These maps show the sea floor topography and backscatter intensity of the Hudson Canyon region on the continental slope and rise offshore of New Jersey and New York (fig. 1 and fig. 2). Sheet 1 shows sea floor topography as shaded relief. Sheet 2 shows sea floor topography as shaded relief with backscatter intensity superimposed in color. Both sheets are at a scale of 1:300,000 and also show smoothed topographic contours at selected intervals. The maps are based on new multibeam echo-sounder data collected on an 18-day cruise carried out aboard the National Oceanic and Atmospheric Administration (NOAA) Ship Ronald H. Brown during August and September 2002. Additional multibeam data of the Hudson Canyon collected by the Woods Hole Oceanographic Institution (WHOI), on the continental shelf collected by the STRATAFORM project (Goff and others, 1999), and a survey of the Hudson Shelf Valley (Butman and others, 2003), and a compilation of bathymetric data from the National Geophysical Data Center (NGDC) Coastal Relief Model provide coverage of areas surrounding Hudson Canyon (fig. 2). Interpretations of the surficial geology also utilize widely spaced 3.5- and 10-kiloHertz (kHz) high-resolution seismic profiles collected by the U.S. Geological Survey (fig.2).

Assessment of the probability for destructive landslide-generated tsunamis depends on the knowledge of the number, size, and frequency of large submarine landslides. This paper investigates the size distribution of submarine landslides along the U.S. Atlantic continental slope and rise using the size of the landslide source regions (landslide failure scars). Landslide scars along the margin identified in a detailed bathymetric Digital Elevation Model (DEM) have areas that range between 0.89 km2 and 2410 km2 and volumes between 0.002 km3 and 179 km3. The area to volume relationship of these failure scars is almost linear (inverse power-law exponent close to 1), suggesting a fairly uniform failure thickness of a few 10s of meters in each event, with only rare, deep excavating landslides. The cumulative volume distribution of the failure scars is very well described by a log�normal distribution rather than by an inverse power-law, the most commonly used distribution for both subaerial and submarine landslides. A log�normal distribution centered on a volume of 0.86 km3 may indicate that landslides preferentially mobilize a moderate amount of material (on the order of 1 km3), rather than large landslides or very small ones. Alternatively, the log�normal distribution may reflect an inverse power law distribution modified by a size-dependent probability of observing landslide scars in the bathymetry data. If the latter is the case, an inverse power-law distribution with an exponent of 1.3 ± 0.3, modified by a size-dependent conditional probability of identifying more failure scars with increasing landslide size, fits the observed size distribution. This exponent value is similar to the predicted exponent of 1.2 ± 0.3 for subaerial landslides in unconsolidated material. Both the log�normal and modified inverse power-law distributions of the observed failure scar volumes suggest that large landslides, which have the greatest potential to generate damaging tsunamis, occur infrequently along the margin.

Landslides are often viewed together with other natural hazards, such as earthquakes and fires, as phenomena whose size distribution obeys an inverse power law. Inverse power law distributions are the result of additive avalanche processes, in which the final size cannot be predicted at the onset of the disturbance. Volume and area distributions of submarine landslides along the U.S. Atlantic continental slope follow a lognormal distribution and not an inverse power law. Using Monte Carlo simulations, we generated area distributions of submarine landslides that show a characteristic size and with few smaller and larger areas, which can be described well by a lognormal distribution. To generate these distributions we assumed that the area of slope failure depends on earthquake magnitude, i.e., that failure occurs simultaneously over the area affected by horizontal ground shaking, and does not cascade from nucleating points. Furthermore, the downslope movement of displaced sediments does not entrain significant amounts of additional material. Our simulations fit well the area distribution of landslide sources along the Atlantic continental margin, if we assume that the slope has been subjected to earthquakes of magnitude ≤ 6.3. Regions of submarine landslides, whose area distributions obey inverse power laws, may be controlled by different generation mechanisms, such as the gradual development of fractures in the headwalls of cliffs. The observation of a large number of small subaerial landslides being triggered by a single earthquake is also compatible with the hypothesis that failure occurs simultaneously in many locations within the area affected by ground shaking. Unlike submarine landslides, which are found on large uniformly-dipping slopes, a single large landslide scarp cannot form on land because of the heterogeneous morphology and short slope distances of tectonically-active subaerial regions. However, for a given earthquake magnitude, the total area affected by subaerial landslides is comparable to that calculated by slope stability analysis for submarine landslides. The area distribution of subaerial landslides from a single event may be determined by the size distribution of the morphology of the affected area, not by the initiation process.

The nearly complete coverage of the U.S. Atlantic continental slope and rise by multibeam bathymetry and backscatter imagery provides an opportunity to reevaluate the distribution of submarine landslides along the margin and reassess the controls on their formation. Landslides can be divided into two categories based on their source areas: those sourced in submarine canyons and those sourced on the open continental slope and rise. Landslide distribution is in part controlled by the Quaternary history of the margin. They cover 33% of the continental slope and rise of the glacially influenced New England margin, 16% of the sea floor offshore of the fluvially dominated Middle Atlantic margin, and 13% of the sea floor south of Cape Hatteras. The headwall scarps of open-slope sourced landslides occur mostly on the lower slope and upper rise while they occur mostly on the upper slope in the canyon-sourced ones. The deposits from both landslide categories are generally thin (mostly 20-40 m thick) and comprised primarily of Quaternary material, but the volumes of the open-slope sourced landslide deposits can be larger (1-392 km3) than the canyon-sourced ones (1-10 km3). The largest failures are located seaward of shelf-edge deltas along the southern New England margin and near salt domes that breach the sea floor south of Cape Hatteras. The spatial distribution of landslides indicates that earthquakes associated with rebound of the glaciated part of the margin or earthquakes associated with salt domes were probably the primary triggering mechanism although other processes may have pre-conditioned sediments for failure. The largest failures and those that have the potential to generate the largest tsunamis are the open-slope sourced landslides.