How do we build DEMs?

DEM developers at CIRES use the best available data to make the best available 3D maps for resilient coastal communities.

DEM Development Process

The DEM build process follows a set of steps when building coastal DEMs (Figure 1):

Figure 1: Diagram illustrating the basic sequence of DEM development procedures.

1. Gather elevation data from multiple sources

Coastal DEMs are developed using all of the best available digital elevation data. Shoreline, topographic, bathymetric, and shoreline-crossing data are obtained from many different government agencies, academic institutions, and private companies (e.g., Figure 2). Many of the datasets retrieved for use in coastal DEM development are already archived at NOAA/NCEI. The original data are collected and compiled in numerous ways (e.g. multibeam swath sonar, lidar, satellite altimetry, USGS quadrangle digitization, etc.), in different terrestrial environments, across any number of time periods, and at various scales and resolutions (e.g., Table 1).

Figure 2: Spatial coverage of datasets used in DEM of Arena Cove, California.

Topographic datasets used in the development of the DEM of Eureka, California. — Table 1: Topographic datasets used in the development of the 1/3 arc-second NAVD 88 DEM of Eureka, California.

Sometimes, important bathymetric or topographic features are not represented by any existing digital data. Such absence of data necessitates the hand digitization of features for inclusion in DEMs (e.g, Figure 3). Satellite imagery, available from ESRI, NASA, or Google is often used to help assess the current morphology of features.

Figure 3: Example of hand-digitized breakwaters near Ocean City, Maryland DEM, shown in red. These breakwaters were not represented in any available topographic datasets.

2. Convert into common file formats and common horizontal and vertical datums (reference frames)

Before generating a DEM, all original datasets are converted into viewable file types and common horizontal (NAD 83 or WGS 84) and vertical (NAVD 88 or MHW) datums. FME or similar software tools are used to shift horizontal datums while VDatum is often used to shift the vertical datum of original datasets. In areas where the VDatum software tool does not provide coverage, vertical datum relationships are calculated based on local tide station values.

3. Visually evaluate and edit data

After the original datasets have been converted, DEM developers assess the datasets for quality and accuracy. It is important that each dataset is accurate both within itself and when compared to and overlaid with other datasets. Datasets must be consistent in order to transition smoothly when their edges meet or overlay. If a dataset contains errors or is overlaid by a more accurate dataset, it is edited and clipped as necessary. DEM developers visually assess data using ESRI's ArcGIS, Blue Marble Geographics' Global Mapper, IVS 3D's Fledermaus, and other software products (Figure 4).

Figure 4: Illustration of errors in a LiDAR dataset.

4. Build and evaluate DEMs

Coastal DEMs are generated using the shareware package MB-System, a National Science Foundation (NSF)-funded software package designed for manipulating multibeam swath sonar data. The edited, clipped, and interpolated digital elevation datasets are converted to ASCII xyz format and assigned a relative gridding weight in a DEM data "hierarchy" (e.g., Table 2). This ensures that datasets with the highest quality and resolution have the greatest impact in determining DEM elevation values. MB-System is used to generate the final DEM ASCII grid, using the weighted xyz datasets. A tight spline tension gridding method is used to interpolate values for DEM grid cells with no real data available; this ensures every cell in the DEM is assigned an elevation value.

Table 2: Sample data heirarchy used to assign gridding weight in MB-System.

Once generated, the DEM has a vertical datum that corresponds to that of its input xyz data. The DEM can be transformed to a new vertical datum, however, to meet the specifications of individual users. The vertical transformation of a DEM can be accomplished by (a) adding a conversion grid of the same extent to the DEM or (b) re-transforming the original data and repeating the DEM generation process in MB-System.

In general, we build DEMs in the North American Vertical Datum of 1988 (NAVD 88) and transform them into new DEMs with a Mean High Water (MHW) vertical datum by adding a conversion grid to the NAVD 88 DEMs. Conversion grids are developed by using the datum relationships from the VDatum software tool, with a thin plate spline interpolation method to estimate inland values. In areas where the VDatum software tool does not provide coverage (e.g., Alaska) we create a conversion grid using the datum relationships from local tide stations with a Kriging interpolation method to estimate offshore and inland values. All DEMs and conversion grids are thoroughly reviewed and assessed for errors throughout the process.

DEMs are evaluated using several different methods. First, the DEMs are visually inspected for anomalous "spikes" and "wells" using GIS software such as ESRI's ArcGIS, Blue Marble Geographics' Global Mapper or IVS 3D's Fledermaus. These software programs render three-dimensional views of the DEM grids that can be rotated, color-coded by depth, and vertically exaggerated.

A "slope" map is also generated from the DEM (e.g., Figure 5). A slope map visually highlights changes in slope throughout the DEM, and should reflect natural morphology. Artificial features, often located at the edge of datasets, can be easily spotted on a slope map. A close visual inspection of DEMs reveals errors, which may necessitate reevaluation of the original data and the regridding of the DEM.

Figure 5: Slope map of a DEM of Craig, Alaska.

Coastal DEM developers assess the horizontal and vertical accuracy of final DEMs, based on the metadata of the original datasets. The final DEM ASCII grid cell values are directly compared with the original data (e.g., Figure 6).

Figure 6: Histogram comparison of an original bathymetric dataset (U.S. Army Corps of Engineers Alabama Hydrographic Surveys) and the Mobile, Alabama DEM.

For a more realistic comparison with true elevation values, NOAA's National Geodetic Survey (NGS) monument locations are extracted from datasheets (e.g., Figure 7). Values from tide stations within the DEM region or values from USGS topographic data can also be used as known elevations that can be compared with the DEM values.

Figure 7: Locations of NGS monuments within the Portland, Maine DEM boundaries.

5. Document DEM development

Following coastal DEM development, the processing procedures, data sources and analyses information are thoroughly documented in a technical report. The technical report allows DEM users to understand the quality and accuracy of the DEM; it also allows anyone to replicate the DEM development process. NCEI creates a detailed metadata record for the DEM, which meets Federal Geographic Data Committee (FGDC) standards.

6. Distribute DEMs on the Internet for public access

When the DEM development process is complete, the DEM is posted online at NCEI. Documentation and metadata accompany each of the DEMs. Information on specific DEM projects: NCEI Coastal Elevation Models.

3D image of the 3 arc-second Pago Pago DEM, American Samoa. — 3-D image of the 3 arc-second Pago Pago DEM, American Samoa.