Beruflich Dokumente
Kultur Dokumente
www.elsevier.com/locate/rse
Preface
Preface
References
Andrefouet, S., Kramer, P., Torres-Pulliza, D., Joyce, K. E., Hochberg,
E. J., Garza-Perez, R., Mumby, P. J., Riegl, B., Yamano, H., White,
W. H., Zubia, M., Brock, J. C., Phinn, S. R., Naseer, A., Hatcher, B.
G., & Muller-Karger, F. E. (2003). Multi-sites evaluation of IKONOS
data for classification of tropical coral reef environments. Remote
Sensing of Environment, 88, 127 142. (this issue).
Birk, R. J., Stanley, T., Snyder, G. I., Hennig, T. A., Fladeland, M. M., &
Policelli, F. (2003). Government programs for research and operational
uses of commercial remote sensing data. Remote Sensing of Environment, 88, 3 16. (this issue).
Dial, G., Bowen, H., Gerlach, F., Grodecki, J., & Oleszczuk, R. (2003).
IIKONOS satellite imagery, and products. Remote Sensing of Environment, 88, 23 36. (this issue).
Goetz, S. J., Wright, R. K., Smith, A. J., Zinecker, E., & Schaub, E. (2003).
IKONOS imagery for resource management: Tree cover, impervious
surfaces and riparian buffer analyses in the Mid-Atlantic region. Remote
Sensing of Environment, 88, 194 207. (this issue).
Goward, S. N., Davis, P. E., Fleming, D., Miller, L., & Townshend, J. R.
(2003). Empirical comparison of Landsat 7 and IKONOS multispectral
measurements for selected Earth Observation System (EOS) validation
sites. Remote Sensing of Environment, 88, 79 98. (this issue).
Goward, S. N., Townshend, J. R. G., Zanoni, V., Pollicelli, F., Stanley, T.,
Ryan, R., Holekamp, K., Underwood, L., Pagnutti, M., & Fletcher, R.
(2003). Acquisition of Earth science remote sensing observations from
commercial sources: Lessons learned from the Space Imaging IKONOS
example. Remote Sensing of Environment, 88, 200 218. (this issue).
Helder, D., Coan, M., Patrick, K., & Gaska, P. (2003). IKONOS geometric characterization. Remote Sensing of Environment, 88, 68 78.
(this issue).
Hurtt, G., Xiao, X., Keller, M., Palace, M., Asner, G. P., Braswell, R.,
Brondzio, E. S., Cardoso, M., Carvalho, C. J. R., Fearon, M. G., Guild,
L., Hagen, S., Hetrick, S., Moore III, B., Nobre, C., Read, J. M., Sa, T.,
Schloss, A., Vourlitis, G., & Wickel, A. J. (2003). IKONOS imagery for
the large scale biosphere atmosphere experiment in Amazonia (LBA).
Remote Sensing of Environment, 88, 110 126. (this issue).
Masuoka, P. M., Claborn, D. M., Andre, R. G., Nigro, J., Gordon, S. W.,
Klein, T. A., & Kim, H. (2003). Use of IKONOS and Landsat for
malaria control in the Republic of Korea. Remote Sensing of Environment, 88, 186 193. (this issue).
Morisette, J. T., Nickeson, J. E., Davis, P., Wang, Y., Tian, Y., Woodcock,
C. E., Shabanov, N., Hansen, M., Cohen, W. B., Oetter, D. R., &
Kennedy, R. E. (2003). High spatial resolution satellite observations
for validation of MODIS land products: IKONOS observations acquired
under the NASA scientific data purchase. Remote Sensing of Environment, 88, 99 109. (this issue).
Pagnutti, M., Ryan, R. E., Kelly, M., Holekamp, K., Zanoni, V., Thome, K.,
& Schiller, S. (2003). Radiometric characterization of IKONOS multispectral imagery. Remote Sensing of Environment, 88, 52 67. (this
issue).
Ryan, R., Baldridge, B., Schowengerdt, R. A., Choi, T., Helder, D. L., &
Blonski, S. (2003). IKONOS spatial resolution and image interpretability characterization. Remote Sensing of Environment, 88, 37 51. (this
issue).
Sawaya, K., Olmanson, L., Heinert, N., Brezonik, P., & Bauer, M. (2003).
Extending satellite remote sensing to local scales: Land and water resource monitoring using high-resolution imagery. Remote Sensing of
Environment, 88, 143 155. (this issue).
Seelan, S. K., Laguette, S., Casady, G. M., & Seielstad, G. A. (2003).
Remote sensing applications for precision agriculture: A learning community approach. Remote Sensing of Environment, 88, 156 168. (this
issue).
Small, C. (2003). High spatial resolution spectral mixture analysis of urban
reflectance. Remote Sensing of Environment, 88, 169 185. (this issue).
Zanoni, V., Stanley, T., Ryan, R., Pagnutti, M., Baldridge, B., Roylance, S.,
Snyder, G., & Lee, G. (2003). The Joint Agency Commercial Imagery
Evaluation (JACIE) team: Overview and IKONOS joint characterization approach. Remote Sensing of Environment, 88, 17 22. (this issue).
Vicki M. Zanoni
NASA Earth Science Applications Directorate,
Mail Code MA20, Stennis Space Center, MS 39529, USA
E-mail address: Vicki.M.Zanoni@nasa.gov
Samuel N. Goward
Department of Geography, University of Maryland,
2181 LeFrak Hall,
College Park, MD 20742, USA
E-mail address: sgoward@umd.edu
NASA Office of Earth Science, 300 E Street SW, Washington, DC 20546, USA
NASA Earth Science Applications Directorate, Stennis Space Center, MS USA
c
USGS National Center, Reston, VA, USA
d
National Imagery and Mapping Agency, Bethesda, MD, USA
e
NASA Ames Research Center, Moffett Field, CA, USA
Received 14 February 2003; received in revised form 6 June 2003; accepted 30 July 2003
Abstract
The private sector is delivering products and services derived from an expanding array of airborne and spaceborne remote sensing
systems. The successful commercial launches of the IKONOS, QuickBird, and OrbView-3 satellites in 1999, 2001, and 2003,
respectively, combined with commercial airborne sensors such as the Positive Systems ADAR 5500 (multispectral), the Intermap
STAR-3i (interferometric synthetic aperture radar), and TerraPoint, LLC, LIDAR System have ushered in an era of expanded capability
and capacity for the field of remote sensing of our Earth. Remote sensing data from commercial sensors offer the public and private
geospatial information communities important new sources of timely and accurate spatial information that can augment data provided
by public-sector remote sensing systems. Several Federal agencies, including the National Aeronautics and Space Administration
(NASA), the U.S. Geological Survey, and the National Imagery and Mapping Agency (NIMA), have established data purchase
programs and related activities to access, evaluate, and assimilate new commercial remote sensing products to serve research and
operational requirements. Plans for future commercial systems and data products indicate an expanding set of data types using
hyperspectral, radar, LIDAR, and microwave technologies. The availability of new data sources has established the basis for Federal
programs to provide for systematic characterization of the products, consistent with the characterization of data products enabled by
traditional sources that include Landsat, SPOT, and the Advanced Very High Resolution Radiometer (AVHRR). An overview of
commercial remote sensing initiatives within the National Aeronautics and Space Administration, the U.S. Department of the Interior,
and the U.S. Department of Defense (DoD), and of their Joint Agency Commercial Imagery Evaluation (JACIE) team, illustrates these
points, highlights lessons learned from these activities, and outlines recommendations for the future.
D 2003 Elsevier Inc. All rights reserved.
Keywords: Earth science; Commercial remote sensing; Multispectral; Imagery; Digital elevation model; Mapping; Monitoring; IKONOS; QuickBird;
OrbView-3; LIDAR; Radar; Hyperspectral
1. Introduction
The United States Government has significant responsibilities in providing mapping and monitoring information to
meet the needs of its citizens. Traditionally, Federal agencies, including the National Aeronautics and Space Admin-
Table 1
Government data purchase program comparisons
Characteristics
Scientific Data
Purchase (Part I)
Scientific Data
Purchase (Part II)
Commercial imagery
data purchase
SeaWiFS
Sponsor
Purpose
NASA
Determine utility of
commercial data for
NASA science and
applications research
NIMA
Acquire high resolution
imagery to support the
warfighter
NASA
Acquire ocean color
imagery for science
research
Funding
Time frame
Centralized tasking/
distribution hub
Payment on data delivery
Up-front cash payment
Centralized verification
and validation effort
Contract mechanism
US$50 million
1997 2003
Yes
NASA
Examine utility of
additional commercial
datasets for NASA
science and
applications research
US$20 million
2000 2003
Yes
US$30 million
2000 2002
Yes
US$43 million
1997 2003
Yes
Yes
No
Yes
Yes
No
Yes
Yes
No
Yes
Yes (20%)
Yes (80%)
Yes
Indefinite delivery,
indefinite quantity/
fixed price
Yes
Small
Indefinite delivery,
indefinite quantity/
fixed price
Yes
Small
Basic ordering
agreement
(fixed price)
Yes
None
Fixed price
No
None
Yes
Yes
Yes
Yes
Yes
Yes
No
Yes
Tasking panel
Deviation from
commercial
specifications
Multiple data sources
Multiple products
Federal government programs and activities that currently acquire commercial data include the USGS The
National Map,1 the Federal Emergency Management Agency (FEMA) Flood Map Modernization Program, the NOAA
Shoreline Mapping program, the U.S. Department of Agriculture Foreign Agriculture Service programs, the National Imagery and Mapping Agency (NIMA) Commercial
Imagery Program (CIP), the NASA Solid Earth and Natural
Hazards program, and the U.S. Global Climate Research
Program (USGCRP). The implementation of U.S. government policies to procure remote sensing data from the
private sector are reflected in specific agency programs
that assess the use of commercial data products to meet
agency mission requirements. NASA, the USGS, and
NIMA have taken different approaches in implementing
commercial data purchase programs and related activities.
Table 1 lists data purchase programs representative of these
agencies and compares several important program features.
Following are descriptions of the different types of data
products available through these programs, recent research
and operational applications, and discussion on current and
future requirements for commercial geospatial data. Lessons learned in the performance of the commercial data
purchase programs to date are also highlighted, along with
remaining challenges and recommended approaches related
to future commercial data purchase opportunities for Federal agencies.
http://www.nationalmap.usgs.gov.
http://seawifs.gsfc.nasa.gov/SEAWIFS.html.
Fig. 1. ESE science questions about how the Earth system functions.
3
Detailed descriptions of each of the products are accessible on the
ESA Directorate Web page at http://www.esad.ssc.nasa.gov/datapurchase/.
Fig. 3. NASAs Scientific Data Purchase project data use statistics on new acquisitions.
Table 2
Four primary ESE science theme areas in 1997
Theme area and purpose
Land-cover and land-use change research
Qualify the past, current and future land cover and land use patterns at
regional and global scales
Understand natural and human-induced influences that lead to changes in
land cover, land use, shorelines, or terrestrial and marine ecosystems
Support the scientific information requirements for national environmental
monitoring and research
Support design of a prototype environmental report on trends in, and the
status of, the U.S. environment
Improve methods for the sustainable management of farmlands, forests,
rangelands, and coastal marine environments
Seasonal-to-interannual climate variability and prediction
Measure globally distributed atmospheric wind profiles
Provide atmospheric soundings from Global Positioning System (GPS)
satellites for weather and climate prediction
Natural hazards research and applications
Improve methods and understanding of how best to characterize and
mitigate the consequences of natural hazards for both managed and
natural ecosystems
Long-term climate: natural variability and change research
Test the utility of new measurements to meet the continuity requirements of
the EOS science program
Table 3
Overview of NASA Scientific Data Purchase Part I products
Data provider and
image product
Data provider: EarthSat
Circa 1975
orthorectified scene
Circa 1990
orthorectified scene
Circa 1990
orthorectified mosaic
Data type
Pixel size
Positional accuracy
Encoding
Radiometric
accuracy
Cloud
%
Satellite, Landsat 1 3,
MSS
Satellite, Landsat 4
and 5, TM
Satellite, Landsat 4
and 5, TM
MSS (4-band)
57 m
F 100 m RMSE
8 bits
NA
< 20%
TM (7-band)
28.5 m
F 50 m RMSE
8 bits
NA
< 20%
TM (3-band)
28.5 m
F 50 m RMSE
8 bits
NA
< 20%
X-SAR
2.5 m
8 bits
NA
NA
DEM
5 10 m
F 2.5 m RMSE
RMSE (1j)
F 2.5 m horizontal
F 1 3 m vertical
RMSE (1j)
32 bits
NA
NA
MS1
0.7 m
8 bits
F 10% absolute,
F 5% relative
< 10%
MS1
0.7 m
Relief dependent,
F 12.2 m (benign),
F 50 m (extreme)
8 bits
NA
< 10%
Pan
1m
F 250 m (std.),
F 3 m (precision)
(CE 90%)
F 250 m (std.),
F 5 m (precision)
(CE 90%)
F 12.2 m (std.),
F 2 m (precision)
(CE 90%)
F 12.2 m (std.),
F 5 m (precision)
(CE 90%)
F 25.4 m horizontal,
7 m vertical (CE 90%)
11 bits
F 10% absolute,
F 5% relative
< 10%
11 bits
F 10% absolute,
F 5% relative
< 10%
11 bits
F 10% absolute,
F 5% relative
< 10%
11 bits
F 10% absolute,
F 5% relative
< 10%
11 bits
NA
< 10%
Satellite, IKONOS-2
MSa
4m
Master Pan
(orthorectified)
Satellite, IKONOS-2
Pan
1m
Master MS
(orthorectified)
Satellite, IKONOS-2
MSa
4m
Satellite, IKONOS-2
Pan/MSa, DEM
1 4 m, 15 m
MS = Multispectral, Pan = panchromatic, RMSE = root mean square error, CE = circular error.
a
Multispectral data emulating first four TM bands.
curement and evaluation of additional sources of commercial remote sensing data. DigitalGlobes QuickBird
high-resolution optical imagery is being evaluated as a
source of information for Earth science research and
applications. NASA contracted with EarthSat to provide
a global database of orthorectified scenes from Landsat
7 and mosaics made from these scenes. These scenes
were orthorectified using the TM dataset procured in the
NASA SDP Part I as a baseline to provide for easy
change detection between the periods. NASA procured
commercial LIDAR data from TerraPoint; unmanned
aerial vehicle based multispectral data from AeroVironment; and additional Intermap Technologies radar data
and Positive Systems multispectral data for a variety of
10
Table 4
Overview of NASA Scientific Data Purchase Part II products
Image data product
Data type
Pixel size
Positional accuracy
Encoding
Radiometric
accuracy
Cloud
%
Satellite, Landsat 7,
ETM +
Pan (1-band)
MS (8-band)
NA
< 20%
Satellite, Landsat 7,
ETM +
PS MS (3-band)
8 bits
NA
< 20%
Satellite, Landsat 7,
ETM +
PS MS (3-band)
14.25 m
F 75 m absolute
F 40 m relative
1990 RMSE
F 75 m absolute
F 40 m relative
1990 RMSE
F 75 m absolute
F 40 m relative
1990 RMSE
8 bits
14.25 m
28.5 m
57 m
14.25 m
8 bits
NA
< 20%
MSa
0.7 m
8 bits
< 10%
MSa
0.7 m
F 12.2 m (benign)
F 50 m
(extreme relief)
RMSE
8 bits
F 10% absolute
F 5% relative
NA
< 10%
X-SAR
2.5 m
8 bits
NA
NA
DEM
5-10 m
F 2.5 m horizontal
RMSE (1j)
F 2.5 m horizontal
F 1 3 m vertical
RMSE (1j)
32 bits
NA
NA
Pan
MSa
Pan
MSa
0.62 0.82
2.48 3.28
0.62 0.82
2.48 3.28
NA
11 bits
< 20%
F 23 m (CE 90%)
11 bits
F 10% absolute
F 5% relative
F 10% absolute
F 5% relative
Pan
MSa
0.62 0.82 m
2.48 3.28 m
11 bits
F 10% absolute
F 5% relative
< 20%
11 bits
NA
< 20%
NA
NA
NA
NA
Satellite, QuickBird
Satellite, QuickBird
Satellite, QuickBird
m
m
m
m
< 20%
Satellite, QuickBird
MS PS
0.62 0.82 m
Airborne, LIDAR
DTM
1.83 m
Airborne, LIDAR
DSM
1.83 m
MS
1m
NA
8 bits
NA
0 10%
MS
0.5 m
NA
8 bits
NA
0 10%
Vertical
profiles
Horizontal
resolutionb
200 km
NA
15
Estimated
retrieval errorsc
NA
11
The USGS needs to understand the performance characteristics of the sensors and geospatial data used to
support its mapping and science missions. For many
years, the USGS has calibrated the Nations analog aerial
mapping cameras. With the advent of a new generation
of digital cameras and high-resolution commercial satellite systems, the USGS is establishing a digital camera
laboratory and is upgrading field-based in-situ test ranges
to allow for the characterization of new products. In
addition, the USGS is contracting with academic institutions and is partnering with NASA to develop new
digital calibration methods and capabilities. For purposes
of assessing the geometric properties of remotely sensed
data, the USGS maintains and utilizes surveyed field test
sites located at or near its regional mapping centers and
has characterized optical, radar, and LIDAR technologies.
The test sites possess a significant number of accurate
ground control points that can be compared to locations
in satellite imagery or to elevation data to substantiate
vendor product specifications. The USGS utilizes partner
sites for testing radiometric properties of remote sensing
data and participates with NASA and NIMA in the Joint
Agency Commercial Imagery Evaluation team.
Notes to Table 4:
MS = Multispectral, Pan = panchromatic, PS MS = pan-sharpened multispectral, CE = circular error, RMSE = root mean square error.
a
Multispectral data emulating first four TM bands.
b
Vertical resolution: O3 = 1 km (10 50 km), NO2 = 1.5 2.5 km (20 40 km), H2O = 1 2 km (10 40 km), aerosols = 1 1.5 km (10 30 km).
c
O3 = 3 5% (10 60 km), NO2 = 5 10% (20 45 km), H2O = 5 8% (10 50 km), aerosols = 10 20% (10 30 km).
12
http://www.mrlc.gov.
A subset of The National Map layers corresponds to the National
Spatial Data Infrastructure primary data layers of orthoimagery, elevation,
hydrology, and transportation.
5
acquiring IKONOS imagery from Space Imaging, QuickBird imagery from DigitalGlobe, and a variety of other
spaceborne and airborne data types in response to diverse
customer requirements.
Since the events of 9/11, emphasis on the use of
commercial imagery has been growing for a variety of
reasons, not the least of which is the need to have the
flexibility of sharing imagery and imagery-derived products
with a diverse group of customers ranging from foreign
coalition partners to domestic first responders. This attribute
of commercial imagery was recognized from the beginning
of NIMAs involvement in the Commercial Imagery Program, but it has become much more important after 9/11 and
the birth of the Homeland Security mission.
Not only has there been increasing demand for unclassified imagery and its derived products, there has also been a
demand for improved spatial and temporal resolution, causing NIMA to expand its use of airborne data, such as
imagery with resolution of less than 1 m and LIDAR data
to capture very high levels of elevation detail over discrete
areas.
Concurrent with these increasing demands has been a
significant increase in funding that will facilitate increased
use and will address a more robust system engineering of
appropriate capabilities across the TPED cycle. In the same
time period, the number of operational spaceborne systems
has grown from one to two; a third system is planned to be
deployed and operational during calendar year 2003.
These factors have all come together to drive a maturing
of the industry and NIMAs usage of commercial remote
sensing data. NIMAs Commercial Imagery Program is
reflecting this maturity by moving each of the functions
gathered under a general program office umbrella out into a
business-as-usual operational scenario.
4.3. NIMA applications of commercial imagery
It is safe to say that commercial imagery is finding usage
in an expanding number of traditional imagery, imagery
intelligence, and geospatial intelligence applications. It is
also true that NIMA has been given clear guidance by those
responsible for the agencys oversight to expand the use of
commercial imagery.
Without going into specific details of where or why
commercial imagery has been collected in support of
Operation Enduring Freedom and the Global War on Terrorism, it is impressive to know that since 9/11, NIMA and
the commercial imagery industry have worked together to
achieve the following:
13
14
6. Lessons learned
The following summarizes findings and lessons learned
by the three agencies in the course of developing and
implementing programs to procure and use geospatial data
products delivered from commercial sources of remote
sensing systems.
an effective solution to the challenge of diverse requirements for verification and validation. The JACIE
approach has been so successful that companies have
proposed using the university teams participating in the
JACIE to support their in-house performance characterizations. The JACIE team recognizes that if university
teams participated in in-house calibrations for companies, that it would severely impact the governments
ability to perform independent assessments of the
companies data. It is apparent that more teams and
methods for calibration and validation are needed.
The commercial sector plans to expand the availability of
data sources and types of remote sensing data products
with the potential to serve even broader needs of Earth
science and mapping communities within the U.S.
Government. The successful launches of Space Imagings IKONOS, DigitalGlobes QuickBird 2, and OrbImages OrbView-3 are providing increasing capacity for
the delivery of high-spatial resolution optical imagery
and derived products.
New licenses issued by the Department of Commerce
authorize on-orbit commercial systems to provide data
with spatial resolutions of 0.5 m. High-spatial resolution
products correlate to higher market shares, which may
correlate to increased interest by remote sensing
companies to provide higher spatial resolution datasets
on future systems.
Remote sensing companies will continue to be a primary
source of high-spatial resolution products available to the
Earth science and mapping communities supported by
federally funded acquisitions and/or by land remote
sensing programs, given current policies restricting the
U.S. Government from competing with the private sector.
The majority of the datasets offered by the commercial
sector have been in the form of terrestrial imagery
products and digital elevation model products. Government policies to populate the National Spatial Data
Infrastructure with the primary framework data layers
and other national priority applications projects continue
to increase demand for these types of commercial
geospatial data products.
Advancements in remote sensing technologies and
systems indicate increased capacity in currently deployed
approaches and project a trend for additional types of
data products from both airborne and spaceborne
platforms.
7. Recommendations
Specific recommendations associated with U.S. Government agencies regarding the future of commercial data
purchases to serve national purposes include the following:
To facilitate widespread use and acceptance of commercial sources of remotely sensed data in the science and
operational communities, government-funded data purchases need to ensure adequate data rights for all
anticipated distribution and sharing.
The continuity of measurements must be considered
when evaluating the use of commercial data for Earth
science, monitoring, and mapping applications. Where
data is deemed necessary for public research or
operations, reliance on commercial assets comes with
the risk that the data stream may not be sustained or
remain consistent. Government policies must ensure the
compatibility of data from different sources through
rigorous calibration and validation activities.
Increase collaboration between and among Federal
agencies to promote data product standards to the
commercial remote sensing industry, facilitating economies of scale.
