Hyperconcentrated floods cause extreme gravel transport through the sandy rivers of the Gangetic Plains

The Gangetic Plains support approximately 10% of the global population and host densely settled, flood-prone communities. Rivers descending from the Himalaya exhibit an abrupt gravel–sand transition (GST) 10–40 km downstream of the mountain front, marked by a tenfold drop in channel gradient and a shift from gravel to sand dominance. Historical floods over the last 10–100 years show no evidence of gravel transport beyond the GST, yet Holocene Kosi megafan cores document intermittent coarse gravel deposition 30–40 km south of the modern GST. Such downstream delivery of coarse clasts onto low-gradient sandy reaches would reduce channel conveyance and promote avulsion and floodplain inundation. Existing paleo-flood research largely reconstructs extreme discharges within confined Himalayan valleys (often linked to glacial lake outburst floods or landslide-dam failures), many of which dissipate before reaching the foreland. Elevated suspended sediment concentrations are known to enhance bedload entrainment, with hyperconcentrated flows defined by ~8–40 vol.% suspended sediment. This study leverages the Siwalik stratigraphic record near the ancient GST, combining sedimentology of isolated conglomerate event beds with entrainment and transport calculations to estimate paleo-discharge, suspended sediment concentrations, and bedload flux. These reconstructions are compared against modern flood return intervals (Karnali River), to test whether extreme, sediment-laden floods can transport coarse gravel far beyond the GST over short timescales relevant to flood hazards.

Prior studies document: (1) the GST as a stable geomorphic boundary with rapid grain-size coarsening at sandstone–conglomerate transitions in Himalayan systems; (2) catastrophic floods from GLOFs and landslide-dam failures producing very large discharges in mountain valleys; (3) enhanced bedload mobilization with increased suspended sediment concentrations and the defining characteristics and triggers of hyperconcentrated flows (e.g., intense rainstorms, lake-breakout floods, landslides, and floodplain recycling); (4) Holocene evidence from the Kosi megafan of distal gravel layers implying episodic downstream transport of coarse material; and (5) methods for estimating paleoslopes, channel geometries, suspended sediment profiles (e.g., Rouse), and bedload transport rates (e.g., Meyer-Peter–Müller and subsequent modifications for highly turbulent, unsteady flows). These works provide the theoretical and empirical foundation for reconstructing transport conditions required to deliver coarse gravels tens of kilometers beyond the GST.

Study area and stratigraphic context: The Miocene Siwalik Group exposed along the Mohand anticline (NW India) was logged, focusing on the upper Middle Siwalik sandstones and the overlying Upper Siwalik conglomerates. The Middle–Upper contact is interpreted as the ancient GST. Isolated conglomerate beds preserved below this contact record episodic coarse gravel transport into distal sandy plains.

Sedimentary logging: A detailed 1:50-scale sedimentary log was measured along the Mohand anticline near the Chakrata River (grid 43R 0757817; 3359903). Grain sizes, sedimentary structures, and bed contacts were recorded. Full logs are in Supplementary Notes 8.

Grain-size analysis of event beds: For two key conglomerate event beds (C1 and C2), high-resolution photographs (5184 × 3888 px) were analyzed (Erdas). Pebble c-axes beneath 100-node grids were measured; a b/c correction factor of 1.5 (from modern Karnali quartzite pebbles) converted c- to b-axis for transport calculations. Conglomerates are predominantly quartzite (C1: 95%; C2: 98%). Median clast sizes: C1 D50 = 52 mm; C2 D50 ≈ 45 mm.

Estimating transport distance beyond GST: The downstream position of event bed deposition relative to its time-equivalent GST was estimated using Himalayan convergence rate (V), foreland sediment accumulation rate (S), and stratigraphic thickness (t) between the event bed and the Middle–Upper Siwalik contact. Assuming steady-state accumulation and an approximately stable mountain front–GST distance, the transport distance l was derived from V, S, and t (details in Methods and Supplementary Figs.). Western foreland V ≈ 12–20 mm/yr; accumulation rates from Quaternary cores and Siwalik data were used to bracket l.

