A seamless, high-resolution, coastal digital elevation model (DEM) for Southern California

ESRI geoprocessing history
  Tool location:
    C:\ArcGIS\ArcToolbox\Toolboxes\Spatial Analyst Tools.tbx\ExtractByMask
  Command issued:
    ExtractByMask la1unclip DEMFullCoverageAreas\la1_fullcoveragearea D:\HazardsModel\DEM\FinalDEMs\CoastalHighRes\DEMs_3m\FinalMosaics\la1final
  _________________
  Spatial data description
  Raster dataset information
  Raster format: GRID
  SDTS raster type: Grid Cell
  Number of raster bands: 1
  Raster properties
  Origin location: Upper Left
  Has pyramids: FALSE
  Has colormap: FALSE
  Data compression type: Default
  Display type: matrix values
  Cell information
  Number of cells on x-axis: 7317
  Number of cells on y-axis: 8160
  Number of cells on z-axis: 1
  Number of bits per cell: 32
  Cell Size
  X distance: 3.000000
  Y distance: 3.000000
  _________________

(Summary) Performed by Patrick Barnard and Dan Hoover:

  ArcGIS was the primary software used for DEM construction.
  For each individual DEM, the native data sets were mosaicked into a single grid
  to preserve the original surfaces as closely as possible.
  Prior to mosaicking, data sets were gridded and/or resampled to 3 m resolution (if necessary),
  and their spatial extents modified according to the following guidelines:

  1. Data sets of comparable quality
     (for example, overlapping multibeam data) or
     where relative data quality could not be determined
     (for example, older multibeam and recent but lower resolution personal-watercraft data,
     were not clipped.
     In these instances the overlapping regions were blended together
     using the Blend algorithm in the Mosaic to New Raster tool in Arc Toolbox.

  2. In overlapping regions where the quality of one data set was clearly inferior
     to the other (for example, IfSAR overlapping with Lidar),
     the spatial extent of the inferior data set was clipped so there was minimal overlap,
     typically about ~10-30 m, with the superior data set.
     The overlapping regions then were smoothed together using the Blend algorithm.
     This range of overlap was found to be the most efficient
     for ensuring a smooth transition between data sets
     while minimizing the use the of data set of inferior quality.

(Detailed) Performed by Patrick Barnard and Dan Hoover:

  1. Divide study area into ~10 km alongshore segments
     A. Define DEM coverage area/polygon that extends
        ~10 km alongshore from -20 m isobath to 20 m topographic contour or
        750 m from back beach, whichever is longer
     B. Ensure that adjacent DEM coverage areas overlap by ~ 250 m
     C. Cut off DEMs at county boundaries with ~500 m overlap

  2. Acquire most recent or highest resolution data sets in DEM coverage areas (Figure 2)
     A. Lidar
     B. Multibeam bathymetry
     C. Local high-resolution beach topography (usually ATV-acquired) and
        nearshore bathymetry (usually PWC-acquired)
     D. IfSAR

  3. Fill gaps with older/lower resolution data sets
     A. Lidar (for example, NOAA Digital Coast, 1997-98)
     B. NOAA lower-resolution multibeam (for example, Los Angeles Harbor entrance)
     C. Regional, lower resolution DEMs (for example, NOAA Santa Barbara Channel 10-m DEM)

  4. Convert all data sets into identical
     horizontal coordinate system, vertical datum, and grid resolution
     A. Horizontal coordinate system: UTM NAD 83 Zone 11 North
     B. Vertical Datum: NAVD88
        -If different (usually MLLW), convert using local NOAA tide station information
        (http://tidesandcurrents.noaa.gov/) based on survey metadata
     C. Grid resolution: 3 m
        -If already gridded at higher or < 10 m resolution,
         resample to 3 m using bilinear interpolation
        -If already gridded at resolution of =10 m, export as xyz,
         re-import as xyz, create TIN (triangular irregular network),
         create 3 m grid from TIN using linear interpolation of the TIN triangles,
         clip to survey extent
        -If ungridded then for:
        -Lidar (topography):
         3 m grid using natural neighbor interpolation
         to preserve abrupt elevation changes (for example, beaches backed by cliffs)
        -Multibeam:
         3 m grid using inverse distance weighting using "Average Gridder"
         in Fledermaus- ideal for data sets >   10 million points
        -Lower resolution surveys (for example, personal watercraft-collected bathymetry):
         create TIN from points then convert to 3 m grid
         using linear interpolation of the TIN triangles

  5. Clip data sets to DEM/coverage needs, if necessary
     A. Useful for data management and processing efficiency
     B. Necessary for very large data sets, such as county-wide IfSAR or
        very large Lidar data sets (for example, Los Angeles County)
     C. Clip ocean and waves from topographic Lidar and IfSAR
        -Clip water level by determining sea level at time of survey then
         using "Extract by Attributes" tool in Arc Toolbox
        -Clip wave crests manually with mask

  6. Manage overlapping data sets
     A. Data sets allowed to overlap extensively only if they are of comparable quality,
        otherwise minimal (~10-30 m) overlap to ensure smooth DEM transitions
     B. Clip IfSAR data (lower quality)
        to minimal overlap with topographic Lidar (better quality)
     C. Clip low-resolution data sets gridded to higher resolution,
        such as Personal Watercraft data and regional DEMs,
        to minimal overlap with adjacent high-resolution data sets
        (usually multibeam and topographic Lidar)
     D. Extensive overlap between adjacent Lidar and multibeam data sets rare but
        allowed as quality is comparable

  7. Fill in data gaps between high-resolution data sets
     A. If no high-resolution data exist between the 10 m isobath and coastal Lidar, or
        in protected harbors/embayments or other areas where interpolation
        from surrounding data sets will create a surface unlikely
        to accurately reflect actual bathymetry/topography,
        fill in with regional DEMs or other low-resolution data sets.
        Otherwise interpolate across gaps.
        -Filling in with regional DEMs/other low-resolution data:
        -Clip best available regional DEM to fill gap with minimal overlap (~10-30 m)
         with adjacent high resolution data sets
        -Export clipped grid as xyz, reimport as points, create tin,
         create 3 m grid from tin, clip to gap extent
        -Interpolation
        -Create preliminary DEM using Mosaic tool with the following settings:
         Coordinate System: UTM Zone 11 North
         Pixel Type: 32_Bit_Float
         Cell Size:  3
         Mosaic Method: Blend
         Mosaic Color Map: Last
        -Create mask of data gap(s) to fill with minimal overlap
         with preliminary DEM surface
        -Clip preliminary DEM with mask, export clipped grid as xyz,
         reimport as points, create tin, create 3 m grid from tin, clip to gap extent

  8. Compile final DEMs
     A. Load all data sets for DEM
     B. Verify all significant data gaps filled (few missing cells OK) in DEM coverage area
     C. Build DEM using Mosaic to New Raster tool in ArcGIS with same setting as in Item #7
     D. Clip DEM to DEM coverage area
     E. Create contours and plot cross-shore profiles to verify data quality and consistency