Maintain coordinated joint activities, such as the JACIE
team, to provide independent performance assessments
that combine the interests of several agencies while
providing a single interface to the commercial remote
sensing industry. Benefits include improved communications with industry, greater government and consumer
confidence in commercial products, and improved
techniques within industry. As the number of sources
increases for any given product type, the importance of
calibration and characterization against a baseline
performance standard becomes a critical element in
establishing the capacity of multiple sources to serve
specific applications on an interchangeable basis.
Increase cooperation with state and local governments in
the area of validating new datasets and sensors by sharing
infrastructures, experiences in methods development, and
empirical knowledge from real-world applications.
Create a national civil remote sensing strategy that
establishes a framework for government requirements
15
References
Andrefouet, S., Muller-Karger, F., Palandro, D., Hu, C., Carder, K., Hochberg, E., & Maeder, J. (2001). High resolution IKONOS data for coral
reefs studies. Proceedings of the 2001 High Spatial Resolution Commercial Imagery Workshop, March 19 21, Greenbelt, MD, USA.
(Sponsored by NASA/NIMA/USGS Joint Agency Commercial Imagery
Evaluation Team, CD-ROM).
Ashton, M. S., Thenkabail, P. S., Enclona, E., Hall, J., Stucky, N., Harris,
D., Van Der Meer, B., Lin, T., & Griscam, B. (2002). Characterization
of Humid-Forest and Savanna Ecoregions of West and Central Africa
Using Satellite Sensor Data of Three Eras: Characterization of Eco
Regions in Africa (CERA) Annual Report for the Period February 28,
2001 to February 18, 2002 (p. 19). New Haven, CT: Center for Earth
Observation and School of Forestry and Environmental Studies, Yale
University.
Capolsini, P., Andrefouet, S., Rion, C., & Payri, C. (2003). A comparison
of Landsat ETM+, SPOT HRV, Ikonos, ASTER, and airborne MASTER data for coral reef habitat mapping in South Pacific islands. Canadian Journal of Remote Sensing, 29(2), 187 200.
Chang, J., Clay, D. E., Dalsted, K., & ONeill, M. (2002). The influence of
remote sensing data on predicting corn yield. Pecora 15/Land Satellite
Information IV Conference in Conjunction with International Society
for Photogrammetry and Remote Sensing (ISPRS) Commission I Midterm Symposium, November 10 15, Denver, CO, USA.
Clinton, W. J. (1994). Coordinating geographic data acquisition and access:
The national spatial data infrastructure. Executive Order 12906, The
White House, April 11. Federal Register, 59(71), 17671 17674.
Available at: http://www.fgdc.gov/publications/documents/geninfo/
execord.html (accessed July 18, 2003).
Driscoll, S., & Lawrence, R. L. (2002). Effects of color balancing of airborne multispectral imagery on invasive plant mapping. 2002 ASPRS/
ACSM Annual Conference, April 22 26, Washington, DC.
Federal Geographic Data Committee (2002). Report to the administration:
Value of civil imagery and remote sensing. FGDC Civil Imagery and
Remote Sensing Task Force. http://www.fgdc.gov/cirs/cirs.pdf.
Gieck, R. E., Kane, D. L., Hinzman, L. D., Overduin, P. P., McNamara,
16
Received 15 July 2002; received in revised form 17 April 2003; accepted 9 July 2003
Abstract
The Joint Agency Commercial Imagery Evaluation (JACIE) team was formed to leverage government agencies capabilities for the
characterization of commercial remote sensing data. The team is composed of staff from the National Aeronautics and Space Administration
(NASA), the National Imagery and Mapping Agency (NIMA), and the U.S. Geological Survey (USGS). Each JACIE agency has a vested
interest in the purchase and use of commercial imagery to support its research and applications. It is critical that this imagery be assessed for
its accuracy and utility. Through JACIE, NASA, NIMA, and USGS jointly characterized image products from Space Imagings IKONOS
satellite. The JACIE team acquired IKONOS imagery of several study sites to perform the assessments. Each JACIE agency performed an
aspect of the characterization according to its area of expertise. NASA and its university partners performed a system characterization
focusing on radiometric calibration, geopositional accuracy, and spatial resolution assessment; NIMA performed image interpretability and
feature extraction evaluations; and the USGS assessed the geopositional accuracy of several IKONOS products. The results of the JACIE
teams IKONOS effort, which were discussed with Space Imaging and then presented at an industry government workshop, ensured and
improved overall product quality and benefited both the commercial industry and the government. Additional JACIE activities include the
characterization of other commercial products such as those from the DigitalGlobe QuickBird and the Orbital Imaging (Orbimage) OrbView3 satellites.
D 2003 Published by Elsevier Inc.
Keywords: Joint Agency Commercial Imagery Evaluation team; IKONOS; Remote sensing
1. Introduction
The purchase of commercial remote sensing data by U.S.
federal agencies is a relatively new way of doing business
that has been and is continuing to be implemented within
government agencies. Commercial data has been purchased
to support the fields of Earth science research, civil government applications, and defense-related intelligence gathering. Strategic partnerships that capitalize on common
commercial data purchase goals can be beneficial. Both
the National Aeronautics and Space Administration (NASA)
2. Background
* Corresponding author. Tel.: +1-228-688-2305; fax: +1-228-6887455.
E-mail address: vicki.m.zanoni@nasa.gov (V. Zanoni).
0034-4257/$ - see front matter D 2003 Published by Elsevier Inc.
doi:10.1016/j.rse.2003.07.005
18
Spatial
X
X
X
X
Geometric
X
X
X
Radiometric
Applications
X
X
X
X
X
X
X
19
20
driven by NIMA/DoD/Intelligence requirements and standards. Commercial imagery can best be used to answer an
intelligence problem if the imagerys quality and utility are
known and trusted. NIMAs quantification of imagery
interpretability is typically expressed through a National
Imagery Interpretability Rating Scale, or NIIRS (Leachtenauer, Malila, Irvine, Colburn, & Salvagio, 1997), a form
of assessment not normally performed by either the civil
community or the vendor. Similarly, commercial imagery
intended for the production of maps must meet NIMA
accuracy standards, requiring an independent verification
of accuracy. Thus, it is critical that commercial data be
independently characterized to ensure that defense and
intelligence decision makers use accurate and reliable
information.
The USGS also has a vital interest in understanding the
qualities and characteristics of the imagery that it applies
in its scientific investigations and in the mapping products
that it provides to the Nation. Furthermore, many Federal
and State government agencies rely on the USGS to
assess commercial products for their mapping applications. All of these uses of IKONOS data strongly justify
the need for an independent commercial data characterization effort.
21
9. Summary
The NASA, NIMA, and USGS JACIE team was established in 2000 and formalized in June 2002 with a goal of
performing independent characterization of commercial data
used by the three agencies. The team has collaborated to
characterize the 1-m panchromatic and 4-m multispectral
image products available from the Space Imaging, IKONOS
satellite. Characterizations focused on radiometric accuracy,
spatial resolution, and geopositional accuracy assessments.
The JACIE effort and collaborative interactions with Space
Imaging have resulted in several improvements to IKONOS
data quality. Both Space Imaging and the government have
benefited from product improvements and from the trust that
has evolved from this government industry collaboration.
NASA, NIMA, and USGS data users have performed a
variety of research and applications projects using IKONOS
data. JACIE characterization and user research results were
presented at JACIE-sponsored workshops in 2001, 2002,
and 2003. JACIE team activities also include characterization of DigitalGlobes QuickBird and Orbimages OrbView3 high-resolution panchromatic and multispectral data.
Acknowledgements
The authors would like to thank Marcia Wise for her
support in the editing of this manuscript.
References
Ager, T., & Bresnahan, P. (2002). IKONOS stereo imagery accuracy assessment. Proceedings of the 2002 High Spatial Resolution Commercial
Imagery Workshop, March 25 27, Reston, VA, USA, sponsored by
NASA/NIMA/USGS Joint Agency Commercial Imagery Evaluation
Team, CD-ROM.
Andrefouet, S., Kramer, P., Torres-Pulliza, D., Joyce, K. E., Hochberg,
E. J., Garza-Perez, R., Mumby, P. J., Riegl, B., Yamano, H., White,
W. H., Zubia, M., Brock, J. C., Phinn, S. R., & Muller-Karger, F. E.
(2003). Multi-sites evaluation of IKONOS data for classification of
tropical coral reef environments. Remote Sensing of Environment, 88,
127 142. (doi:10.1016/j.rse.2003.04.005)
Andrefouet, S., & Muller-Karger, F. (2003). Influence of pre-processing of
Ikonos images on texture characterization of bottom types in shallow
coral reefs. IEEE Transactions on Geosciences and Remote Sensing,
(submitted for publication).
Bindschadler, R. (2001). Ice dolines on Larsen Ice Shelf, Antarctica: An
application of IKONOS data. Proceedings of the 2001 High Spatial
Resolution Commercial Imagery Workshop, March 19 21, Greenbelt,
MD, USA, sponsored by NASA/NIMA/USGS Joint Agency Commercial
Imagery Evaluation Team, CD-ROM.
22
Birk, R. J., Stanley, T., Snyder, G. I., Hennig, T. A., Fladeland, M. M., &
Policelli, F. (2003). Government programs for research and operational
uses of commercial remote sensing data. Remote Sensing of Environment, 88, 3 16. (doi:10.1016/j.rse.2003.07.007)
Chen, X., Vierling, L., Rowell, E., Dykstra, D., Capehart, W., & DeFelice,
T. (2002). Relationships among IKONOS imagery, airborne scanning
LIDAR, and ground-based tree inventory data in a Ponderosa Pine
forest: A multiple endmember approach. Proceedings of the 2001 High
Spatial Resolution Commercial Imagery Workshop, March 19 21,
Greenbelt, MD, USA, sponsored by NASA/NIMA/USGS Joint Agency
Commercial Imagery Evaluation Team, CD-ROM.
Constance, E. (2002). IKONOS DEM evaluation. Proceedings of the 2002
High Spatial Resolution Commercial Imagery Workshop, March 25
27, Reston, VA, USA, sponsored by NASA/NIMA/USGS Joint Agency
Commercial Imagery Evaluation Team, CD-ROM.
Dial, G., & Grodecki, J. (2002). Block adjustment with rational polynomial
camera models. ACSM-ASPRS 2002 Conference and Technology Exhibition Proceedings, April 22 26, Washington, DC.
Ecuyer, D. (2001). Use of IKONOS data in littoral battlespace. Proceedings
of the 2001 High Spatial Resolution Commercial Imagery Workshop,
March 19 21, Greenbelt, MD, USA, sponsored by NASA/NIMA/USGS
Joint Agency Commercial Imagery Evaluation Team, CD-ROM.
Leachtenauer, J. C., Malila, W., Irvine, J., Colburn, L., & Salvagio, N.
(1997). General Image-Quality Equation: GIQE. Applied Optics: Information Processing, 36(32), 8322 8328.
NASA (2002). Domestic Non-reimbursable Space Act Agreement between
National Aeronautics and Space Administration John C. Stennis Space
Center, National Imagery and Mapping Agency Commercial Imagery
Program, and United States Geological Survey Land Remote Sensing
Program for the Joint Agency Commercial Imagery Evaluation (JACIE)
Team. SSC/FED-02-001-06, John C. Stennis Space Center, MI.
NASA, NIMA, & USGS (2001). Proceedings of the 2001 High Spatial
Resolution Commercial Imagery Workshop, March 19 21, Greenbelt,
MD, USA, sponsored by NASA/NIMA/USGS Joint Agency Commercial
Imagery Evaluation Team, CD-ROM.
NASA, NIMA, & USGS (2002). Proceedings of the 2002 High Spatial
Resolution Commercial Imagery Workshop, March 25 27, Reston, VA,
USA, sponsored by NASA/NIMA/USGS Joint Agency Commercial Imagery Evaluation Team, CD-ROM.
NASA/NIMA/USGS (2003). Proceedings of the 2003 High Spatial Resolution Commercial Imagery Workshop, May 19 21, Reston, VA, USA,
sponsored by NASA/NIMA/USGS Joint Agency Commercial Imagery
Evaluation Team, CD-ROM.
National Research Council (2002). Toward New Partnerships in Remote
Sensing: Government, the Private Sector and Earth Science Research.
Washington, DC: National Academy Press.
NIMA (2000). Commercial Imagery Program Brochure.
NIMA (2001). Commercial Imagery Uses Brochure, Version 2.0.
OSTP (2003). U.S. Commercial Remote Sensing Policy: Fact Sheet. Executive Office of the President of the United States, Office of Science and
Technology Policy, April 25.
Pagnutti, M., Ryan, R., Kelly, M., Holekamp, K., Zanoni, V., Thome, K., &
Schiller, S. (2003). Radiometric characterization of IKONOS multispectral imagery. Remote Sensing of Environment, 88, 52 67. (doi:10.1016/
j.rse.2003.07.008)
Peterson, B. (2001). IKONOS Relative Spectral Response and Radiometric
Calibration Coefficients, Space Imaging. Doc. #SE-REF-016, Revision
A, April 26.
Ryan, R., Baldridge, B., Schowengerdt, R., Choi, T., Helder, D., & Blonski, S. (2003). IKONOS spatial resolution and image interpretability.
Remote Sensing of Environment, 88, 37 51. (doi:10.1016/j.rse.2003.
07.006)
Seelan, S. K., Laguette, S., Casady, G. M., & Seielstad, G. A. (2003).
Remote sensing applications for precision agriculture: A learning community approach. Remote Sensing of Environment, 88, 156 168.
(doi:10.1016/j.rse.2003.04.007)
Sidor, J. (2001). Utility of IKONOS data for disaster response. Unpublished presentation at the 2001 High Spatial Resolution Commercial
Imagery Workshop, March 19 21, Greenbelt, MD, USA, sponsored
by NASA/NIMA/USGS Joint Agency Commercial Imagery Evaluation
Team.
Small, C. (2003). High resolution spectral mixture analysis of urban reflectance. Remote Sensing of Environment, 88, 169 185. (doi:10.1016/
j.rse.2003.04.008)
Yang, L., Huang, C., Homer, C. G., Wylie, B. K., & Coan, M. J. (2003). An
approach for mapping large-area impervious surfaces: Synergistic use
of Landsat-7 ETM+ and high spatial resolution imagery. Canadian
Journal of Remote Sensing, 29(2), 230 240.
Abstract
The IKONOS satellite simultaneously collects 1-m panchromatic and 4-m multispectral images, providing the commercial and
scientific community with a dramatic improvement in spatial resolution over previously available satellite imagery. The sun-synchronous
IKONOS orbit provides global coverage, consistent access times, and near-nadir viewing angles. The system is capable of 1:10,000 scale
mapping without ground control and 1:2400 scale mapping with ground control. The IKONOS ground station produces radiometrically
corrected images, georectified images, orthorectified images, stereo pairs, and digital elevation models (DEMs) for image analysis,
photogrammetric, and cartographic applications. This article provides an overview of the IKONOS satellite, ground systems, products, and
applications.
D 2003 Elsevier Inc. All rights reserved.
Keywords: IKONOS satellite; Panchromatic; Imagery
1. Introduction
Space Imaging was formed to explore the commercial
prospects of high-resolution satellite imagery. Space Imaging contracted Lockheed Martin to develop the Space
Segment and Raytheon to develop the Ground Segment of
the Commercial Remote Sensing Satellite (CRSS) now
known as IKONOSR. The IKONOS satellite was launched
September 24th, 1999 to provide global, accurate, highresolution imagery to individuals, organizations, and governments for mapping, monitoring, and development.
The design of a high-resolution satellite imaging system
proceeds from a series of design trades between spatial
resolution and swath width, revisit time and off-nadir
viewing angle, image compression and data transmission
rates, and other performance specifications. Higher resolution results in a narrower swath width. Waiting for nearnadir opportunities results in long revisit times but short
revisit times can be obtained with off-nadir viewing geometries. The vast quantity of collected data and available
communications bandwidth require that the imagery be
compressed for transmission.
Space Imaging provided high-resolution imagery to
NASA under the Scientific Data Purchase (SDP) contract
* Corresponding author. Tel.: +1-303-254-2058; fax: +1-303-254-2214.
E-mail address: gdial@spaceimaging.com (G. Dial).
0034-4257/$ - see front matter D 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2003.08.014
24
Table 1
IKONOS performance summary
Specification
Value
Orbit height
Orbit inclination
Descending node time
Field of regard
Revisit time at mid-latitude
681 km
98.1j, sun synchronous
f 10:30 a.m., local solar time
Up to 45j off nadir
3 days at 60j elevation
11 days at 72j elevation
141 days at 89j elevation
Panchromatic and multispectral
13,816 pixels
3454 pixels
11 km at nadir
11 km
1000 km
400 km
11 bits
0.82 m at nadir
17%
Blue, green, red, NIR
3.28 m at nadir
445 516 nm
506 595 nm
632 698 nm
757 853 nm
Image sensors
Width of panchromatic arrays
Width of multispectral arrays
Field of view
Minimum image length
Maximum mono image length
Maximum stereo image length
Radiometric resolution
Panchromatic ground sample distance
MTF at Nyquist
Multispectral bands
Multispectral GSD
Blue bandpass
Green bandpass
Red bandpass
NIR bandpass
Fig. 1. Relationship between IKONOS revisit intervals, latitude, sensor elevation angles, and GSD.
25
that differentiates IKONOS in the high-resolution, commercial satellite-imaging arena. Fig. 2 also shows the stereo area
rate capability. Note that IKONOS can image a strip as long
as 450 km and then rotate back to collect the entire area in
stereo on the same pass.
From launch to August of 2003, Space Imaging collected
over 162,000,000 km2 of the earths surface and archived
over 400 terabytes of data.
3.2. Image quality
Image quality can be described by SNR, MTF, and
National Imagery Interpretability Rating Scale (NIIRS)
ratings (Ryan et al., 2003). SNR and MTF were measured
during On Orbit Acceptance Tests (OOAT) under conditions
that approximated a 10% reflective target with a 30j sun
elevation angle. Table 2 shows the OOAT measurements of
SNR and MTF. As expected, the system was found to be
shot-noise limited. MTF was measured with the edge target
shown in Fig. 3 with the results shown in Fig. 4. Exoatmospheric measurements of IKONOS MTF are discussed
in Bowen and Dial (2002).
Table 2
IKONOS SNR and MTF measurements
Band
SNR
Nyquist MTF
Pan
Blue
Green
Red
NIR
89
94
143
103
67
17%
26%
28%
29%
28%
26
Fig. 3. IKONOS image of edge target used for modulation transfer function
measurements (Imageryn Space Imaging).
Description
Requirement
OOAT
results
Post OOAT
results
F 10%
F 10%
F 10%
F 5%
F 5%
F 5%
F 10%
Compensation
F 5%
F 5%
F 5%
NA
None
< 1%
< 1%
F 1%, 1-r
F 3%, 1-r
None
Production date
Bands
Blue
Green
Red
NIR
11-bit
633
728
79
91
649
727
81
91
840
949
105
119
746
843
93
105
8-bit
27
Fig. 5. Relative radiometric uniformity requirements demonstrated on Antarctic snowfield uniform scene. Left: Visible banding artifacts still visible after
radiometric correction to better than 10 counts or 0.8%. Right: Radiometric uniformity is now better than 0.5% over the dynamic range of the detectors
(Imageryn Space Imaging).
28
3.3.3. Linearity
The response of the IKONOS system has been shown to
be linear throughout the dynamic range of the detectors. The
technique used to demonstrate this capability is to image
several stable, radiometrically characterized stars. The ataperture radiance for each star is calculated. The star is
imaged and the total DN generated by the star is measured.
Fig. 6 shows the calculated versus measured response
values. This technique shows the linearity of the IKONOS
system to be better than 1% of full scale.
Requirement
OOAT results
25.4 m CE90,
22.0 m LE90
2.0 m CE90,
3.0 m LE90
25.4 m CE90
12.2 m CE90
2.0 m CE90
1.4 m CE90
30 m CE90
5.8 m LE90
7 m RMSE
4.3 m RMSE
3 m LE90
29
Fig. 8. IKONOS geometric accuracy testing range over San Diego consisting of 140 ground control points over a 22 22-km area (Imageryn Space Imaging).
30
Fig. 9. Post-FAM-calibration residual errors over the Denver test range displayed against the GCP layout; no significant systematic error pattern.
Fig. 10. Post-FAM-calibration residual errors over the southern test range
displayed against the GCP layout; no significant systematic error pattern.
Average
error / (m)
2.9
1.5
Average
error k (m)
Average
error h (m)
CE90
(m)
LE90
(m)
2.5
0.7
1.0
0.0
4.4
2.7
2.7
2.5
each image were determined by block adjusting the IKONOS images in our ground station and later used to compute
the mean interlock angle corrections. Fig. 11 shows the error
distribution after applying the interlock angle corrections. It
is seen that the resulting geometric accuracy of an uncontrolled IKONOS imagery is better than 10-m CE90.
3.4.7. Validation
A Space Imaging project located in a relatively flat area
of Mississippi, with six stereo strips and a large number of
well-distributed GCPs, was used to validate the FAM and
interlock calibrations. Each of the 12 source images was
produced as a georectified image with RPC camera model
data. The images were then loaded onto a SOCET SETR
workstation running a Space Imaging developed RPC block
adjustment model, described in more detail in Dial and
Grodecki (2002a) and Grodecki and Dial (2002a). To
quantify block adjustment accuracy, GCPs were selectively
changed between control and check points. Horizontal and
vertical accuracy for the case with no GCP and the case with
one GCP are given in Table 6. Vertical and horizontal error
distribution for a case with a single GCP is shown in Fig.
12. More detailed results can be found in Grodecki and Dial
(2002b).
Several statistical tests of IKONOS accuracy have been
performed since the completion of OOAT. IKONOS mono
accuracy, exclusive of terrain displacement, was measured
31
at 4-m RMS per axis (9-m CE90) with RMS relative errors
of 50 parts per million (Dial & Grodecki, 2002b). IKONOS
stereo accuracy without control has been reported at 6.2-m
CE90 horizontal with 10.1-m LE90 vertical by Dial and
Grodecki (2003a) and 7.9-m CE90 horizontal with 7.6-m
LE90 vertical by Ager (2003).
Description
Nearest commercial
product name
Standard
Original
Geo
Precision
Original
Standard
Master
Precision
Master
Model
Fig. 12. Validation of FAM and interlock calibrations, and quantification of
block adjustment accuracy: vertical and horizontal errors shown by dotted
and solid arrows, respectively, for a case with a single GCP.
None
Reference Stereo +
IKONOS Terrain
Model
32
Table 8
Scientific data purchase parameters
Parameter
Value
Bits
DRA
MTFC
Mosaic
Pan sharpen
Resampling
GSD
Bands
Datum
Projection
Format
Media
11
No
Yes
No
No
Cubic convolution
1 m pan, 4 m MSI
Pan + blue, green, red, and NIR
WGS84
UTM
GeoTIFF
CD-ROM
Fig. 13. Extracting detail from shadows in 11-bit imagery. Left: Image displayed with full radiometric range. Right: Same image stretched to show detail in
shadows.
