Optimizing an Inner-Continental Shelf Geologic Framework Investigation through Data Repurposing and Machine Learning

Elizabeth A. Pendleton; Edward M. Sweeney; Laura L. Brothers

doi:10.3390/geosciences9050231

Pendleton, Elizabeth A.;Sweeney, Edward M.;Brothers, Laura L.;

2019-05-21 00:00:00

geosciences Article Optimizing an Inner-Continental Shelf Geologic Framework Investigation through Data Repurposing and Machine Learning 1, 2 1 Elizabeth A. Pendleton * , Edward M. Sweeney and Laura L. Brothers Woods Hole Coastal and Marine Science Center, U.S. Geological Survey, Woods Hole, MA 02543, USA; [email protected] Santa Barbara Museum of Natural History Sea Center, Santa Barbara, CA 93101, USA; [email protected] * Correspondence: [email protected] Received: 30 March 2019; Accepted: 15 May 2019; Published: 21 May 2019 Abstract: The U.S. Geological Survey (USGS) and the National Oceanic Atmospheric Administration (NOAA) have collected approximately 5400 km of geophysical and hydrographic data on the Atlantic continental shelf between Delaware and Virginia over the past decade and a half. Although originally acquired for dierent objectives, the comprehensive coverage and variety of data (bathymetry, backscatter, imagery and physical samples) presents an opportunity to merge collections and create high-resolution, broad-scale geologic maps of the seaﬂoor. This compilation of data repurposes hydrographic data, expands the area of geologic investigation, highlights the versatility of mapping data, and creates new geologic products that would not have been independently possible. The data are classiﬁed using a variety of machine learning algorithms, including unsupervised and supervised methods. Four unique classes were targeted for classiﬁcation, and source data include bathymetry, backscatter, slope, curvature, and shaded-relief. A random forest classiﬁer used on all ﬁve source data layers was found to be the most accurate method for these data. Geomorphologic and sediment texture maps are derived from the classiﬁed acoustic data using over 200 ground truth samples. The geologic data products can be used to identify sediment sources, inform resource management, link seaﬂoor environments to sediment texture, improve our understanding of the seaﬂoor structure and sediment pathways, and demonstrate how ocean mapping resources can be useful beyond their original intent to maximize the footprint and scientiﬁc impact of a study. Keywords: hydrographic data; geophysical data; machine learning; geologic maps; seaﬂoor geology; MBES data; backscatter 1. Introduction The U.S. Geological Survey (USGS) began as a collaborative project with the University of Delaware, the National Park Service, the Mid-Atlantic Coastal Resilience Institute, and the National Oceanographic and Atmospheric Administration (NOAA) in 2014 to deﬁne the geologic framework of the Delmarva Peninsula coastal system. This mapping eort builds on recent and ongoing hydrographic, geologic, and ecological studies in the area and represents an opportunity to construct a geospatial framework around existing datasets, and acquire new data to ﬁll knowledge gaps. Characterizations and classiﬁcations are a key component of all geologic seaﬂoor studies, and machine learning is a powerful tool for creating classiﬁcations through the assimilation and interpretation of large hydrographic and geophysical datasets. Geologic studies of the seaﬂoor support coastal and ocean science and management by providing valuable baseline information to monitor seaﬂoor change, identify benthic habitats and sediment resources, understand sediment transport pathways, and bolster Geosciences 2019, 9, 231; doi:10.3390/geosciences9050231 www.mdpi.com/journal/geosciences Geosciences 2019, 9, 231 2 of 24 risk assessments and geohazard investigations, such as vulnerability to storms, sea-level rise, shoreline erosion, earthquakes, and tsunamis [1,2]. NOAA carried out 31 hydrographic surveys between 2006 and 2013 at 40-m line spacing using multibeam echosounders and sidescan sonars, over more than 5000 square kilometers of the mid-Atlantic inner-continental shelf adjacent to the Delmarva Peninsula, in water depths of 2.5 to 37.5 m, for the purposes of updating nautical charts (Figures 1 and 2; Table 1) [3–33]. In 2014 and 2015, the U.S. Geological Survey collected swath bathymetry, sidescan sonar, chirp and multi-channel boomer seismic-reﬂection proﬁles, bottom photographs, and sediment samples at 200-m line spacing over some of the same area, for the primary purpose of acquiring seismic data [34,35]. Coincident swath bathymetric and backscatter data provided the opportunity for change-detection comparisons with the hydrographic surveys [36,37], while bottom photographs and samples, collected with a GoPro modiﬁed, SEABed Observation and Sampling System (SEABOSS), provided validation [38]. These two regional mapping eorts by the USGS and NOAA provided a wealth of seaﬂoor data and an opportunity to combine datasets and use the hydrographic data beyond its original intent while at the same time increasing the resolution and extent of the USGS framework study. This data collaboration optimizes resources, expands coverage, and is an example of the “map once, use many times” mantra that exempliﬁes interagency and interdisciplinary cooperation. Machine learning has been applied to bathymetric and backscatter datasets to classify the seaﬂoor and benthic communities [39–45] for over two decades. Previous studies have tested the ability and performance of dierent and increasingly sophisticated machine learning algorithms on these types of data [41,42,44,46]. Supervised, object-based techniques (i.e., random forest) typically produce classiﬁcations with better accuracy than unsupervised (e.g., iso cluster) or simpler supervised methods (i.e., maximum likelihood classiﬁcation (MLC)) [41]. Here we tested the ability of simple and more complex classiﬁcation methods to discriminate classes of the seaﬂoor using bathymetric and backscatter datasets and derivatives from 33 surveys on the continental shelf oshore of the Delmarva Peninsula. Since the hydrographic data were not acquired with geologic products as an objective and, as such, contained numerous artifacts associated with acquisition that typically pose a challenge to machine learning, the ﬁrst goal of this study was to determine if the repurposed hydrographic data could be used to conduct an automated seaﬂoor classiﬁcation, given the artifacts and acquisition dierences among surveys. With a modest goal established, unsupervised and supervised methods were tested for feasibility on a subset of the data, then applied to the entire 33-survey dataset following a successful feasibility test. This study utilizes repurposed hydrographic data, applies machine learning algorithms for ecient and objective classiﬁcation, and produces geologic seaﬂoor data products for a large area of the mid-Atlantic shelf. Geosciences 2019, 9, 231 3 of 24 Geosciences 2019, 8, x FOR PEER REVIEW 3 of 24 Figure 1. Map showing location of the Delmarva Peninsula and a hillslope shaded relief map [3–36] Figure 1. Map showing location of the Delmarva Peninsula and a hillslope shaded relief map [3–36] of of the seafloor area to be interpreted in this study. Basemap from Esri, DeLorme, The General the seaﬂoor area to be interpreted in this study. Basemap from Esri, DeLorme, The General Bathymetric Bathymetric Chart of the Oceans (GEBCO), National Oceanographic and Atmospheric Chart of the Oceans (GEBCO), National Oceanographic and Atmospheric Administration (NOAA), Administration (NOAA), National Geographic, HERE technologies, and GeoNames. National Geographic, HERE technologies, and GeoNames. Geosciences 2019, 9, 231 4 of 24 Geosciences 2019, 8, x FOR PEER REVIEW 4 of 24 Figure 2. Map showing the location of NOAA surveys collected in 2006–2013 and U.S. Geological Figure 2. Map showing the location of NOAA surveys collected in 2006–2013 and U.S. Geological Survey (USGS) surveys collected in 2014–2015 along with sample and bottom photo locations. Survey (USGS) surveys collected in 2014–2015 along with sample and bottom photo locations. Geosciences 2019, 9, 231 5 of 24 Table 1. National Oceanic and Atmospheric Administration and National Ocean Service Hydrographic surveys and USGS geophysical surveys used in this study [3–35]. Multibeam Sidescan Sonar Survey Number Year Collected Survey Area (km ) System System H11554 2006 Reson 8101 Klein 3000 155 H11555 2006 Reson 8101 Klein 3000 240 H11647 2007 Reson 8101 Klein 3000 121 H11648 2007 Reson 8101 Klein 3000 223 H11649 2007 Reson 8101 Klein 3000 200 H11650 2007 Reson 8101 Klein 3000 180 H11872 2008 Reson 8101 Klein 3000 263 H11873 2008 Reson 8101 Klein 3000 290 H11874 2008 Reson 8101 Klein 3000 248 H11992 2008 Reson 8101 Klein 3000 142 H12001 2009 Reson 7125 Klein 3000 77 H12002 2010 Reson 7125 Klein 3000 203 H12003 2010 Reson 7125 Klein 3000 220 H12091 2010 Reson 7125 Klein 3000 164 H12092 2010 Reson 7125 Klein 3000 225 H12093 2010 Reson 7125 Klein 3000 182 H12094 2010 Reson 7125 Klein 3000 196 H12160 2011 Reson 7125 Klein 3000 127 H12161 2011 Reson 7125 Klein 3000 119 H12336 2011 Reson 7125 Klein 3000 103 H12337 2011 Reson 7125 Klein 3000 118 H12338 2011 Reson 7125 Klein 3000 163 H12339 2011 Reson 7125 Klein 3000 154 H12394 2012 Reson 7125 Klein 3000 115 H12395 2012 Reson 7125 Klein 3000 121 H12396 2012 Reson 7125 Klein 3000 116 H12397 2012 Reson 7125 Klein 3000 60 H12559 2013 Reson 7125 Klein 3000 126 H12560 2013 Reson 7125 Klein 3000 123 H12561 2013 Reson 7125 Klein 3000 144 H12668 2013 Reson 7125 Klein 3000 152 2014-002-FA 2014 SwathPlus Klein 3000 535 2015-002-FA 2015 SwathPlus Edgetech 4200 808 2. Geologic Setting The Delmarva Peninsula is situated along the mid-Atlantic coast between Delaware and the Chesapeake Bay (Figure 1). The Peninsula is a megaspit that formed in the Mio-Pliocene from material delivered by rivers and braided streams [47–49]. The continental shelf oshore of the Delmarva Peninsula is just over 100 km wide. Shelf slopes, tidal range, wave height, littoral transport, barrier island morphology, and historical erosion rates dier between the northern half of the Delmarva coast, from Cape Henlopen to the southern tip of Assateague Island (mostly stable over the long-term, except near Ocean City Inlet), and the southern half, from Chincoteague Bight to Fisherman’s Island (mostly erosional) [49,50]. These dierences along the coast create two distinct coastal compartments, such that the northern half of the Delmarva Peninsula is considered wave dominated, while the southern section of the Delmarva Peninsula is considered a mixed energy coast [51,52]. The surficial sediments of the inner-continental shelf from Cape Henlopen to the mouth of Chesapeake Bay have been described as arkosic to subarkosic sands [53–56]. Sand ridges interspersed by zones of flat seafloor characterize much of the northern half of the Delmarva inner-continental shelf [37,57,58] (Figure 1). The sand ridges are mobile and locally bury a transgressive surface [37,53,59–62], which results in local differences in grain size [53]. Fewer sand ridges are found on the inner shelf off Chincoteague Bight and the Virginia barrier islands (Figure 1). Here the less-undulating seafloor is composed of a Geosciences 2019, 9, 231 6 of 24 discontinuous sheet of medium to fine sand and gravelly sand, with a muddy sand area located within Chincoteague Bight [38,63,64]. 