High
Resolution Satellite Imagery
for Mapping Invasive Aquatic Plants
Introduction
Thousands of reservoirs and lakes are used for
water supplies and recreation. The past several decades
have seen an invasion of aquatic plants that can rapidly
degrade a water body and the ability of people to use the
resource. These plants include Hydrilla, Elodea, Eurasion
water milfoil, water Hyacinth and many others. The
economic impact of invasive aquatic plants on our water
bodies is measured in the hundreds of millions of dollars.
Specialty companies have been established who work
year-round to limit the impact of these invasive aquatic
plants
(Figure 1) on our reservoirs and
lakes. This case history documents the efforts of a
progressive company, Aquatic Nuisance Plant Control, Inc.
(ANPC) of Littleton, North Carolina, who wanted to
determine if high-resolution satellite imagery could
improve the effectiveness of its remediation service to
property owners and agencies. The company acquired new
imagery, had enhanced imagery and maps generated, and
integrated the results with its new GIS and extensive
field database. ANPC found that the focused application of
satellite imagery does improve planning, productivity, and
presentations.
Managing Invasive Aquatic
Plants
ANPC provides remediation programs to federal,
state and local agencies, lake associations and homeowners
associations along the shoreline of Gaston Reservoir on
the Virginia-North Carolina border (Figure 2), to control
the extent of invasive aquatic plants. Biological-,
physical-, and chemical-remediation programs are applied
to areas where invasive plants impede access to the
reservoir, clog waterways, and impact water intake
facilities. Timely and accurate knowledge of the location
and acreage of invasive plants are required to properly
and efficiently carry out the remediation programs.
Prior to implementing remote sensing, ANPC
employees flew in a helicopter, collecting data and
photographs, and noting the extent of aquatic plants on
maps and in a notebook. This method was useful for
providing a timely overview of vegetation in the
reservoir, but the images and photographs were difficult
to integrate with company maps and field operations. In
addition, statistics were difficult to compile from the
helicopter data and ANPC did not have a complete inventory
of invasive plants across the reservoir to show potential
clients and agencies.
The bulk of the company’s database on extent,
type, and depth of invasive aquatic plants has been
collected during boat surveys to 500 sites located around
the reservoir. This point data has been compiled into
detailed tables that characterize the vegetation at each
site. After inventorying the aquatic plants, ANPC works
with the property owners and agencies to physically
remove, chemically treat, or introduce biological
solutions to mitigate the impact of these invasive plants.
These remediation activities are documented in
spreadsheets.
Can Remote Sensing Improve the
Work Process and Productivity?
This past year ANPC implemented GIS technology to
improve display and understanding of its field
observations and remediation programs. However, current
and complete digital maps of the shoreline, docks, and
development are not available. It became apparent that
having an up-to-date image of the reservoir in the GIS
would improve planning for fieldwork and presentations to
property owners and agencies. In addition, appropriate
imagery would enable ANPC to map the spatial extent of
aquatic plants directly into the GIS and to integrate its
GPS-controlled field observations with the new imagery and
maps. The imagery would also enable ANPC to establish a
database for future use and comparison.
Acquiring the Imagery
ANPC contacted Ellis GeoSpatial through its
business partner, Clean Lakes, Inc. of Martinez,
California, to evaluate remote sensing alternatives. The
requirements included:
-
imagery collection in October
2003, coverage of 375 km2,
-
derived maps within two weeks
of acquisition,
-
spatial accuracy equivalent to
USGS 1:24,000 topographic maps,
-
detection of docks,
-
clear delineation of dense
mats and growing vegetation, and
-
some degree of water
penetration.
DigitalGlobe’s multispectral QuickBird satellite
with its 2.6-meter (8-foot) spatial resolution was chosen
as the optimum data capture tool for this application.
AeroMap U.S. (a DigitalGlobe Reseller) refined the
acquisition parameters and prepared DigitalGlobe for the
acquisition of three scenes during September. Approval to
start the acquisition on a priority basis was given on
September 26 and the first image was acquired on October
2.
The new image was delivered via ftp to Ellis
GeoSpatial for processing and mapping. By October 8, ANPC
was receiving enhanced color-infrared and natural color
plots of the first scene for its GIS. Derived maps of
shoreline and floating vegetation began to be delivered to
ANPC on schedule. The final image was acquired on October
16.
Enhanced Imagery
The QuickBird imagery was initially processed to
provide presentation materials of developed areas and the
nearshore. Unfortunately, turbidity in the Gaston
Reservoir was higher than normal due to the massive
run-off associated with Hurricane Isabel in late
September. The turbidity significantly degraded water
penetration with band 1 (blue light—the shortest
wavelength band). Natural color composites of Quickbird
bands 1, 2, 3 as blue, green, red contained minimal
information on submerged vegetation. In addition,
grayscale images of the individual bands had little value
for mapping aquatic vegetation.
Color-infrared composites displayed the onshore
vegetation as shades of bright red, while the offshore
floating vegetation was displayed as shades of gray. To
maximize the visual appearance of the aquatic vegetation
on the imagery, a mask was created over the land area, and
a color-infrared composite made of only the water in the
Gaston Reservoir. This was a highly informative image as
the floating vegetation was displayed as shades of pink
and was clearly differentiated from the water (Figure 3).
