Improvements to a Methodology for the Derivation of Shallow-water Bathymetry from High Resolution Multispectral Imagery

2010 Ocean Sciences Meeting
February 22-26, 2010, Portland, OR

Special Session MT02: Marine Geodesy and Geographic Information Systems

KR Hogrefe

USGS

Alaska Science Center

Anchorage, AK

khogrefe@usgs.gov

DJ Wright

Dept of Geosciences, Oregon State Univ

Corvallis, OR 97331-5506

dawn@dusk.geo.orst.edu

JE Miller

Pacific Islands Benthic Habitat Mapping Center

Honolulu, HI

T Battista

NOAA National Centers for Coastal Ocean Science

Center for Coastal Monitoring and Assessment

Silver Spring, MD

Since the launch of the first Earth observation satellites, scientists have understood the potential for determining water depth from spectral data based on the knowledge that longer wavelengths of light lose energy more quickly as they pass through water. Several efforts to realize this potential have been made, and each has focused on defining the differential attenuation rates between spectral bands as water depth increases. A recent study analyzed seven published methods at multiple sites across the Pacific and found the Lyzenga (1985) approach to yield the most accurate results. The Lyzenga (1985) methodology was then used to derive nearshore bathymetry to fill the gap between terrestrial DEM data and the shallowest reaches of gridded multibeam sonar data in the creation of a coastal terrain model of Tutuila, American Samoa. A recent error analysis of the derived bathymetry found that the derived depths were statistically accurate to a depth of 15 m and that the bathymetric surface was suitable for terrain modeling.
     Due to these promising results, a project was launched to derive shallow-water bathymetry at twelve sites across the Pacific. As the work progressed, significant improvements to the statistical accuracy and spatial coverage of the product were achieved. Initial advances were due to increased accuracy of the IKONOS imagery realized through enhanced georectification using ground control points collected on site in the study areas. Further improvements resulted from increased familiarity with sea surface glint correction process. The most significant innovation impacting accuracy and spatial coverage was the manipulation of the original multiple linear regression variables, calculated by plotting blue band and green band radiance against depth, used to derive depth. Due to an inversion of the relationship between blue and green band radiance that occurs in less than 3 m of water, and a lack of multibeam sonar data for control in depths less than 5 m deep, data derived using the original variables did not extend into shallow water. The spatial coverage of the derived bathymetry was expanded into this range by either decreasing the value of the original Y-intercept value or simultaneously increasing the slope values of the blue and green bands. However, these changes caused a reduction in the statistical accuracy of the derived product. The solution to this trade-off was to derive multiple products using incrementally adjusted variables, and then integrating them in a final mosaic that prioritized the most accurate data. Thus, spatial coverage was expanded while maintaining, and sometimes improving, the statistical accuracy of the derived bathymetry product. These methodological improvements resulted in data with an increased depth range of up to 25 m and superior statistical accuracy. These derived bathymetry products are currently being used to assess and monitor the health of coral reefs in United States holdings across the Pacific, and tsunami run--up modeling for the island of Tutuila, American Samoa. They have the potential for additional uses such as habitat analysis based on bathymetric position indices or depth stratification for fisheries analysis.

Keywords: derived bathymetry; IKONOS satellite imagery; multibeam sonar; coral reef remote sensing; coastal terrain model

Download PowerPoint file of poster (X Mb)


Return to Dawn's Vita