3. Methods 3.1. Data Sources Thirty-three hydrographic and geophysical datasets were collected between 2006 and 2015 on the inner-continental shelf oshore of the Delmarva Peninsula (Table 1; Figure 2). These data can be broken into two groups. The ﬁrst group of data consists of NOAA hydrographic surveys collected by Leidos (formerly Science Applications International Corporation (SAIC)) between 2006 and 2013 with 40-m line spacing, using a RESON 8101 (240 kHz) or a RESON 7125 (200 or 400 kHz) multibeam echosounder (MBES) and a Klein 3000K sidescan sonar system, for the purposes of updating nautical charts [3–33]. These data were acquired to IHO S-44 5th edition Order 1a speciﬁcations, and as such are accurate within 0.5 m vertically and 2 m horizontally. Small gaps, typically less than ﬁve% of the swath width in water depths less than 12 m in the NOAA bathymetric data, were ﬁlled using an inverse distance weighed (IDW) interpolant, then the data were resampled to 25 m per pixel (mpp). Backscatter intensity derived from the GSF (generic sensor format) MBES data was processed with the Fledermaus Geocoder Toolbox. Backscatter intensity derived from sidescan sonar data was processed with Chesapeake Technology’s SonarWiz. All backscatter imagery (MBES and sidescan sonar) was ﬁltered to reduce speckling and noise and resampled to 25 mpp. The MBES data contained only beam-average backscatter intensity. Beam-average data can produce high-quality backscatter imagery, however, the full-time series has a more dynamic range in a ﬁnal backscatter mosaic when compared with beam-average data. For this reason and reasons related to acquisition artifacts, the Klein sidescan sonar data were usually preferred in this study over the beam-averaged MBES backscatter data, especially in areas where the Reson 8101 was used. The second group of data consists of two USGS geophysical surveys conducted in 2014 and 2015 [34,35]. SwathPlus interferometric (234 kHz) bathymetric data were processed with BathySwath and Caris. Bathymetric data were interpolated (with IDW) to ﬁll gaps, typically less than 30% of the swath width, then resampled to 25 mpp. The higher resolution NOAA MBES data were given priority over the more widely spaced interferometric bathymetric data when creating the merged bathymetry grid, which has an assumed accuracy of 1 m vertically and 4 m horizontally. Sidescan sonar data were acquired in 2014 with a Klein 3000 and in 2015 with an Edgetech 4200. Sidescan sonar data were processed using SonarWiz. Filtered backscatter mosaics were resampled to 25 mpp. A suite of 227 sediment samples were collected during the USGS surveys and were analyzed for grain-size statistics and carbonate content by the USGS Woods Hole sediment analysis lab following methods outlined by McMullen et al. [65]. This study uses the Barnhardt classiﬁcation [66] (Figure 3) to characterize sediment texture, which is based on four end-member sediment units [67]: rock (R; grain size greater than 64 mm (mm) or less than 6 phi), gravel (G; grain size 2 to 64 mm or 1 to 6 phi), sand (S; grain size 0.062 to 2 mm or 4 to 1 phi), and mud (M; grain size less than 0.062 mm or greater than 4 phi). The classiﬁcation is further divided into 12 composite units, which are two-part combinations of the four end-member units. This classiﬁcation is deﬁned such that the primary texture, representing more than 50% of an area’s texture, is given an uppercase letter, and the secondary texture, representing less than 50% of an area’s texture, is given a lowercase letter. If one of the basic sediment units represents more than 90% of the texture, only its uppercase letter is used (Figure 3). Geosciences 2019, 9, 231 7 of 24 Geosciences 2019, 8, x FOR PEER REVIEW 7 of 24 Figure 3. Barnhardt and others (1998) [66] bottom-type classification based on four end-member Figure 3. Barnhardt and others (1998) [66] bottom-type classiﬁcation based on four end-member sediment units: rock (R), gravel (G), sand (S), and mud (M). Twelve additional two-part units sediment units: rock (R), gravel (G), sand (S), and mud (M). Twelve additional two-part units represent represent combinations of the four basic units. In the two-part (composite) units, the primary texture combinations of the four basic units. In the two-part (composite) units, the primary texture (greater (greater than 50% of the area) is given an uppercase letter and the secondary texture (less than 50% of than 50% of the area) is given an uppercase letter and the secondary texture (less than 50% of the area) the area) is given a lowercase letter. Figure is modified from [66]. is given a lowercase letter. Figure is modiﬁed from [66]. 3.2. Image Analysis Approach and Testing 3.2. Image Analysis Approach and Testing The primary objectives of the initial analysis were: (1) determine if these data were of sucient The primary objectives of the initial analysis were: (1) determine if these data were of sufficient quality to conduct a seaﬂoor classiﬁcation, and (2) identify a simple representation of seaﬂoor variability quality to conduct a seafloor classification, and (2) identify a simple representation of seafloor based on bathymetric and backscatter data. To achieve these goals, a relatively high quality single variability based on bathymetric and backscatter data. To achieve these goals, a relatively high quality dataset was chosen from among the 33 surveys to use as a subset for sensitivity and accuracy testing. single dataset was chosen from among the 33 surveys to use as a subset for sensitivity and accuracy Four distinct classes were identiﬁed using backscatter and seaﬂoor slope. Sediment samples provided testing. Four distinct classes were identified using backscatter and seafloor slope. Sediment samples the ground truth and texture characterization for the classes (Figure 4). Unsupervised (iso cluster) provided the ground truth and texture characterization for the classes (Figure 4). Unsupervised (iso and simple supervised (maximum likelihood) classiﬁcations were conducted on the test area to see cluster) and simple supervised (maximum likelihood) classifications were conducted on the test area if machine learning algorithms could capture the user identiﬁed classes from backscatter and slope. to see if machine learning algorithms could capture the user identified classes from backscatter and The accuracy of each test classiﬁcation was assessed using total accuracy and Cohen’s kappa coecient slope. The accuracy of each test classification was assessed using total accuracy and Cohen’s kappa (see Section 3.3). Based on initial tests, classiﬁcation was then expanded to the entire dataset. coefficient (see Section 3.3). Based on initial tests, classification was then expanded to the entire The relative hardness or softness of the seaﬂoor, which is closely related to sediment texture dataset. and cohesion, is distinguished using changes in backscatter intensity as an indicator of dierences The relative hardness or softness of the seafloor, which is closely related to sediment texture and in particle size and composition [68]. Dierences in the geomorphologic structure of the seaﬂoor is cohesion, is distinguished using changes in backscatter intensity as an indicator of differences in readily identiﬁed using seaﬂoor slope, calculated using a three by three cell window in ArcGIS 10.5.1. particle size and composition [68]. Differences in the geomorphologic structure of the seafloor is Sand ridges have relatively steep slopes (here, greater than ~1.75 degrees) and ﬂat areas of seaﬂoor readily identified using seafloor slope, calculated using a three by three cell window in ArcGIS 10.5.1. have low slopes. The four target classes, based on observations of the acoustic data and grain-size Sand ridges have relatively steep slopes (here, greater than ~1.75 degrees) and flat areas of seafloor statistics from sediment samples were: (1) high backscatter areas with low seaﬂoor slopes (HBLS), have low slopes. The four target classes, based on observations of the acoustic data and grain-size which are primarily composed of sand with shell and gravel; (2) high backscatter areas with steep statistics from sediment samples were: 1) high backscatter areas with low seafloor slopes (HBLS), slopes (HBSS) associated with the stoss side of sand ridges, also composed of sand with shell and which are primarily composed of sand with shell and gravel; 2) high backscatter areas with steep gravel; (3) low backscatter areas with low slopes (LBLS), composed primarily of sand with some mud; slopes (HBSS) associated with the stoss side of sand ridges, also composed of sand with shell and and (4) low backscatter areas with steep slopes (LBSS), comprising the tops and lee sides of sand ridges gravel; 3) low backscatter areas with low slopes (LBLS), composed primarily of sand with some mud; and composed of nearly 100% sand (Figure 4; Table 2). and 4) low backscatter areas with steep slopes (LBSS), comprising the tops and lee sides of sand ridges and composed of nearly 100% sand (Figure 4; Table 