33
Product name
CE90 (m)
RMS (m)
Scale
GCP
Standard
Reference
Pro
Precision
Precision Plus
50
25
10
4
2
25
12
5
2
1
1:100,000
1:50,000
1:12,000
1:4800
1:2400
No
No
Optional
Yes
Yes
34
Fig. 16. Anaglyph visualization made from an IKONOS stereo pair (Imageryn Space Imaging).
35
References
Ager, T. (2003). Evaluation of the geometric accuracy of Ikonos imagery.
SPIE 2003 AeroSense Conference, April 21 25, Orlando, FL.
Baldridge, B. (2002). Civil and Commercial Applications Project (CCAP):
Evaluation of imagery interpretability for IKONOS pan, MSI, and pansharpened imagery. Proceedings of the 2002 High Spatial Resolution
Commercial Imagery Workshop, March 25 27, Reston, VA, USA.
(CD-ROM; sponsored by NASA/NIMA/USGS Joint Agency Commercial Imagery Evaluation Team).
Bowen, H. S. (2002). Absolute radiometric calibration of the IKONOS
sensor using radiometrically characterized stellar sources. Proceedings of the ISPRS Commission I Mid-Term Symposium/Pecora 15Land Satellite Information IV Conference, November 10 14, Denver, CO.
Bowen, H. S., & Dial, G. (2002). IKONOS calculation of MTF using
stellar images. Proceedings of the 2002 High Spatial Resolution Commercial Imagery Workshop, March 25 27, Reston, VA, USA. (CDROM; sponsored by NASA/NIMA/USGS Joint Agency Commercial
Imagery Evaluation Team).
Bowen, H. S., & Oleszczuk, R. (2002). IKONOS radiometric stability and
relative calibration. Proceedings of the 2002 High Spatial Resolution
Commercial Imagery Workshop, March 25 27, Reston, VA, USA.
(CD-ROM; sponsored by NASA/NIMA/USGS Joint Agency Commercial Imagery Evaluation Team).
Cook, M. K., Peterson, B. A., Dial, G., Gibson, L., Gerlach, F. W., Hutchins, K. S., Kudola, R. S., & Bowen, H. S. (2001). IKONOS technical
36
performance assessment. In S. S. Shen, & M. R. Descour (Eds.), Proceedings of SPIE: Algorithms for Multispectral, Hyperspectral, and
Ultraspectral Imagery VII, 4381(10), pp. 94 108.
Dial, G., Gibson, L., & Poulsen, R. (2001). IKONOS satellite imagery and
its use in automated road extraction. In Baltsavias, Gruen, & Gool
(Eds.), Automatic extraction of man-made objects from aerial and space
images (III). A.A. Balkema Publishers.
Dial, G., & Grodecki, J. (2002a). Block adjustment with Rational Polynomial Camera models. Proceedings of ASPRS 2002 Conference, April
22 26, Washington, DC.
Dial, G., & Grodecki, J. (2002b). IKONOS accuracy without ground
control. Proceedings of ISPRS Commission I Mid-Term Symposium,
November 10 15, Denver, CO.
Dial, G., & Grodecki, J. (2003a). IKONOS stereo accuracy without ground
control. Proceedings of ASPRS 2003 Conference, May 5 9, Anchorage, Alaska.
Dial, G., & Grodecki, J. (2003b). IKONOS applications. Proceedings of
ASPRS 2003 Conference, May 5 9, Anchorage, Alaska.
Fraser, C., Hanley, H., & Yamakawa, T. (2002). High-precision geopositioning from IKONOS satellite imagery. Proceedings of ASPRS 2002
Conference, April 22 26, Washington, DC.
Gerlach, F. (2001). How collection geometry affects specular reflections.
Imaging NOTES, March/April.
Grodecki, J. (2001). IKONOS stereo feature extraction-RPC approach.
Proceedings of ASPRS 2001 Conference, April 23 27, St. Louis, MO.
Grodecki, J., & Dial, G. (2001). IKONOS geometric accuracy. Proceedings
of Joint Workshop of ISPRS Working Groups I/2, I/5 and IV/7 on High
Lockheed Martin Space Operations-Stennis Programs, Remote Sensing Directorate, Bldg. 1105, John C. Stennis Space Center, MS 39529, USA
b
Booz Allen Hamilton, now with Research Systems, Inc., Vienna, VA, USA
c
Department of Electrical and Computer Engineering, University of Arizona, Tucson, AZ, USA
d
Department of Electrical Engineering, South Dakota State University, Brookings, SD, USA
Received 7 January 2003; received in revised form 18 July 2003; accepted 30 July 2003
Abstract
Five individual projects characterized the spatial performance of the IKONOS commercial imaging sensor. The result was determination
of the spatial image quality of IKONOS data products in terms of the National Imagery Interpretability Rating Scale (NIIRS), the system
Modulation Transfer Function (MTF), the system stability over its first year of operation, the characteristics of the Space Imaging MTF
Compensation (MTFC) procedure, and the application-specific capabilities of IKONOS imagery. Both panchromatic and multispectral
imagery were evaluated. Major conclusions of this work are that the system was stable in imaging performance during the first year of
operation, that its MTF meets the specification for the NASA Scientific Data Purchase program, that the initial MTFC processing appears to
be transposed in the in-track and the cross-track directions, that the MTFC results in a noise amplification of 2 to 4 in addition to
sharpening the imagery, and that IKONOS panchromatic imagery achieves an average NIIRS rating of 4.5.
D 2003 Elsevier Inc. All rights reserved.
Keywords: IKONOS; Modulation Transfer Function; National Imagery Interpretability Rating Scale
1. Introduction
The pre-eminent characteristic of IKONOS is its significantly higher spatial resolution compared to other nonmilitary satellite remote sensing systems. Both the National
Aeronautics and Space Administration (NASA) and the
National Imagery and Mapping Agency (NIMA) have purchased IKONOS 1- and 4-m ground sample distance (GSD)
imagery for various purposes. NIMA acquires commercial
satellite imagery as part of its mission to provide geospatial
information to the Department of Defense (DoD) and to the
national intelligence community. NIMAs vehicle for assessing the image quality and utility of commercial imagery is
the Civil and Commercial Applications Project (CCAP).
NASA, through its Scientific Data Purchase (SDP), purchased imagery primarily for land use research. NASA
Stennis Space Center and its academic partners at the
University of Arizona and South Dakota State University
38
0:344G
SNR
N X
N
X
G
MTFCKernelij 2
# 12
2
i1 j1
Table 1
Noise gain associated with MTFC processing
Band
Noise gain
Blue
Green
Red
NIR
Pan
1.59
1.63
1.68
1.81
4.16
39
40
41
Fig. 1. Space Imaging target (August 5, 2001) with MTFC-Off (left) and with MTFC-On (right). The Roberts gradient magnitude images are shown directly
below. Note the sharper gradient in the MTFC-On case. The amplitude of the gradient is greater in the latter case.
Table 2
Reference image comparisons for sharpness quality metric
Location
Processing
Average DN
gradient
Average %
difference
Space Imaging
target
Tucson, AZ
MTFC-Off
MTFC-On
MTFC-Off
MTFC-On
54.2
78.2
60.97
80.38
44.3
31.8
Table 3
Anniversary image pair solar and sensor angles
Date
Solar
azimuth (j)
Solar
elevation (j)
Sensor
azimuth (j)
Sensor
elevation (j)
113.8
117.7
65.4
70.2
136.1
276.8
84.2
84.1
42
Table 4
Average DN gradient magnitudes for Tucson sites
Date
Area 1
(residential)
499 406 pixels
Area 2
(road/industrial)
424 451 pixels
167.2
159.5
129.7
118.1
Fig. 2. Illustration of tarp orientation with respect to true north and corresponding example image.
43
5.3. Results
Fig. 4. A blurred edge and differentiation of the edge to locate the point of maximum slope.
44
Fig. 5. IKONOS multispectral image of blue tarps deployed at Brookings, SD on August 13, 2001.
Fig. 6. Pulse response function obtained from blue tarps on August 13,
2001.
Fig. 9. MTF function for the blue tarp target on August 13, 2001.
45
Fig. 10. Overplot of four estimates of blue band pulse response functions
from blue tarps.
Fig. 11. Overplot of four estimates of blue band MTF functions from blue
tarps.
46
6.2. Method
Full width at half maximum of a LSF is used as a measure
of spatial resolution of the images. Before LSFs are derived
from edge responses by numerical differentiation, the edge
responses are measured and analyzed using a modified knifeedge technique (Tzannes & Mooney, 1995). Adjacent black
and white square panels, either painted on a flat surface or
deployed as tarps, form a ground-based edge target used in
the tests. During the measurements, the edge target is intentionally oriented so the image of the edge is aligned slightly
off-perpendicular to a pixel grid direction. The tilted-edge
modification to the original knife-edge method allows properly sampled edges to be obtained, minimizing aliasing
(Reichenbach, Park, & Narayanswamy, 1991). To measure
an edge response, a rectangular region containing the tilted
edge is extracted from an image of the edge target as shown in
Fig. 12. In such a region, each line across the edge forms an
approximate edge response. Exact edge responses (in the
direction perpendicular to the edge) are obtained when
distances are additionally scaled by cosine of the tilt angle.
The distance correction is usually small, but it becomes
important when results from measurements with different
edge orientations are to be compared.
Size of the edge target panels is a critical factor in spatial
resolution measurements of satellite images. To accommodate the 1-m GSD of the IKONOS panchromatic images,
panels 20 20 m in size are used. Larger panels would
provide even more accurate measurements, but deployment
of such large targets becomes extremely difficult. Therefore,
the number of edge response samples available for analysis is
still limited by the target size even with the additional
sampling provided by the edge tilt. For panels of given size,
sampling of the spatial response is also affected by the edge
tilt angle. When deployable tarps are used in the measure-
3
X
a
k
x b1 Di b2
k1 1 exp
ck
Fig. 12. IKONOS panchromatic images of the edge target tarps deployed at Stennis Space Center, MS, for the easting direction measurement on January 15,
2002 (left) and for the northing direction measurement on February 17, 2002 (right). Gray rectangular frames overlaid on the tarp images show the areas
selected for the edge response analyses.
47
Fig. 13. Measured edge responses (left column) and the best fits to them with superposition of three sigmoidal functions (right column). Data are for the
measurements in the northing direction using the IKONOS image acquired on February 17, 2002 and processed either with MTFC-On (bottom row) or with
MTFC-Off (top row).
Fig. 14. Superimposed edge responses and the fitted sigmoidal functions for the IKONOS images processed either with MTFC-On (bottom row) or MTFC-Off
(top row). Northing direction (February 17, 2002) is the left column and easting direction (January 15, 2002) is the right column.
48
Fig. 15. Line spread functions derived from the fitted edge responses for the IKONOS images processed either with MTFC-On (bottom row) or with MTFCOff (top row). Northing direction (February 17, 2002) is the left column and easting direction (January 15, 2002) is the right column.
Table 5
Line spread function FWHM for IKONOS panchromatic images (January/
February 2002)
MTFC-Off
MTFC-On
Easting (m)
Northing (m)
1.45
0.85
1.27
0.52
49
for each chip using nearest-neighbor resampling and applying an identical histogram stretch as the 1 version. Each
chip pair (1 and 2 ) was rotated to the appropriate
cardinal direction that best aligned image obliquity to the
top of the display.
The evaluation was conducted at the NIMA Imagery
Support and Assessment Branchs softcopy evaluation
facility. All evaluation participants used the same workstation with a precision color monitor, which was calibrated
before the start of the evaluation. The monitors were set to
a minimum luminance response of 0.10 fl and a maximum
luminance response of 35.0 fl. Evaluation participants were
free to roam and zoom at 1 or 2 magnification within
the image. All ratings were made at 2 . No interactive
enhancement of the imagery was allowed and image chips
were rendered with no additional processing.
Participants for this evaluation consisted of eight NIMA
Visible NIIRS-certified imagery analysts. Evaluation participants were experienced with assessing Order of Battle as
well as natural and cultural features on panchromatic
imagery. The evaluation procedure consisted of each participants reviewing a sequence of panchromatic (pan)
scenes on the softcopy workstation and responding with
both NIIRS ratings and task satisfaction confidence ratings
for each scene. Examples of specific questions used are as
follows:
50
image was removed from the NIIRS dataset and two images
were removed from the EEI confidence ratings. These
cleaned datasets were used for all further analyses. The
average NIIRS value for TIFF pan images was 4.65 based
on a population of 46 image chips obtained from 24 image
products. The average NIIRS value for GeoTIFF pan images
was 4.41 based on a population of 26 chip sets from 15
image products.
For these levels of NIIRS for the TIFF imagery, the RER
is approximately 0.7. Approximately a 0.4 NIIRS improvement is expected with MTFC processing based on the
GIQE.
7.4. NIIRS ratings
The first step in obtaining NIIRS ratings was to calculate
descriptive statistics. Then an analysis of variances was
conducted to examine the difference in ratings for format
(GeoTIFF vs. TIFF) and climate (Arid, Tropical, Temp (N),
Temp (S)), including GSD as a covariate. For all analyses
that include GSD, the collected GSD was used (as opposed
to the resampled GSD) and was transformed to log10.
Climate was not found to be significantly different, so the
variable was dropped from further analyses. An Analysis of
Covariance was conducted to determine the impact of
format on mean NIIRS. The analysis included log10GSD
as a covariate, as well as an interaction term between
log10GSD and format. The analysis revealed that format
was not a significant main effect, but log10GSD ( p = 0.02)
and the interaction term ( p = 0.07) were significant predictors of NIIRS (R2 = 0.19). These results indicate that format
does not directly affect NIIRS ratings. The 0.24 difference
in mean NIIRS between TIFF and GeoTIFF is based on the
average differences in GSD and the interaction between
GSD and format. TIFF imagery has a lower average GSD
(0.931 m) than GeoTIFF (0.993 m), thus accounting for
higher NIIRS ratings. These results were used to derive
simplified regression Image Quality Equations, predicting
NIIRS from log10GSD for both formats:
GeoTIFF predicted NIIRS 4:41 0:77 log10 GSD
7.6. Conclusion
An average NIIRS rating of 4.5 was achieved with
IKONOS pan imagery, so this imagery should be able to
satisfy intelligence tasks for NIIRS levels 3 and 4. Although
the highest NIIRS and EEI values were obtained with the
TIFF products, the level of processing was not found to be a
significant predictor of NIIRS in the sample size used. As
might be expected, GSD was found to be significant
predictor of NIIRS for both formats. The range in average
NIIRS values for the entire sample of TIFF and GeoTIFF
images was 3.61 5.28, with NIIRS values increasing as
collected GSD decreased. A statistical analysis shows that
the IQE for two formats had statistically different slopes. It
is suspected that a broken-line IQE is more appropriate for
the GeoTIFF images, but such a model could not be
distinguished from a single linear equation with the data
available. Because TIFF images are higher NIIRS than
GeoTIFF images when the collected GSD is actually less
than 1 m, the TIFF image format is recommended for
intelligence EEI-type application if the user has the option.
This study indicates that IKONOS panchromatic products
can satisfy requirements for NIIRS level 4 imagery. Where
the best possible satisfaction of image tasks is required, the
highest collected resolution (i.e., at nadir) IKONOS panchromatic imagery should be acquired. For imagery where
coordinate embedded pixels are not required (as in GeoTIFF) and submeter GSD is assured, TIFF products should
be obtained.
51
Acknowledgements
This work was supported by the NASA Earth Science
Applications Directorate under contract number NAS 13650 at the John C. Stennis Space Center, MS and by NIMA.
The authors thank Marcia Wise and Denise Jarrell for their
assistance in preparing this manuscript. Special thanks go to
Vicki Zanoni for her support, careful reading, and comments
on this manuscript.
References
Castleman, K. R. (1996). Digital image processing ( p. 667). Upper Saddle
River, NJ: Prentice Hall.
Garvin, J. B., Mahmood, A., & Yates, J. (2002). Monitoring landscapes on
oceanic islands: Sub-meter imaging from IKONOS in the context of
RADARSAT SAR. Proceedings of the 2002 High Spatial Resolution
Commercial Imagery Workshop, March 25 27, Reston, VA, USA (sponsored by NASA/NIMA/USGS Joint Agency Commercial Imagery Evaluation Team, CD-ROM).
Holst, G. C. (1995). Electro-optical imaging system performance
( pp. 146 147). Bellingham, WA: SPIE Optical Engineering Press.
IRARS (1996). Civil NIIRS Reference Guide. Imagery Resolution Assessment and Reporting Standards Committee, http://www.fas.org/irp/imint/
niirsM_c/guide.htm, accessed June 12, 2003.
Leachtenauer, J. C., Malila, W., Irvine, J. M., Colburn, L. P., & Salvaggio,
N. L. (1997). General Image-Quality Equation: GIQE. Applied Optics,
36(32), 8322 8328.
NIMA (1996). General Image Quality Equation Users Guide, Version 4.0.
National Imagery and Mapping Agency.
52
Pagnutti, M., Ryan, R., Kelly, M., Holekamp, K., Zanoni, V., Thome, K.,
& Schiller, S. (2003). Radiometric characterization of IKONOS multispectral imagery. Remote Sensing of Environment. (doi:10.1016/
jrse.2003.07.008)
Rauchmiller, R. F., & Schowengerdt, R. A. (1988). Measurement of the
Landsat Thematic Mapper MTF using an array of point sources. Optical
Engineering, 27(4), 334 343.
Reichenbach, S. E., Park, S. K., & Narayanswamy, R. (1991). Characterizing digital image acquisition devices. Optical Engineering, 30,
170 177.
Schowengerdt, R. A. (1997). Remote sensingmodels and methods for
image processing ( p. 522). San Diego, CA: Academic Press.
Schowengerdt, R. A., Archwamety, C., & Wrigley, R. C. (1985). Landsat
Thematic Mapper image-derived MTF. Photogrammetric Engineering
and Remote Sensing, 51(9), 1395 1406.
Storey, J. C. (2001). Landsat 7 on-orbit modulation transfer function estimation. In H. Fujisada, J. B. Lurie, & K. Weber (Eds.), Proceedings of SPIE:
Sensors, Systems, and Next-Generation Satellites, 4540 (pp. 50 61).
Tucker, C. J. (2002). From Agamemnon to IKONOS In search of the
Trojan War. Proceedings of the 2002 High Spatial Resolution Commercial Imagery Workshop, March 25 27, Reston, VA, USA (sponsored by
NASA/NIMA/USGS Joint Agency Commercial Imagery Evaluation
Team, CD-ROM).
Tzannes, A. P., & Mooney, J. M. (1995). Measurement of the modulation transfer function of infrared cameras. Optical Engineering, 34,
1808 1817.
Wigglesworth, J. K. (2002). What is a Post-Hoc Analysis? Ithaca College,
Survey of Statistical Methods, http://www.ithaca.edu/jwiggles/stats/
notes/notes23.htm, accessed July 30, 2003.
Remote Sensing Directorate, Lockheed Martin Space Operations-Stennis Programs, Building 1105, John C. Stennis Space Center, MS 39529, USA
b
Lockheed Martin Space Operations-Stennis Programs, USA
c
NASA Earth Science Applications Directorate, John C. Stennis Space Center, USA
d
Remote Sensing Group, Optical Sciences Center, University of Arizona, USA
e
Physics Department, South Dakota State University, USA
Received 2 May 2002; received in revised form 25 July 2003; accepted 31 July 2003
Abstract
A radiometric characterization of Space Imagings IKONOS 4-m multispectral imagery has been performed by a National Aeronautics
and Space Administration (NASA) funded team from NASA Stennis Space Center, the University of Arizona Remote Sensing Group (U of A
RSG), and South Dakota State University (SDSU). Both intrinsic radiometry and the effects of Space Imaging processing on radiometry were
investigated. Relative radiometry was examined with uniform Antarctic and Saharan sites. Absolute radiometric calibration was performed
using reflectance-based vicarious calibration methods on several uniform sites imaged by IKONOS coincident with ground-based surface and
atmospheric measurements. Ground-based data and the IKONOS spectral response function served as input to radiative transfer codes to
generate a top-of-atmosphere (TOA) radiance estimate. Calibration coefficients derived from each vicarious calibration were combined to
generate an IKONOS radiometric gain coefficient for each multispectral band assuming a linear response over the full dynamic range of the
instrument. These calibration coefficients were made available to Space Imaging, which subsequently adopted them by updating its initial set
of calibration coefficients. IKONOS imagery procured through the NASA Scientific Data Purchase was processed with or without a
Modulation Transfer Function Compensation (MTFC) kernel. The radiometric effects of this kernel on various scene types were also
investigated.
D 2003 Elsevier Inc. All rights reserved.
Keywords: IKONOS; Space Imaging; Multispectral imagery
1. Introduction
The National Aeronautics and Space Administration
(NASA) Earth Science Enterprise (ESE) initiated the
Scientific Data Purchase (SDP) to evaluate a new way
of obtaining remote sensing datasets for meeting its
research objectives. In the SDP, NASA purchased data
without the up-front investment required with a traditional
approach where a remote sensing system is built and
operated by the U.S. Government. The NASA Earth
Science Applications (ESA) Directorate at Stennis Space
Center (SSC) implemented the SDP and administered the
contracts with the commercial companies that were competitively selected to provide the programs remotely
sensed image products: Earth Satellite (EarthSat), Space
Imaging, DigitalGlobe (formerly EarthWatch Incorporated)/Intermap Technologies, Positive Systems, and AstroVision International. Image products are also being
procured from DigitalGlobe with additional SDP funds.
The ESA Directorate was also responsible for verification
and validation of products delivered under the SDP (Birk
et al., 2003).
The U.S. Government is prohibited from competing
with U.S. private industry (National Research Council,
2002); therefore, several federal government agencies,
including NASA, the National Imagery and Mapping
Agency (NIMA), the U.S. Geological Survey (USGS),
and the National Oceanic and Atmospheric Administration (NOAA), as well as state and local governments, all
54
55
Fig. 1. IKONOS spectral response curves for panchromatic and multispectral bands.
56
Fig. 3. Cross-sectional plot of pixels summed in the in-track direction of the uniform Saharan scene.
minimize deleterious image artifacts. Therefore, to visualize array and detector defects better, columns of pixels in
the image were summed and normalized. The summing
process, in essence, increases the signal-to-noise ratio and
amplifies any image defects. Fig. 3 shows a cross-sectional
plot of this summing, indicating that the differences
between arrays (banding) are less than 1.5% for each
band, which is well within the NASA SDP IKONOS
relative radiometric accuracy specification of 5%. The plot
also shows that a few pixels have DN values slightly less
than their adjacent pixels. This is most clearly seen in the
blue band on the first detector array. The single white
stripe shown in Fig. 2 appears as a single spike in the
cross-sectional summation plot (Fig. 3), and the array
transitions appear as step discontinuities. Similar results
were obtained with the other Saharan and Antarctic scenes.
