immediately downstream of the lake (at cross-section CS-1 shown in Fig. 3A) peaked at 4.85 × 10⁴ m³ s^?1 (Fig. 3D), with the eroded sediment discharge (solid phase, P_s) from the frontal moraine peaking at 1.03 × 10⁴ m³ s^?1. The GLOF peak discharge at CS-1 vastly exceeds meteorological flood magnitudes, suggesting that it is a rare event in the historical context of this region, equivalent to a return period exceeding 200 years (3) (fig. S9). The outflow hydrograph (at CS-1) revealed the GLOF process chain, where the initial impulse wave immediately after impact lasted for ~3 min, causing progressive erosion of the frontal moraine. This was followed by slow breaching of the moraine for ~13 min, revealed by the decreasing sediment discharge of the eroded frontal moraine (P_s) and gradually increasing water discharge (P_L). Thereafter, erosion further slowed until full breaching of the moraine was reached. The water discharge from the lake became constant after ~18 min. Modeled maximum flow depth and velocity of P_L at CS-2 (located 1 km downstream of the breached moraine) are 11 m and 26 m s^?1 respectively (Fig. 3E). The reconstructed GLOF reached the Indo-Tibetan Border Police (ITBP) camp, 7.12 km downstream, at ~22.30:00 IST, consistent with the reported arrival time from ITBP officials (personal communication). The GLOF reached Chungthang (location of the 1200 MW Teesta-III hydropower, at CS-6, Fig. 3F) at ~00:30:00 IST (4 October), in line with the reported arrival time (~00:35:00 IST on 4 October). The discharge peaked at 5340 m³ s^?1 at Chungthang within ~6 min of the GLOF’s first arrival with ~1.0 × 10⁶ m³ and ~3.5 × 10⁶ m³ of water accumulating in the first 5 and 10 min, respectively (Fig. 3F). The flow depth and velocity at Chungthang reached a maximum of 9 m and 9 m s^?1 respectively (Fig. 3). The GLOF inundation reconstruction showed good agreement with observations mapped using 2 m resolution Pléiades multi-spectral post-GLOF imagery acquired between 21 - 31 October 2023 and seismic flood signals (Fig. 3, A and F, and fig. S6).

TO BE CONTINUED

(www.science.org/doi/10.1126/science.ads2659?fbclid=IwZXh0bgNhZW0CMTEAAR2Xx8WLq9ZqVg_UhG3M2O2UZCFz5kBGOtSJ0TLxsZhNeYJytkyNaH7bH8M_aem_XDjyoEsPLbbclUL2C8-G3w)