2). Table 2. Table showing mean grain-size statistics for each class. *Sediment samples in column two are reported as a ratio of sample agreement with mean texture. For example, for the low backscatter Table 2. Table showing mean grain-size statistics for each class. *Sediment samples in column two areas with low slopes (LBLS) class, 99 out of 120 samples were classiﬁed as ‘S’ based on [66], leaving are reported as a ratio of sample agreement with mean texture. For example, for the low backscatter 21 samples that were not ‘S’ texture. **Shell content is shown as a percent of the total sample. areas with low slopes (LBLS) class, 99 out of 120 samples were classified as ‘S’ based on [66], leaving 21 samples that were not ‘S’ texture. **Shell content is shown as a percent of the total sample. Class *Samples Mean F % Gravel % Sand % Mud **% Shell Texture Geomorphology LBLS 99/120 2.83 1.28 91.89 6.82 3.03 S Flat Class *Samples Mean ɸ % Gravel % Sand % Mud **% Shell Texture Geomorphology HBLS 14/19 0.51 22.89 74.64 2.55 28.36 Sg Flat LBLS 99/120 2.83 1.28 91.89 6.82 3.03 S Flat LBSS 62/67 1.82 1.99 97.00 1.00 2.65 S Ridged HBLS 14/19 0.51 22.89 74.64 2.55 28.36 Sg Flat HBSS 14/22 0.59 13.96 84.40 1.63 18.23 Sg Ridged LBSS 62/67 1.82 1.99 97.00 1.00 2.65 S Ridged HBSS 14/22 0.59 13.96 84.40 1.63 18.23 Sg Ridged Geosciences 2019, 9, 231 8 of 24 Geosciences 2019, 8, x FOR PEER REVIEW 8 of 24 Figure 4. Ternary diagram depicts the sediment samples within each class. Ellipses are drawn around Figure 4. Ternary diagram depicts the sediment samples within each class. Ellipses are drawn around the majority of samples within each class to highlight the representative sediment signature the majority of samples within each class to highlight the representative sediment signature associated associated with each. The four pie charts represent the average sediment composition in each class: with each. The four pie charts represent the average sediment composition in each class: LBLS, high LBLS, high backscatter areas with low seafloor slopes (HBLS), low backscatter areas with steep slopes backscatter areas with low seaﬂoor slopes (HBLS), low backscatter areas with steep slopes (LBSS), and (LBSS), and high backscatter areas with steep slopes (HBSS). The map and photos on the right-hand high backscatter areas with steep slopes (HBSS). The map and photos on the right-hand side show the side show the sediment sample and photo locations with representative seafloor photos, one for each sediment class. sample and photo locations with representative seaﬂoor photos, one for each class. Two classifications were evaluated on the test dataset using the image classification toolbox in Two classiﬁcations were evaluated on the test dataset using the image classiﬁcation toolbox in Esri’s ArcGIS. The iso cluster (ISO) classification was chosen as the simplest and first method. In ISO Esri’s ArcGIS. The iso cluster (ISO) classiﬁcation was chosen as the simplest and ﬁrst method. In ISO analysis, the user needs no prior knowledge of the seafloor or the input data, and the number of analysis, the user needs no prior knowledge of the seaﬂoor or the input data, and the number of classes classes is often unknown. As it is applied in ArcGIS, the ISO algorithm uses a modified iterative is often unknown. As it is applied in ArcGIS, the ISO algorithm uses a modiﬁed iterative optimization optimization clustering procedure known as migrating means to separate cells into a user-specified clustering procedure known as migrating means to separate cells into a user-speciﬁed number of number of classes. For our purposes, we wanted to see if ISO could identify our target classes based classes. For our purposes, we wanted to see if ISO could identify our target classes based on statistical on statistical relationships between slope and backscatter for each pixel, without supplying any user knowledge except for the number of classes. ISO distinguished three of the four target classes in relationships between slope and backscatter for each pixel, without supplying any user knowledge backscatter and slope data: high backscatter intensity, low backscatter intensity, and steep slopes, but except for the number of classes. ISO distinguished three of the four target classes in backscatter could not further discriminate a meaningful fourth class for sloping areas with high backscatter. For and slope data: high backscatter intensity, low backscatter intensity, and steep slopes, but could not example, sand ridges often have areas of high backscatter associated with their stoss side, while the further discriminate a meaningful fourth class for sloping areas with high backscatter. For example, lee side is consistently low backscatter. sand ridges often have areas of high backscatter associated with their stoss side, while the lee side is The second method evaluated was a supervised maximum likelihood classification. In MLC, consistently representalow tive pixe backscatter ls are ident . ified for each class by a user with prior knowledge of the classes to be identified. This step creates a training dataset that is then used to determine the class of all other The second method evaluated was a supervised maximum likelihood classiﬁcation. In MLC, pixels based on the variance and covariance of the class signatures. The training dataset contained 20 representative pixels are identiﬁed for each class by a user with prior knowledge of the classes to polygons drawn around representative pixels for each class, ranging in size from 20 to greater than be identiﬁed. This step creates a training dataset that is then used to determine the class of all other 1000 pixels. Pixels in the classified raster are assigned to the class that they have the highest pixels based on the variance and covariance of the class signatures. The training dataset contained probability of being a member. The target classes were the same as described above: LBLS, HBLS, 20 polygons drawn around representative pixels for each class, ranging in size from 20 to greater than LBSS, and HBSS. 1000 pixels. Pixels in the classiﬁed raster are assigned to the class that they have the highest probability 3.3. Accuracy Evaluation of being a member. The target classes were the same as described above: LBLS, HBLS, LBSS, and HBSS. Following initial image analysis on the test dataset using unsupervised ISO, and supervised 3.3. Accuracy Evaluation MLC, two hundred randomly-located, reference points (RPs) were created within the test dataset area. The RPs were user-assigned to one of the four classes defined for seafloor types (LBLS, HBLS, Following initial image analysis on the test dataset using unsupervised ISO, and supervised LBSS, and HBSS). User-defined RPs were compared to the classified pixels for each test method, again MLC, two hundred randomly-located, reference points (RPs) were created within the test dataset area. The RPs were user-assigned to one of the four classes deﬁned for seaﬂoor types (LBLS, HBLS, LBSS, and HBSS). User-deﬁned RPs were compared to the classiﬁed pixels for each test method, again using ArcGIS. A confusion matrix was created from the RPs and the classiﬁed points were used to test the validity of the classiﬁcation method in the test area. Total accuracy, or the percentage of Geosciences 2019, 9, 231 9 of 24 correctly classiﬁed pixels, was calculated, as well as the Cohen’s kappa coecient, which indicates the agreement between classiﬁed pixels and RPs and adjusts for inﬂated accuracy due to chance. 3.4. Applying Classiﬁcation to the Entire Dataset Based on initial feasibility and accuracy ﬁndings (see the results section for test area accuracies or Table 3), it was determined that these data could support machine learning classiﬁcation, using only two inputs, backscatter and slope. In initial testing, we found that the iso cluster performed well at distinguishing high and low backscatter, but was inadequate at dierentiating slope dierences associated with high and low backscatter. With that knowledge, moving forward with classiﬁcation of the entire study area, and in order to address relative backscatter value dierences introduced by combining survey data with dierent instrumentation and acquisition parameters, we utilized the ISO’s ability to quickly, consistently, and objectively dierentiate high and low backscatter, prior to supervised classiﬁcation. To achieve this, an 8-bit backscatter image of each of the 33 survey areas was passed through ISO analysis to produce a two-class outcome: high and low backscatter. This additional step to pre-classify backscatter reduced uncertainty associated with user-deﬁned high and low backscatter decisions, minimized the eects of artifacts on the classiﬁcation, and normalized the backscatter range values among surveys and instruments. Table 3. Table showing accuracy and kappa coecient results for each classiﬁcation method that was conducted on the test area and the whole study area. Location Method % Accuracy Kappa ISO 78.1 0.682 Test Area MLC 2 inputs 92.7 0.892 MLC 2 inputs 73.0 0.603 Whole Area MLC 5 inputs 76.0 0.645 RF 5 inputs 77.5 0.667 Following ISO classiﬁcation of the backscatter data for all surveys, pre-classiﬁed backscatter for all 33 surveys was combined with slope, and an MLC was done using two inputs. Additional data layers were introduced, including curvature (the second derivative of bathymetry, which determines if a surface is concave or convex), shaded-relief (which helps captures the slipface and apex of sand ridges with a common orientation), and bathymetry. These secondary inputs would help determine if introducing more data complexity, dimensionality, redundancy, and relationships improved classiﬁcation results. Another MLC was conducted on the ﬁve data inputs (ISO classiﬁed backscatter (high and low), slope, bathymetry in meters below the North American Datum of 1988, shaded-relief, and curvature), using 112 polygons as the training dataset. Finally, a random forest (RF) classiﬁcation in ArcPro was conducted for all 33 surveys and ﬁve inputs. RF classiﬁcation is a supervised method that is an ensemble classiﬁcation based on several individual decision trees. Decision trees have been shown to achieve good results for MBES-type data classiﬁcations [41]. In ArcGIS forest-based classiﬁcations are an adaptation of Breiman’s random forest algorithm, in which decision trees are created using the training data, then trees generate predictions, and the ﬁnal outcome is decided based on a voting scheme. The same training data set from MLC, consisting of 112 polygons, was used for RF with the same four target classes of LBLS, HBLS, LBSS, and HBSS. The workﬂow from data source to classiﬁcation output is summarized in Figure 5. All analysis, including merging, interpolating, ﬁltering, and resampling the data in preparation for machine learning classiﬁcation was conducted in ArcGIS. Following image analysis of the entire dataset using the three dierent methodologies, another 200 randomly-located, reference points (RPs) were created within the classiﬁed area. Once again, confusion matrices were created from the RPs and the classiﬁed points were used to test the validity of the classiﬁed seaﬂoor. Geosciences 2019, 9, 231 10 of 24 Geosciences 2019, 8, x FOR PEER REVIEW 10 of 24 Figure Figure 5. 