Overall, the imagery is well corrected and our analysis
illustrates how Space Imaging can provide a single calibration coefficient instead of a set of coefficients, one for
each detector.
3.2. Absolute radiometric accuracy assessment
NASA SSC, the U of A RSG, and SDSU all used a
reflectance-based vicarious calibration method to evaluate
the absolute IKONOS multispectral imagery radiometry
(Dinguirard & Slater, 1999; Kaufman & Holben, 1993;
Meygret, Briottet, Henry, & Hagolle, 2000; Slater et al.,
1987; Slater, Biggar, Thome, Gellman, & Spyak, 1994;
Slater, Biggar, Thome, Gellman, & Spyak, 1996; Thome
et al., 1993). This method involves imaging a relatively
uniform area of interest coincident with taking ground-based
measurements. TOA spectral radiance is estimated from
ground reflectance, atmospheric pressure, water, and aerosol
profile measurements. The sensor spectral response function, provided by Space Imaging from laboratory measurements, is used to estimate the measured in-band radiance.
The sites used in this investigation are listed in Table 1,
which summarizes the location, collection dates and times,
and satellite elevation and azimuth angle. Each teams
approach and results are discussed and a compilation of
57
all the measurements is presented. These results are compared to the initial Space Imaging calibration coefficients
that were provided with each image. Space Imaging
obtained its calibration coefficients by imaging well-known
stellar sources primarily found in the Gunn and Stryker
catalog (Bowen, 2002). Measured integrated DN values
were compared against these known sources to determine
these initial calibration coefficients (Peterson, 2001). This
procedure was repeated several times over several years to
determine radiometric stability.
Table 1
Vicarious calibration acquisition summary
Date
Location
Overpass
time (UTC)
Solar
zenith (j)
Solar
azimuth (j)
Sensor
nadir (j)
Sensor
azimuth (j)
Nominal
GSD (m)
White Sands
Missile Range, NM
Lunar Lake, NV
Railroad Valley, NV
Lunar Lake, NV
Railroad Valley, NV
Ivanpah, CA
Brookings, SD
Brookings, SD
17:38
21.9
116.9
12.1
346
0.85
18:16
18:14
18:23
18:23
18:20
17:12
17:13
24.5
24.3
22.9
22.8
24.5
26.5
25.8
123.2
123.8
126.3
126.9
122.6
136.1
135.9
19.0
11.6
17.0
9.5
6.4
17.9
4.4
158
144
351
267
284
0.45
302
0.90
0.85
0.89
0.84
0.83
0.87
0.83
June 7, 2000
June 7, 2000
June 10, 2000
June 10, 2000
July 21, 2000
June 30, 2000
June 30, 2000
58
Fig. 4. IKONOS image of Lunar Lake Playa (RGB) acquired on June 7, 2000.
59
60
Fig. 6. Comparison of solar radiometer measurements and MODTRAN prediction for June 10, 2000.
Table 2
NASA SSC TOA radiance results
Site
Date
Band
DN
TOA
radiance
[W/(m2 sr)]
IKONOS
radiance
[W/(m2 sr)]
Lunar
Lake
6/7/00
Lunar
Lake
6/10/00
Blue
Green
Red
NIR
Blue
Green
Red
NIR
828
1178
1174
1083
811
1093
1108
1028
12.56
16.987
13.531
14.153
12.841
17.28
13.642
14.053
12.991
20.551
17.711
21.536
12.029
19.075
16.712
20.442
Percent
difference
3.43
20.98
30.89
52.17
6.32
10.39
22.50
45.46
than 30 min to collect the data when the reference measurements are included. The reference measurements were taken
at every two transects and at the beginning and end of the
collection, giving a total of 9 reference sets (or 72 reference
samples).
The reflectance of the site was determined by ratioing
the measurements described above to those of the reference panel for which the bidirectional reflectance factor
has been determined in the laboratory. The sampling level
of the reference, relative to the test site surface, reduces
uncertainties due to changes in instrument response with
time and changing atmospheric conditions while keeping
the data collection time to a reasonable level. Knowing the
BRDF of the reference allows the reflectance of each
sample to be computed, taking into account effects due
to sun-angle changes and reflectance panel BRDF over the
30 min of data collection. Once each of the spectral
samples of the site was converted to reflectance, and all
480 data points were averaged to give a single spectral
reflectance for the entire site. The same sampling approach
and test site size were used for all four sites used by the U
of A RSG.
A critical part of this reflectance retrieval is the characterization of the reference panel in the laboratory. The
calibration of this panel was done with reference to a
standard made from a pressed polytetrafluoroethylene
panel based on a prescribed approach defined by NIST.
The calibration reference is a directional-to-hemispheric
reflectance standard provided by NIST. Polynomial fits
were made to the measured data to calculate the reflectance of the field standard for the sun-view geometry and
wavelengths for a given set of field measurements. The
SpectralonR sample was calibrated for a nadir viewing
sensor with a variety of solar illumination angles. The
measurements were made in only one plane, but this does
not create problems in the field because the SpectralonR
panel was aligned in the field such that the solar incident
angle stayed in this same plane. Two different 18-in.,
monolithic SpectralonR panels and two ASDs were used
for this work (U of A RSG serial numbers 6 and 7 for the
SpectralonR panels with the #6 panel paired with ASD687
and the #7 with the ASD614).
Rayleigh scattering, water extinction, and ozone and
aerosol distribution were derived from the solar radiometer data and from the meteorological data. The aerosol
distribution was defined in terms of a Junge distribution
(Biggar, Gellman, & Slater, 1990). These atmospheric
and surface data were inputs to a radiative transfer code
that computes hyperspectral, at-sensor radiances based on
a Gauss-Seidel iteration approach (Thome, GustafsonBold, Slater, & Farrand, 1996). The code assumes a
plane-parallel, homogeneous atmosphere and divides this
atmosphere into layers to account for the vertical distribution of scatterers and weak absorption due to ozone in
the visible and near infrared (approximately the 400 800nm spectral range known as the Chappuis absorption
61
Table 3
U of A RSG TOA radiance results
Site
Date
Band
DN
TOA
IKONOS
Percent
radiance
radiance
difference
[W/(m2 sr)] [W/(m2 sr)]
Lunar
Lake
6/7/00
Lunar
Lake
6/10/00
Railroad
Valley
6/7/00
Railroad
Valley
6/10/00
Blue
Green
Red
NIR
Blue
Green
Red
NIR
Blue
Green
Red
NIR
Blue
Green
Red
NIR
Blue
Green
Red
NIR
Blue
Green
Red
NIR
851
1216
1216
1123
791
1137
1151
1062
734
1012
968
875
667
914
876
787
1422
1826
1694
1564
642
935
999
929
13.155
18.287
13.509
15.293
12.884
17.738
13.272
15.140
9.953
13.431
9.258
10.513
10.352
13.990
9.718
11.009
22.160
28.236
18.641
20.749
9.668
13.946
10.956
12.745
White
5/26/00
Sands
Missile
Range
Ivanpah
7/21/00
13.359
21.222
18.341
22.326
12.418
19.843
17.360
21.113
11.523
17.661
14.600
17.400
10.471
15.951
13.212
15.646
22.323
31.867
25.551
31.093
10.078
16.318
15.068
18.469
1.56
16.05
35.77
45.99
3.62
11.87
30.81
39.45
15.77
31.49
57.70
65.47
1.14
14.02
35.95
42.12
0.74
12.86
37.07
49.85
4.24
17.01
37.53
44.91
62
Table 4
SDSU TOA radiance results
Site
Date
Band
DN TOA
IKONOS
Percent
radiance
radiance
difference
[W/(m2 sr)] [W/(m2 sr)]
3.618
4.168
2.7
7.72
2.439
1.806
0.652
0.654
3.679
4.219
2.718
7.741
2.488
1.839
0.660
0.651
3.910
5.044
3.394
11.471
2.889
2.600
1.071
1.193
4.097
5.305
3.605
11.710
2.920
2.566
1.056
0.795
8.04
21.01
25.69
48.59
18.43
43.98
64.25
82.39
11.37
25.75
32.63
51.27
17.36
39.50
59.97
22.15
63
counts to subtract out the offset, the offset was set to zero
and the least squares approach generating only a gain
parameter was used.
The composite radiometric calibration curves and the
Space Imaging calibration curve are shown in Fig. 7. Table
5 lists both the Space Imaging and the NASA team gain
coefficients for each of the multispectral bands and the ratio
of the two sets of coefficients. Only the Space Imaging
IKONOS blue band calibration was in excellent agreement
64
Table 5
NASA team gain coefficients
Table 7
U of A RSG TOA radiance results using updated SI calibration coefficients
Band
Space
Imaging
gain
NASA team
gain
Site
Date
Band
DN
TOA
IKONOS
Percent
radiance
radiance
difference
[W/(m2 sr)] [W/(m2 sr)]
Blue
Green
Red
NIR
63.7
57.3
66.3
50.3
64.1 F 2.6
65.4 F 2.7
87.7 F 3.1
75.8 F 2.9
1.01
1.14
1.32
1.51
Lunar
Lake
6/7/00
Lunar
Lake
6/10/00
Railroad
Valley
6/7/00
Railroad
Valley
6/10/00
Blue
Green
Red
NIR
Blue
Green
Red
NIR
Blue
Green
Red
NIR
Blue
Green
Red
NIR
Blue
Green
Red
NIR
Blue
Green
Red
NIR
851
1216
1216
1123
791
1137
1151
1062
734
1012
968
875
667
914
876
787
1422
1826
1694
1564
642
935
999
929
13.155
18.287
13.509
15.293
12.884
17.738
13.272
15.14
9.953
13.431
9.258
10.513
10.352
13.99
9.718
11.009
22.16
28.236
18.641
20.749
9.668
13.946
10.956
12.745
White
5/26/00
Sands
Missile
Range
Ivanpah
7/21/00
8. MTFC processing
The effect of MTFC on radiometry for vicarious calibrations and for imagery in general was evaluated as part of this
investigation. To facilitate this effort, Space Imaging provided a 5 5-km scene over Phoenix, AZ, with and without
MTFC processing taken on October 12, 1999 at 17:46 UTC.
The image contained a large urban area having a high spatial
frequency content because of the many edges associated
with buildings and roads. The image also contained some
vegetative areas with relatively low spatial frequency content. First-order statistics, means, standard deviations, histograms, RMSs, and peak differences between MTFC-On and
MTFC-Off imagery were calculated for both the panchromatic and the multispectral radiometrically corrected, georeferenced imagery. In addition, all image pairs (processed
with and without MTFC) were visually inspected and the
differences between them were displayed visually as a
difference image. MTFC estimates were made, and
inverses to these functions were performed to restore
compensated imagery back to its MTFC pre-processed state.
Table 6
NASA SSC TOA radiance results using updated SI calibration coefficients
Site
Date
Band
DN
TOA
radiance
[W/(m2 sr)]
IKONOS
radiance
[W/(m2 sr)]
Lunar
Lake
6/7/00
Lunar
Lake
6/10/00
Blue
Green
Red
NIR
Blue
Green
Red
NIR
828
1178
1174
1083
811
1093
1108
1028
12.56
16.987
13.531
14.153
12.841
17.28
13.642
14.053
13.081
18.151
13.976
14.517
12.812
16.841
13.19
13.78
Percent
difference
4.148
6.852
3.289
2.572
0.226
2.541
3.313
1.943
13.444
18.737
14.476
15.054
12.496
17.519
13.702
14.236
11.596
15.593
11.524
11.729
10.537
14.083
10.429
10.550
22.464
28.136
20.167
20.965
10.142
14.407
11.893
12.453
2.20
2.46
7.16
1.57
3.01
1.23
3.24
5.97
16.50
16.10
24.47
11.57
1.79
0.67
7.31
4.17
1.37
0.36
8.18
1.04
4.90
3.31
8.55
2.29
Date
Band
DN TOA
IKONOS
Percent
radiance
radiance
difference
[W/(m2 sr)] [W/(m2 sr)]
3.618
4.168
2.7
7.72
2.439
1.806
0.652
0.654
3.679
4.219
2.718
7.741
2.488
1.839
0.66
0.651
3.934
4.453
2.679
7.735
2.907
2.296
0.845
0.804
4.123
4.684
2.845
7.895
2.938
2.265
0.833
0.536
8.72
6.84
0.79
0.19
19.18
27.12
29.64
22.98
12.07
11.02
4.68
2.00
18.10
23.17
26.26
17.64
Panchromatic
MTFC-Off
Panchromatic
MTFC-On
598.2
183.6
149
1983
844
781
0.085
598.2
205.2
0
2047
844
781
0.085
65
Table 10
First-order multispectral statistics summary of Phoenix site
Blue
MTFC-Off
Mean DN
S.D. DN
Minimum DN
Maximum DN
Maximum difference DN
Minimum difference DN
RMS difference DN
328.3
89.46
153
2047
425
327
0.047
Green
MTFC-On
328.3
97.89
0
2047
MTFC-Off
412.5
133.45
135
2047
444
307
0.059
Red
MTFC-On
412.5
147.29
0
2047
MTFC-Off
397.8
146.0
83
2047
454
327
0.070
NIR
MTFC-On
397.8
161.5
0
2047
MTFC-Off
426.9
126.2
65
2047
427
378
0.0761
MTFC-On
426.8
145.84
0
2047
66
Fig. 8. Phoenix NIR MTFC-On image (top) and difference between MTFC-On and MTFC-Off NIR imagery (bottom).
difference in pixel values between MTFC-Off and MTFCOn imagery in the NIR band. Simulations of ideal uniform
scenes that do not contain noise show no difference in pixel
value between MTFC-Off and MTFC-On imagery.
Acknowledgements
This work was supported by the NASA Earth Science
Applications Directorate under contract number NAS 13650 at the John C. Stennis Space Center, MS.
References
Berk, A., Anderson, G. P., Acharya, P. K., Chetwynd, J. H., Bernstein, L.
S., Shettle, E. P., Matthew, M. W., & Adler-Golden, S. M. (1999).
MODTRAN4 users manual. Hanscom AFB, MA: Air Force Research Laboratory, Space Vehicles Directorate, Air Force Materiel
Command.
Biggar, S. F., Gellman, D. I., & Slater, P. N. (1990). Improved evaluation of
optical depth components from Langley plot data. Remote Sensing of
Environment, 32, 91 101.
Biggar, S. F., Slater, P. N., & Gellman, D. I. (1994). Uncertainties in the inflight calibration of sensors with reference to measured ground sites in
the 0.4 1.1 Am range. Remote Sensing of Environment, 48, 245 252.
Birk, R. J., Stanley, T., Snyder, G. I., Hennig, T. A., Fladeland, M. M., &
Policelli, F. (2003). Government programs for research and operational
67
68
Sawaya, K., Olmanson, L., Heinert, N., Loeffelholz, B., Rich, R., Brezonik,
P., & Bauer, M. (2003). Extending satellite remote sensing to local
scales: Land and water resource management using high-resolution
imagery. Remote Sensing of Environment, 88, 143 155. (do:10.1016/
j.rse.2003.04.006)
Slater, P. N., Biggar, S. F., Holm, R. G., Jackson, R. D., Mao, Y., Moran,
M. S., Plamer, J. M., & Yuan, B. (1987). Reflectance- and radiancebased methods for the in-flight absolute calibration of multispectral
sensors. Remote Sensing of Environment, 22, 11 37.
Slater, P. N., Biggar, S. F., Thome, K. J., Gellman, D. I., & Spyak, P. R.
(1994). In-flight radiometric calibration of ASTER by reference to wellcharacterized scenes. In W. L. Barnes, & B. J. Horais (Eds.), Proceedings of SPIE: Platforms and systems, vol. 2317 (pp. 49 60).
Slater, P. N., Biggar, S. F., Thome, K. J., Gellman, D. I., & Spyak, P. R.
(1996). Vicarious radiometric calibration of EOS sensors. Journal of
Atmospheric and Oceanic Technology, 13, 349 359.
Thome, K., & Nandy, P. (2000). Accuracy of ground-reference calibration
Electrical Engineering Department, South Dakota State University, PO Box 2220, Brookings, SD 57007, USA
b
USGS EROS Data Center, Sioux Falls, SD 57105, USA
c
NIMA, 3838 Vogel Road, L-85, Arnold, MO 63010, USA
Received 2 June 2002; received in revised form 24 January 2003; accepted 22 April 2003
Abstract
The IKONOS spacecraft acquired images on July 3, 17, and 25, and August 13, 2001 of Brookings SD, a small city in east central South
Dakota, and on May 22, June 30, and July 30, 2000, of the rural area around the EROS Data Center. South Dakota State University (SDSU)
evaluated the Brookings scenes and the USGS EROS Data Center (EDC) evaluated the other scenes. The images evaluated by SDSU utilized
various natural objects and man-made features as identifiable targets randomly distribution throughout the scenes, while the images evaluated
by EDC utilized pre-marked artificial points (panel points) to provide the best possible targets distributed in a grid pattern. Space Imaging
provided products at different processing levels to each institution. For each scene, the pixel (line, sample) locations of the various targets
were compared to field observed, survey-grade Global Positioning System locations. Patterns of error distribution for each product were
plotted, and a variety of statistical statements of accuracy are made. The IKONOS sensor also acquired 12 pairs of stereo images of globally
distributed scenes between April 2000 and April 2001. For each scene, analysts at the National Imagery and Mapping Agency (NIMA)
compared derived photogrammetric coordinates to their corresponding NIMA field-surveyed ground control point (GCPs). NIMA analysts
determined horizontal and vertical accuracies by averaging the differences between the derived photogrammetric points and the fieldsurveyed GCPs for all 12 stereo pairs. Patterns of error distribution for each scene are presented.
D 2003 Elsevier Inc. All rights reserved.
Keywords: IKONOS; Brookings; Space Imaging
1. Introduction
As part of the Joint Agency Commercial Imaging Evaluation (JACIE, which includes NASA, NIMA, and the
USGS), several institutions evaluated the geometric registration of Space Imagings IKONOS imagery using data
purchased under the NASA Scientific Data Purchase (SDP)
Program and the NIMA Commercial Imagery Program
(CIP). The results of the various registration efforts, using
definitions set up for the SDP over a year prior to the launch
of IKONOS, are reported in this article. This paper is neither
an endorsement nor promotion of Space Imaging products
by any of the agencies listed above. It is a report that
summarizes evaluations of registration accuracy of the
images purchased from Space Imaging by the JACIE
agencies. The current data products publicly available from
* Corresponding author. Tel.: +1-605-688-4526; fax: +1-605-6885880.
E-mail address: dennis_helder@sdstate.edu (D. Helder).
0034-4257/$ - see front matter D 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2003.04.002
70
Verification and validation analyses involved the assessment of geolocation, spatial resolution, radiometric
response functions, and the creation of a digital elevation
model (DEM) using multiple IKONOS data sets prepared
by Space Imaging. The geometric evaluation sites at
Brookings, SD, the EDC near Sioux Falls, SD, and the
12 NIMA sites distributed worldwide comprise the test
site locations where the registration evaluations were
performed.
The Space Imaging Carterra commercial products and
their relevant registration accuracies are:
Product
Accuracy
Geo
Reference
Precision
Accuracy
Original Standard
Original Precision
Master Standard
Master Precision
Band
Spectral bandwidth
(Am)
Spatial resolution
(m)
PAN
Blue
Green
Red
Near IR
0.45 0.90
0.45 0.52
0.52 0.60
0.63 0.69
0.76 0.90
1
4
4
4
4
71
(Strange, 2000). This was accomplished using NGS-supplied software for horizontal time dependent position
(HTDP) giving correction factors near scene center of
+ 0.733m UTM X and 0.838m UTM Y.
2.2. EROS Data Center site
The USGS/EROS Data Center is located about 15 miles
northeast of Sioux Falls, SD (Fig. 1). The area around the
center is mostly gently rolling farmland, planted with corn,
soybeans, or left fallow for grazing, and intersected by a
rectangular system of straight, unpaved roads. The roads are
a result of the Public Lands Surveys, typically of 1-mile2
sections, used to establish land ownership in this region.
Over an approximately 10 by 10 km2 area suitable for a
single IKONOS scene acquisition, a system of artificially
marked points were established to augment the few existing photo-identifiable man-made features proximal to the
EROS Data Center. Marker panels were constructed and
placed near the road intersections to create a nearly linear
grid of points. The panels were nominally aligned with the
orbital path of the spacecraft such that the long axis
approximated the along-track direction of the satellite.
The middle of the center panel, a 3-ft2, was the registration
point (Fig. 2).
These panel markers provided clearly identifiable points
for which to compare the image pixel locations to the GPS
coordinates. The temporary targeted panels on the EDC site
served as highly accurate GCPs. The EDC team provided
the Stennis GRIT crew with two National Geodetic Survey
Cooperative Base Network (CBN) stations (PIDs OQ1033
and OQ1245) as the basis for establishing onsite control for
registration of the target ground control points. The simultaneous observations between the two NGS CBN stations
Fig. 1. Location of Brookings and Sioux Falls, South Dakota test sites.
72
Fig. 2. EDC artificial GCP panels. The long, rectangular panel legs, and
spacing were designed for accurate location of the center square.
and the two new onsite control points agreed within 0.020
m. These new control points were simultaneously occupied
and used to locate all panel points, within agreements of
0.050 m. Because the initial control of the field survey was
based on the NAD83 datum, all of the resulting GPS point
locations supplied by the Stennis crew were in the NAD83
datum. However, the coordinate system of the images
supplied by Space Imaging is based on the WGS84 datum.
The NGS software was used to compute the required datum
transformation from WGS84 to NAD83. For a point near
the center of the scene acquired over EDC, these values
were computed to be: + 0.756m UTM X, 0.808 m UTM
Y. All data points were transformed to NAD83 prior to
calculation of RMSE.
2.3. NIMA sites
The Precision Engagement Divisions Office of Targeting
and Navigation (PTNT) at NIMA ordered IKONOS stereo
pairs over Test and Evaluation (T&E) sites. The T&E
feature coordinates provide highly accurate photogrammetric control data for several locations worldwide for use in
metric assessment and calibration activities for National
Technical Means. The control data are derived from GPS
ground truth surveys. Each T&E site contains at least 10
distinct photo-identifiable features. Test and Evaluation
graphics depict each of the cluster points along with their
associated geodetic coordinates and accuracies.
3. Procedures
3.1. Brookings/SDSU
During the summer of 2001, four IKONOS images were
collected of the Brookings site. All imagery was the Space
Imaging Standard Original product, i.e. no ground control
Date
Azimuth (j)
Elevation (j)
July 3
July 17
July 25
August 13
281.85
289.94
269.50
353.76
76.19
60.80
73.20
82.54
Table 2
Geometric accuracy results for all four dates, panchromatic band,
Brookings site
RMSEx
RMSEy
RMSEmin/RMSEmax
RMSE vector
Accuracy
Circular standard error
(39%)
Circular standard error
probable (50%)
Circular map accuracy
standard (90%)
July 3
July 17
July 25
August 13
3.458
0.804
0.233
3.551
5.217
2.131
15.663
1.774
0.113
15.763
21.341
8.719
12.755
1.288
0.101
12.820
17.186
7.021
2.681
1.693
0.632
3.171
5.353
2.187
2.509
10.265
8.267
2.575
5.388
23.921
19.454
4.812
73
Fig. 3. Displacement vectors indicating GCP errors for the panchromatic band on July 3.