Paleo- and modern channel geometries: Modern bankfull depths and widths for rivers downstream of the GST (Karnali and Ganga) were taken from ADCP cross-sections. Ancient bankfull depths were estimated from preserved barform thicknesses in the Mohand, Surai Khola, and Karnali Siwalik sections, acknowledging preservation and compaction biases. Channel slopes were obtained from SRTM DEM profiles with 10-km moving averages and separated into upstream/downstream GST reaches. Paleoslopes were estimated via force-balance relationships: for sandy reaches (below GST) using Ganti et al. (2019) with Monte Carlo sampling of Shields stress (θ) over dune to upper-plane bed regimes and D50 ranges (0.25–1 mm), and for gravelly reaches (above GST) using Paola & Mohrig (1996), Sest = 0.094 D50/h with Monte Carlo sampling of D50 (40–60 mm) and bedform heights (h).

Suspended sediment concentration modeling: Depth-averaged suspended sediment concentration c was estimated by integrating a Rouse-type profile, with reference concentration ca and reference level za functions of transport parameter Ts, shear stress components, and bed roughness (z0 = 25 D50 for sand; 6.8 D50 for gravel). Mixture density ρ = ρw(1−c)+ρc c (ρw = 1000 kg/m3; ρc = 2650 kg/m3) informed shear velocity and transport estimates. Grain settling velocities and shear parameters followed standard formulations (Soulsby, van Rijn; detailed in Methods and Supplementary Notes).

Bedload transport calculations: Bedload transport per unit width q was computed via the dimensionless transport rate Φ using the Meyer-Peter–Müller (MPM) relation Φ = (θ − 0.047)2 and a modified form for highly turbulent, erosive, unsteady flows, Φ = 9 θ (θ − 0.047)2, incorporating a turbulence modification coefficient (φ) following Cao et al. (2011), with φ spanning 1 (steady) to 6 (highly turbulent). Threshold for motion was evaluated from u2 = 0.047 g (s−1) D50 2 / cf using water depth h = 11 m and appropriate skin-friction coefficients. Channel cross-section was idealized as rectangular; discharge Q = u A with A = 0.9 B h (B from geomorphic constraints). Sensitivity tests varied D50 (±10 mm), φ (1–6), and c as a function of flow velocity.

Event reconstruction and return intervals: Using C2 geometry (2 m gravel base; 11.3 m total thickness; D50 ≈ 45 mm) and a minimum river depth >11 m during the event, scenarios were explored to transport sufficient gravel to deposit a 2 m-thick bed 10 km downstream of the GST within 12–24 hours. Suspended sediment concentration–discharge combinations were compared to return-interval estimates for the Karnali River at Chisapani (1962–2014), modeled with a Gumbel distribution and 95% confidence limits.

• The Siwalik stratigraphy at Mohand records two thick conglomerate event beds within Middle Siwalik sandstones: C1 (2 m conglomerate, D50 = 52 mm; total 6.1 m with overlying sandstone) and C2 (2 m conglomerate, D50 ≈ 45 mm; total 11.3 m). Seventeen thinner pebble layers (20–50 mm basal lags) also occur, with total bed thicknesses 2–14.2 m.
• Lack of internal sorting/imbrication in the 2 m conglomerates and massive overlying sandstones indicate rapid deposition from turbulent, sediment-laden flows consistent with hyperconcentrated conditions.
• The C2 event bed is estimated to have been deposited 7–25 km downstream of its time-equivalent GST, consistent with Holocene Kosi megafan gravel layers 30–40 km south of the modern GST.
• Threshold bulk velocity to mobilize 45 mm gravel at h = 11 m is ~3.75 m/s, corresponding to Q ≈ 12,550 m3/s. Under clear-water, low suspended-sediment conditions, moving gravel 10 km beyond the GST would require months of sustained discharge exceeding twice average monsoon levels—improbable given observed flood durations (hours to a day).
• Hyperconcentrated flows (≈8–40 vol.% suspended sediment; occasionally >50 vol.% observed elsewhere) damp turbulence locally, increase mixture density, reduce settling velocities, and can narrow/deepen channels, enhancing downstream gravel transport as wash/load coupling.
• Using modified MPM with φ = 3–6 and hyperconcentrated suspended loads, sufficient gravel to form a 2 m-thick bed can be transported 10 km beyond the GST: in under 24 h by a 200–1000 yr flood with Q ≈ 23,200–27,500 m3/s; in under 12 h by a 500–2000 yr flood with Q ≈ 26,200–30,500 m3/s. Transit time sensitivity to D50 variations (±10 mm) is up to ~6%.
• Contemporary observations in the Karnali downstream of the GST during moderate monsoon show near-bed concentrations ~6 vol.% and depth-averaged ~1 vol.%; the C2 event implies substantially higher concentrations and deeper flow (>11 m), likely achieved by entrainment of bed/bank sediments during hyperconcentrated flood passage.
• Climate projections indicate up to ~60% increases in the intensity of extreme rainfall in western Nepal, implying higher likelihood of such extreme, sediment-laden floods in coming decades.

