John W. Jones
US Geological Survey, Eastern Geographic Science Center, Reston, VA, USA

In addition to water surfaces interpolated from Everglades Depth Estimation Network (EDEN) stage data, modeled spatially distributed water depths require the development of a system-wide digital elevation model (DEM). Using a system called the “Airborne Height Finder” (AHF), the U.S. Geological Survey (USGS) collected over 43,000 highly accurate (RMSE < 15cm) elevation data points at an approximate spatial sampling distance of 400m. These data were added to approximately 11,000 elevation points collected via airboat to create an elevation data set for the Everglades region. Over the 10-year period of data collection, the specific attribute information collected for points increased. Also, some AHF data were collected to replace some elevation values measured using airboats or LIDAR technology. As a result, disparate data files were created. For EDEN development, the entire data set of more than 70 files was first mosaicked into a single Geographic Information System (GIS) data file and processed through a final quality assurance and quality control process. Using GIS, these data were then segregated by Water Conservation Areas, National Park boundaries, and landscape units so that local trends could be isolated, sub-region specific interpolation models could be developed, and realistic breaks in elevation along sub-region boundaries could be created in the final region-wide DEM. The data were further segregated for model development. For example, 15% of the data points in each sub-region were randomly selected and withheld for use in evaluating DEM production techniques for their respective area. Numerous interpolation methods and parameters within interpolation methods were specified using the remaining 85% of AHF data points. Then simulated elevation values where compared with the 15% of points held-back. Cross-validations using all data points within sub-regions were also employed to further evaluate and document model performance. Models were generated with various resolutions for use with satellite image products and to match the spatial sampling of the EDEN grid. Differences in errors produced as a function of model spatial resolution were also documented. Through this evaluation process, krigging was selected as the interpolation technique for initial EDEN DEM development. This method consistently produced the lowest error for the 15% of held-back points and during cross-validations. Changes in error produced as a function of resolution were also lowest with krigging. While errors produced by radial based functions were sometimes comparable to those from krigging for some sub-regions, krigging has the added advantage of providing additional statistical diagnostics like standard error surfaces that identify areas where less confidence in depth estimates is appropriate and more elevation data collection may be necessary. Because we are interested in simulating water depths at the sub-400m resolution, future plans include the development of pseudo-topography using statistical examination of more than 54,000 highly accurate elevation observations on ground elevation as a function of vegetation type.

This work was funded by the USGS Greater Everglades Priority Ecosystem Science and Land Remote Sensing Programs.

Contact Information: John W. Jones, USGS, 521 National Center, Reston, VA, 20192 USA, Phone: 703-648-5543, Fax: 703-648-4165, Email: jwjones@usgs.gov

(This abstract is from the 2006 Greater Everglades Ecosystem Restoration Conference.)

Back to Project Homepage