3.2. USGS/EDC
Both the Original Standard and Original Precision products were initially evaluated by the EDC using the same
image acquisition. The 11-bit image of each Space Imaging
panchromatic product contained a great deal of information
at the high end of the dynamic range, enabling distinction
between bright objects such as graveled roads, concrete, and
metal roofs, as well as the Tyvek target material. In normal
Fig. 4. Displacement vectors indicating GCP errors for the panchromatic band on July 17.
74
Table 3
Geometric accuracy results for all four dates, panchromatic band,
Brookings site, after terrain correction
RMSEx
RMSEy
RMSEmin/RMSEmax
RMSE vector
Accuracy
Circular standard error
(39%)
Circular standard error
probable (50%)
Circular map accuracy
standard (90%)
July 3
July 17
July 25
August 13
4.011
0.879
0.219
4.107
5.986
2.445
9.249
2.370
0.256
9.548
14.219
5.809
6.191
3.063
0.495
6.907
11.326
4.627
4.281
1.842
0.430
4.661
7.494
3.062
2.879
6.840
5.448
3.605
6.232
14.488
10.482
7.073
5.885 m
3.819 m
0.649 m
7.015 m
11.876 m
4.852 m
5.712 m
10.646 m
4. Results
4.1. Brookings/SDSU
Table 2 shows the accuracy results for the panchromatic
band on all four dates. For each date, the observer whose
75
RMSE vector result was the median value (third largest for
the case of six observers) is reported. All other values were
within F 0.5 m. These results indicate excellent geometric
accuracy in the northing direction, down to the 1-m level for
the July 3 date. However, in the easting direction, accuracies
are significantly larger on two dates. For the July 3 and
August 13 dates, easting RMSEs are at the 3-m level.
However, on the July 17 and July 25 dates, easting RMSEs
increase dramatically to 13 m and more. It is clear that there
is a strong correlation between the RMSE vector magnitude
and the satellite elevation angle. Higher elevation angles give
smaller RMSE vector magnitudes. When RMSEx does not
equal RMSEy, geolocational errors are not circular. Since,
from a practical point of view, this will rarely occur in
practice, a threshold was established for determining if errors
are indeed circular. Established by the NSSDA, a threshold
of RMSEmin/RMSEmax > 0.6 is used as an indicator of
circularity, where RMSEmin is the smaller of RMSEx or
RMSEy, and RMSEmax is the larger of the two. For the
present results, the ratio of RMSEmin/RMSEmax is only
greater than 0.6 on August 13 indicating that this is the only
date where the errors can be considered circular in nature.
Even though these ratios do not satisfy the criteria in FGDCSTD-007.3 Appendix A, circular errors were calculated for
the sake of completeness using the methods outlined in this
appendix. On the July 3 and August 13 dates CE90 hovers
around 5 m. However, on the other two dates, values of 20 m
or more were obtained. These results indicate the presence of
a systematic error. But, it should be noted that all CE90
values are still within specification for this data product.
Vector representation of these results (panchromatic
band) for each GCP is shown for the July 3 and 17 images
Fig. 6. Displacement vectors of the Standard Original product of the EDC scene.
76
Table 5
Three observers of the EDC Precision Original product
NSSDA: RMSEx
NSSDA: RMSEy
NSSDA: RMSEmin/RMSEmax
NSSDA: RMSEr
NSSDA: accuracy
(95% confidence)
NMAS: circular standard
error (39%)
NMAS: circular error
probable (50%)
NMAS: circular map accuracy
standard (90%)
Table 6
Height adjusted values of EDC Precision Original product
Obs1 (m)
Obs2 (m)
Obs3 (m)
3.297
0.750
0.227
3.381
4.952
3.235
0.794
0.245
3.331
4.930
3.292
0.722
0.219
3.371
4.914
2.023
2.014
2.007
2.382
2.372
2.364
5.130
5.055
5.115
NSSDA: RMSEx
NSSDA: RMSEy
NSSDA: RMSEmin/RMSEmax
NSSDA: RMSEr
NSSDA: accuracy
(95% confidence)
NMAS: circular standard
error (39%)
NMAS: circular error
probable (50%)
NMAS: circular map accuracy
standard (90%)
Obs1 (m)
Obs2 (m)
Obs3 (m)
1.930
0.605
0.313
2.023
3.103
1.996
0.636
0.319
2.095
3.221
1.930
0.585
0.303
2.017
3.079
1.268
1.316
1.258
1.493
1.549
1.481
3.070
3.179
3.061
Fig. 7. Displacement vectors of EDC scene in the Precision Original product with terrain effects.
77
Fig. 8. Terrain corrected displacement vectors of EDC scene in the Precision Original product.
4.2. USGS/EDC
The equations in FDGC-STD-007.3 Appendix A were
applied to the observers tagging of the paneled points in the
panchromatic image. One experienced observers results for
the Standard Original product are shown in Table 4. Notice
the ratio of the root-mean-square errors, RMSEmin/RMSEmax
is greater than 0.6 indicating, as mentioned above, the
suitability of the data for approximation of circular errors.
The drift pattern for this image is show in Fig. 6.
For the EDC scene processed to a Precision Original
product, three observers were used. The first two observers
had no prior knowledge of the experimental procedures, and
after a brief introduction, were left alone to generate their
results. The third observer was the same individual who
evaluated the Standard Original image in its delivered form.
Table 5 shows their individual findings. Notice the remarkTable 7
Results for stereo pairs provided in RPC00A format
Stereo pair accuracy
Antananarivo, MA
Miami, FL
Sioux City, IA
Fallon, NV
Abu Musa, TC
Hickam AFB, HI
St. Simons Is., GA
Villa Delores, AR
Utapao, TH
Acquisition date
HA (CE90)
(m)
VA (LE90)
(m)
29 April 2000
28 May 2000
28 May 2000
27 April 2000
3 October 2000
15 May 2000
3 June 2000
23 September 2000
16 May 2000
6.27
2.57
4.60
7.98
10.32
10.95
11.15
7.08
8.92
8.58
8.21
15.19
2.00
3.28
16.11
5.90
5.00
13.39
Abu Musa, TC, and Villa Delores, AR, repeats in both tables having
different RPC tags.
ably high agreement of RMSE components. These components are of the most significance because they are used to
derive all other statistical statements. The maximum difference between observers is expressed by the differences in
the RMSEr statistic (between observers 1 and 2) as (3.381
3.331 m) = 0.050 m. This is believed to be a direct consequence of using the pre-marked panels of known geometric
and spectral characteristics. Also notice the difference of the
Precision Original product from the Standard Original
products ratio of RMSE. Here, in all three cases, it falls
below the guidelines for FGDC case 2 by being less than the
threshold of 0.6. This raises a flag: errors may not be
normally distributed, or independent in x and y components.
Despite the indicated unsuitability of these data for generating further circular error statistics as per the FGDC
standard, circular error statistics were generated for consistency in reporting. The implications of this are noticeable by
looking at the drift patterns (Fig. 7). The drifts all tend to be
horizontal in nature. In other words, these errors may be
systematic.
It was not until after these results were reported at the
first JACIE Workshop on Commercial High-Resolution
Table 8
Results for stereo pairs provided in RPC00B format
Stereo pair accuracy
Abu Musa, TC
Villa Delores, AR
Christchurch, NZ
Keflavik, IC
Sunnyvale, CA
Acquisition date
HA (CE90)
(m)
VA (LE90)
(m)
3 October 2000
23 September 2000
27 April 2000
13 Mar 2001
17 April 2001
10.09
6.79
6.18
8.35
9.11
2.69
5.36
9.80
3.79
6.16
78
Table 9
Summary of accuracies for stereo products
Stereo RPC00A
Horizontal
accuracy (m)
Vertical
accuracy (m)
(n = 9)
Maximum
difference
Minimum
difference
Mean
11.15
16.11
2.57
2.00
7.76
8.63
Maximum
difference
Minimum
difference
Mean
Stereo RPC00B
Horizontal
accuracy
(n = 5)
4.3. NIMA
Maximum
difference
Minimum
difference
Mean
10.09
9.80
6.18
2.69
8.10
5.56
Maximum
difference
Minimum
difference
Mean
Vertical
accuracy
5. Conclusions
In all cases reported, the Space Imaging products met
specifications. Excellent geometric accuracy can be obtained with the Standard Original producton the order
of 5 7 m RMSE for relatively flat terrain. There appears to
be a relationship between satellite elevation angle and
geometric accuracy. It was found that higher elevation
angles tend to result in lower RMSE. The relationship
suggests that satellite elevation angles above 75j tend to
Fig. 9. Stereo pair circular error (CE) and linear error (LE).
79
References
Baltsavias, E., Pateraki, M., & Zhang, L. (2001). Radiometric and geometric evaluation of IKONOS GEO images and their use for 3D building modeling. Proc. Joint ISPRS Workshop High Resolution Mapping
from Space 2001, Hannover, 19 21 September 2001. Pages on CDROM.
Davis, C. H., & Wang, W. (2001). Planimetric accuracy of IKONOS 1-m
panchromatic image products. Proc. ASPRS Annual Conference, St.
Louis, 23 27 April 2001 (p. 14), On CD-ROM.
Dial, G. (2000). IKONOS satellite mapping accuracy. Proc. ASPRS
Annual Conference, Washington, DC, 22 26 May 2000 (p. 8), On
CD-ROM.
Fraser, C. S., Hanley, H. B., & Yamakawa, T. (2001). Sub-metre geopositioning with IKONOS GEO imagery. Proc. Joint ISPRS Workshop High
Resolution Mapping from Space 2001, Hannover, 19 21 September
2001 (p. 8), On CD-ROM.
Geospatial Positioning Accuracy Standards Part 3: National Standard for
Spatial Data Accuracy, FGDC-STD-007.3-1998.
Gerlach, F. (2000). Characteristics of Space Imagings one-meter resolution
satellite imagery products. International Archives of Photogrammetry
and Remote Sensing, vol. 33 (Part B1) (pp. 128 135).
Hanley, H. B., & Fraser, C. S. (2001). Geopositioning accuracy of IKONOS imagery: Indications from 2D transformations. Photogrammetric
Record, 17(98), 317 329.
Strange, W. (2000 (October)). Do you really have WGS 84 coordinates?
Professional Surveyor, 20(9), 37 38.
Abstract
The Space Imaging IKONOS observatory may provide an important benefit in terrestrial scientific research. The five-band, 1 m
panchromatic and 4 m multispectral measurements have the potential to provide a source of measurements to evaluate subpixel land cover
variability in measurements from observatories such as Landsat 7 Enhanced Thematic Mapper Plus (ETM+) and Terra MODIS sensor. The
IKONOS observations are at a spatial scale equivalent to field measurements typically carried out in ecological and land cover research. As
such, the IKONOS observations may serve as a source of virtual ground measurements, for the lower spatial resolution, global
observatories.
In this study we examine how well IKONOS sensor observations replicate Landsat 7 ETM+ visible/near infrared observations for selected
Earth Observation System (EOS) validation sites in the United States. The sites examinedBeltsville, MD, Konza Prairie, KS, and Sevilleta,
NMsample the east west moisture gradient across the United States. Observations for each site were acquired, as nearly time-coincident as
possible, from ETM+ and the IKONOS sensor, several times over the growing season. This was done to insure that we compared these two
sensors over the widest range of observing conditions possible.
We also examined IKONOS imagery from Phoenix, AZ, where Space Imaging had and had not applied a modulation transfer function
compensation (MTFC) process. The MTFC is their standard product. We found that this product, at the original 4 m spatial resolution,
appears to have minor radiometric artifacts as a result of the process. When the IKONOS observations were aggregated to 30 m, this problem
was essentially absent, allowing us to proceed with the remainder of our study.
We processed the IKONOS sensor and ETM+ measurements to produce close approximates of each other. Our processing steps included
ortho-rectification, calibration to planetary reflectance, pixel alignment and pixel aggregation. We initially found radiometric differences
between the two sensors that increased with increasing wavelength. Space Imaging updated their calibration information, based on analyses
from NASA Stennis Space Center staff, which removed much of this discrepancy. We now find that the IKONOS red and near infrared
measurements differ between the two sensors, with IKONOS generally producing higher reflectance in the red band and lower reflectance in
the near infrared band than the Landsat 7 ETM+ sensor. This results in the IKONOS sensor producing lower spectral vegetation index
measurements, for the same target, than ETM+, a measurement variation that has been observed between other sensors.
We also encountered far more cirrus cloud (and shadow) contamination in these paired observations that we had expected. After careful
initial selection, we lost over half of our image pairs from the analysis because of cirrus cloud contamination. We do not know whether this is
simply because of the paired, comparative design of this study or whether it relates to the increased spatial and radiometric resolution of the
IKONOS sensor.
The results of this study not only provide a baseline assessment of IKONOS versus Landsat 7 ETM+ visible and near infrared
measurements but also suggest some of the issues that need more attention when comparing other sensor systems as well as developing the
design of future land observatories.
D 2003 Elsevier Inc. All rights reserved.
Keywords: Landsat 7; IKONOS; ETM+
1. Introduction
The value of multi-scale land imaging is well understood
(Colwell, 1960; Reeves, 1975). Examining land heterogeneity at spatial resolutions between a meter and a kilometer
reveals many of the artifacts that originate from natural
processes and human activities (Hall, Strebel, & Sellers,
1988; Miller, 1978; Woodcock & Strahler, 1987). For example, such multi-scale observations will permit relating regional and global trends in vegetation growth to local scale
processes such as fire and forest clearing by local human
populations (Kasischke, French, & Bourgeau-Chavez, 1999).
Such observations are now available from spacecraftdeployed, digital sensors. The recent innovation of high
spatial resolution observations from the Space Imaging
IKONOS1 sensor, combined with imagery from sensors such
as Moderate Resolution Imaging Spectrometer (MODIS) on
the NASA Terra platform and the Enhanced Thematic
Mapper Plus (ETM+) onboard Landsat 7, provides multiscale measurements from 1 m to 1 km.
Unlike previous satellite land observatories, which have
predominately been government funded, the recent high
spatial resolution systems, such as Space Imaging s IKONOS satellite and Digital Globes Quickbird, are commercially supported observatories. This provides a novel
opportunity for US federal agencies such as NASA, to explore the use of commercial vendors to support their scientific
goals. To explore this potential, NASAs Earth Science
Enterprise conducted a Scientific Data Purchase that included
IKONOS observations from Space Imaging. This study was
undertaken in support of this evaluation process. That is: do
the commercially supplied IKONOS sensor observations
compliment and compare favorably with current NASAfunded observatories such as the ETM+ on Landsat 7?
81
2. Approach
Over the last half century, terrestrial applications of
remote sensing have shifted from qualitative, visual analysis
(Colwell, 1960) to quantitative, spectro-radiometric evaluations (Swain & Davis, 1978). Derivation of similar quantitative spectro-radiometric measurements from two or more
different sensors can be difficult. When multiple sensors are
used to examine land characteristics, a number of uncertainties can be introduced into the analysis problem (Duggin
& Philipson, 1985; Pease & Pease, 1972; Williams, 1991).
Factors such as time of year, time of day, solar illumination
conditions, viewing angle, spectral band passes, and radiometric precision can and do vary between comparative
observations. Typically, instrument observations cannot be
compared at the simple digital number (DN) level but
require careful attention to instrument characterization and
1
Despite reviewer concerns, Space Imaging states that IKONOS rather
than Ikonos is the correct form.
IKONOS spectral
range (nm)
Landsat spectral
range (nm)
(1)
(2)
(3)
(4)
445 516
506 595
632 698
757 853
450 515
525 605
630 690
775 900
Blue
Green
Red
NIR
82
Fig. 1. Study site locations relative to MODIS June 9 25, 2000 spectral vegetation index measurements (Normalized Difference Vegetation Index).
Table 2
Scene pairs examined at primary and secondary study sitesa
Scene pair
Abbreviation
IK date
ETM+
date
Days
apart
IK
VAA
(j)
ETM+
VAA
(j)
IK
VZA
(j)
ETM+
VZA
(j)
IK
SAA
(j)
ETM+
SAA
(j)
IK
SZA
(j)
ETM+
SZA
(j)
Vegetation
type
Pair
accepted
Beltsville 1
Beltsville 2
Beltsville 3
Konza 1
Konza 2
Konza 3
Sevilleta 1
Sevilleta 2
Sevilleta 3
Sevilleta 4
Sevilleta 5
Phoenix 1
BEL1
BEL2
BEL3
KON1
KON2
KON3
SEV1
SEV2
SEV3
SEV4
SEV5
PHO1
06/03/00
09/29/00
09/10/00
05/11/00
07/16/00
09/15/00
05/12/00
07/23/00
09/18/00
10/02/00
12/04/00
10/12/99
06/12/00
10/02/00
10/02/00
06/07/00
07/09/00
09/11/00
05/09/00
07/28/00
09/14/00
09/30/00
12/03/00
10/10/99
9
4
23
27
7
4
3
5
4
3
1
3
339.59
333.39
264.03
68.51
151.25
268.25
104.28
258.94
76.70
19.32
137.72
27.82
98.2
98.2
98.2
98.2
98.2
98.2
278.2
278.2
278.2
278.2
278.2
98.2
22.46
15.96
25.97
24.68
19.69
28.18
20.04
27.32
23.46
15.00
14.67
29.28
0.13
0.13
0.13
1.13
1.13
1.13
2.27
2.27
2.27
2.27
2.27
1.80
130.1
157.2
153.7
129.5
124.5
156.2
120.7
124.2
143.4
152.4
159.8
148.5
123.1
152.0
152.0
124.5
122.5
144.9
124.9
119.0
141.6
148.0
157.4
150.7
66.99
46.02
53.03
61.08
63.25
51.46
63.56
67.45
51.57
48.33
30.76
44.06
65.39
43.43
43.43
65.30
63.81
50.14
63.83
62.74
52.23
47.38
29.74
45.83
Forest, crops
Forest, crops
Forest, crops
Grass, crops
Grass, crops
Grass, crops
Semi-desert
Semi-desert
Semi-desert
Semi-desert
Semi-desert
Urban
No
Yes
No
No
No
Yes
No
No
Yes
No
Yes
Yes
Scene pair includes both an IKONOS and Landsat image. Dates are for acquisition; time of day is always late morning. VAA is the sensor viewing
azimuth angle, VZA is the sensor viewing zenith angle, and SZA is the solar zenith angle. Azimuths are solar angles at time of acquisition; sensor azimuths not
listed here. Pair accepted indicates whether scene pair was used in final study (Yes) or was disqualified (No).
83
Fig. 2. Scatter plots at (a) 30 m (b) 60 m, (c) 90 m and (d) 120 m of IKONOS versus ETM+ measurements from Konza Prairie site for the green spectral band.
84
3. Methods
Effective comparison of the images required several
processing steps noted in Table 3. These steps were carried
out to insure that the measurements from each sensor were
as closely comparable as possible.
3.1. Ortho-rectification
Examining the relative performance of IKONOS and
ETM+ requires, as much as possible, that individual compared pixels are for the same location on the Earths surface.
One of the largest potential sources of error is relief displacement caused by the two sensors viewing the same variable
elevation terrain from two different look angles (Bernstein,
1983; Slama, 1980). We found, particularly for the Konza
Prairie site, that relief displacement created significant registration problems. Note that, on average, the difference in view
zenith angle between the two sensors was approximately 25j
(Table 2). To address this problem, we subjected our primary
site observations to consistent ortho-rectification procedures.
Geographic reference points were extracted by locating
clearly observed points in the Landsat image, IKONOS
image, and US Geological Survey digital orthoquadrangles
(DOQs) (USGS, 1996). The US Geological Survey 1:24,000
DEM data, available for the United States, provided the
terrain information (USGS, 1993) and ortho-rectification
was carried out for both sensor data sets using the PCI
Geomatics OrthoEngineR IKONOS and Landsat 7 Models
(Lee & Bethel, 2001). Nearest neighbor re-sampling was
employed in the ortho-rectification process to preserve as
much as possible the radiometry of the imagery. Some
residual misalignment between Landsat and IKONOS is still
visibly evident in the resultant products, probably the result of
the differing spatial resolution of the IKONOS sensor and
Landsat 7 ETM+ instrument.
3.2. Spatial aggregation
Comparison of the observations radiometry required
aggregation of the IKONOS measurements to spatial resolution equivalent to the Landsat spectral measurements.
Numerous aggregation methods have been used in remote
Table 3
Processing methods
Process
Goal
Topographic ortho-rectification
Masking clouds and cloud shadow
Radiometric calibration
Pixel aggregation
Pixel-to-pixel registration
View of same target
Equivalent physical units
Measurement from same
ground area.
85
the IKONOS and ETM+ observations would closely approximate each other at 60 m. Other factors such as variable
atmospheric attenuation and scene differences would come
into play beyond this aggregation level.
3.3. Cloud exclusion
Some scenes we selected for analysis had clearly visible
cloud and cloud shadow, typical when cumulus-type clouds
are present. Pixels from such targets, which are not shared
across the images pairs, confuse comparative analysis and
therefore needed to be excluded. We visually identified
these features in the images and produced masks to exclude
them from analysis. This task was harder than we had
anticipated, requiring reassessment several times before
we were satisfied with the results. This masking effort did
not extend to our discovery of considerable cirrus cloud
contamination in half of the scene pairs we had originally
selected for analysis. There was simply too much uncertainty in how we might approach this problem. We simply
rejected such image pairs from further analysis.
3.4. Radiometric calibration
Although digital remotely sensed data are widely distributed and analyzed in the form of digital numbers (DN), DNs
are not suitable for comparative analysis, particularly when
comparing the output from two sensors. The IKONOS/
ETM+ is a case in point. For the ETM+ sensor, the DNs
are the same values that were produced by the A-to-D
converter onboard the satellite. For IKONOS observations,
the DNs made available to the public are produced as the
result of internal Space Imaging pre-processing. Space Imaging re-maps the original instrument DN values so that the
calibration-offset term is removed from the consumer-supplied values. The ETM+ DN values are not comparable to the
IKONOS DN values. In general, the use of DNs for comparative analysis purposes is not valid and can produce to
seriously misleading conclusions (Goward, Dye, Turner, &
Yang, 1993). Conversion to calibrated physical units is
required.