The study demonstrates that transporting coarse gravel far into the low-gradient sandy reaches of the Gangetic Plains is feasible during rare, extreme monsoon floods only when hyperconcentrated conditions are triggered. Stratigraphic evidence (C1, C2, and thin gravel lags) indicates short-lived, high-energy events capable of bypassing the GST and depositing coarse material 7–25 km downstream, aligning with Holocene Kosi core records. Purely hydrodynamic (clear-water) floods of realistic duration cannot account for these deposits; elevated suspended sediment concentrations fundamentally alter flow rheology and bed–suspension coupling, enabling rapid gravel transport at discharges consistent with 200–1000 yr (or rarer) events. Potential triggers include landslide-dam failures and/or GLOFs that deliver abundant fines combined with intense monsoon rainfall and high floodplain/bed recycling near the mountain front. The implications for hazard are substantial: coarse deposition on low-gradient channels reduces conveyance, increases flood stage for moderate future events, promotes super-elevation and avulsion, and deposits persistent gravel on agricultural land. With projected increases in extreme rainfall intensity and persistent seismicity, the frequency of hyperconcentrated, gravel-transporting floods may increase, challenging current flood risk management paradigms.

By integrating Siwalik stratigraphic observations with transport modeling, the study shows that hyperconcentrated, sediment-laden floods during extreme monsoon events can transport coarse gravel tens of kilometers beyond the GST into the sandy rivers of the Gangetic Plains within hours. This reconciles distal gravel layers in both Miocene Siwalik and Holocene Kosi records with plausible hydraulics and sediment concentrations. The work quantifies discharge and concentration thresholds (e.g., Q ≈ 23,200–27,500 m3/s for 24 h transport over 10 km; higher for 12 h scenarios) and underscores the pivotal role of suspended sediment in modulating bedload entrainment and transit times. Future research should: (1) monitor and model hyperconcentrated flows in Himalayan foreland channels, including real-time suspended sediment concentration and rheology; (2) refine probabilistic flood hazard models to include sediment-laden flood scenarios and channel capacity feedbacks; (3) improve constraints on source-to-sink sediment supply during compound events (monsoon + landslides/GLOFs); and (4) integrate stratigraphic archives with modern observations to calibrate event frequencies and magnitudes for risk management.

Key limitations include: (1) Assumption of an approximately stable mountain front–GST distance through time; significant progradation/retrogradation would alter inferred transport distances, though the lack of thick conglomerates in Middle Siwalik supports relative stability. (2) Idealized channel geometry (rectangular cross-section) and the assumption that gravel is transported across the full channel width; in reality, transport likely localizes in the thalweg, which would reduce required discharge thresholds. (3) Use of empirical bedload relations (MPM) and a turbulence modification coefficient (φ) derived from dam-break/landslide-lake experiments; natural flow complexities and unsteadiness introduce uncertainty of up to orders of magnitude in q estimates. (4) Grain-size uncertainties (D50 ± ~10 mm) modestly affect transit times (~6%), while roughness and concentration profiles under hyperconcentrated conditions are simplified via adapted Rouse formulations. (5) Lack of direct contemporary observations of hyperconcentrated flows in the distal Nepal plains; inference relies on analogs (e.g., Chinese rivers) and stratigraphic proxies. (6) Suspended sediment data used for calibration are mostly from moderate floods; extreme-event concentrations were estimated from theory and analog observations. Together these make the discharge–concentration thresholds conservative upper bounds, with real events potentially achieving transport at lower discharges if gravel transport is confined to the deepest, fastest parts of the channel.