South
Carolina Leverages New Aerial Imaging
Technique to Map Oyster Beds
Kevin P. Corbley
Harvesting shellfish is a favorite pastime—and a big
business—in coastal South Carolina. Especially popular
are intertidal oysters, which lay exposed above the water
line at low tide, making them easily accessible to both
commercial shellfishermen and recreational shellfish
harvesters. Last year, this local delicacy pumped more
than $1 million into the state economy through commercial
sales.
Intertidal oysters attach themselves to shellfish reefs
and other hard surfaces in the shallow bays and estuaries
of South Carolina’s extensive coastal zone. Teeming with
marine life, these areas comprise a complex network of
organisms often under stress from pollution, development,
and—in the case of oysters—the threat of
over-harvesting. Managing this valuable resource is the
responsibility of the South Carolina Department of Natural
Resources (DNR).
As geospatial technology became popular in the 1980s, DNR
determined that it could manage shellfish more effectively
by mapping the locations and conditions of intertidal
oyster beds using a geographic information system (GIS).
Management practices were ultimately enhanced through the
GIS project, but the actual mapping effort proved too
expensive and time-consuming to repeat with any frequency.
“Surveying the sites in boats or on foot was the only
viable mapping technique at the time and literally took
years to complete,” said Bob Van Dolah, Oyster Survey
Project Manager and Director of DNR’s Marine Resources
Research Institute in Charleston.
DNR often collaborates on marine resource management
programs with the National Oceanic and Atmospheric
Administration’s (NOAA) Coastal Services Center (CSC) in
Charleston. Both organizations are savvy users of GIS and
remote sensing technologies, routinely acquiring satellite
imagery and aerial photography to map upland, coastal, and
marine habitats. With the exception of an experimental
false-color infrared imaging flight in the 1970s, neither
had attempted to use remote sensing techniques for
intertidal oyster mapping until 2002. DNR and CSC
collaborated on a pilot project that demonstrated the
utility of high-resolution aerial imagery for oyster reef
mapping.
“A primary challenge of this approach is that imagery
must be acquired during a three-hour negative tidal window
that only happens about twice a month,” said Steve Raber,
CSC’s Coastal Remote Sensing Program Manager. “Very
few aerial or satellite imaging platforms have the
flexibility and scalability to exploit that small a window
and cover the entire South Carolina coastal area in a
timely fashion.”
A practical and affordable solution was offered in 2003 by
GeoVantage Inc., a digital aerial imaging firm in
Swampscott, Massachusetts. The company had developed and
built 22 lightweight, high-resolution multispectral
cameras that could be transported inexpensively to the
project site and mounted on leased aircraft. The
company’s unique solution was to put six to 10 planes in
the air during each tidal window and map 1,500 square
miles in less than two flying seasons.
“The pilot project demonstrated that GeoVantage could
exploit the short acquisition window with multiple
aircraft and that its GeoScanner cameras were capable of
capturing the oyster reefs in digital imagery,” said Van
Dolah. “With a potential for application in numerous
other intertidal resource management programs, the South
Carolina project was initiated to collect digital imagery
throughout the coast, with funding provided by NOAA’s
National Marine Fisheries Service (NMFS).”
Exploiting the Window
Oysters are considered “keystone” species because the
reefs they form benefit numerous organisms. Built just
below the surface in shallow marine environments, these
reefs serve as substrates for other shellfish to dwell
upon, hiding places for crabs and fish, and protective
near-shore barriers for fragile wetland vegetation. In
South Carolina, 95 percent of oysters live on intertidal
reefs composed of mud, dead oysters, and discarded shells.
Mapping the entire reef structure from the air is
difficult because turbid tidal waters obscure the
formations when submerged, and the reef is only fully
exposed during extremely low tides, called lunar lows,
which occur five or more days each month. Complicating
matters is a typical reef’s narrow footprint and
proximity to shore, where the oysters can be hidden in the
shadows of marsh grasses much of the day. They can only be
photographed from above during times of high sun angle,
leaving just a few three-hour windows per month when the
sun and moon cooperate.