5. ((A A– –E E) ). . A work A workﬂow flow a at t the top depi the top depicts cts the step the steps s u used sed to prepare the data sou to prepare the data sour rces ces f for or m machine achine learning learning (ML) (ML). . Orange and gr Orange and gr een cells in een cells in the the chart re chart r present epresent the the two two and five and ﬁve source sour inputs f ce inputs or the for iso the iso cluscluster ter (ISO) (ISO), , maxmaximum imum likelihood likelihood classif classiﬁcation ication (MLC (MLC), ), and random and random forest (RF forest ) cla (RF) ssifica classiﬁcat tions. Gions. reen Gr cell een s were cells onl werye u only sed used in the in five the inpu ﬁve input t classif classiﬁcations. ications. Orang Orange e cell cells s wer wer e used in e used in the thetwo input two input classiﬁcations. classifications. Inpu Inputs ts for ML for ML a ar re shown in ( e shown in (A A) bathymetry, ) bathymetry, ((B B) slope ) slope, , ((C C) shaded-relief, ) shaded-relief, ((D D) curvature, ) curvature, and (E) backscatter. and (E) backscatter. 4. Results 4. Results 4.1. Mapped Accuracy 4.1. Mapped Accuracy The best accuracy results in the test area were achieved with the supervised maximum likelihood The best accuracy results in the test area were achieved with the supervised maximum likelihood classiﬁcation (Table 3). The ISO classiﬁcation performed with the lowest total accuracy due to classification (Table 3). The ISO classification performed with the lowest total accuracy due to its its inability to distinguish all four target classes (Table 3), however ISO performed at 100% when inability to distinguish all four target classes (Table 3), however ISO performed at 100% when distinguishing between high and low backscatter areas. distinguishing between high and low backscatter areas. Accuracy and kappa values were lower for classiﬁcations evaluated over the whole study area Accuracy and kappa values were lower for classifications evaluated over the whole study area (all 33 surveys) than on the test area (Table 3). The MLC performed on two inputs (MLC 2 inputs) had (all 33 surveys) than on the test area (Table 3). The MLC performed on two inputs (MLC 2 inputs) the lowest accuracy and kappa coecient, while the MLC performed on ﬁve input data layers (MLC had the lowest accuracy and kappa coefficient, while the MLC performed on five input data layers 5 inputs) achieved slightly better results, and the random forest (RF 5 inputs) achieved the best results (MLC 5 inputs) achieved slightly better results, and the random forest (RF 5 inputs) achieved the best for the entire study area. results for the entire study area. Geosciences 2019, 8, x FOR PEER REVIEW 11 of 24 Table 3. Table showing accuracy and kappa coefficient results for each classification method that was conducted on the test area and the whole study area. Location Method % Accuracy Kappa ISO 78.1 0.682 Test Area MLC 2 inputs 92.7 0.892 MLC 2 inputs 73.0 0.603 Whole Area MLC 5 inputs 76.0 0.645 Geosciences 2019, 9, 231 11 of 24 RF 5 inputs 77.5 0.667 4.2. Classiﬁed Maps 4.2. Classified Maps The classiﬁed maps for the test area are displayed in Figure 6A,B. The ISO analysis only The classified maps for the test area are displayed in Figure 6A,B. The ISO analysis only distinguished three of the four classes (Figure 6A). MLC captures the four target classes, but contains distinguished three of the four classes (Figure 6A). MLC captures the four target classes, but contains more striping introduced by artifacts in the backscatter data layer (Figure 6B). more striping introduced by artifacts in the backscatter data layer (Figure 6B). Figure 6. (A) ISO classiﬁcation on the test area, using slope and backscatter intensity. (B) MLC on the Figure 6. (A) ISO classification on the test area, using slope and backscatter intensity. (B) MLC on the test area using slope and backscatter intensity. test area using slope and backscatter intensity. The classiﬁed maps over the entire study for the 33 individual hydrographic and geophysical The classified maps over the entire study for the 33 individual hydrographic and geophysical surveys are shown in Figure 7A–C. At a glance, results appear fairly similar, however, the best accuracy surveys are shown in Figure 7A–C. At a glance, results appear fairly similar, however, the best and kappa coecient were achieved for the whole study area using random forest with ﬁve input data accuracy and kappa coefficient were achieved for the whole study area using random forest with five grids (Table 3). The initial ISO analysis of backscatter prior to the supervised classiﬁcations over the input data grids (Table 3). The initial ISO analysis of backscatter prior to the supervised classifications whole study area was advantageous in that it reduced stripiness by normalizing relative backscatter over the whole study area was advantageous in that it reduced stripiness by normalizing relative values among surveys and instruments. The classiﬁed area was 5394 km . LBLS and LBSS classes backscatter values among surveys and instruments. The classified area was 5394 km . LBLS and LBSS comprise the largest percentages of the classiﬁed seaﬂoor by area at 39% and 38%, respectively. High classes comprise the largest percentages of the classified seafloor by area at 39% and 38%, backscatter areas make up 23% of the seaﬂoor by area with HBLS accounting for 15%, and HBSS respectively. High backscatter areas make up 23% of the seafloor by area with HBLS accounting for accounting for 8%. 15%, and HBSS accounting for 8%. We generated the geomorphic structure and sediment-texture maps by integrating the results of the random forest analysis with seaﬂoor sample data. Sand ridges are a prominent geomorphic feature in the study area (Figure 1) that are deﬁned primarily by relief. A meaningful geomorphologic structure map can be derived from the four classes by ignoring the backscatter dierences and focusing on the steeper slope classes. Figure 8 shows the areas of the seaﬂoor that are classiﬁed as HBSS or LBSS, and thus can be considered ridges. Approximately 46% of the classiﬁed seaﬂoor can be considered sand ridge, while the remaining 54% is relatively ﬂat. Geosciences 2019, 9, 231 12 of 24 Geosciences 2019, 8, x FOR PEER REVIEW 12 of 24 Figure 7. (A) MLC with two inputs, slope, and ISO-classified backscatter for the whole survey area. Figure 7. (A) MLC with two inputs, slope, and ISO-classiﬁed backscatter for the whole survey area. ((B B)) MLC with MLC with ﬁve five inputs, inputs, slope,slope, ISO-cla ISO-classiﬁedssifie backscatter d back,scat bathymetry ter, bathy , hillshaded-r metry, hillelief, shaded-r and elief curvatur , and e. ( curvature. ( C) RF withC ﬁve ) RF inputs, with fiv slope, e inputs, slope, ISO-classiﬁed ISO-clas backscatter sified ba , bathymetry ckscatter, bath , relief, yme and try, relief, curvatur and cu e. rvature. Grain-size statistics of sediment samples were averaged by class (Table 2). LBLS and LBSS classes We generated the geomorphic structure and sediment-texture maps by integrating the results of have high sand content (>90%) and low carbonate concentrations, whereas sediments in the high the random forest analysis with seafloor sample data. Sand ridges are a prominent geomorphic backscatter classes (HBSS and HBLS) have a higher gravel-sized component (~18%) and higher shell feature in the study area (Figure 1) that are defined primarily by relief. A meaningful geomorphologic content (18–28%; Table 2 and Figure 4), but are still primarily sand. Grain size data also indicate slightly structure map can be derived from the four classes by ignoring the backscatter differences and higher mud content within the low slope classes (3% to 7%, versus 1% to 2% in steeper sloping areas). focusing on the steeper slope classes. Figure 8 shows the areas of the seafloor that are classified as Applying the Barnhardt classiﬁcation to the mean grain-size statistics for each class produces two HBSS or LBSS, and thus can be considered ridges. Approximately 46% of the classified seafloor can unique sediment texture categories within the classiﬁed area: sand (S) and sand with gravel (Sg). Both be considered sand ridge, while the remaining 54% is relatively flat. low-backscatter classes (LBLS and LBSS) have mean sand contents over 90%, so they are classiﬁed as sand. The two high-backscatter classes (HBLS and HBSS) have sand contents between 75% and 84% and gravel contents between 14% and 25%, which makes them Sg (or sand with gravel; Figure 9). The S class represents roughly 77% of the seaﬂoor by area, suggesting that three-quarters of the mapped seaﬂoor may be greater than 90% sand. The Sg class represents 23% of the seaﬂoor, suggesting that the seaﬂoor is still primarily sand in these areas, but also contains more than 10% gravel and/or more abundant shells. Displaying individual samples over the sediment texture map (Figure 10) shows where mean grain-size statistics in each class dier from the sediment texture classiﬁcation assigned based on RF classiﬁcation. Four mud-rich samples (greater than 50% mud) were collected from areas that were not acoustically dierent from adjacent areas of sandier sediment