Roads, developments, and docks can be seen in the
enhanced imagery. GPS-controlled field sites located along
the shoreline were superimposed on the “ortho-ready”
Quickbird imagery in the GIS. The fit between the field
sites and imagery was good. Large masses of growing and
topped out mixed vegetation were immediately recognized on
the enhanced imagery.
Maps Derived from Imagery
Generating maps that provided location and acreage
of heavily infested invasive vegetation was a primary
objective. The maps were developed with the same map
projection/datum as the imagery—North Carolina State
Plane, NAD-83 (meters)—to enable area (acres and
hectares) to be calculated by the GIS.
In areas where the turbidity and aquatic plant
conditions were appropriate, classification of the imagery
could be based on spectral characteristics of the pixels.
In areas where the signal
from the vegetation was minimal due to water, atmospheric,
or plant conditions, the enhanced imagery was interpreted
using GIS digitizing tools. Field observations made on the
same day as the satellite acquisitions confirmed that the
masses of vegetation were largely Hydrilla mixed with
Elodea and milfoil. Knowledge
about the spatial extent of Hydrilla at a few sites
visited on the day of image acquisition provided the
“ground truth” or training sites for interpreting the
plants on the imagery.
The spectral classification evaluated both
supervised and unsupervised methods. It was found that
unsupervised classification with 15 classes and three
iterations provided good results. The 15 classes were
grouped into three—land, water, and aquatic vegetation.
These three classes were converted to vector polygons (shapefiles).
Statistics were derived from these three classes,
including acreage of Hydrilla (Figure 4).
The heads-up interpretation of the enhanced imagery
used the color infrared images (masked and unmasked). The
500 field sites were displayed in the GIS and the
attributes posted on the screen during interpretation. As
the polygons were being mapped, they were assigned to
specific field sites. This correlation has significant
advantages as acreage of Hydrilla and other vegetation was
immediately assigned to each field site. The new attribute
table being developed with heads-up interpretation was
joined in the GIS to the extensive attributes documented
in the field for each site.
Products for Presentations,
Reports, and Marketing
Color plots and maps are easily generated from the
Quickbird images and derived maps using GIS. Enlargements
to 1:10,000 could be plotted, enabling ANPC to show
property owners and agencies their areas of interest. The
color polygons of vegetation were made transparent and
superimposed over the imagery to facilitate communication
and understanding of the new interpretation and acreage.
The GIS attribute tables derived from the imagery
were exported to Microsoft Excel for easier summation of
acreage and integration into reports. For each of the 500
field sites, acreage of dense mats of growing vegetation
as of October 2003 is now known.
Priorities can now be more efficiently established
using the October 2003 vegetation maps. Marketing of
remediation services can first target those areas with
extensive growth and a large number of impacted property
owners. Impacted docks are now documented with the imagery
and maps, providing a convincing story to property owners
of the need to manage the invasive vegetation (Figure 5).
Agencies responsible for the watershed and reservoir may
find the imagery and maps of interest, opening
opportunities for partnering and cost sharing. Areas that
were remediated in the past by ANPC can be clearly seen as
having little or no invasive vegetation. These successful
examples are proving to be very effective for
presentations and reports.
Managing Remediation
The enhanced imagery and derived maps enable ANPC
to more accurately model the level of effort needed to
clean up an impacted site. The spatial extent and shape of
the invasive vegetation patches improves the company’s
knowledge base for making decisions about the optimum
remediation method to employ. GPS receivers are being
evaluated that would enable the remote sensing-based maps
to be downloaded and deployed on the company’s boats to
directly support fieldwork.
The methods of remediation and level of effort are
being documented in the GIS for the 500 field sites. This
growing database will enable ANPC to easily review and
summarize its efforts for different parts of the
reservoir.
Issues and Future Work
High turbidity limited water penetration by the
shorter wavelength bands (visible blue and green light) at
Lake Gaston during early October 2003. Most of the
invasive aquatic plant biomass is submerged, so having the
ability to detect and map the outlines of submerged plants
is important for ANPC’s remote sensing business model.
Differentiation of invasive species using Quickbird
imagery was not done with this project. The differences
may be too subtle for Quickbird’s limited spectral
depth. For high value assets, airborne hyperspectral
imagery may be required to identify the different aquatic
plants. Correlation of the spatial extent of invasive
plants with bathymetry was also not done, but it seems
reasonable that with further experience such correlation
would improve the accuracy of derived maps.
ANPC has ordered new imagery acquisition in late
2004 to document changes of Hydrilla acreage as well as
other invasive plants across the large reservoir. Areas
under remediation should show significant decrease in
invasive vegetation, while those areas that are being
neglected could show significant increase in invasive
vegetation. Using imagery to document the difference in
invasive plant acreage over time will reveal the
effectiveness of treatments, show how different lake
environments affect treatment success, and provide
excellent marketing material.
About the Authors
James Ellis, Ph.D., is Principal of Ellis
GeoSpatial (www.ellis-geospatial.com).
He derives new GIS maps and databases for environmental
applications from satellite and airborne imagery. He can
be reached at [email protected].
Skip Wiegersma is General Manager of Aquatic
Nuisance Plant Control. He has 24 years of experience with
aquatic plant management and is a member of several
professional societies in the discipline, including the
National Aquatic Plant Management Society. He can be
reached at [email protected].
Acknowledgments
The authors would like to acknowledge Stan Moll and
AeroMap US for designing a very effective Quickbird
acquisition program.
Back
|