USGS Home
Contact USGS
Search USGS

Construction of a 3-Arcsecond Digital Elevation Model for the Gulf of Maine

Common Processing Steps and Data

Title Page List of Figures Conversion Factors Abstract Introduction Common Processing Steps and Data Bathymetry Gulf of Maine Mapping Initiative Swath Bathymetry National Ocean Service NGDC/NOAA Data System Bathymetry Northwest Atlantic Bathymetry Olex Data Smith Sandwell Global Topography U.S. Naval Oceanographic Office: Bathymetry UNH/NOAA Joint Hydrographic Center: Bathymetry and Acoustic Backscatter USGS Massachusetts Bay Bathymetric Data USGS Massachusetts CZM Geologic Mapping Data USGS Middle Ground Shoal Bathymetric Data Topography National Elevation Data NOAA CSC Coastal LIDAR Shuttle Radar Mission Topography Grid Assembly Data Products Acknowledgments References Cited Data Processing Scripts and Software Glossary

All data were converted to ASCII XYZ format. Z-values are negative for bathymetric data and positive for topographic data. The first step in dealing with data from various sources was to ensure a common horizontal and vertical datum. The second step was merging datasets that have varying resolution and accuracy. Higher resolution and more accurate surveys were used where they were available, with lower resolution, less accurate surveys filling in the gaps. We approached this problem by using a combination of buffering and blockmedian techniques. Horizontal Datum All data were converted to North American Datum of 1983 (NAD 83), except for data in World Geodetic System of 1984 (WGS 84), where it was determined that horizontal differences (approximately 1 meter) between NAD83 and WGS84 were insignificant at the resolution of the 3-arcsecond (approximately 90-m) grid. The horizontal accuracy of the original sounding data varies depending on the source and year of collection; for recent swath surveys that used differential Global Positioning System (GPS), the accuracy is within a few meters, but with older surveys, the horizontal accuracy is not known. These horizontal errors likely translated into vertical errors in the estimated elevations. Random errors in the original data were reduced by the averaging process onto the 3-arcsecond grid cells, while biased errors remained. Vertical Datum All depth values were converted to NAVD88 using information from the Oceanic and Atmospheric Administration’s (NOAA) VDatum project. Because the VDatum regional grid for the Gulf of Maine did not extend into Canadian waters, we interpolated mean lower low water (MLLW) and mean low water (MLW) corrections from the VDatum tidal model in the Gulf of Maine, which did include Canadian waters. This transformed depths from the MLW or MLLW datum into the tidal model datum. The depths were further transformed into the NAVD88 datum by subtracting 0.09 m, the median value of NAVD88 values above the tidal model datum obtained at five long-term tide gauges in the Gulf of Maine. The maximum cumulative uncertainty from these transformations is estimated to be 0.13 m (National Oceanic and Atmospheric Administration, 2011b). Buffering and Manual Clipping Soundings from difference sources were mapped in ArcGIS to visually assess the resolution and overlap regions. In some regions, buffers of 1,000 m were applied to the higher resolution data (any points from the coarser dataset that fell within 1,000 m of the higher resolution data were removed). This was accomplished using the "near" tool from the ArcMap 9.2 ArcToolbox analysis tools. In other regions, overlapping survey data were removed with manual polygon selection. Blockmedian All data were processed using a Generic Mapping Tools (GMT) blockmedian filter command over the 3-arcsecond (approximately 90-m) grid cells. Using a blockmedian filter removes outlier points without modifying the average depth of the cell. For high-resolution (approximately 1-m) swath bathymetry, the data were preprocessed using a 1-arcsecond (approximately 30-m) blockmedian filter before combining the filtered data with data from other datasets to be processed with the 3-arcsecond blockmedian filter. This technique allowed for more efficient processing while still allowing the high-resolution data to dominate the median value of the resulting 3-arcsecond cell. For example, if a stray point from a low-resolution survey fell within the 90-m cell, the median would still be determined by the nine points derived from the high-resolution survey. Details about individual input datasets and processing that was performed on those data are discussed by dataset in the subsequent sections.