texture. Nine samples considered gravel-rich (greater than 50% gravel) or shell-rich (greater than 50% shell) were also collected, and fell mostly within the HBLS or HBSS classes. For areas classiﬁed as S in the sediment texture map, 86% of the samples were at least 90% sand, and therefore in agreement with the S classiﬁcation. The 14% that diered were samples classiﬁed as Sm (7%), Sg (5%), or the previously mentioned mud-rich samples (<2%). For areas classiﬁed as Sg in the sediment texture map, 55% of the samples were also Sg, while 36% were S, 7% were shell or gravel-rich (Gs), and 2% was Sm. Sampling positional accuracy is assumed to be +/–50 m, and could be erroneously classiﬁed if sampled near a class boundary. Geosciences 2019, 8, x FOR PEER REVIEW 13 of 24 Geosciences 2019, 9, 231 13 of 24 Figure 8. Geomorphologic classes derived from the four classes by combining the low slope (LBLS and HBLS) and steep slope units (LBSS and HBSS). Figure 8. Geomorphologic classes derived from the four classes by combining the low slope (LBLS and HBLS) and steep slope units (LBSS and HBSS). Grain-size statistics of sediment samples were averaged by class (Table 2). LBLS and LBSS classes have high sand content (>90%) and low carbonate concentrations, whereas sediments in the high backscatter classes (HBSS and HBLS) have a higher gravel-sized component (~18%) and higher shell content (18–28%; Table 2 and Figure 4), but are still primarily sand. Grain size data also indicate slightly higher mud content within the low slope classes (3% to 7%, versus 1% to 2% in steeper sloping areas). Applying the Barnhardt classification to the mean grain-size statistics for each class produces Geosciences 2019, 8, x FOR PEER REVIEW 14 of 24 two unique sediment texture categories within the classified area: sand (S) and sand with gravel (Sg). Both low-backscatter classes (LBLS and LBSS) have mean sand contents over 90%, so they are classified as sand. The two high-backscatter classes (HBLS and HBSS) have sand contents between 75% and 84% and gravel contents between 14% and 25%, which makes them Sg (or sand with gravel; Figure 9). The S class represents roughly 77% of the seafloor by area, suggesting that three-quarters of the mapped seafloor may be greater than 90% sand. The Sg class represents 23% of the seafloor, suggesting that the seafloor is still primarily sand in these areas, but also contains more than 10% Geosciences 2019, 9, 231 14 of 24 gravel and/or more abundant shells. Figure 9. Sediment samples (small circles) overlay a classiﬁed sediment texture map derived from the Figure 9. Sediment samples (small circles) overlay a classified sediment texture map derived from the RF classiﬁcation by combining HBLS and HBSS to create a sand with gravel (Sg) texture class, and RF classification by combining HBLS and HBSS to create a sand with gravel (Sg) texture class, and combining LBLS and LBSS to create a sand (S) texture class. The small circles are coded by the sample’s combining LBLS and LBSS to create a sand (S) texture class. The small circles are coded by the Barnhardt texture. Large circles around some samples illustrate where sediment texture is more than sample’s 50% Barnhardt tex shell, gravel, tor urmud, e. Larg which e circ diles ers arou signiﬁcantly nd somfr e sa omm the ples il meanlu sediment strate where se texture assigned diment tex to ture is the classes. more than 50% shell, gravel, or mud, which differs significantly from the mean sediment texture assigned to the classes. Geosciences 2019, 8, x FOR PEER REVIEW 15 of 24 Displaying individual samples over the sediment texture map (Figure 10) shows where mean grain-size statistics in each class differ from the sediment texture classification assigned based on RF classification. Four mud-rich samples (greater than 50% mud) were collected from areas that were not acoustically different from adjacent areas of sandier sediment texture. Nine samples considered gravel-rich (greater than 50% gravel) or shell-rich (greater than 50% shell) were also collected, and fell mostly within the HBLS or HBSS classes. For areas classified as S in the sediment texture map, 86% of the samples were at least 90% sand, and therefore in agreement with the S classification. The 14% that differed were samples classified as Sm (7%), Sg (5%), or the previously mentioned mud-rich samples (<2%). For areas classified as Sg in the sediment texture map, 55% of the samples were also Sg, while 36% were S, 7% were shell or gravel-rich (Gs), and 2% was Sm. Sampling positional accuracy is assumed to be +/–50 m, and could be erroneously classified if sampled near a class Geosciences 2019, 9, 231 15 of 24 boundary. Figure 10. Figure 10. Map of the USGS Map of the USGS backscatter im backscatter imager agery acquired in 2014. Inset y acquired in 2014. Inset (A( )A and ) and ( (B) show B) shdetailed ow detailed maps of sample locations (pink dot) collected within a sub-circular feature (A) and a linear feature maps of sample locations (pink dot) collected within a sub-circular feature (A) and a linear feature (B) (B) in the backscatter data at 2 mpp resolution. Linear and sub-circular pits or features observed on in the backscatter data at 2 mpp resolution. Linear and sub-circular pits or features observed on the the seaﬂoor in geophysical data often have a ﬁner sediment texture than the surrounding seaﬂoor seafloor in geophysical data often have a finer sediment texture than the surrounding seafloor and and occur where ﬁner grained material is cropping out at the seaﬂoor. These features often are not occur where finer grained material is cropping out at the seafloor. These features often are not representative of sediment texture adjacent to the feature. representative of sediment texture adjacent to the feature. 5. Discussion Hydrographic and geophysical datasets collected with dierent instrumentation over a 10-year period pose challenges for seaﬂoor classiﬁcation. Ideally, the hydrographic data would have been time-series MBES data and would have lacked artifacts and range dierences that came with combining sidescan sonar and beam-averaged MBES backscatter across surveys. Due to these challenges in the backscatter data, we adopted ISO classiﬁcation to standardize and normalize the high and low backscatter values among surveys. Previous studies informed our selection of methods [41,43,63–65] Geosciences 2019, 9, 231 16 of 24 and initial tests helped address challenges associated with the backscatter data repurposing and conglomeration. ISO has been shown as a useful ﬁrst step when evaluating the spatial variability of seaﬂoor facies [43,69], but typically requires samples or subsequent reclassiﬁcation to achieve desired results. Here we found that ISO lacked the ability to distinguish all four classes of interest; however, it performed with perfect accuracy on the test area dataset when distinguishing only between high and low backscatter. This particular strength of ISO analysis, and the fact that it is very simple to execute and requires no prior user knowledge of the data, made it a simple and valuable ﬁrst step in preparing the entire 33-survey data collection for supervised classiﬁcation. By performing a simple ISO analysis on individual surveys, we were able to normalize relative changes in backscatter among surveys with dierent instruments and acquisition parameters, and thus overcome some of the challenges associated with repurposing these types of data for automated classiﬁcation. This initial ISO step also minimized striping and minor artifacts within some surveys (Figure 4), and consistently distinguished high and low backscatter in agreement with a user ’s discrimination. The two surveys with the most abundant artifacts that interfered with high and low backscatter determination were H12160 and H12161. A reasonably accurate, general sediment texture map of the study area could be generated using only ISO, which would capture the dierences in high and low backscatter and corresponds to the dierence in S and Sg within this study area. Because the ﬁrst goal of this study was to determine the feasibility of conducting a machine learning classiﬁcation using repurposed data, we adhered to relatively simple initial testing and methods, from the classiﬁcation techniques used (ISO and MLC), to data sources (only two initial inputs), and the four target classes. Results revealed that these data were usable for auto-classiﬁcation, as long as the expectations of classiﬁed outcomes were not too complex. Two data layers could provide enough seaﬂoor variability to deﬁne four classes, but accuracy was improved by adding additional data inputs and increasing the complexity of the classiﬁcation algorithm, which is consistent with ﬁndings from previous studies using MBES datasets [41,70,71]. Sediment samples with grain-size statistics used for validation were averaged within each class. Averaging sediment statistics has the eect of removing anomalous samples or patches of seaﬂoor that may exist within or in addition to the deﬁned four classes. Some anomalous samples or seaﬂoor areas could not be distinguished in the geophysical and hydrographic data as it is processed and analyzed in this study. For example, samples that are far muddier than the average sediments within the LBLS class exist, but they are not captured by this classiﬁcation (Figures 9 and 10), and cannot be resolved at the 25 mpp resolution of the source data. Sub-circular and linear seaﬂoor features, such as those observed in Figure 10 are shown at 2 mpp. Such features are often depressions associated with outcropping units of ﬁner grained material here [38,55] and in other study areas [72]. These relatively small features (typically less than 100 m) are not representative of the surrounding seaﬂoor and not identiﬁable at the 25 mpp scale. The scale at which this study is conducted cannot resolve the linear and sub-circular features identiﬁed in the backscatter at higher resolution (2 mpp). The sediment samples associated with these features are not representative of the larger LBLS class, and averaging out their impact on a class is therefore acceptable for this regional-scale classiﬁcation. Other samples that dier from the average sediment statistics for each