“These windows are predictable, but a suitable window
can easily be disrupted by high winds pushing tide water
into the estuaries or clouds casting shadows over
acquisition targets,” said Mark Finkbeiner, CSC’s
Benthic Mapping Project Leader. “Acquisition flexibility
is a critical requirement for these aerial photo
missions.”
Another significant requirement was high-resolution,
multispectral imaging. DNR and CSC knew that sub-meter
spatial detail would be needed to adequately map the
extent of the reefs. And just as significantly, a
combination of spectral and spatial data would be
necessary to differentiate healthy live oyster clusters
from the muddy, shell-covered reef substrate. Derived with
advanced image processing techniques at CSC, this
information would be crucial in assessing oyster
populations.
Fortunately, along with GeoVantage’s ability to deploy a
squadron of sensors over South Carolina during the desired
acquisition windows, its GeoScanners were also able to
satisfy the rigorous imaging specifications of the
project. Each GeoScanner consists of four coregistered
sensors operating in the red, green, blue and
near-infrared portions of the spectrum. With a spatial
resolution of 0.25 meters, a single data set allows users
to create a multitude of panchromatic, visible, and
false-color band combinations.
Imaging Oyster Beds
For planning purposes, project participants divided South
Carolina’s coastal region according to USGS digital
ortho quarter-quad (DOQQ) coverage areas. A total of 122
DOQQs would have to be imaged to map all of the state’s
shellfish beds. Because of the limited acquisition
windows, flights would be spread over two years.
In preparation for the 2003 flights, GeoVantage contracted
with aviation rental services at several local airports to
reserve up to 10 Cessna 152, 172, and 182 aircraft on the
pre-determined target dates. A significant advantage of
the GeoVantage camera design is that nearly any high-wing
aircraft can be used because no camera port is needed. The
GeoScanner bolts onto the undercarriage of most small
airplanes with fixed landing gear.
“Another cost-saving benefit for a project like this is
that we can hire local pilots with experience doing aerial
photography,” said Matthew Herring, GeoVantage Vice
President, Sales and Marketing. “It takes just an hour
or two for them to learn our highly automated flight
management system.”
GeoVantage planned the flight lines for each mission in
ESRI’s ArcView software. This process was extremely
complicated because of the number of planes involved and
the delimiting acquisition conditions. Safety and
flexibility were key considerations. With up to 10
aircraft flying over the South Carolina coast
simultaneously at the same 2,500-foot altitude, pilots
were each assigned separate DOQQs on a given day to
minimize any chance of collision.
Clouds and wind both presented unpredictable challenges
that had to be considered. GeoVantage built flexibility
into the acquisition program by planning primary and
alternative missions for each aircraft in one day. Flight
lines for each mission were uploaded as Shape Files from
ArcView into GeoPlan, the firm’s mission management
software.
Running on a laptop computer with a touch screen in the
cockpit, GeoPlan linked with an onboard GPS and inertial
measurement unit (IMU), developed by GeoVantage using
military technology, to display the exact location of the
aircraft and its intended flight lines. The software
precisely guided the pilot onto the correct heading and
triggered the camera to collect images at pre-assigned
points. If the aircraft deviated from the flight line, the
software instantly notified the pilot to fly it again.
When bad weather moved into an area, the pilot simply
clicked to an alternative mission plan and headed for that
target area. The software kept track of collected vectors,
even if they were partial flight lines. For instance,
pilots sometimes had to manually turn the cameras off in
the middle of a flight line due to encroaching clouds.
Rather than lose the work completed, the software saved it
and guided the pilot back later to pick up the flight line
where it had been interrupted.
“The pilots are very good at predicting cloud movements
and getting shadow-free imagery before the clouds moved
into their areas,” said Herring.
With up to 10 planes in the air at once and each one
guaranteed to collect useable data regardless of weather
due to the multiple mission plans, GeoVantage collected
data over 100 DOQQs in 2003, about 70 of which were
covered completely. After each day’s flights, digital
files were downloaded from the onboard systems and sent to
GeoVantage’s Swampscott, Massachusetts headquarters for
processing.
