Data Processing

Navigation

Position and elevation data associated with each sounding were post-processed using differential corrections derived from the base/rover setup. Three GPS base stations located at benchmarks east of the wilderness breach (U374), near the Wilderness Visitor Center (VC), and near the town of Robins Rest (REST; fig. 2) were used for the survey. Base station coordinates were verified with Continuously Operating Reference Stations (CORS) using the Online Positioning User Service (OPUS; http://www.ngs.noaa.gov/OPUS/). OPUS computed that the reference stations had a horizontal and vertical error of 0.7 centimeter (cm). Applying the reference station coordinates, GPS data acquired from the rover were processed to the concurrent GPS session data at the base station using GrafNav version 8.5 software (Waypoint Product Group). The horizontal and vertical coordinates were recorded in the World Geodetic System of 1984 (WGS84) reference frame and exported as American Standard Code for Information Interchange (ASCII) files. Backpack and wheel GPS trajectory data were converted within GrafNav© to North American Datum of 1983 (NAD83; 2011) and North American Vertical Datum of 1988 (NAVD88; GEOID12A) and exported as ASCII files.

Single-Beam Soundings

Soundings were merged with differentially processed GPS data and sound velocity profiles, using Matlab. Each transect was visually inspected for elevation outliers, and dropouts associated with wave breaking in the surf zone were manually corrected. Typically, the highest intensity return is generated by the seafloor surface. Breaking waves in the surf zone may create air bubbles in the water column and create an erroneous peak in backscatter intensity. This creates errors in the interpreted seafloor reflection. When this type of error was suspected, a corrected seafloor elevation was manually digitized by analyzing the complete waveform signal recorded by the Odom. The soundings were then corrected for the speed of sound (table 2) associated with the mean water temperature and salinity. A moving average filter was then applied to the soundings in order to reduce instrument noise and depth variations associated with the pitch and roll of the PWC. The soundings were referenced to the height of the GPS antenna and subsequently to the WGS84 ellipsoid. Referencing to the ellipsoid reduces error associated with determining mean water level and (or) tidal elevations. National Oceanic and Atmospheric Administration’s VDatum software was used to convert the sounding positions and elevations to NAD83 and NAVD88 (GEOID12A), respectively.

The accuracy of the single-beam soundings was evaluated by identifying locations where survey track lines crossed and where soundings from each line were within a horizontal distance of 0.25 m of each other. Any track line with an error greater than 0.6 m along multiple crossings was removed. Evaluation of the track line crossings not removed indicated there was a root mean square (RMS) vertical uncertainty of 16.5 cm with 3.8 cm bias between the two PWCs. Large elevation differences often appear near channels where elevations vary rapidly over short distances. The RMS error for PWC2 crossing a line it previously surveyed was 15.4 cm. When PWC1 crossed a line it previously surveyed, the RMS error was 18.1 cm. The larger error for PWC1 may be attributed to the narrower four-degree transducer beam. For consistency between the PWC elevations, PWC1 elevations were adjusted to that of PWC2. There is a total of 5.4 cm error in the transformation using the conversion from the WGS84 ellipsoid to NAVD88 (GEOID12A; http://vdatum.noaa.gov/docs/est_uncertainties.html). Applying the square root of the sum of the datum conversion uncertainty and the sounding uncertainty results in a combined vertical error of 17.4 cm. Horizontal uncertainty is assumed to be, at most, half of the vertical uncertainty (8.7 cm).

Ground Based Elevations

Ground-based elevation errors were calculated by computing the vertical differences at crossings occurring at least 1 minute apart. The calculated RMS error is 12.5 cm. Elevation differences between the backpack and wheel data points and PWC data points indicate the backpack and wheel elevations were 5 cm higher than elevations recorded using PWCs. Given the high degree of uncertainty arising from variations of the backpack surveyor stride over a subaqueous shoal surface and tilting of the wheel mounted GPS, the data were adjusted to the PWC elevation at the crossings.

Digital Elevation Model

The fully processed and corrected bathymetric data points (XYZ) were separated into six regions: Fire Island Inlet, Great South Bay, Bellport Bay, Narrow Bay, the wilderness breach, and the shoreface, with the wilderness breach ebb-tidal delta also located in the shoreface region. Overlap also occurs between Great South Bay and Bellport Bay, and Bellport Bay and Narrow Bay to ensure consistency between boundaries. Fire Island Inlet, Great South Bay, and Narrow Bay data points consist solely of PWC data while the wilderness breach, Bellport Bay, and the shoreface data points include PWC, backpack, and (or) wheel GPS data. The bathymetric data points were exported as ASCII files from Matlab. A digital elevation model (DEM) was not created for Great South Bay, Bellport Bay, Fire Island Inlet, and Narrow Bay due to large spacing between transects. Data points for the wilderness breach and the shoreface were converted to an ArcGIS shapefile using the “Create Feature Class from xy table” tool in ArcCatalog. The resulting point shapefiles were then used to create a triangulated irregular network (TIN) dataset for the shoreface and the wilderness breach using the “Create TIN” tool. A raster dataset was created from the TIN dataset using a 50-m cell size with a “natural neighbors” interpolation for the shoreface (fig. 12) and the wilderness breach bathymetry (fig. 13) using the “TIN to Raster” tool. Due to the removal of erroneous lines and inability to survey certain lines in the field, some interpolated grid cells are more than 2 cell sizes (100 m) away from a sounding. When a cell was more than 2 cell sizes away from a data point, the cell was set to null and appears as a gap in the bathymetry. In order to determine the uncertainty associated with the rasters, we used the method of Lentz and Hapke (2011) and Lentz and others (2013), and withheld 10 percent of the data points when creating the TIN using the ArcGIS “subset features” tool. The raster was then sampled using ArcGIS tool “Extract Values to Points” tool at the corresponding eastings and northings of the withheld data, and a RMS error between the sampled bathymetry and the measured value was found to be 15.3 cm vertically for the shoreface bathymetry and 31.1 cm vertically for the wilderness breach bathymetry.

Figure 12. A 50-meter grid of October 2014 bathymetry of the shoreface of Fire Island, N.Y. [Click image to enlarge]

Figure 13. A 50-meter grid of October 2014 bathymetry of the wilderness breach, Fire Island, N.Y. [Click image to enlarge]