Voids in the data are found in the Strait of Juan de Fuca at approximately the northern extent of U.S. territorial waters. Additional voids may result due to the orientation of the bounding box of the data relative to the Washington State Plane North (WA SPN) projection grid. No other voids are intentional.
The horizontal accuracy is a function of the accuracy of the underlying data plus the errors introduced during the production of this DEM (including reprojecting, resampling and adjusting the horizontal datum).
The vertical accuracy is a function of the accuracy of the underlying data (see data sources section) plus the errors introduced during the production of this DEM. Production of the DEM included reprojecting the data to Washington State Plane North; converting units from meters to feet; adjusting the vertical datum to NAVD88 using CORPSCON 5.11.08 (for terrestrial data); VDatum 1.06 (marine data south of 48 10'); or by adding a value from a NAVD88 correction surface developed from NOS tidal benchmarks for soundings north of 48 10' (see processing steps section for details), and finally resampling the data to a 30-foot raster resolution.
In addition to the transformation errors described above, bathymetry-bathymetry and terrestrial-terrestrial overlapping data sets were merged together by using the ArcGIS 9.0 "Mosaic to New Raster" command with the "Blend" option. This proprietary algorithm feathers overlapping datasets into one another to minimize edge artifacts. It will also lower the fidelity of accurate datasets when they are "blended" with lower fidelity data.
For submittal to:
The US Environmental Protection Agency, Region 10, Seattle, WA
The Washington State Department of Ecology, Northwest Regional Office, Bellevue, WA
Finally, I merged all of these datasets into the Master DEM by allowing the bathy LIDAR to fill NODATA gaps in the Master DEM.
First the data was divided into 10 tiles of 0.5 x 0.5 degree extent (this was necessary to save memory while interpolating at 30-foot resolution). All of the points within the tile plus a 1 km overlap were gridded using a TIN interpolator. This TIN included a thin ribbon of points taken from the shoreline of the MASTER DEM to ensure that the TIN merged perfectly at the shoreline. The TIN was converted to a raster and a shaded relief map was produced from the raster. Obvious errors in the data were edited out of the point file. This process was repeated until all obvious errors in the bathymetry were eliminated.
The resulting raster was subtracted from the appropriate NOS hydrographic tile(s) and a mean difference was established between the two elevation datasets. The correction was applied to the Swath Bathymetry (since these data were not vertically controlled to survey grade) so that the two datasets approximately matched. The corrections applied were: -4.16 ft (Duwamish Delta), -7.3 ft (Puyallup), -5.14 ft (Nisqually Delta), +6 ft (Possession Sound), +5.6 (Admiralty Inlet).
The low-resolution NOS bathymetry was trimmed around the swath bathymetry with a 300 ft overlap. Then these datasets were merged together using ArcGIS's "Mosaic to New Raster" with the "Blend" option, which is a proprietary algorithm that feathers the two data sets into one another reducing the margin artifacts.
The Admiralty Inlet swath bathymetry from the UW overlapped three tiles. Since the low-resolution NOS bathymetry was mathematically identical in each of the three tiles (where they overlapped) only a single correction was necessary (+5.6 ft).
I used ArcGIS's proprietary "Mosaic to New Raster" with the "Blend" option to smooth the transition between the low-quality elevation data and the high-quality LIDAR-derived data. The overlap between the two datasets was hand-drawn and varied from a few hundred feet in the eastern Puget Lowlands, to many hundreds of feet over the Olympics.
A few NODATA gaps remained in River Channels. And I set back to NODATA a few small areas that had obvious errors (like the seam between the Lidar Bathymetry and Multibeam bathymetry off Alki Point), a mountain in Elliott Bay that doesn't exist, etc. I then interpolated over the gaps using a TIN made from the pixels surrounding each data gap. This formed the final DEM.