class are shown in Figure 9. Most mud-rich samples are explained by their occurrence in small-scale sub-circular and linear features, except for the mud-rich samples in Chincoteague Bight. These muddier than average samples reﬂect the sediment texture across a fairly large area of muddy sand that is associated with dewatered backbarrier muds or an ebb plume of suspended sediment [51]. However, this change in seaﬂoor texture is not recognizable in the acoustic data [38] and, therefore, would not be identiﬁable by a classiﬁcation relying on acoustic data. Gravel and shell-rich samples exist in a few locations throughout the survey and are often associated with bioherms as indicated in the bottom photographs [34,35], or occasional coarser-grained (than average HBLS) ﬂat areas of high backscatter. Geosciences 2019, 9, 231 17 of 24 It would be misleading to provide these generalized and regional-scale textural and geomorphologic classifications without outlining the limitations and assumptions that resulted in the classiﬁcation outcomes presented here. A good way to illustrate application scale and areas where generalized geologic maps may not capture all the seaﬂoor variability for evaluation by managers and other data users is to display the sediment sample locations along with the classiﬁcation, thereby identifying the areas where dierences in classiﬁed texture and actual sediment texture may exist, and where further discrimination or more detailed analysis may be required. Previous regional geologic studies of the Delmarva inner-continental shelf have deﬁned the seaﬂoor as relatively homogenous, consisting of primarily sand [52,55], and this study conducted using simple classes with mean grain-size statistics would seem to conﬁrm this description, such that the majority of textural variability along this inner-continental shelf can be explained by relatively small changes in sand concentration, identiﬁed using only two sediment texture classes (Figure 9). However, a more heterogeneous sediment environment, especially one in a periglacial environment, where all sediment types are possible and often adjacent, would require more complex substrate classes and input data types, such as the secondary features deﬁned by [41]. Pendleton et al. [38], created sediment texture and geomorphology maps of the area covered by USGS surveys with user-deﬁned polygons [34,35,38]. The sediment texture classiﬁcation used the same sediment samples as the current study and was classiﬁed according to the Barnhardt classiﬁcation (Figure 11). In general, the sediment texture and geomorphology maps produced here are in good agreement with the manual classiﬁcation study (Figure 11). Manual classiﬁcation oers the ﬂexibility of identifying features at multiple resolutions, so capturing the muddy pits and linear features that are not visible at 25 mpp is possible, but user digitization is time consuming and not as objective as machine learning techniques. Producing regional geologic information maps from repurposed hydrographic data is cost eective for programs, resourceful, ecient, collaborative, inter-disciplinary, and can maximize the scientiﬁc impact and footprint of a geologic investigation. The methodology used here required generalization of more complex sediment texture and geomorphology, but the beneﬁts of creating these interpretations outweighs the limitations. A more robust interpretation of seaﬂoor geology would include surﬁcial geologic units derived from shallow sub-bottom interpretations and could be integrated with the sediment texture and geomorphology maps produced here. Those maps and interpretations will be part of the next phase of the regional geologic framework study along the Delmarva Peninsula that aims to link coastal and shelf processes with vulnerability. The recommendations from this study are that any mapping eort, regardless of scope or budget, can beneﬁt from a “map once, use many times” attitude, and repurposing and machine learning classiﬁcation can create new geologic products quickly and inexpensively that can support science objectives as well as student opportunities. Geosciences 2019, 9, 231 18 of 24 Geosciences 2019, 8, x FOR PEER REVIEW 18 of 24 Figure 11. Sediment texture (A,B) and geomorphic (C,D) maps generated from machine-learning (this Figure 11. Sediment texture (A,B) and geomorphic (C,D) maps generated from machine-learning (this study, A,C) and from manual digitization ([38], B,D). Dierences observed in the sediment texture study, A,C) and from manual digitization ([38], B,D). Differences observed in the sediment texture maps (C,D) are attributable to the scales of the studies and subjectivity and ﬂexibility that inherently maps (C,D) are attributable to the scales of the studies and subjectivity and flexibility that inherently exists in manual classiﬁcations. exists in manual classifications. Geosciences 2019, 9, 231 19 of 24 6. Conclusions This study evaluated the feasibility of using repurposed hydrographic data to produce classiﬁed geologic maps of the seaﬂoor, and conﬁrms that it is possible but requires modest expectations and sensitivity testing. This classiﬁcation of the sea ﬂoor at 25-m resolution over a nearly 5400 km area of the Delmarva inner-shelf would not have been possible without the repurposing of hydrographic survey data in conjunction with new geophysical data and samples. The classiﬁcation produced here can be further divided into sediment texture and geomorphologic maps. The use of automated classiﬁcation techniques makes interpretation of such a large and diverse dataset feasible. However, there are limitations to these types of interpretations. Acquisition artifacts, especially in the MBES backscatter mosaics, often creates striping in the imagery that can be erroneously interpreted as high backscatter. Filtering, smoothing, and resampling the data prior to classiﬁcation can help reduce the artifacts, but some large artifacts can persist in the interpretations. The seaﬂoor along the Delmarva inner-continental shelf is more complex than four acoustic classes, two sediment texture classes, and two geomorphologies, but the available data, resolution, and the methodology used here do not support further discrimination. As such, they are meant as a ﬁrst step to geologic characterization of the continental shelf. Based on accuracy tests of the classiﬁcation evaluated here, the random forest classiﬁer produced the best results when compared at reference points. ISO was useful for distinguishing high and low backscatter areas, but was not capable of identifying all four target acoustic classes. ISO can be a useful starting place for these types of studies. Relatively simple and regional-scale classiﬁcations can be used to advise ﬁner-scale or more targeted follow-up studies. Additionally, large volumes of hydrographic data that have been collected on the continental shelf for charting purposes can be repurposed to support geologic seaﬂoor investigations without the added expense of new data acquisition. The beneﬁts of classiﬁed seaﬂoor datasets are numerous and include providing baseline science for ocean and coastal resource management, facilitating and enabling related research, and raising public awareness and engagement in ocean and coastal science, and management [1]. Finally, this classiﬁcation contributes to the growing body of research that is providing schemes and methods for the classiﬁcation of seaﬂoor data sets. An important takeaway from this study is that it is essential to know the limitations of the source data, know whether the discriminations you want to make in the data are possible given data inputs, and choose methods and classes according to what can be supported by the data quality. Patience, testing, and method assimilation can improve outcomes. Author Contributions: Conceptualization, E.A.P., E.M.S., and L.L.B.; Methodology, E.A.P. and E.M.S.; Validation, E.A.P. and E.M.S.; Formal analysis, E.A.P. and E.M.S.; Investigation, E.A.P. and E.M.S.; Writing—original draft preparation, E.A.P. and E.M.S.; Writing—review and editing, E.A.P. and L.L.B.; Visualization, E.A.P.; Supervision, E.A.P. and L.L.B.; Project administration, L.L.B.; Funding acquisition, L.L.B. Funding: This research was funded by the U.S. Department of the Interior Hurricane Sandy Recovery program and This study was part of the Linking Coastal Processes and Vulnerability—Assateague Island Regional Study project (GS2-2C). Acknowledgments: We thank Gene Parker (Hydrographic Team Lead, NOAA Oce of Coast Survey, Hydrographic Surveys Division, Atlantic Hydro Branch), LTJG David Rodziewicz (NOS Hydrographic Data manager at NGDC, NOAA Corps) and Jason Baillio (NOS Hydrographic Data manager at NCEI, NOAA) for providing the majority of the Reson and Klein raw data ﬁles for the hydrographic surveys used in this study. We thank Rebecca T. Quintal and Rex LeBeau of Leidos (formerly SAIC) for providing, with NOAA’s permission, Reson data ﬁles for hydrographic surveys H12092, H12093, H12094. Gene, David, Jason, Rebecca, and Rex were extremely helpful in communicating information and providing data to aid this eort, and we sincerely wish to thank them for their patience and cooperation with our large data requests. This manuscript beneﬁted from review by Chris Sherwood (of the USGS) and anonymous journal peer reviews. Any use of trade, ﬁrm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. Conﬂicts of Interest: The authors declare no conﬂict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results. Geosciences 2019, 9, 231 20 of 24 References 1. Johnson, S.Y.; Cochrane, G.R.; Golden, N.E.; Dartnell, P.; Hartwell, S.R.; Cochran, S.A.; Watt, J.T. The California seaﬂoor and coastal mapping program—Providing science and geospatial data for California’s State Waters. Ocean Coastal Manag. 2017, 140, 88–104. [CrossRef] 2. United States Geological Survey. Geologic Mapping of the Massachusetts Sea Floor. Available online: https://woodshole.er.usgs.gov/project-pages/coastal_mass/index.html (accessed on 14 March 2019). 3. National Oceanic and Atmospheric Administration. Survey H11554—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 06–TR–015; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2016; p. 29. