Environmental modelling · Landslide tsunami twin

Tracy Arm, 10 August 2025: a 481 m tsunami, rebuilt from the physics up

At 05:26 AKDT about 64 million m³ of rock fell from the wall above the retreating South Sawyer Glacier into Tracy Arm, south of Juneau. The wave ran 481 m up the opposite wall, the second-highest tsunami run-up on record. This twin reproduces the event with laboratory-derived impulse-wave theory and a 2D shallow-water solver on the real fjord terrain, then checks itself against what was observed.

Heller–Hager impulse theory2D NLSWE · wet/dryMonte CarloImpulse theory benchmarked vs Lituya 1958
Run the simulation See the validation
Loading terrain…

Upper Tracy Arm on USGS 3DEP terrain: slide scar, 481 m trimline and Sawyer Island (≥30 m wave).

Observed run-up

481m

2nd highest on record

Slide volume

64M m³

USGS preliminary: up to 100 M m³

Seismic equivalent

M5.4

Globally recorded long-period waves

Wave at Sawyer Island

≥30m

No vessels in the arm at 05:26

What it does

Three questions a hazard modeller asks of an event like this

Generates the wave

Turns slide volume, speed, thickness and impact angle into the impulse wave using the Heller–Hager impulse product parameter, checked against the Lituya Bay benchmark.

Propagates it through the fjord

A kinematic slide displaces water on the real Tracy Arm terrain; a well-balanced shallow-water solver carries the wave past Sawyer Island and the bends toward the cruise-ship lane.

Quantifies what it does not know

Monte Carlo over uncertain slide inputs, a grid-convergence study and a modelled-vs-observed scorecard, so every number comes with its error bars.

Why it matters

Glacier retreat is loading the gun in Alaska's fjords

South Sawyer Glacier retreated rapidly between 2 and 5 August 2025, leaving an over-steepened wall without the ice that had buttressed it. Microearthquakes accelerated for more than a day before the slope failed. The same chain (retreat, debuttressing, failure into deep water) produced the 2015 Taan Fiord tsunami (193 m run-up), and Tracy Arm sees more than 20 tour vessels a day in summer. At 05:26 on a Sunday none were in the arm. Exposure, not only hazard, sets the outcome, so the simulator reports wave arrival and height at the mid-arm vessel anchorage.

Recent Alaskan landslide tsunamis

  • Lituya Bay 1958524 m
  • Taan Fiord 2015193 m
  • Pedersen Lagoon 202417 m wave
  • Tracy Arm 2025481 m

How it works

From a DEM to a run-up map in five steps

  1. 01

    Terrain

    USGS 3DEP DEM + NOAA multibeam (2005–2018) → UTM 8N grids at 60 m (fjord) and 30 m (upper arm)

  2. 02

    Slide

    Energy-line impact velocity; 64 M m³ block moving as a bed deformation

  3. 03

    Impulse

    Heller–Hager P → a_M, H_M, T_M and decay laws

  4. 04

    Propagation

    2D non-linear shallow water, MUSCL–Rusanov, wetting/drying

  5. 05

    Run-up & risk

    Run-up laws, shoreline water level, gauges, arrival at the vessel lane

Platform

Three modules

2D Simulator

Kinematic slide + shallow-water solver on the real fjord

Impulse Lab

Heller–Hager wave theory, run-up laws and Monte Carlo

Validation

Benchmarks, grid convergence and observed vs modelled

Methodology

The physics, the numerics and the assumptions

01Terrain and bathymetry

Land

USGS 3DEP 1/3 arc-second DEM, reprojected to NAD83 / UTM 8N and block-averaged to 60 m over Holkham Bay → Sawyer glaciers (40 × 26 km).

Water

Five NOAA NOS multibeam surveys (BAG, 2005–2018) cover 84% of the fjord, shifted from MLLW to MSL (+2.62 m, Juneau). The fjord the glacier uncovered after 2018 is extended to the 2025 terminus with a 30° trough capped at the nearest surveyed depth.

z = DEM on land z = survey(MLLW) − 2.62 m in surveyed water z = −min(d·tan30°, max survey within 1 km) in gaps
02Slide kinematics

Impact velocity

Energy balance along the slide path with a dynamic bed friction angle δ (Heller & Hager 2014). Δz_sc = 550 m, δ = 20°, α = 40° gives 78 m/s, in line with the >70 m/s flow speeds reported for 2025.

Kinematic slide in the 2D model

A smooth block (length V/(s·b)) leaves the scar along a fixed azimuth at Vs and decelerates uniformly over its runout. Its thickness is added to the bed each step, so the water above is lifted and pushed aside.

Vs = [ 2 g Δz_sc (1 − tanδ · cotα) ]^½ z(x,y,t) = z₀(x,y) + s·f(ξ − ξc(t))·f(ζ)
03Impulse wave (Heller & Hager 2010)

Impulse product parameter

Slide Froude number F = Vs/√(gh), relative thickness S = s/h and relative mass M = mₛ/(ρw·b·h²) combine into one parameter that collapses 211 wave-channel experiments.

Generated and propagated wave

P gives the impact-zone maximum amplitude, height and period, and their decay with distance x. Validity: 0.17 ≤ P ≤ 8.13, 30° ≤ α ≤ 90°; the Impulse Lab flags inputs outside the tested range.

