Modelling large tsunamis generated by Cascadia Subduction Zone earthquakes Josef Cherniawsky1, Fred Stephenson1, Kelin Wang2, Bodo de Lange Boom1 and Vasily Titov3 1Institute of Ocean Sciences and 2Pacific Geoscience Centre, Sidney, B.C., 3PMEL, NOAA, Seattle, WA THE CASCADIA SUBDUCTION ZONE The Cascadia Subduction Zone is a very long sloping fault that stretches from mid-Vancouver Island to Northern California. It separates the Juan de Fuca and North America plates. New ocean floor is being created offshore of British Columbia, Washington and Oregon. As more material wells up along the ocean ridge, the ocean floor is pushed toward and beneath the continent. The Cascadia Subduction Zone is where the two plates meet. The diagram on the right shows subduction of the oceanic Juan de Fuca plate beneath the continental crust of North America. Geophysical evidence demonstrates that the subduction zone is locked and accumulating strain that will be released in future great (magnitude 8 or larger) earthquakes. The diagram also shows the locations of large historic, crustal and subcrustal earthquakes in southern British Columbia and northern Washington. HISTORICAL EVIDENCE There is abundant geological evidence in tidal marshes along the Pacific coast from Vancouver Island to northern California for repeated, historically unprecedented great earthquakes in the recent past. A layer of clean sand in the picture on the right (photo by John Clague, SFU) is sharply bounded by peat and mud in a pit dug at a marsh just east of Tofino on the west coast of Vancouver Island. The sand occurs as a sheet that thins and fines landward and contains marine microfossils. It was deposited by a landward surge of seawater at the time of the last great earthquake at the Cascadia subduction zone in CE (Common Era) 1700. The figure on the right shows a 6-m long core sample taken from the deep sea floor showing fine grained mud layers alternating with sandier layers. The latter are interpreted to have been deposited from submarine landslides triggered by great earthquakes. The mud layers were formed by the slow continuous rain of finer sediment settling from the ocean. The volcanic ash at the bottom of the sample is dated as about 7800 years old (Adams, 1990). This computer-generated image (courtesy of Kenji Satake) shows the tsunami produced by the great (magnitude 9) Cascadia earthquake of January 26, 1700, six hours after initiation. The tsunami moved across the Pacific Ocean and produced destructive waves up to several metres high along a 1000-km length of the coast of Honshu in Japan. Much larger waves struck the west coast of North America less than 30 minutes after the shaking stopped. Deposits from this tsunami are preserved in tidal marshes and low-elevation coastal lakes on the Pacific coast. Their distribution provides information on the wave run-up that can be expected from future Cascadia tsunamis. TSUNAMI MODEL (MOST3) Numerical model MOST3 (Method of Splitting Tsunamis, version 3) computes generation, propagation and runup of tsunami waves and currents (Titov and Synolakis, 1995, 1996, 1997). Wave propagation is accomplished with a numerical dispersion scheme and non-linear shallow-water wave equations in spherical coordinates. We present here the results from numerical experiments designed to compute waves and currents in Victoria and Esquimalt Harbours after a large Cascadia earthquake. Three nested grids are used in the model and are shown on the right: Nx Ny latitude range longitude range A 351x700 43.000-50.000N 122.000-129.000W B 870x900 47.750-49.250N 122.100-125.000W C 1200x675 48.400-48.461N 123.368-123.476W Each finer grid communicates with a coarser grid through common open boundaries. The model was designed to compute shoreline wetting (wave runup) and drying (wave retreat) in its finest grid (C), which in our case has a grid size of about 10 m. However, due to lack of detailed digital elevation data on land, the coast was assumed to be a steep wall and only the drying in very shallow seas was computed here. MODEL RESULTS The following plots (and animations) show the results from numerical simulations on all three grids, using one particular earthquake scenario, that is an initial deformation of the sea floor (at time = 0). We also run experiments for other deformation scenarios, calculated by the Theoretical Modelling Group of the Pacific Geoscience Centre. The three plots below show initial propagation (first 40 minutes) of the tsunami wave, starting from time zero. The wave propagates both west and east. First West Coast communities are “hit” after 20 minutes. After 40 minutes, most of the outer coast is affected and the first wave has entered Juan de Fuca Strait. Note the typical dispersion pattern behind each wavefront, as short waves propagate slower than the longer ones. Because of the length of the source (> 1000 km), each coastal irregularity acts a secondary “source” and the waves persist longer than 12 hours. The plots below show four snapshots of propagation of tsunami waves inside Juan de Fuca and Georgia Straits. The first positive wave enters Esquimalt Harbour after about 1 hour and 20 minutes, Victoria Harbour after 1 hour and 25 minutes, and continues through Haro Strait, approaching Sidney after 1 hour and 50 minutes. Saanich Inlet and White Rock “see” their first wave after about 2.5 hours, after a second positive wave enters Esquimalt and Victoria Harbours. The plots above show snapshots of waves and currents inside Esquimalt (top row of panels) and Victoria (second row) Harbours at four different times, starting from 1 hour and 25 mins and ending at 2 hours and 25 minutes. The total duration of the numerical simulation was 12 hours and even after this relatively long time, the waves continue to arrive (albeit smaller in magnitude) from the entrance of Juan de Fuca Strait, or reflecting from other places in the strait. Depending on the earthquake scenario used, the maximum of simulated wave amplitude was between 4.1 and 4.5 m in Esquimalt Harbour and somewhat smaller in Victoria Harbour. Values of maximum water currents inside these two harbours are shown on the left. Occasionally these maximum speeds exceed 25 knots, though more typical maximum values approach 15 knots. C A B 1