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H10001-H12000/H11554/DR/H11554.pdf (accessed on 14 March 2019). 4. National Oceanic and Atmospheric Administration. Survey H11555—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 06–TR–0; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2016; p. 86. Available online: https://data.ngdc.noaa. gov/platforms/ocean/nos/coast/H10001-H12000/H11555/DR/H11555.pdf (accessed on 14 March 2019). 5. National Oceanic and Atmospheric Administration. Survey H11650—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 07–TR–011; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2017; p. 40. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H10001-H12000/H11650/DR/H11650.pdf (accessed on 14 March 2019). 6. National Oceanic and Atmospheric Administration. Survey H11647—Multibeam and Side Scan sonar descriptive report: National Oceanic and Atmospheric Administration SAIC Doc.; 07–TR–008; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2007; p. 35. Available online: https://data.ngdc.noaa. gov/platforms/ocean/nos/DAPRs/OPR-D302-KR-307.pdf (accessed on 14 March 2019). 7. National Oceanic and Atmospheric Administration. Survey H11648—Multibeam and Side Scan Sonar Data Acquistion & Processing Report; OPR–D302–KR–07; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2007; p. 35. Available online: https://data.ngdc.noaa.gov/platforms/ocean/nos/DAPRs/ OPR-D302-KR-307.pdf (accessed on 14 March 2019). 8. National Oceanic and Atmospheric Administration. Survey H11872—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 09–TR–035; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2008; p. 221. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H10001-H12000/H11872/DR/H11872.pdf (accessed on 14 March 2019). 9. National Oceanic and Atmospheric Administration. Survey H11873—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 09–TR–043; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2008; p. 159. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H10001-H12000/H11873/DR/H11873.pdf (accessed on 14 March 2019). 10. National Oceanic and Atmospheric Administration. Survey H11992—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 09–TR–045; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2008; p. 105. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H10001-H12000/H11992/DR/H11992.pdf (accessed on 14 March 2019). 11. National Oceanic and Atmospheric Administration. Survey H11874—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 09–TR–044; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2008; p. 152. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H10001-H12000/H11874/DR/H11874.pdf (accessed on 14 March 2019). 12. National Oceanic and Atmospheric Administration. Survey H12001—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–001; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2009; p. 65. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H12001-H14000/H12001/DR/H12001.pdf (accessed on 14 March 2019). 13. National Oceanic and Atmospheric Administration. Survey H12093—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 11–TR–001; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2010; p. 35. Available online: http://surveys. ngdc.noaa.gov/mgg/NOS/coast/H12001-H14000/H12093/DR/H12093.pdf (accessed on 14 March 2019). Geosciences 2019, 9, 231 21 of 24 14. National Oceanic and Atmospheric Administration. Survey H12094—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–040; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2010; p. 113. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H12001-H14000/H12094/DR/H12094.pdf (accessed on 14 March 2019). 15. National Oceanic and Atmospheric Administration. Survey H12002—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–002; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2010; p. 74. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H12001-H14000/H12002/DR/H12002.pdf (accessed on 14 March 2019). 16. National Oceanic and Atmospheric Administration. Survey H12003—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–003; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2010; p. 31. Available online: http://surveys. ngdc.noaa.gov/mgg/NOS/coast/H12001-H14000/H12003/DR/H12003.pdf (accessed on 14 March 2019). 17. National Oceanic and Atmospheric Administration. Survey H12091—Multibeam and Side Scan Sonar Data Acquistion and Processing Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–010; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2010; p. 73. Available online: https://data.ngdc.noaa.gov/ platforms/ocean/nos/DAPRs/OPR-D302-SA-309_DAPR.pdf (accessed on 14 March 2019). 18. National Oceanic and Atmospheric Administration. Survey H12092—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–030; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2010; p. 156. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H12001-H14000/H12092/DR/H12092.pdf (accessed on 14 March 2019). 19. National Oceanic and Atmospheric Administration. Survey H12093—Multibeam and Side Scan Sonar Data Acquisition and Processing Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–010; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2010; p. 73. Available online: https://data.ngdc.noaa.gov/platforms/ocean/nos/DAPRs/OPR-D302-SA-309_DAPR.pdf (accessed on 14 March 2019). 20. National Oceanic and Atmospheric Administration. Survey H12095—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–040; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2010; p. 98. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H12001-H14000/H12095/DR/H12095.pdf (accessed on 14 March 2019). 21. National Oceanic and Atmospheric Administration. Survey H12336—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–030; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2011; p. 53. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H12001-H14000/H12336/DR/H12336.pdf (accessed on 14 March 2019). 22. National Oceanic and Atmospheric Administration. Survey H12338—Multibeam and Sidescan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–030; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2011; p. 81. Available online: https://data.ngdc.noaa. gov/platforms/ocean/nos/coast/H12001-H14000/H12338/DR/H12338.pdf (accessed on 14 March 2019). 23. National Oceanic and Atmospheric Administration. Surveys H12160, H12161—Multibeam and Sidescan Sonar Data Acquisition and Processing Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–038; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2011; p. 62. Available online: https://data.ngdc.noaa.gov/platforms/ocean/nos/DAPRs/OPR-D302-KR-310_DAPR.pdf (accessed on 14 March 2019). 24. National Oceanic and Atmospheric Administration. Survey H12337—Multibeam and Side Scan sonar descriptive report: National Oceanic and Atmospheric Administration SAIC Doc.; 10–TR–030; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2011; p. 41. Available online: http://surveys.ngdc. noaa.gov/mgg/NOS/coast/H12001-H14000/H12092/DR/H12337.pdf (accessed on 14 March 2019). 25. National Oceanic and Atmospheric Administration. Survey H12339—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration Saic Doc.; 10–TR–030; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2011; p. 41. Available online: http://surveys. ngdc.noaa.gov/mgg/NOS/coast/H12001-H14000/H12092/DR/H12339.pdf (accessed on 14 March 2019). Geosciences 2019, 9, 231 22 of 24 26. National Oceanic and Atmospheric Administration. Survey H12394—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 13–TR–002; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2012; p. 56. Available online: https://data.ngdc.noaa.gov/platforms/ ocean/nos/coast/H12001-H14000/H12394/DR/H12394_DR.pdf (accessed on 14 March 2019). 27. National Oceanic and Atmospheric Administration. Survey H12395—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 13–TR–003; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2012; p. 49. Available online: https://data.ngdc. noaa.gov/platforms/ocean/nos/coast/H12001-H14000/H12395/DR/H12395.pdf (accessed on 14 March 2019). 28. National Oceanic and Atmospheric Administration. Survey H12396—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 13–TR–004; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2012; p. 55. Available online: https://data.ngdc.noaa.gov/platforms/ ocean/nos/coast/H12001-H14000/H12396/DR/H12396_DR.pdf (accessed on 14 March 2019). 29. National Oceanic and Atmospheric Administration. Survey H12397—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration SAIC Doc.; 10-TR-030; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2012; p. 41. Available online: http://surveys. ngdc.noaa.gov/mgg/NOS/coast/H12001-H14000/H12092/DR/H12397.pdf (accessed on 14 March 2019). 30. National Oceanic and Atmospheric Administration. Survey H12559—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration Leidos Doc.; 14-TR-017; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2013; p. 66. Available online: https://data.ngdc.noaa.gov/platforms/ ocean/nos/coast/H12001-H14000/H12559/DR/H12559_DR.pdf (accessed on 14 March 2019). 31. National Oceanic and Atmospheric Administration. Survey H12560—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration Leidos Doc.; 14-TR-018; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2012; p. 44. Available online: https://data.ngdc.noaa.gov/platforms/ ocean/nos/coast/H12001-H14000/H12560/DR/H12560_DR.pdf (accessed on 14 March 2019). 32. National Oceanic and Atmospheric Administration. Survey H12561—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration Leidos Doc.; 14-TR-019; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2013; p. 52. Available online: https://data.ngdc.noaa.gov/platforms/ ocean/nos/coast/H12001-H14000/H12561/DR/H12561_DR.pdf (accessed on 14 March 2019). 33. National Oceanic and Atmospheric Administration. Survey H12668—Multibeam and Side Scan Sonar Descriptive Report: National Oceanic and Atmospheric Administration; National Oceanic and Atmospheric Administration: Silver Spring, MD, USA, 2014; p. 59. Available online: https://data.ngdc.noaa.gov/platforms/ocean/nos/coast/ H12001-H14000/H12668/DR/H12668_DR.pdf (accessed on 14 March 2019). 