3.4.1. Spectral radiance
Conversion to spectral radiance is a substantial improvement over use of DNs in comparative analysis. When transformed, all individual sensor measurements are in comparable physical units. This is generally accomplished through
information supplied by the instrument developer in the form:
Lk gain DN offset
1
2
86
Table 4
ETM+ spectral radiance range (W/m2-sr-Am)
Band
number
Low gain
High gain
LMIN
LMAX
LMIN
LMAX
1
2
3
4
6.2
6.4
5.0
5.1
293.7
300.9
234.4
241.1
6.2
6.4
5.0
5.1
191.6
196.5
152.9
157.4
Handbook (LPSO, 1998). Space Imaging supplies equivalent information for the IKONOS sensor (Peterson, Gerlach,
& Hutchins, 2001).
There are limitations with use of spectral radiance measurements for comparative purposes. If two sensors do not
Fig. 3. Relative spectral response functions for Landsat 7 ETM+ and IKONOS (a) blue spectral band, (b) green spectral band, (c) red spectral band and (d) near
infrared spectral band. Note the differences particularly for the red and near infrared bands.
pLk d 2
ESUNk coshs
87
Table 6
IKONOS bandwidth characteristics ( 016, 1999)
Band
Pan
MS-1
MS-2
MS-3
MS-4
(Blue)
(Green)
(Red)
(VNIR)
Lower
50% (nm)
Upper
50% (nm)
Bandwidth
(nm)
Center
(nm)
525.8
444.7
506.4
631.9
757.3
928.5
516.0
595.0
697.7
852.7
403
71.3
88.6
65.8
95.4
727.1
480.3
550.7
664.8
805.0
Originala
CalCoefk,
DN [mW/
cm2-sr] 1
Pre 2/22/01b
CalCoefk,
DN [mW/
cm2-sr] 1
Post 2/22/01b
CalCoefk,
DN [mW/
cm2-sr] 1
Full-scale
dynamic
range (mW/
cm2-sr)
MS-1
(Blue)
MS-2
(Green)
MS-3
(Red)
MS-4
(VNIR)
637
633
728
2.98
573
649
727
3.32
663
840
949
2.87
503
746
843
3.75
(3) The pre and post CalCoefk values are to be used respectively for images
that have image production (creation) dates before and after February 22,
2001. The original CalCoefk values for the blue, green, red, and visible near
infrared are no longer used for calibration.
a
Space Imaging Document Number SE-REF-016, N/C.
b
Space Imaging Document Number SE-REF-016, Rev. A.
for visible-near infrared spectral measurements for monitoring of vegetation (Deering, Rouse, Haas, & Schell, 1975;
Huete, 1988; Jackson, 1983; Kaufman & Tanre, 1992; Kauth
& Thomas, 1976; Wiegand & Richardson, 1984). The purpose of these indices is to compensate for variable background (e.g. soil and litter) reflectance and some forms of
atmospheric attenuation, while emphasizing vegetation spectral features (Trishchenko et al., 2002).
In this study, we examined three alternative SVIs that
combine the spectral measurements in differing algebraic
forms:
1. The simple ratio (SR, Eq. (6)) originally used by Jordan
(1969) to examine the ratio of light at 800 Am to that at 675
Am to measure leaf area index underneath forest canopies.
SR
qnir
qvis
qnir qvis
qnir qvis
Landsat 7
IKONOS
Blue
Green
Red
NIR
1970
1843
1555
1047
1939
1847
1536
1148
88
Fig. 4. Scatter plots of blue (a), green (b), red (c) and near infrared (d) planetary reflectance, from 4 m measurements with MTFC-on versus MTFC-off for the
October 12, 1999 scene. Note the reflectance bias introduced by brightening the bright pixels and darkening the dark pixels.
89
Fig. 5. Scatter plots of NDVI (a), SR (b) and ARVI (c) spectral vegetation indices from 4 m measurements with MTFC-on versus MTFC-off for the October 12,
1999 scene. Note the outlier measurements in the MTFC-on data set.
Target
IKONOS
Landsat
Forest
8.2 4.2 1.95
Forest
9.1 5.9 1.54
Dark
12.9 11.3 1.14
vegetation
Sevilleta, NM (Sept.) Shadow
9.5 5.2 1.83
Sevilleta, NM (Dec.) Shadow
10.0 4.4 2.27
9.0 4.1
9.8 5.6
11.8 9.6
2.20
1.75
1.23
8.3 4.5
9.1 4.3
1.84
2.11
90
Fig. 6. Scatter plots of blue (a), green (b), red (c) and near infrared (d) planetary reflectance measurements from aggregated 30 m measurements with MTFC-on
versus MTFC-off for the October 12, 1999 scene. Note that minor residual differences are still present at this nearly factor of 10 averaging.
Fig. 7. Scatter plots of NDVI (a), SR (b) and ARVI (c) spectral vegetation indices from the aggregated 30 m measurements with MTFC-on versus MTFC-off
for the October 12, 1999 scene.
91
Fig. 8. Sevilleta, NM May 2000 scene pair revealing cirrus cloud presence in the IKONOS image. Obvious cloud contamination is present in the upper left corner
of the IKONOS image (a) versus no cloud contamination in the ETM+ image (b). A ratio of the IKONOS scene/ETM+ scene (c) reveals further cloud (cirrus) and
shadow contamination in the IKONOS scene stretching diagonally across the center of the IKONOS image. The lower right corner is mountain and its shadows.
92
S.N. Goward et al. / Remote Sensing of Environment 88 (2003) 8099
Fig. 9. Comparison of IKONOS and ETM+ 120 m aggregations for five study images (Table 2) (a) blue band, (b) green band, (c) red band and (d) near infrared band. Note that the Sevilleta results deviate from the
other sites. Why this is the case is not certain.
93
94
95
Fig. 11. Comparative blue/red spectral band ratios for five image pairs examined in this study. Note that the band ratios vary between 1.2 and 2.3, suggesting
wide variations in atmospheric attenuation between scene pair. However, for any given scene pair, the atmospheric conditions are more closely related.
NDVI measurements from ETM+ versus NDVI measurements from IKONOS. This is the equivalent of f 12%
ground cover (or absorbed PAR) (Goward & Huemmrich,
1992; Huemmrich & Goward, 1992). Similarly, The simple
ratio results indicate that IKONOS proportionately underestimates Landsat measurements by f 28%, approximately
the same as underestimating leaf area index (LAI) by the
same proportion (e.g. Landsat LAI of 3.0 would be 2.2 in the
IKONOS measurements). The ARVI results are a bit more
complicated because of the blue band inclusion. However,
they, in general, confirm the NDVI results and further suggest
problems in the December Sevilleta image pair.
These results confirm that the spectral band passes of
the two sensors differ (Fig. 3). Basically, the IKONOS
measures lower SVI measurements for the same land
surface conditions than ETM+. One interesting conclusion
is that despite the differences noted in the individual
spectral band measurements, particularly between sites,
the spectral vegetation index calculations show considerable consistency between all the scene pairs, across all
sites. This may be important in ultimately explaining the
potential source of the observed variations noted within
specific spectral wavelengths (Fig. 9). For example, perhaps some calibration variation over time has not been yet
identified.
Fig. 10. Scatter plots of modeled IKONOS and ETM+ measurements for the (a) blue, (b) green, (c) red, and (d) near infrared spectral regions based on the
relevant sensor, solar and ground information. These results confirm the proposed impact of the IKONOS relative spectral response functions versus the
Landsat 7 ETM+ RSR functions.
96
97
Fig. 12. Scatter plots of NDVI (a), SR (b) and ARVI (c) of IKONOS and ETM+ aggregated to 120 m for five study sites. Note that in all cases the IKONOS
measurements are lower than the same Landsat 7 ETM+ measurement, but that they vary as a function of the SVI used.
98
99
NASA Goddard Space Flight Center, Code 923, Greenbelt, MD 20771, USA
b
Science Systems and Applications, Incorporated, USA
c
Department of Geography, University of Maryland, College Park, MD, USA
d
Department of Geography, Boston University, Boston, MA, USA
e
USDA Forest Service, USA
f
Forest Science Department, Oregon State University, Corvallis, OR, USA
Received 10 May 2002; received in revised form 26 February 2003; accepted 25 April 2003
Abstract
Phase II of the Scientific Data Purchase (SDP) has provided NASA investigators access to data from four different satellite and airborne
data sources. The Moderate Resolution Imaging Spectrometer (MODIS) land discipline team (MODLAND) sought to utilize these data in
support of land product validation activities with a focus on the EOS Land Validation Core Sites. These sites provide a globally distributed
network of sites where field, aircraft, and satellite data are being collected. As much as possible, uniform data sets useful for validation are
being made available for the core sites. The globally consistent, high-resolution imagery available from IKONOS are being used for their
geolocation accuracy and ability to characterize the landscape at the 1- to 4-m spatial resolution. This paper provides an overview of the
MODIS Land Teams validation strategy to incorporate high-resolution imagery and presents three case studies as examples of the use of
IKONOS data for MODIS land validation activities. We conclude that the globally consistent data from IKONOS, available through NASAs
SDP, have supplied critical validation data sets at a reasonable cost.
D 2003 Published by Elsevier Inc.
Keywords: MODIS; IKONOS; NASA
1. Introduction
The Moderate Resolution Imaging Spectrometer
(MODIS) is a key instrument onboard the EOS Terra
(formerly EOS-AM) satellite, successfully launched in December 1999. An additional MODIS sensor aboard the EOS
Aqua (formerly EOS-PM) platform, launched in May 2002,
now complements the first MODIS sensor. MODIS instrument data are converted on a systematic basis into derived
atmospheric, terrestrial, and oceanic products. Scientists
101
Product suite
Product
DAAC
MOD09
MOD11
MOD43
MOD10
MOD29
MOD13
MOD15
MOD17
MOD12
MOD14
MOD44
Surface reflectance
Surface temperature and emissivity
BRDF/albedo
Snow cover
Sea ice extent
Vegetation indices
LAI/FPAR
Net primary vegetation production
Land cover and change
Thermal anomalies and fire
Vegetation cover conversion/continuous fields
EDC
EDC
EDC
NSIDC
NSIDC
EDC
EDC
EDC
EDC
EDC
EDC
Products related to the case studies presented in this paper are shown in
bold font. The Distributed Active Archive Center (DAAC) column
represents the distribution point, either Land Processes (LP) or National
Snow and Ice Data Center (NSIDC).
Additional details available for LP DAAC (http://edcdaac.usgs.gov/modis/
dataprod.html) and for NSIDC DAAC (http://nsidc.org/).
global change research questions associated with biogeochemical cycling, energy balance, land cover change, and
ecosystem health (Table 1).
Lessons learned from the previous generation of global
land imaging systems indicate that product validation is
critical for accurate and credible product usage (Cihlar,
Chen, & Li, 1997; Justice & Townshend, 1994). The
Committee on Earth Observing Satellites (CEOS) Working
Group on Calibration and Validation (WGCV) defines
validation as the process of assessing, by independent
means, the quality of the data products derived from the
system outputs (Justice et al., 2000). In this context, the
MODLAND validation activities are a means by which
independent field, airborne, and other satellite data are
collected and used to assess the accuracy of MODLAND
products. These validation activities will provide the user
community with quantitative estimates of uncertainty for
MODLAND products.
The objective of this paper is to demonstrate how
NASAs Scientific Data Purchase (SDP) has been used
to support MODIS land product validation. We start with
a description of the EOS Core Sites and the information
gathered to characterize these sites. The NASA SDP
program was utilized to acquire at least one fine-resolution
(1 4 m) IKONOS image per core site. These fine spatial
resolution images complement the high-resolution (15 30
m) Landsat 7 and ASTER data available for the EOS
Core Sites (Morisette, Privette, & Justice, 2002). To
provide specific examples of the use of IKONOS data,
we present three case studies. The case studies demonstrate the utility of fine-resolution data for validation of
moderate spatial resolution products. The MODLAND
products related to these case studies are listed in bold
font in Table 1. Lastly, we discuss the experience of
coordinating commercial data for the needs of a large
science team.
102
Under research activities being carried out at the University of Maryland with the support of the NASA Stennis
Space Center, comparative analysis of observations from the
EOS Landsat 7s Enhanced Thematic Mapper plus (ETM+)
and IKONOS has been conducted (Goward et al., 2004). To
support this Commercial Remote Sensing for Earth Science
(CRESS) activity, the principal investigators contributed to
the development of the EOS Land Validation Core Site
information database to assist with comparative analyses.
This resulted in CRESS staff developing site characterization layers for each of the EOS Core Sites. The layers
include maps and tables of land cover, soils, elevation,
climate, and population, as well as a list of contacts and
in situ measurements taken at each site. Since the core sites
are located globally in a variety of biomes (Morisette et al.,
2002), it was imperative to utilize standardized data sets
with global coverage to generate consistent characterization
layers. Thus, data sources were acquired from scientifically
acceptable data sets subject to: (1) availability in digital
form, and (2) consistent global coverage. The characterization output GIS layers, integrated reports, and JPEG maps
for each core site are available on-line or on CD-ROM
through the Global Land Cover Facility at the University of
Maryland (http://glcf.umiacs.umd.edu/; search under data). A summary of characterization layers is given in Table
2 with values for each layer given for the three sites
presented in the following three case studies.
The first case study, over Senanga, Zambia, demonstrates
the use of paired IKONOS and ETM+ to classify the
landscape into either crown cover or no crown cover.
Table 2
Characterization layers available for each site
Global data layer
Source
Resolution
Measurement at
Maun
Measurement at
Senanga
Measurement at
Sevilleta
Reference
(ETM+ path/row
footprint, major
roads and water
bodies
Elevation
UMD land cover
Various
Scale varies
19.92 latitude,
23.59 longitude
15.86 latitude,
23.34 longitude
34.32 latitude,
106.8 longitude
1 km
1 km
929 m
Wooded grassland
1019 m
Cropland
1925 m
Open shrubland
1 km
Savanna
Open shrubland
1 km
Savanna
Cropland/natural
mosaic
Broadleaf crops
Shrubs
1 km
9%
14%
0.00%
Alluvial subtropicals
Original, varies by
country and unit
Subtropical tropical
podzols
4 persons per square
mile
Calcichromic
cambisols
1 person per
square mile
NA
Jan pr: NA
Jul pr: NA
Jan temp: 25.2jC
Jul temp: 19.48jC
Jan pr: NA
Jul pr: NA
Jan temp: 25jC
Jul temp: 20.14jC
Climate; monthly
means
103
Fig. 1. Example of progression from IKONOS to MODIS for the Senanga IKONOS acquisition. (a) False-color infrared composite of IKONOS bands 4, 3, and
2. (b) Crown/no crown classified map. (c) Percent crown cover training from IKONOS aggregated to 30-m cells. (d) Result of ETM+ characterization of
percent crown cover. (e) ETM+ result aggregated to 250-m MODIS resolution cells. (f) Crown cover map from MODIS data using data from ETM+-aggregated
250-m map. Includes material from Space Imagingn.
104
was true for sites in four separate biomes, where the ranges
(or scales of variation) were between 125 and 500 m.
Assuming that image data can adequately track general
spatial patterns of a biophysical variable, image data can be
used to ask whether there are scales of variation finer than
25 m. This can be determined using IKONOS data. Fig. 3
shows IKONOS imagery degraded from 4 to 512 m and
corresponding semivariograms calculated from that imagery
at a desert core site, the Sevilleta National Wildlife Refuge
in New Mexico. These semivariograms are derived from the
first principal component of the four IKONOS multispectral
bands. As theory suggests (Milne & Cohen, 1999), the
shape of the semivariogram changes with the support (or
pixel size) of the imagery used. In particular, the sill is
reduced as image grain size increases, due to the reduction
of image variance with each successive coarsening of the
imagery. Determination of the range of a semivariogram is
generally considered to be the lag at which 95% of the sill is
reached (Deutsch & Journel, 1998). Using this rule, the
105
Fig. 3. First principal component of a four-band IKONOS image over a 3 3-km area of the Sevilleta National Wildlife Refuge in New Mexico spatially
aggregated from 4 to 512 m (top). Semivariograms calculated from the imagery at each spatial resolution or grain size (bottom). Includes material from Space
Imagingn.
106
Table 3
Semivariogram ranges for the IKONOS image of Fig. 3 spatially
aggregated from a support of 4 m to a support of 512 m
Support (m)
Range (m)
4
8
16
25
32
64
128
256
512
504
527
559
580
600
638
690
756
800
The range of 25-m support was calculated from Landsat ETM+ data.
Fig. 4. Normalized semivariograms for the first principal component and the NDVI of 4-m IKONOS imagery and the NDVI and Wetness of Landsat ETM+
imagery from a 3 3-km area of the Sevilleta National Wildlife Refuge in New Mexico.
107
108
Fig. 7. (a) Color ETM+ RGB image from bands 4, 3, and 2 of a 1 1-km
region of the Maun site. (b) The patch map after image segmentation.
Patches 1, 2, 4, 7, 8, 9, 12, 13, and 15 are savannas. Patches 3, 5, 6, 10, 11,
and 14 are shrubs.
Fig. 9. Mean LAI over patches derived from field measurements and ETM+
data.
7. Conclusion
The validation of global products requires consistent
data with which to compare the global products. A main
concern for validating the relatively coarse-resolution products from the MODLAND team is that of relating field or
ground-based point measurements with the large area
represented by a MODIS pixel. The approach used by the
MODLAND team was to use imagery at intermediate
scales to bridge between the two. The 1-m panchromatic
and 4-m multispectral data from IKONOS, available in a
consistent format and quality throughout the globe, have
provided a complement to the 30-m Landsat 7 ETM+ data.
Without the high-resolution imagery available through the
SDP, the only other option for such data would have been
airborne imagery. Flight planning, digital image acquisition, and post-processing are extremely time-consuming
and expensive. Furthermore, companies or agencies supplying such imagery typically work within a finite range,
thus making it difficult (or impossible) to acquire consistent data over the globally distributed EOS Land Validation
Core Sites. Lastly, there are some regions of the world
where airborne image acquisition is either impossible or
unsafe. In light of these considerations, NASAs investment
in high-resolution imagery available through the SDP has
supplied the EOS Land Validation Core Sites with unique,
globally consistent, critical validation data sets at a reasonable cost.
One of the biggest concerns in the use of IKONOS from
the SDP was the uncertainty in the timing of acquisitions.
The SDP contract with Space Imaging required that we
accept data acquired within F 2 weeks of the requested
acquisition date. Attempting to acquire a scene for a narrow
date window to coincide with fieldwork was nearly impossible. In many instances, tasks required a formal extension
of the original date window specification in order to accept
data acquired outside of that original window. Tasking
priorities and overpass timing were largely unknown to
us. There is no doubt there are competing requests for
IKONOS acquisitions, and cloud cover also has to be taken
into account; however, explicit knowledge of how NASAs
SDP requests fit within Space Imagings tasking priority
system was never known. The uncertainty of acquisition
timing was difficult for those trying to couple IKONOS data
acquisitions with field activities. When relating satellite
spectral data and field measurements, it is desired that
ground data collection be nearly simultaneous with the
satellite overpass. Currently, for EOS platforms and several
other satellites, the EOS program provides a web-based tool
to obtain overpass information (http://earthobservatory.nasa.
gov/MissionControl/overpass.html). Such information for
IKONOS would be helpful to ascertain at least the potential
dates and times for IKONOS acquisitions. Knowledge of
the priority of SDP requests, and the likelihood of actual
collection from a list of possible acquisitions would be very
useful information to have.
109
Acknowledgements
MODIS land validation activities are primarily supported
through the MODIS land discipline team and EOS
validation, with special thanks to C. Justice, J. Privette,
and D. Starr. The MODLAND allocation of SDP resources
was made possible through NASA Headquarters and NASA
Stennis Space Center, with special thanks to Diane Wickland
and Fritz Policelli. Thanks also to NASA Code YS, through
Martha Maiden, for providing the ETM+ data used to scale
between IKONOS and MODIS. Case study 2 and three were
part of the Southern African Regional Science InitiativeSAFARI 2000. Thanks to three anonymous reviewers who
provided helpful suggestions that improved the paper.
References
Burrows, S., Gower, S., Clayton, M., Mackay, D., Ahl, D., Norman, J., &
Diak, G. (2002). Application of geostatistics to characterize LAI from
flux tower to landscape scales using a cyclic sampling design. Ecosystems 5, 667 679.
Campbell, J. C., Burrows, S., Gower, S. T., & Cohen, W. B. (1999). BigFoot: Characterizing land cover, LAI, and NPP at the Landscape Scale
for EOS/MODIS Validation. Field manual 2.1. Environmental Science
Division, vol. 4937. Oak Ridge, TN: Oak Ridge National Laboratory,
103 pp.
Cihlar, J. (1997). GCOS/GTOS plan for terrestrial climate related obser-
110
vations. Version 2. GCOS Report 32 (130 pp.). Geneva: World Meteorological Organisation.
Cihlar, J., Chen, J., & Li, Z. (1997). On the validation of satellite-derived
products for land applications. Canadian Journal of Remote Sensing,
23(4), 381 389.
Cohen, W., Spies, T., & Bradshaw, G. (1990). Semivariograms of digital
imagery for analysis of conifer canopy structure. Remote Sensing of
Environment, 34, 167 178.
Crist, E. P., & Cicone, R. C. (1984). A physically-based transformation of
thematic mapper datathe TM tasseled cap. IEEE Transactions on
Geoscience and Remote Sensing, GE-22, 256 263.
Curran, P. J. (1988). The semivariogram in remote sensing: An introduction. Remote Sensing of Environment, 24, 493 507.
DeFries, R. S., Hansen, M. C., & Townshend, J. R. G. (2000). Global
continuous fields of vegetation characteristics: A linear mixture model
applied to multi-year 8 km AVHRR data. International Journal of
Remote Sensing, 21(6 7), 1389 1414.
Deutsch, C. V., & Journel, A. G. (1998). GSLIB geostatistical software
library and users guide. New York: Oxford University Press, 369 pp.
Dowty, D., Frost, P., Lesolle, P., Midgley, G., Mukelabai, M., Otter, L.,
Privette, J., Ringrose, S., Scholes, B., & Wang, Y. (2000). Summary of
the SAFARI 2000 wet season field campaign along the Kalahari Transect. Earth Observer, 12, 29 34.
ESRI (1998). ArcAtlas. 380 New York Street, Redlands, CA 92373-8100,
USA: ESRI.
GOFC Design Team (1999). A strategy for global observation of
forest cover: Version 1.2, http://www.fao.org/gtos/gofc-gold/docs/
execsum.pdf.
Goward, S. N., David, P. E., Fleming, D., Miller, L., & Townshend, J. R.
(2004). Empirical comparison of Landsat 7 and IKONOS multispectral
measurements for selected earth observation system (EOS) validation
sites. Remote Sensing of Environment, 88, 79 98 (this issue).
Hame, T., Stenberg, P., Andersson, K., Rauste, Y., Kennedy, P., Folving, S.,
& Sarkeala, J. (2001). AVHRR-based forest proportion map of the PanEuropean area. Remote Sensing of Environment, 77, 76 91.
