Automated feature extraction and spatial organization of seaoor pockmarks, Belfast Bay, Maine, USA Brian D. Andrews a, , Laura L. Brothers b , Walter A. Barnhardt a a U.S. Geological Survey, Woods Hole Coastal and Marine Science Center, 384 Woods Hole Rd. Woods Hole, MA 02543, USA b University of Maine, Department of Earth Sciences, Orono, ME, 04469, USA abstract article info Article history: Received 29 January 2010 Received in revised form 5 August 2010 Accepted 16 August 2010 Available online 20 August 2010 Keywords: Pockmarks Geomorphometry Marine geology Methane Seaoor pockmarks occur worldwide and may represent millions of m 3 of continental shelf erosion, but few numerical analyses of their morphology and spatial distribution of pockmarks exist. We introduce a quantitative denition of pockmark morphology and, based on this denition, propose a three-step geomorphometric method to identify and extract pockmarks from high-resolution swath bathymetry. We apply this GIS-implemented approach to 25 km 2 of bathymetry collected in the Belfast Bay, Maine USA pockmark eld. Our model extracted 1767 pockmarks and found a linear pockmark depth-to-diameter ratio for pockmarks eld-wide. Mean pockmark depth is 7.6 m and mean diameter is 84.8 m. Pockmark distribution is non-random, and nearly half of the eld's pockmarks occur in chains. The most prominent chains are oriented semi-normal to the steepest gradient in Holocene sediment thickness. A descriptive model yields eld-wide spatial statistics indicating that pockmarks are distributed in non-random clusters. Results enable quantitative comparison of pockmarks in elds worldwide as well as similar concave features, such as impact craters, dolines, or salt pools. Published by Elsevier B.V. 1. Introduction First identied in muddy sediments of the Scotian Shelf (King and MacLean, 1970), pockmarks are seaoor depressions that are found worldwide in a variety of geologic settings (Hovland and Judd, 1988; Judd and Hovland, 2007). These craters can measure hundreds of meters in diameter, may occur in chains kilometers long and, where present in extensive elds, may dominate the seaoor surface (Fader, 1991; Rogers et al., 2006; Pilcher and Argent, 2007). Despite global distribution and general association with seaoor uid escape, the mechanisms for pockmark formation and evolution remain uncertain (Ussler et al., 2003). Analysis of pockmark morphology and spatial distribution relative to antecedent geology and subsurface uids (e.g., methane) can provide insight into uid-migration pathways, pockmark eld evolution, and possible mechanisms for pockmark generation and maintenance. In the absence of high-resolution seaoor bathymetry data, previous characterizations of entire pockmark elds relied upon visual interpretation of acoustic backscatter data for pockmark delineation, size statistics and spatial distribution (Fader, 1991; Kelley et al., 1994; Gontz et al., 2002; Rogers et al., 2006). Although acoustic backscatter data were the best available in the cited studies, interpreting size dimensions of concave features, such as pockmarks from these data is often ambiguous (Song, 2007). High-resolution bathymetry data collected by multibeam echosounder and swath sonar technologies enable the study of seaoor morphology to reach scales and resolutions similar to studies in subaerial geomorphology based on digital elevation models (DEMs) (Hughes Clarke et al., 1996). With one exception (Webb et al., 2009) these new technologies have not been applied to an entire pockmark eld. Instead, whole-eld spatial and morphologic analysis has ebbed, replaced by characterization of small portions of a eld or focus on individual pockmarks (Pilcher and Argent, 2007; Wildish et al., 2008). This change in emphasis is in part due to differing research objectives, but also is related to the challenges associated with feature extraction from acoustic data, particularly of features exhibiting complex morphologies that exist over an entire pockmark eld. This paper presents a novel method for pockmark feature extraction based upon the principals of geomorphometry and established algorithms for surface analysis (Pike, 1995). We apply this approach to a well-studied pockmark eld located in Belfast Bay, Maine (Fig. 1). Our objectives are to: 1) develop a technique for pockmark identication and extraction, which may be applied to similar landforms/objects (e.g. planetary craters, thermokarst lakes, dolines, and salt pools); 2) apply this technique to high-resolution swath bathymetric data collected for the Belfast Bay eld; and 3) characterize the spatial and morphological distribution of pockmarks in relation to the eld's subsurface geology as determined from concurrently collected CHIRP seismic data, providing results that can Geomorphology 124 (2010) 5564 Corresponding author. Tel.: + 1 508 548 8700x2348; fax: + 1 508 457 2310. E-mail address: bandrews@usgs.gov (B.D. Andrews). 0169-555X/$ see front matter. Published by Elsevier B.V. doi:10.1016/j.geomorph.2010.08.009 Contents lists available at ScienceDirect Geomorphology journal homepage: www.elsevier.com/locate/geomorph