Customized processing software for the GeoScanner data
uses automated algorithms to mosaic and color-balance the
flight lines and orthorectify the imagery into a four-band
product ready for GIS analysis and enhancement. At CSC’s
request, the South Carolina data sets are being delivered
in the IMG format compatible with ERDAS IMAGINE software.
Analyzing the Imagery
With the 2004 acquisitions set to begin in May, DNR and
CSC spent much of the winter focusing on developing and
refining image analysis techniques to extract information
from the data quickly and accurately. The CSC remote
sensing facility in Charleston operates primarily in an
ESRI and ERDAS (now Leica Geosystems) environment and is
assessing the utility of automated and semi-automated
feature extraction tools for enhancement and analysis of
satellite and aerial imagery.
Geospatial technicians found that oyster beds were visible
to the naked eye in the processed imagery, and the extent
of the reefs could be readily measured and mapped using
automated routines in Feature Analyst software.
Identifying clusters of live oysters and assessing their
condition, however, is a more complex process.
“These reefs are muddy mixtures of oysters, dead shells,
and other materials, and the color of the oysters is very
similar to the mud,” said Bill Stevenson, a CSC Image
Analyst. “The challenging task is to classify each
quarter-meter pixel into one of four categories—mud,
high-relief oysters, low-relief oysters, or dead shells.
Once this is accomplished, each reef can be characterized
on a continuum from a robust, healthy reef to a remnant
unproductive reef.”
The high-relief class is the most important because it
contains live oysters. In general, healthy, live
intertidal oysters sit upright in clusters with the edges
of their shells sticking into the air. Oysters that are
lying down in low-relief may be dead or healthy single
oysters. Dead shells have been emptied of their content by
predators or left on the reef by harvesters.
“The key to classification is interpreting the texture
of the reefs. The vertical, high-relief clusters have a
different texture from the other feature classes,” said
CSC’s Finkbeiner. “We are using automated algorithms
in the software that combine spatial and multispectral
data to differentiate classes and characterize oyster
reefs by texture.”
The process is still being refined, and technicians can
currently perform about 70 percent of the individual
oyster reef delineation work using these automated
routines. The remainder of the image interpretation must
be carried out by human eyes, but CSC believes that a
greater percentage of classification will eventually be
accomplished with a high level of accuracy by automated
techniques.
“Automated classification will never be perfect because
every reef will have slightly different characteristics
due to sun angle or mud composition,” said Finkbeiner.
The production-level oyster mapping work is planned to be
carried out by private industry, where the methodologies
developed in this project and subsequent investigations
will be available for use.
Applying Information
Once completed, South Carolina DNR will input the oyster
bed maps into their shellfish management GIS to update and
possibly compare with the 1980 data. DNR will use the new
information to calculate commercial harvesting lease
acreages and look for signs of over-harvesting and reef
building. The GIS maps will be accessible to recreational
users on the Web and shared with the state agency that
decides where new docks can be built. Most importantly,
DNR will examine the imagery for oyster population
changes.
The funding agency, NOAA’s NMFS, sees enormous potential
in the new mapping solution. Its Restoration Center in
Silver Spring, Maryland, hopes to one day incorporate the
oyster maps into a centralized Web-accessible GIS that
will include details on thousands of marine restoration
projects across the country. Other coastal states, such as
North Carolina, Georgia, and Florida, are carefully
watching the South Carolina project in hopes it will
eventually help them better manage their intertidal
oysters.
In terms of cross-over application, DNR already is
considering using the new acquisition technique to map
marsh grasses, hammock islands and other fragile resources
in the intertidal zone.
“This is a new application of remote sensing, and the
fact that GeoVantage could put 10 cameras in the air for
this sort of mission was a decisive factor in making the
application a reality,” said Finkbeiner.
For more information, visit www. geovantage.com.
About the Author
Kevin Corbley is the principal in Corbley
Communications Inc. (CCI) of Fort Collins, Colorado. He
may be reached at [email protected].
Acknowledgment
The author would like to acknowledge Kurt Allen and Photo
Science Inc. for their efforts on this project. Photo
Science is the prime contractor for the SCDNR Oyster Bed
Mapping Project. Contact Photo Science in Greenbelt,
Maryland at (301) 345-4488.
Back
|