Hansen, M. C., DeFries, R. S., Townshend, J. R. G., & Sohlberg, R. (2000).
Global land cover classification at 1 km spatial resolution using a classification tree approach. International Journal of Remote Sensing,
21(6/7), 1331 1364.
Hansen, M. C., DeFries, R. S., Townshend, J. R. G., Sohlberg, R., Dimiceli, C., & Carroll, M. (2002). Toward an operational MODIS continuous field of percent tree cover algorithm: Examples using AVHRR and
MODIS data. Remote Sensing of Environment, 83(1 2), 303 319.
Justice, C., Belward, A., Morisette, J., Lewis, P., Privette, J., & Baret, F.
(2000). Developments in the validation of satellite sensor products for
the study of land surface. International Journal of Remote Sensing,
21(17), 3383 3390.
Justice, C. O., & Townshend, J. R. G. (1994). Data sets for global remote
sensing: Lessons learnt. International Journal of Remote Sensing,
15(17), 3621 3639.
Justice, C., Townshend, J., Vermote, E., Masuoka, E., Wolfe, R., Saleous,
N., Roy, D., & Morisette, J. (2002). An overview of MODIS, its data
processing and products for terrestrial science applications. Remote
Sensing of Environment, 83(1 2), 3 15.
Knyazikhin, Y., Martonchik, J. V., Diner, D. J., Myneni, R. B., Verstraete,
M., Pinty, B., & Gorbon, N. (1998). Estimation of vegetation canopy
leaf area index and fraction of absorbed photosynthetically active radiation from atmosphere-corrected MISR data. Journal of Geophysical
Research, 103, 32239 32256.
Knyazikhin, Y., Martonchik, J. V., Myneni, R. B., Diner, D. J., & Runing,
S. (1998). Synergistic algorithm for estimating vegetation canopy leaf
area index and fraction of absorbed photosynthetically active radiation
from MODIS and MISR data. Journal of Geophysical Research, 103,
32257 32275.
Loveland, T. R., Reed, B. C., Brown, J. F., Ohlen, D. O., Zhu, Z., Yang, L.,
& Merchant, J. W. (2000). Development of a global land cover characteristics database and IGBP DISCover from 1 km AVHRR data. International Journal of Remote Sensing, 21(6/7), 1303 1330.
Milne, B. T., & Cohen, W. B. (1990). Multiscale Assessment of Binary
and Continuous Land cover Variables for MODIS Validation Mapping, and Modeling Applications. Remote Sensing of Environment,
70, 82 98.
Morisette, J. T., Privette, J. L., & Justice, C. O. (2002). A framework for the
validation of MODIS land products. Remote Sensing of Environment,
83(1 2), 77 96.
Myneni, R. B., Nemani, R. R., & Running, S. W. (1997). Estimation of
global leaf area index and absorbed par using radiative transfer models. IEEE Transactions on Geoscience and Remote Sensing, 35,
1380 1393.
NASA (2001). Exploring our home planet: Earth Science Enterprise
strategic plan, http://www.earth.nasa.gov/visions/stratplan/
ese_strategic_plan.pdf.
NASA CIDC (1999). Climatology interdisciplinary data center.
NASA. Atmospheric soundings from TOVS, http://daac.gsfc.nasa.
gov/CAMPAIGN_DOCS/FTP_SITE/INT_DIS/readmes/tovs.html.
Privette, J. L., Myneni, R. B., Knyazikhin, Y., Mukufute, M., Robert, G.,
Tian, Y., Wang, Y., & Leblanc, S. G. (2002). Early spatial and temporal
validation of MODIS LAI product in Africa. Remote Sensing of Environment, 83, 232 243.
Rahman, H., & Didieu, G. (1994). SMAC: A simplified method for the
atmospheric correction of satellite measurements in the solar spectrum.
International Journal of Remote Sensing, 15, 123 143.
Shabanov, N. V., Wang, Y., Buermann, W., Dong, J., Hoffman, S., Smith,
G. R., Tian, Y., Knyazikhin, Y., Gower, S. T., & Myneni, R. B. (2003).
Effect of foliage spatial heterogeneity in the MODIS LAI and FPAR
algorithm over broadleaf forests. Remote Sensing of Environment 85(4),
410 423.
Tian, Y., Woodcock, C. E., Wang, Y., Privette, J. L., Shabanov, N. V., Zhou,
L., Zhang, Y., Buermann, W., Dong, J., Veikkanen, B., Hame, T., Andersson, K., Ozdogan, M., Knyazikhin, Y., & Myneni, R. B. (2002a).
Multiscale analysis and validation of the MODIS LAI product over
Maun, Botswana: II. Sampling strategy. Remote Sensing of Environment, 83, 431 441.
Tobler, W., Deichman, U., Gottsegen, J., & Maloy, K. (1995). The global
demography project. UC Santa Barbara: NCGIA.
USGS, United States Geological Survey (1996). GTOPO30: Global 30 Arc
second elevation data set. Sioux Fallo, South Dakota, USA: Land
Processes Distributed Active Archive Centre, http://edcdaac.usgs.gov/
gtopo30/gtopo30.html.
Wang, Y., Woodcock, C. E., Buermann, W., Stenberg, P., Voipio, P., Smolander, H., Hame, T., Tian, Y., Hu, J., Knyazikhin, Y., & Myneni, R. B.
(in review). Validation of the MODIS LAI product in coniferous forest
of Ruokolahti, Finland. Remote Sensing of Environment.
Woodcock, C. E., Strahler, A. H., & Jupp, D. L. (1988). The use of variograms in remote sensing: I. Scene models and simulated images. Remote Sensing of Environment, 25, 323 343.
Institute for the Study of Earth Oceans and Space, University of New Hampshire, Durham, NH 03824, USA
b
Department of Natural Resources, University of New Hampshire, Durham, NH 03824, USA
c
USDA Forest Service, International Institute of Tropical Forestry, Rio Piedras, PR, USA
d
Department of Global Ecology, Carnegie Institution of Washington, Stanford University, Stanford, CA, USA
e
Department of Anthropology, Indiana University, Student Building 130 Bloomington, IN 47405, USA
f
EMBRAPA Amazonia Oriental, Belem, PA 66095-100, Brazil
g
Ecosystem Science and Technology, NASA Ames Research Center, Moffett Field, CA 94035, USA
h
Anthropological Center for Training and Research on Global Environmental Change Indiana University, Bloomington, IN 47405, USA
i
CPTEC/INPE Instituto Nacional de Pesquisas Espaciais, Cachoeira Paulista, SP 12630-000 Brazil
j
Department of Geography, Maxwell School, Syracuse University, Syracuse, NY 13244, USA
k
Biological Sciences Department, California State University, San Marcos, CA 92096, USA
l
Center for Development Research (ZEF), Department of Ecology and Resource Management, University of Bonn, D-53113 Bonn, Germany
Received 19 July 2002; received in revised form 12 September 2002; accepted 22 April 2003
Abstract
The LBA-ECO program is one of several international research components under the Brazilian-led Large Scale Biosphere Atmosphere
Experiment in Amazonia (LBA). The field-oriented research activities of this study are organized along transects and include a set of primary
field sites, where the major objective is to study land-use change and ecosystem dynamics, and a smaller set of 15 operational eddy flux tower
sites, where the major objective is to quantify net exchange of CO2 with the atmosphere. To supplement these studies and help to address
issues of fine-scale spatial heterogeneity and scaling, high-resolution satellite imagery (IKONOS, 1 4 m) have been acquired over some of
these study sites. This paper begins with a description of the acquisition strategy and IKONOS holdings for LBA. This section is followed
with a review of some of the most promising new applications of these data in LBA.
D 2003 Elsevier Inc. All rights reserved.
Keywords: IKONOS; Remote sensing; Spatial heterogeneity; Land use; Land cover; LBA
1. Introduction
Tropical deforestation in Amazonia has been a growing
ecological and climatological concern for several years
(Fearnside, 1990; Nepstad et al., 1999; Skole & Tucker,
1993). In 1997, the Large Scale Biosphere Atmosphere
* Corresponding author. Institute for the Study of Earth Oceans and
Space, University of New Hampshire, Durham, NH 03824, USA. Tel.: +1603-862-4185; fax: +1-603-862-0188.
E-mail address: george.hurtt@unh.edu (G. Hurtt).
0034-4257/$ - see front matter D 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2003.04.004
Experiment in Amazonia (LBA) was initiated as an international research initiative led by Brazil with strong United
States and European Union participation. LBA is designed
to understand the climatological, ecological, biogeochemical, and hydrological functioning of Amazonia, the impact
of land-use change on these functions, and the interactions
between Amazonia and the earth system. The scientific
questions of LBA are: (1) How does Amazonia currently
function as a regional entity? (2) How will change in land
use and climate affect the biological, chemical and physical
functions of Amazonia, including the sustainability of
112
Table 1
Summary statistics on IKONOS holdings for LBA
Total
Total
Total
Total
Total
Total
Final
Beta
Complete
Final Unique
Beta Unique
Unique
f Area (km2)
5268
1540
6808
3761
1491
5252
71
19
90
46
18
64
Beta and Final are the two data versions. Beta is a preliminary data product.
Fig. 1. Locations of IKONOS images for LBA-Ecology study sites acquired. Open stars represent locations of eddy-flux tower sites, and solid stars represent
field study sites. Map-ID labels for each star are used to cross reference site names listed in Appendices A and B.
113
Fig. 2. Examples of IKONOS imagery for two important LBA study regions near the cities of Manaus and Santarem. The images displayed are false color
representations of multispectral (4 m) imagery and, with the exception of Santarem Kilo83 11.14.01, cover a domain size of 7 7 km. Greater detail is
available in larger-scale (i.e. smaller domain) subsets of these images. White patches, which are clouds, can be seen in several images along with cloud
shadows. Standard tasking criteria require images to have less than 10% cloud coverage. This figure includes materialn Space Imaging L.P.
114
Table 2
Cumulative statistics from EOS-WEBSTER (http://www.eos-webster.sr.unh.
edu) pertaining to IKONOS holdings for LBA
Number of investigators/team requests
Number of registered IKONOS users
Number of visitors to IKONOS subsystem
Number of IKONOS products ordered
IKONOS data available to users (GB)
IKONOS data distributed (GB)
19
47
1248
671
16.6
87.6
surface studies. Medium- to coarse-resolution optical sensors (e.g., AVHRR, Terra-1 MODIS, SPOT-4 VEGETATION) provide daily observations of land cover for the
globe. However, the land surface in the Amazon is often a
mixture of land-cover types at the spatial resolution of these
sensors (0.5 1 km). Quantifying the fractional coverage of
broad land-cover types within large pixels by extrapolating
knowledge that is obtainable by visual or automated
interpretation of IKONOS is a remote sensing problem of
current interest to LBA research. All of the approaches
discussed below rely on the development of scaling relationships between the spectral information in regional-scale
data and the spatial information in high resolution data
(IKONOS in this case), a process known as spectral
unmixing.
A number of spectral unmixing approaches have been
developed to estimate subgrid-scale land cover fractions. In
general, the methods are either based either on linear
mixture modeling (Boardman, 1989; Cross, Settle, Drake,
& Paivinen, 1991; Smith, 1990), or on nonlinear regression
using artificial neural networks (e.g. Foody, Lucas, Curran,
& Honzak, 1997). In traditional spectral mixture modeling,
the surface reflectance at each pixel of the image is assumed
to be a linear combination of the reflectance of each endmember present within the pixel. For an image pixel that has
M spectral bands and N end-members, the spectral linear
unmixing model is described in the following equation:
Ri
N
X
rki fk ei
k1
where Ri is the reflectance of spectral band i in a multispectral image (i = 1, 2, . . ., M), fk is the fractional cover of
end-member k (k = 1, 2, . . ., N) within a pixel, rki is the
reflectance of end-member k at spectral band i, and ei is the
residual term. Nonlinear regression takes the converse
approach, directly modeling the fractions as a function of
reflectance fk = Gk(R), where G is normally defined by a
simple back-propagation neural network (Foody et al.,
1997). To avoid overfitting, a Bayesian modification can
be made to these algorithms (Bishop, 1995; Braswell et al.,
2000) so that all the available high-resolution data can be
used in training the model.
Some studies have explored the use of linear spectral
unmixing of AVHRR data for sub-pixel characterization of
cropland (Quarmby, Townshend, Settle, & White, 1992) and
tropical forests (Cross et al., 1991). These studies have had
some success, but have been limited by sensor resolution
and the lack of relevant spectral bands in AVHRR data (only
red and near infrared bands for vegetation). The MODIS
sensor has higher resolution and more spectral bands related
to land cover and vegetation (2 bands at 250 m resolution
plus 5 bands at 500 m resolution). Because of the increased
resolution and added bands, research using MODIS data has
the potential to provide improved estimates of land cover
and land-cover changes.
115
Fig. 3. A comparison of IKONOS, Landsat 7 ETM+ and MODIS images. Upper left panel-IKONOS image acquired on April 6, 2000, false color composite
(band 4-3-2). Lower left panel-Landsat 7 ETM+ image acquired on September 6, 1999, false color composite (band 4-5-3). The white box illustrates the
domain of the IKONOS image in the upper left. Upper right panelMODIS image (8-day composite of surface reflectance, July 20 26, 2000), false color
composite (band 2-6-1). Lower right panelFractional cover of forest within 500 m pixels, derived from spectral mixture analysis of MODIS data. This figure
includes materialn Space Imaging L.P.
116
Fig. 4. Histogram of shadow fractions from 29 IKONOS pan-chromatic images of Amazon tropical forests. Descriptive statistics are provided in upper right.
From Asner and Warner (submitted for publication).
117
118
Fig. 5. (a) Cumulative frequency distributions of crown area detected by various methods. IKONOS_current_paper refers to manual interpretation. VB
Logic Routine refers to automated method. (b d) Automated crown detection algorithm of an IKONOS image. (e) Number of trees vs. DBH class. Analyses
of Tapajos km 83 Tower area. This figure includes materialn Space Imaging L.P.
bance gradient comprised three treatments: areas of oldgrowth (undisturbed) forest, logged forest excluding major
roads and logging decks, and logged forest including major
roads and decks. In all cases the spatial methods were
successful at discriminating between the three treatments.
These results indicate that it may be possible to combine
spatial analyses of higher-resolution datasets over small
areas with coarser resolution datasets over large areas, for
assessing basin-wide fine-scale canopy disturbances such as
those resulting from selective logging.
Using 4 m multispectral IKONOS data, Hagen et al. have
been examining a section of the Tapajos National Forest,
south of Santarem, a Brazilian city in central Amazonia.
This field research area has been separated into 1 km plots,
some of which were selectively logged at various times in
the last f5 years. A spatial analysis of IKONOS data was
conducted in an attempt to automate the discrimination of
logged and intact forests. As an alternative to analyzing the
raw IKONOS panchromatic brightness data, spectral bands
were combined to calculate the normalized vegetation index
119
Fine-scale heterogeneity in land cover, such as is introduced by fine-scale disturbance and selective logging, is
hypothesized to be important to the carbon balance of
ecosystems (Moorcroft, Hurtt, & Pacala, 2001; Nepstad et
al., 1999). Work by Vourlitis and Priante-Filho is expanding
measurements of the net CO2 exchange, evapotranspiration,
and energy balance of intact tropical transitional forest to
Table 3
The number of clusters identified in each of the 19 plots (9 logged and 10
not logged) using the Threshold-Binary Analysis with an optimal threshold
value (T ) of 0.41 and group size (GS) of three pixels
Threshold-binary analysis (T = 0.41, GS> = 3)
Mean
S.D.
Logged
Not logged
4
5
7
3
0
6
4
1
7
0
0
0
1
0
0
0
0
0
1
0.20
0.42
4.11
2.47
The logged plots have more clusters than the not logged plots.
120
Fig. 6. The ratio between NIR/red for April 2000 (top panel) and May 2001 (bottom panel) for the transitional forest site near Sinop, Mato Grosso. The lighter
surfaces (NIR/red = 0 1) show the recently cleared areas while darker areas (NIR/red > 3) show forest. This figure includes materialn Space Imaging L.P.
121
Fig. 7. Multi-scale analysis of Aca palm agroforestry system and floodplain forests in the Amazon estuary. Classified Landsat TM over IKONOS multispectral image (top), aerial view of agroforestry canopy (middle), ground view of intensively managed agroforestry (bottom). This figure includes materialn
Space Imaging L.P.
122
123
Acknowledgements
Much of the material for this paper was organized for
a series of lectures given at the High-Spatial Resolution
Commercial Imagery Workshops held at USGS Headquarters in 2001 and 2002. Support for this research was
provided by NASA through the Terrestrial Ecology, LandUse Land-Cover Change, New Millennium, Interdisciplinary Science, and Scientific Data Purchase Programs. D.
Clark and three anonymous reviewers provided suggestions that greatly improved this manuscript. D. Blaha, M.
Routhier, and S. Spencer provided assistance with
graphics and data management. The IKONOS data
referenced in this paper are hosted and made available
to the scientific community via the NASA Earth Science
Information Partner EOS-WEBSTER. (http://www.eoswebster.sr.unh.edu). This paper includes materialn Space
Imaging L.P.