34. Pendleton, E.A.; Ackerman, S.D.; Baldwin, W.E.; Danforth, W.W.; Foster, D.S.; Thieler, E.R.; Brothers, L.L. High-Resolution Geophysical DATA Collected Along the Delmarva Peninsula, 2014, USGS Field Activity 2014-002-FA; U.S. Geological Survey Data Release; U.S. Geological Survey: Reston, VA, USA, 2015; Version 3.0. [CrossRef] 35. Sweeney, E.M.; Pendleton, E.A.; Ackerman, S.D.; Andrews, B.D.; Baldwin, W.E.; Danforth, W.W.; Foster, D.S.; Thieler, E.R.; Brothers, L.L. High-Resolution Geophysical Data Collected Along the Delmarva Peninsula 2015, U.S; Geological Survey Field Activity 2015-001-FA; U.S. Geological Survey: Reston, VA, USA, 2016; Version 3.0. [CrossRef] 36. Pendleton, E.A.; Brothers, L.L.; Thieler, E.R.; Danforth, W.W.; Parker, C.E. National Oceanic and Atmospheric Administration Hydrographic Survey Data Used in a U.S. Geological Survey Regional Geologic Framework Study along the Delmarva Peninsula; Geological Survey Open-File Report 2014-1262; U.S. Geological Survey: Reston, VA, USA, 2015; p. 18. [CrossRef] 37. Pendleton, E.A.; Brothers, L.L.; Thieler, E.R.; Sweeney, E.M. Sand ridge morphology and bedform migration patterns derived from bathymetry and backscatter on the inner-continental shelf oshore of Assateague Island, USA. Cont. Shelf Res. 2017, 144, 80–97. [CrossRef] 38. Pendleton, E.A.; Brothers, L.L.; Sweeney, E.M.; Thieler, E.R.; Foster, D.S. Sediment Texture and Geomorphology of the Sea Floor from Fenwick Island, Maryland to Fisherman’s Island; U.S. Geological Survey Data Release; U.S. Geological Survey: Reston, VA, USA, 2017. [CrossRef] 39. Buhl-Mortensen, P.; Dolan, M.; Buhl-Mortensen, L. Prediction of benthic biotopes on a Norwegian oshore bank using a combination of multivariate analysis and GIS classiﬁcation. Ices J. Mar. Sci. 2009, 66, 2026–2032. [CrossRef] Geosciences 2019, 9, 231 23 of 24 40. Dartnell, P.; Gardner, J.V. Predicting seaﬂoor facies from multibeam bathymetry and backscatter data. Photogramm. Eng. Remote Sens. 2004, 70, 1081–1091. [CrossRef] 41. Stephens, D.; Diesing, M. A Comparison of supervised classiﬁcation methods for the prediction of substrate type using multibeam acoustic and legacy grain-size data. PLoS ONE 2014, 9, e93950. [CrossRef] 42. Marsh, I.; Brown, C. Neural network classiﬁcation of multibeam backscatter and bathymetry data from Stanton Bank (Area IV). Appl. Acoust. 2009, 70, 1269–1276. [CrossRef] 43. Galparsoro, I.; Agrafo, X.; Roche, M.; Degrendele, K. Comparison of supervised and unsupervised automatic classiﬁcation methods for sediment types mapping using multibeam echosounder and grab sampling. Ital. J. Geosci. 2015, 134, 41–49. [CrossRef] 44. Simons, D.G.; Snellen, M. A Bayesian approach to seaﬂoor classiﬁcation using multi-beam echo-sounder backscatter data. Appl. Acoust. 2009, 70, 1258–1268. [CrossRef] 45. Mitchell, N.C.; Hughes-Clarke, J.E. Classiﬁcation of seaﬂoor geology using multibeam sonar data from the Scotian Shelf. Mar. Geol. 1994, 121, 143–160. [CrossRef] 46. Buscomb, D.; Grams, P.E. Probabilistic substrate classiﬁcation with multispectral acoustic backscatter: A comparison of discriminative and generative models. Geosciences 2018, 8, 395. [CrossRef] 47. Ramsey, K.W. Response to Late Pliocene climate chang: Middle Atlantic Coastal Plain, Virginia and Delaware. In Quaternary Coasts of the United States: Marine and Lacustrine Systems; Fletcher, C.H., III, Wehmiller, J.F., Eds.; Society for Sedimentary Geology—SEPM: Tulsa, OK, USA, 1992; Volume 48. 48. Pazzaglia, F.J. Stratigraphy, petrography, and correlation of late Cenozoic middle Atlantic Coastal Plain deposits: Implications for late-stage passive-mamrgin geologic evolution. Geol. Soc. Am. Bull. 1993, 105, 1617–1634. [CrossRef] 49. Mixon, R.B. Stratigraphic and Geomorphic Framework of Uppermost Cenozoic Deposits in the Southern Delmarva Peninsula, Virginia and Maryland; United States Government Publishing Oce: Washington, DC, USA, 1985; p. 59. 50. McBride, R.A.; Fenster, M.S.; Seminack, C.T.; Richardson, T.M.; Sepanik, J.M.; Hanley, J.T.; Bundick, J.A.; Tedder, E. Holocene barrier-island geology and morphodynamics of the Maryland and Virginia open-ocean coasts: Fenwick, Assateague, Chincoteague, Wallops, Cedar, and Parramore Islands. In Tripping from the Fall Line: Field Excursions for the GSA Annual Meeting, Baltimore, 2015: Geological Society of America Field Guide 40; Brezinski, D.K., Halka, J.P., Ortt, R.A., Jr., Eds.; Geological Society of America: Boulder, CO, USA, 2015; Volume 40, pp. 309–424. 51. Oertel, G.F.; Kraft, J.C. New Jersey and Delmarva Barrier Islands. In Geology of Holocene Barrier Island Systems; Davis, R.A., Ed.; Springer: Berlin/Heidelberg, Germany, 1994; pp. 207–232. 52. Fisher, J.J. Origin of barrier chain shorelines: Middle Atlantic Bight. Geol. Soc. Am. Ann. Program 1967, 115, 66–67. 53. Toscano, M.A.; Kerhin, R.T.; York, L.L.; Cronin, T.M.; Williams, S.J. Quaternary Stratigraphy of the Inner Continental Shelf of Maryland; Maryland Geological Survey: Baltimore, MD, USA, 1989; p. 119. 54. Milliman, J.D.; Pilkey, O.H.; Ross, D.A. Sediments of the continental margin o the Eastern United States. Geol. Soc. Am. Bull. 1972, 83, 1315–1334. [CrossRef] 55. Milliman, J.D. Atlantic Continental Shelf and Slope of the United States—Petrology of the Sand Fraction of Sediments, Northern New Jersey to Southern Florida; US Department of Interior: Washington, DC, USA, 1972; Volume 529-J. 56. Kehrin, R.T.; Williams, S.J. Surﬁcial sediments and later Quaternary sedimentary framework of the Maryland inner continental shelf: Proceedings, Coastal Sediments’ 87. In Proceedings of the American Society of Civil Engineers, New Orleans, LA, USA, 13 May 1987; pp. 2126–2140. 57. Swift, D.J.P.; Field, M.E. Field Evolution of a classic sand ridge ﬁeld: Maryland sector, North American inner shelf Sedimentology. Sedimentology 1981, 28, 461–482. [CrossRef] 58. Uchupi, E. Atlantic Continental Shelf and Slope of the United States Physiography; U.S. Geological Survey: Reston, VA, USA, 1968; Volume 529-C, pp. 1–30. 59. Toscano, M.A.; York, L.L. Quaternary stratigraphy and sea-level history of the U.S. Middle Atlantic Coastal Plain. Quat. Sci. Rev. 1992, 11, 301–328. [CrossRef] 60. Wells, D.V. Non-Energy Resources and Shallow Geological Framework of the Inner Continental Margin o Ocean City, Maryland; US Department of the Interior, Minerals Management Service: Baltimore, MD, USA, 1994; p. 107. 61. Field, M.E. Sand bodies on coastal plain shelves: Holocene record of the U.S. Atlantic inner shelf o Maryland. J. Sediment. Petrol. 1980, 50, 505–528. Geosciences 2019, 9, 231 24 of 24 62. Field, M.E.; Duane, D.B. Post-Pleistocene history of the United States inner continental shelf: Signiﬁcance to origin of barrier islands. Geol. Soc. Am. Bull. 1976, 87, 691–702. [CrossRef] 63. Krantz, D.E.; Hobbs III, C.H.; Wikel, G.L. Coastal processes and oshore geology. In The Geology of Virginia; Bailey, C., Ed.; College of William and Mary: Williamsburg, VA, USA, 2015; p. 44. 64. Fenster, M.S.; Dolan, R.; Smith, J.J. Grain-size distributions and coastal morphodynamics along the southern Maryland and Virginia barrier islands. Sedimentology 2016, 63, 809–823. [CrossRef] 65. McMullen, K.Y.; Paskevich, V.F.; Poppe, L.J. USGS east-coast sediment analysis—Procedures, database, and GIS data: U.S. In GIS Data Catalog; Poppe, L.J., Williams, S.J., Paskevich, V.F., Eds.; Geological Survey Open-File Report 2005–1001; U.S. Department of the Interior: Washington, DC, USA, 2005; Version 2.2. Available online: http://woodshole.er.usgs.gov/openﬁle/of2005-1001/htmldocs/datacatalog.htm (accessed on 20 May 2019). 66. Barnhardt, W.A.; Kelley, J.T.; Dickson, S.M.; Belknap, D.F. Mapping the Gulf of Maine with side-scan sonar: A new bottom-type classiﬁcation for complex seaﬂoors. J. Coastal Res. 1998, 14, 646–659. 67. Wenworth, C.K. A scale of grade and class terms for clastic sediments. J. Geol. 1922, 30, 377–392. [CrossRef] 68. Go, J.A.; Olson, H.C.; Duncan, C.S. Correlation of side-scan backscatter intensity with grain-size distribution of shelf sediments, New Jersey margin. Geo-Mar. Lett. 2000, 20, 43–49. [CrossRef] 69. Barnhardt, W.A.; Andrews, B.D.; Ackerman, S.; Baldwin, W.; Hein, C. High Resolution Geologic Mapping of the Inner Continental Shelf: Cape Ann to Salisbury Beach Massachusetts; Open-File Report 2007-1373; U.S. Geological Survey: Reston, VA, USA, 2009; p. 50. 70. Che Hasan, R.; Ierodiaconou, D.; Monk, J. Evaluation of four supervised learning methods for benthic habitat mapping using backscatter from multi-beam sonar. Remote Sens. 2012, 4, 3427–3443. [CrossRef] 71. Ierodiaconou, D.; Monk, J.; Rattray, A.; Laurenson, L.; Versace, V.L. Comparison of automated classiﬁcation techniques for predicting benthic biological communities using hydroacoustics and video observations. Cont. Shelf Res. 2011, 31, S28–S38. [CrossRef] 72. Twichell, D.C.; Pendleton, E.A.; Baldwin, W.; Flocks, J. Subsurface control on seaﬂoor erosional processes oshore of the Chandeleur Islands, Louisiana. Geo-Mar. Lett. 2009, 29, 349–358. [CrossRef] © 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png

Geosciences Unpaywall

http://www.deepdyve.com/lp/unpaywall/optimizing-an-inner-continental-shelf-geologic-framework-investigation-Ea9YVn18pc

Optimizing an Inner-Continental Shelf Geologic Framework Investigation through Data Repurposing and Machine Learning

Loading next page...

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher: Unpaywall
ISSN: 2076-3263
DOI: 10.3390/geosciences9050231
Publisher site: See Article on Publisher Site

Abstract

Journal

Geosciences – Unpaywall

Published: May 21, 2019

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Optimizing an Inner-Continental Shelf Geologic Framework Investigation through Data Repurposing and Machine Learning

Optimizing an Inner-Continental Shelf Geologic Framework Investigation through Data Repurposing and Machine Learning

Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Optimizing an Inner-Continental Shelf Geologic Framework Investigation through Data Repurposing and Machine Learning

Optimizing an Inner-Continental Shelf Geologic Framework Investigation through Data Repurposing and Machine Learning

References

Abstract

Journal

Recommended Articles

There are no references for this article.

Our policy towards the use of cookies