P = F · S^½ · M^¼ · [cos(6α/7)]^½ a_M = (4/9) P^⅘ h H_M = (5/9) P^⅘ h H(x) = (3/4) [P (x/h)^−⅓]^⅘ h T(x) = 9 [P (x/h)^5/4]^¼ (h/g)^½
04Shallow water solver

Equations

Depth-averaged mass and momentum with bed slope and Manning friction (n = 0.03). Hydrostatic, so a breaking 100 m wave becomes a shock (bore) that the scheme captures and dissipates.

Numerics

Finite volumes, Rusanov fluxes, MUSCL/minmod reconstruction of h, η, u, v, Heun RK2, CFL 0.45. Audusse hydrostatic reconstruction keeps a lake at rest over 40° walls and depths positive, so the walls wet and dry. Runs in a Web Worker.

∂h/∂t + ∇·(h u) = 0 ∂(h u)/∂t + ∇·(h u⊗u + ½ g h² I) = −g h ∇z − g n² u|u| / h^⅓
05Run-up

Empirical

Müller (1995) for impulse waves and Synolakis (1987) for solitary waves on a plane slope, each fed the decayed wave H(x) and the impact-zone wave H_M. The four results form the reported envelope.

2D

Highest water surface reached on shoreline or flooded cells, excluding cells the slide covered. On a coarse grid this is the resolvable proxy for run-up on a steep wall.

Müller: R/h = 1.25 (H/h)^5/4 (H/L)^−3/20 (90°/β)^⅕ Synolakis: R/h = 2.831 √cotβ (H/h)^5/4
06Uncertainty

Monte Carlo

4,000 seeded draws over volume, drop height, friction, slope, thickness, width, depth and run-up distance propagate through the empirical chain to a run-up distribution and P(R ≥ 481 m).

Numerical

A three-level grid study (180, 120, 60 m) quantifies discretisation error in the 2D model; it is reported, not hidden, on the Validation page.

X ~ U(lo, hi) for each input → R_lower, R_upper P(R ≥ 481 m) = #{R ≥ 481} / N

Data sources

Every input, with its data tier

Sourced: published observation or dataset. Modelled: an assumption of this twin, stated and varied in the Monte Carlo.

DatasetProviderUsed forTier
USGS 3DEP 1/3″ DEMU.S. Geological SurveyFjord walls, shoreline, run-up surfacesSourced
Event page: 2025 Tracy Arm landslide-generated tsunamiUSGS Landslide Hazards ProgramTiming, volume range, run-up 470–500 m, ≥30 m at Sawyer IslandSourced
Shugar et al. (2026), ScienceUniversity of Calgary + 18 co-authors64 M m³ volume, 481 m run-up, ~100 m breaking waveSourced
Alaska Earthquake CenterUniversity of Alaska FairbanksPrecursory seismicity, M5.4-equivalent signalSourced
Place names and positionsOpenStreetMap / GNISSawyer Island, Holkham Bay, glacier locationsSourced
NOS hydrographic surveys H13006, H13007 (2018), H11998, H11760, F00503NOAA NCEI (BAG)Bathymetry for 84% of the fjordSourced
Post-2018 fjord head depthThis twin (trough fill)Water the 2018 surveys could not reachModelled
Slide geometry and kinematicsThis twin, bracketed by Monte CarloThickness, width, drop height, azimuthModelled

References

Literature

  1. Heller, V. & Hager, W.H. (2010). Impulse product parameter in landslide generated impulse waves. J. Waterway, Port, Coastal, Ocean Eng. 136(3), 145–155.
  2. Heller, V. & Hager, W.H. (2014). A universal parameter to predict subaerial landslide tsunamis? J. Mar. Sci. Eng. 2(2), 400–412.
  3. Fritz, H.M., Hager, W.H. & Minor, H.-E. (2001). Lituya Bay case: rockslide impact and wave run-up. Science of Tsunami Hazards 19(1), 3–22.
  4. Fritz, H.M., Hager, W.H. & Minor, H.-E. (2004). Near field characteristics of landslide generated impulse waves. J. Waterway, Port, Coastal, Ocean Eng. 130(6), 287–302.
  5. Müller, D.R. (1995). Auflaufen und Überschwappen von Impulswellen an Talsperren. VAW Mitteilungen 137, ETH Zürich.
  6. Synolakis, C.E. (1987). The runup of solitary waves. J. Fluid Mech. 185, 523–545.
  7. Audusse, E., Bouchut, F., Bristeau, M.-O., Klein, R. & Perthame, B. (2004). A fast and stable well-balanced scheme with hydrostatic reconstruction for shallow water flows. SIAM J. Sci. Comput. 25(6), 2050–2065.
  8. Lynett, P. & Liu, P.L.-F. (2002). A numerical study of submarine-landslide-generated waves and run-up. Proc. R. Soc. Lond. A 458, 2885–2910.
  9. Shugar, D.H. and 18 co-authors (2026). Study of the 10 August 2025 Tracy Arm landslide and tsunami. Science, published 6 May 2026. doi:10.1126/science.aec3187.
  10. U.S. Geological Survey (2025). 2025 Tracy Arm landslide-generated tsunami. Landslide Hazards Program.