124
Map
ID
Name
Country
lat nw
lat sw
lat ne
lat se
lon nw
lon sw
lon ne
lon se
area
(km2)
Date
acquired
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
guas Emendadas
A
guas Emendadas
A
guas Emendadas
A
Braslia East
Braslia West
Caxiuana
Caxiuana
Cerrado
Cerrado
Fazenda Nossa Senhora
Fazenda Nossa Senhora
IBGE Campo Sujo
IBGE Campo Sujo
Jaru Tower
Jaru Tower
Jaru Tower
Manaus 1
Manaus 1
Manaus 2
Manaus 2
Santarem Kilo 67
Santarem Kilo 67
Santarem Kilo 83
Santarem Kilo 83
Santarem Kilo 83
Santarem Kilo 83
Santarem Kilo 83
Santarem Pasture
Santarem Pasture
Sinop Mato Grosso
Sinop Mato Grosso
Sinop Mato Grosso
Sugarcane
Sugarcane
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
15.52
15.52
15.50
15.62
15.58
1.72
1.72
21.63
21.63
10.73
10.73
15.92
15.92
10.03
10.05
10.05
2.56
2.56
2.58
2.58
2.83
2.83
2.99
3.01
2.99
3.04
3.05
2.99
2.99
11.38
11.38
11.38
21.07
21.07
15.58
15.58
15.60
15.72
15.68
1.78
1.78
21.70
21.70
10.79
10.79
15.98
15.98
10.13
10.11
10.11
2.62
2.62
2.64
2.64
2.89
2.89
3.05
3.05
3.05
3.07
3.08
3.05
3.05
11.44
11.44
11.44
21.13
21.13
15.52
15.52
15.50
15.62
15.58
1.72
1.72
21.64
21.63
10.73
10.73
15.92
15.92
10.03
10.05
10.05
2.56
2.56
2.58
2.58
2.83
2.83
2.99
3.01
2.99
3.04
3.05
2.99
2.99
11.38
11.38
11.38
21.07
21.07
15.58
15.58
15.60
15.73
15.68
1.78
1.78
21.70
21.70
10.79
10.79
15.98
15.99
10.13
10.11
10.11
2.62
2.62
2.64
2.64
2.89
2.89
3.05
3.05
3.05
3.07
3.08
3.05
3.05
11.44
11.44
11.44
21.13
21.13
47.63
47.63
47.65
47.99
48.08
51.49
51.49
47.65
47.65
62.39
62.39
47.90
47.90
61.98
61.97
61.96
60.15
60.15
60.24
60.24
54.99
54.99
55.00
55.00
55.00
55.01
55.00
54.92
54.92
55.36
55.36
55.36
48.10
48.10
47.63
47.63
47.65
47.99
48.07
51.49
51.49
47.65
47.65
62.39
62.39
47.90
47.90
61.98
61.96
61.96
60.15
60.15
60.24
60.24
54.99
54.99
55.00
55.00
55.00
55.01
55.00
54.92
54.92
55.36
55.36
55.36
48.10
48.10
47.57
47.57
47.54
47.89
47.97
51.42
51.42
47.58
47.58
62.33
62.33
47.84
47.84
61.88
61.90
61.90
60.08
60.08
60.18
60.18
54.93
54.93
54.94
54.94
54.94
54.94
54.94
54.86
54.86
55.29
55.29
55.29
48.03
48.03
47.57
47.57
47.55
47.89
47.97
51.42
51.42
47.58
47.58
62.33
62.33
47.84
47.84
61.88
61.90
61.90
60.08
60.08
60.18
60.18
54.93
54.93
54.94
54.94
54.94
54.94
54.94
54.86
54.86
55.29
55.29
55.29
48.03
48.03
49.06
49.57
123.71
124.10
124.17
49.06
48.83
49.03
50.43
49.03
49.43
49.06
49.64
110.29
49.06
48.43
49.06
49.13
49.03
49.11
49.06
49.03
49.03
33.06
49.03
25.86
23.54
49.00
49.02
49.03
49.73
49.42
49.03
50.80
6/23/2000
6/23/2001
6/1/2001
6/1/2001
6/1/2001
6/7/2000
9/3/2001
5/29/2000
5/29/2001
8/2/2000
8/2/2001
5/29/2000
5/29/2001
4/6/2000
5/17/2000
5/28/2001
7/22/2000
8/2/2001
8/24/2000
10/7/2001
6/21/2000
11/3/2001
8/29/2000
11/14/2001
11/19/2001
8/29/2000
7/5/2002
6/13/2000
7/27/2001
4/30/2000
5/19/2001
7/5/2002
6/12/2000
6/12/2001
Name
Country
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
Altamira East
Altamira West
Alto Paraso 1
Alto Paraso 2
Alto Paraso 3
Apeu
Cangussu
Cauaxi
Cauaxi
Cauaxi C
Cauaxi R (Scene A)
Cauaxi R (Scene B)
Cauaxi R (Scene C)
Ecuador 1 East
Ecuador 1 West
Ecuador 2 East
Ecuador 2 West
Ecuador 3 East
Ecuador 3 West
Ecuador 4
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Ecuador
Ecuador
Ecuador
Ecuador
Ecuador
Ecuador
Ecuador
lat nw
3.17
3.17
9.65
9.42
9.61
1.27
9.95
3.69
3.69
3.61
3.71
3.71
3.71
0.21
0.21
0.19
0.19
0.19
0.19
0.21
lat sw
3.31
3.31
9.73
9.50
9.70
1.34
10.01
3.82
3.82
3.65
3.75
3.75
3.75
0.09
0.09
0.07
0.07
0.31
0.31
0.33
lat ne
3.17
3.17
9.65
9.42
9.61
1.27
9.95
3.69
3.69
3.61
3.71
3.71
3.71
0.21
0.21
0.19
0.19
0.19
0.19
0.21
lat se
lon nw
lon sw
lon ne
lon se
Area
(km2)
Date
acquired
3.31
3.31
9.73
9.50
9.70
1.34
10.01
3.82
3.82
3.65
3.75
3.75
3.75
0.09
0.09
0.07
0.07
0.31
0.31
0.33
52.26
52.35
63.19
63.35
63.29
48.00
50.04
48.33
48.33
48.50
48.43
48.43
48.43
76.84
76.94
76.28
76.30
76.92
76.94
76.60
52.26
52.35
63.19
63.35
63.29
48.01
50.04
48.33
48.33
48.50
48.43
48.43
48.43
76.84
76.94
76.28
76.30
76.92
76.94
76.60
52.17
52.26
63.15
63.30
63.25
47.94
49.98
48.27
48.27
48.39
48.37
48.37
48.37
76.82
76.82
76.18
76.20
76.82
76.91
76.49
52.17
52.26
63.15
63.30
63.25
47.94
49.97
48.27
48.27
48.39
48.37
48.37
48.37
76.82
76.82
76.18
76.20
76.82
76.91
76.49
171.32
162.14
45.03
50.63
50.19
49.06
49.00
91.51
91.51
48.76
28.10
28.10
28.10
23.74
76.97
154.42
152.19
146.04
34.84
160.59
10/14/2000
10/14/2000
7/25/2002
7/14/2002
7/17/2002
7/10/2001
7/7/2002
11/2/2000
6/18/2002
7/7/2002
8/9/2002
8/20/2002
8/23/2002
1/13/2002
1/5/2002
1/10/2002
2/15/2002
10/25/2000
10/25/2000
8/26/2001
125
Appendix B (continued)
Map
ID
Name
Country
lat nw
lat sw
lat ne
lat se
lon nw
lon sw
lon ne
lon se
Area
(km2)
Date
acquired
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
Ecuador 4 East
Ecuador 4 West
Fazenda Nova Vida
Fazenda Vitoria
Igarape-Acu 2
Igarape-Acu 2
Igarape-Acu 2
Igarape-Acu 2
Igarape-Acu 2
Ituqui, Brazil scene A
Ituqui, Brazil scene B
Ituqui, Brazil scene C
Ituqui, Brazil scene D
Ituqui, Brazil scene E
Machadinho
Mato Grosso
Medicilandia
Mil Madeiras
Pastaza
Pilche
Ponta de Pedras East
Ponta de Pedras West
Rohden East
Rohden Northeast
Rohden West
Ruropolis East
Ruropolis West
Santarem FLONA
Tapajos
Sao Francisco do
Para East
Sao Francisco do
Para East
Sao Francisco do
Para West
Sewaya
Tiguano
Tome-Acu East
Tome-Acu West
Zabalo
Ecuador
Ecuador
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Ecuador
Ecuador
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
Brazil
0.21
0.21
10.11
2.93
1.15
1.15
1.15
1.15
1.12
2.51
2.51
2.51
2.51
2.51
9.43
9.57
3.42
2.79
0.07
0.26
0.19
1.34
10.53
10.44
10.53
4.06
4.06
2.99
0.33
0.33
10.17
3.00
1.22
1.22
1.22
1.22
1.21
2.68
2.68
2.68
2.68
2.68
9.54
9.63
3.47
2.89
0.05
0.32
0.31
1.45
10.56
10.52
10.56
4.09
4.09
3.08
0.21
0.21
10.11
2.93
1.15
1.15
1.15
1.15
1.12
2.51
2.51
2.51
2.51
2.51
9.43
9.57
3.42
2.79
0.07
0.26
0.19
1.34
10.53
10.45
10.53
4.06
4.06
2.99
0.33
0.33
10.17
3.00
1.22
1.22
1.22
1.22
1.21
2.68
2.68
2.68
2.68
2.68
9.54
9.63
3.47
2.89
0.05
0.32
0.31
1.45
10.56
10.52
10.56
4.09
4.09
3.08
76.51
76.61
62.84
47.45
47.61
47.61
47.61
47.61
47.62
54.33
54.24
54.39
54.50
54.50
62.11
55.97
52.92
58.84
77.14
76.27
76.92
48.92
58.52
58.47
58.59
54.83
54.85
55.00
76.51
76.61
62.84
47.45
47.61
47.61
47.61
47.61
47.63
54.32
54.24
54.39
54.50
54.50
62.11
55.97
52.92
58.84
77.14
76.27
76.92
48.92
58.52
58.48
58.59
54.83
54.85
55.00
76.49
76.51
62.75
47.37
47.54
47.54
47.54
47.54
47.54
54.22
54.19
54.28
54.37
54.49
62.00
55.90
52.86
58.76
77.07
76.22
76.82
48.82
58.42
58.40
58.50
54.72
54.74
54.94
76.49
76.51
62.75
47.37
47.54
47.54
47.54
47.54
47.54
54.22
54.19
54.28
54.37
54.48
62.00
55.90
52.86
58.76
77.07
76.22
76.82
48.82
58.42
58.40
58.50
54.72
54.74
54.94
34.35
154.75
60.64
74.22
55.00
55.00
55.00
55.00
92.92
211.80
100.94
227.15
261.12
35.40
154.47
49.00
36.02
96.99
105.22
40.63
146.04
138.07
41.63
68.79
35.60
39.90
40.95
76.55
8/26/2001
8/26/2001
8/16/2001
7/29/2001
9/19/2001
10/3/2001
8/31/2002
9/19/2002
10/22/2002
7/27/2001
8/7/2001
8/7/2001
10/6/2001
10/9/2001
5/28/2001
7/5/2002
10/17/2002
7/25/2002
1/16/2002
1/10/2002
12/19/2000
12/19/2000
7/8/2002
10/1/2002
7/8/2002
7/8/2002
7/5/2002
7/5/2002
Brazil
1.07
1.20
1.07
1.20
47.82
47.82
47.72
47.72
175.55
7/18/2001
Brazil
1.07
1.20
1.07
1.20
47.83
47.83
47.72
47.72
191.10
12/30/2001
Brazil
1.07
1.20
1.07
1.20
47.83
47.83
47.80
47.80
53.28
7/18/2001
Ecuador
Ecuador
Brazil
Brazil
Ecuador
0.15
0.42
2.34
2.34
0.19
0.20
0.48
2.46
2.46
0.24
0.15
0.42
2.34
2.34
0.19
0.20
0.48
2.46
2.46
0.24
76.20
76.51
54.21
54.29
75.43
76.20
76.51
54.20
54.29
75.43
76.14
76.45
54.17
54.18
75.37
76.14
76.45
54.17
54.18
75.37
35.86
35.86
49.60
151.69
35.86
2/4/2002
1/13/2002
7/27/2001
7/27/2001
2/9/2002
83
84
85
86
87
88
89
90
References
Araujo, T. M., Higuchi, N., & Carvalho, J. A. (1999). Comparison of
formulae for biomass content determination in a tropical rainforest site
in the state of Para, Brazil. Forest Ecology and Management, 117,
43 52.
Asner, G. P., Keller, M., Pereira, R., & Zweede, J. (2002). Remote sensing
of selective logging in Amazonia: Assessing limitations based on detailed field measurements, Landsat ETM+ and textural analysis. Remote
Sensing of Environment, 80(3), 483 496.
Asner, G. P., Palace, M., Keller, M., Pereira Jr., R., Silva, J. N. M., &
Zweede, J. C. (submitted for publication). Estimating canopy structure
in an Amazon forest from laser rangefinder and IKONOS satellite
observations. Biotropica.
Asner, G. P., & Warner, A. S. (submitted for publication). Canopy shadow
in IKONOS satellite observations of tropical forests and savannas. Remote Sensing of Environment.
Bishop, C. (1995). Neural networks for pattern recognition. New York:
Oxford University Press (504 pp.).
Boardman, J. W. (1989). Inversion of imaging spectrometry data using
singular value decomposition. IEEE International Geoscience and Remote Sensing Symposium, Vancouver, BC ( pp. 2069 2072).
Boardman, J. W. (1992). Sedimentary facies analysis using imaging spectrometry: A geophysical inverse problem. PhD Thesis, University of
Colorado, Boulder, CO, 212 pp.
Braswell, B. H., Linder, E., Hagen, S., Xiao, X., Frolking, S., Moore III, B.,
& Liu, J. (2000). A Bayesian unmixing algorithm for retrieving landcover distributions using global reflectance data. American Geophysical
Union Fall Meeting, San Francisco, CA.
Bradshaw, G. A., & Spies, T. A. (1992). Characterizing canopy gap structure in forests using wavelet analysis. Journal of Ecology, 80, 205 215.
Brienza Junior, S. (1999). Biomass dynamic of fallow vegetation enriched
with leguminous trees in the Eastern Amazon of Brazil. PhD Thesis,
Goettingen Goettinger Beitrage zur Land- und Forstwirtschaft in den
Tropen und Subtropen, 133 pp.
Brondzio, E. S., Moran, E. F., Mausel, P., & Wu, Y. (1994). Land use
change in the Amazon estuary: Patterns of Caboclo settlement and
landscape management. Human Ecology, 22(3), 249 278.
Brondzio, E. S., Moran, E. F., Mausel, P., & Wu, Y. (1996). Changes in
land cover in the Amazon estuary: Integration of thematic mapper with
126
botanical and historical data. Photogrammetric Engineering and Remote Sensing, 62(8), 921 929.
Brown, S. (1997). Estimating biomass and biomass change of tropical
forests: A primer. FAO Forestry Paper 134, United Nations Food and
Agriculture Organization, Rome, Italy.
Cerri, C., Niro, H., Melillo, J., Krug, T., Fernandes, E., Forsberg, B.,
Houghton, R., Keller, M., Martinelli, L., Nepstad, D., Nobre, A., Richey, J., Reynaldo, V., Crill, P., Davidson, E., De Mello, W., Melack, J.,
Mozeto, A., Skole, D., Soares, J. V., Sterberg, L., & Trumbore, S.
(1995). The ecological component of an integrated Amazon study (also
known as LBA): The effects of forest conversion. Workshop Report,
Manaus, Brazil, 55 pp.
Clark, D. B., Read, J. M., Clark, M., Cruz, A. M., Dotti, M. F., & Clark,
D. A. (submitted for publication). Application of 1-m and 4-m resolution satellite data to studies of tree demography, stand structure and
land-use classification in tropical rain forest landscapes. Ecological
Applications.
Cohen, W. B., & Spies, T. A. (1992). Estimating structural attributes of
Douglas-fir/western hemlock forests stands from Landsat and SPOT
imagery. Remote Sensing of the Environment, 41, 1 17.
Cohen, W. B., Spies, T. A., & Bradshaw, G. A. (1990). Semivariograms of
digital imagery for analysis of conifer forest structure. Remote Sensing
of the Environment, 34, 167 178.
Cressie, N. (1991). Statistics for spatial data. USA: Wiley-Interscience
(900 pp.).
Cross, A. M., Settle, J. J., Drake, N. A., & Paivinen, R. T. M. (1991). Subpixel measurement of tropical forest cover using AVHRR data. International Journal of Remote Sensing, 5, 1119 1129.
de Sa, T. D. A. & Alegre, J. (2001). Praticas agroflorestais visando o
manejo de vegetac oes secundarias: Uma abordagem com enfase em
experiencias amazonicas. In CONGRESSO BRASILEIRO DE SISTEMAS AGROFLORESTAIS, 3., 2000, Manaus, AM. Embrapa Amazonia Ocidental. Documentos, 17, 102 115.
de Sa, T. D. A., de Oliveira, V. C., Coimbra, H. M., de Carvalho, C. J. R.,
Dias-Filho, M., Sommer, R., & Brienza Junior, S. (1998). Diurnal and
seasonal patterns of leaf water relations in spontaneous and enriched
secondary vegetation components: Tools to understand water exchange
and water stress behavior, Abstracts of the Third SHIFT Workshop,
ManausAM. March 15 19, p. A14.
de Sa, T. D. A., de Oliveira, V. C., de Araujo, A. C., & Brienza Junior, S.
(1997). Biophysical aspects of enriched fallow with leguminous fast
growing trees in Northeastern Para, Brazil: Spectral composition of
light and water vapour. In International Symposium on the Science
and Practice of Short-term Improved Fallows, Lilongwe, Malawi,
1997, Abstracts, Lilongwe, IUFRO/ISSS-AISS-IBG/ICRAF.
de Sa, T. D. A., Vielhauer, K., Denich, M., Kanashiro, M., & Vlek, P. L. G.
(1998). Towards improving natural resources use in Eastern Amazonia
through a modified sequential agroforestry system. In Resumos Expandidos: II Congresso Brasileiro em Sistemas Agroflorestais, 25.11.1998,
pp. 95 100.
Drake, J. B., Dubayah, R. O., Clark, D. B., Knox, R. G., & Blair, J. B.
(2002). Sensitivity of large-footprint lidar to canopy structure and biomass in a neotropical rainforest. Remote Sensing of Environment, 81,
378 392.
Drake, J. B., Dubayah, R. O., Clark, D. B., Knox, R. G., Blair, J. B.,
Hofton, M. A., Chazdon, R. L., Weishampel, J. F., & Prince, S.
(2002). Estimation of tropical forest structural characteristics using
large-footprint lidar. Remote Sensing of Environment, 79, 305 319.
Ehleringer, J. R., & Field, C. B. (Eds.) (1993). Scaling physiological processes leaf to globe. San Diego, CA: Academic Press, 388 pp.
Fearnside, P. M. (1990). The rate and extent of deforestation in Brazilian
Amazon. Environmental Conservation, 17, 213 226.
Foody, G. M., Lucas, R. M., Curran, P. J., & Honzak, M. (1997). Nonlinear mixture modeling without end-members using an artificial neural
network. International Journal of Remote Sensing, 18, 937 953.
Harding, D. J., Lefsky, M. A., Parker, G. G., & Blair, J. B. (2001).
Laser altimeter canopy height profiles: Methods and validation for
127
The LBA Science Planning Group (1996). The Large Scale Biosphere
Atmosphere Experiment in Amazonia (LBA): Concise Experimental
Plan. LBA Project Office, CPTEC, Brazil, http://yabae.cptec.inpe.br/lba.
USDA Forest Service (1992). Forest service resource inventories: An overview, Washington, DC.
Vielhauer, K., Denich, M., de Sa, T. D. A., Kato, O. R., Kato, M. d. S. A.,
Brienza Jr., S., & Vlek, P. L. G. (2001). Land-use in a mulch-based
farming system of small holders in the Eastern Amazon. Proceedings of
the Deutscher Tropentag (Conference on International Agricultural Research for Development) One World: Research for a better quality of
life, Bonn, Germany, October 9 11 2001, pp. 1 9.
Vourlitis, G. L., Priante-Filho, N., Hayashi, M. M. S., de Souza Nogueira,
J., Caseiro, F. T., & Campelo Jr., J. H. (2002). Seasonal variations in
the evapotranspiration of a transitional tropical forest of Mato Grosso,
Brazil. Water Resources Research, 38(6), 30 31.
Vourlitis, G. L., Priante-Filho, N., Hayashi, M. M. S., de Souza Nogueira,
J., Caseiro, F. T., Raiter, F., & Campelo Jr., J. H. (in press). The role of
seasonal variations in meteorology on the net CO2 exchange of a Brazilian transitional tropical forest. Ecological Applications.
Institute for Marine Remote Sensing, College of Marine Science, University of South Florida, 140 7th Avenue S., St. Petersburg, FL 33701, USA
b
Rosenstiel School of Marine and Atmospheric Science, University of Miami, Miami, FL, USA
c
Center for Coastal and Regional Marine Studies, United States Geological Survey, St. Petersburg, FL, USA
d
Biophysical Remote Sensing Group, Department of Geographical Sciences and Planning, University of Queensland, St. Lucia, Australia
e
Hawaii Institute of Marine Biology, University of Hawaii, Honolulu, Kaneohe, USA
f
Coral Reef Ecosystems Ecology Laboratory, Marine Resources Department, CINVESTAV-I.P.N. Unidad Merida, Merida, Mexico
g
Marine Spatial Ecology Laboratory, University of Exeter, Exeter, UK
h
Oceanographic Center, National Coral Reef Institute, Nova Southeastern University, Miami, FL, USA
i
Social and Environmental Systems Division, National Institute for Environmental Studies, Onogawa, Tsukuba, Ibaraki, Japan
j
Department of Marine Science and Coastal Management, The University of Newcastle, Newcastle upon Tyne, UK
k
Laboratoire Terre-Oceans, Universite de la Polynesie Francaise, Tahiti, French Polynesia
l
Department of Biology, Dalhousie University, Halifax, Nova Scotia, Canada
Received 3 June 2002; received in revised form 5 December 2002; accepted 22 April 2003
Abstract
Ten IKONOS images of different coral reef sites distributed around the world were processed to assess the potential of 4-m resolution
multispectral data for coral reef habitat mapping. Complexity of reef environments, established by field observation, ranged from 3 to 15
classes of benthic habitats containing various combinations of sediments, carbonate pavement, seagrass, algae, and corals in different
geomorphologic zones (forereef, lagoon, patch reef, reef flats). Processing included corrections for sea surface roughness and bathymetry,
unsupervised or supervised classification, and accuracy assessment based on ground-truth data. IKONOS classification results were
compared with classified Landsat 7 imagery for simple to moderate complexity of reef habitats (5 11 classes). For both sensors, overall
accuracies of the classifications show a general linear trend of decreasing accuracy with increasing habitat complexity. The IKONOS sensor
performed better, with a 15 20% improvement in accuracy compared to Landsat. For IKONOS, overall accuracy was 77% for 4 5 classes,
71% for 7 8 classes, 65% in 9 11 classes, and 53% for more than 13 classes. The Landsat classification accuracy was systematically lower,
with an average of 56% for 5 10 classes. Within this general trend, inter-site comparisons and specificities demonstrate the benefits of
different approaches. Pre-segmentation of the different geomorphologic zones and depth correction provided different advantages in different
environments. Our results help guide scientists and managers in applying IKONOS-class data for coral reef mapping applications.
D 2003 Elsevier Inc. All rights reserved.
Keywords: Landsat; Bathymetric correction; Glint; Accuracy; Habitat mapping; Seagrass
1. Introduction
Remote sensing provides an effective way to observe
and monitor shallow coral reefs worldwide, to characterize
129
130
131
Fig. 1. Location of the study sites. The map indicates the main coral biogeographic areas according to Veron (1995). Our sites represent the Arabian Gulf,
Indian Ocean, Indo Pacific, Pacific, and Caribbean biogeographic zones. Each site is presented using a RGB color composite based on the red, green, and blue
bands of the IKONOS sensor. Size (in km) of the image showing the most characteristic zone is indicated, though the actual processed area may be wider.
Includes material Space Imagingn.
132
Table 1
Site information and characteristics of the IKONOS and Landsat images
Latitudeb
Longitudeb
IKONOS
Landsat 7 ETM +
Acquisition date
Product
Resampling
Path/row
Acquisition date
Format
Ground-truth
Depth (m)
References
Stoddart (1964)
Kramer, Kramer,
and Ginsburg (1998)
Frouin and Hutchings (2001)
Jaap (1984)
Kramer and Kramer (2000)c
Riegl (1999)
McClanahan and
Muthiga (1998)
Rogers (1997),
Smith et al. (1998)
0.6149
24.5833
73.1201
77.7833
14 March 2000
12 March 2001
MS
OS
CC
NN
145/60
13/43
20 December 2000
26 March 2000
HDF
EDC-Geotif
March 2002
2000/2001
0 15
0 30
Arue SA, MZ
Biscayne DTP, JB, SA
Boca Paila RG
Dubai BR
Glovers PJM, WHW
17.5750
25.3500
19.9166
24.9400
16.8166
149.6000
80.2166
87.5000
54.8900
87.8000
11 March 2000
18 March 2001
20 July 2000
2 May 2001
12 April 2001
OS
OS
MS
MS
OS
CC
CC
CC
NN
NN
53/72
15/42
19/46
160/42
18/48
6 June 2000
5 February 2000
24 July 1999
N/A
8 November 2000
Earthsat-Fast
EDC-HDF
EDC-Geotif
N/A
EDC-Geotif
June 2000
2001/2002
1999/2000
Fall 1995
1999 2001
0 12
2 12
0 30
09
0 18
23.5258
151.8900
7 May 2001
MS
NN
90/77
14 November 1999
EDC-HDF
2001/2002
0 15
Mayotte SA
Shiraho HY
12.8933
24.3166
45.2180
124.2333
31 August 2000
28 March 2002
OS
OS
NN
NN
91/77
161/69
115/43
18 September 1999
30 August 2000
23 February 2002
EDC-HDF
EDC-Geotif
Fast-L7A EROSd
Dec. 2000
1999
0 15
03
OS: Original Standard, MS: Master Standard, CC: cubic convolution, NN: nearest neighbor convolution, N/A: not available.
a
Investigators: initials from author list.
b
Lower left corner of the IKONOS scene. Negative latitude for South, negative longitude for West.
c
Describe Yucatan reefs.
d
Provided by National Space Development Agency of Japan.
Site and
investigatora
133
Table 2
Classification scheme for the 10 sites (Boca Paila 15 classes and Mayotte 14 classes are not shown)
Site
Class label
Site
Class label
Addu
Classes: 8
Sand/rubble, backreef
Sand, backreef
Sand, lagoon floor
Coral/algae, patch reef
Algae, crest
Coral, forereef
Seagrass, backreef
Coral, backreef
Dense brown algae (>80%), crest
High density coral heads on heterogeneous pavement
Low density coral heads on sandy floor
Dense brown algae (>50%) on reef flat, heterogeneous floor
Moderate brown algae (15 25%) on reef flat,
heterogeneous floor
Sparse brown algae ( < 5%) on reef flat, heterogenous floor
Sand and ruble (>90%)
Deep (>10 m) lagoon floor
Dense coral (>25%)
Moderate coral ( < 25%)
Seagrass/algae
Heterogeneous lagoon floor
Pavement
Shallow sand, lagoon floor
Deep sand, forereef
Andros
Classes: 8
Arue
Classes: 8
Boca Paila
Classes: 7
Glovers
Classes: 5
Heron
Classes: 13
Forereef
Brown algae
Seagrass/Lobophora sp.
Seagrass
Sand
Branching corals (>75%) on forereef and crests
Multi-growth forms, dense corals (>50%) on reef flat
Multi-growth forms, moderate corals (25 50%) on reef flat
Coral pavement (>75%) on crest and reef flats
Coral head, forereef
Heterogeneous reef flat (coral < 10%, sand, rocks,
fleshy algae, coralline)
Dead coral (>80%) coated by encrusting coralline
Dead coral (>80%) covered by fleshy algae
Sand with rocks ( < 15%) covered by fleshy algae
Sand with scattered dead coral heads ( < 15%),
coated by encrusting coralline
Sand with dense dead coral heads (>50%),
covered by fleshy algae
Coral sand and mud (>90%)
Pavement and ruble (>90%)
Shiraho
Classes: 4
Biscayne
Classes: 8
Dubai
Classes: 8
Heron
Classes: 5
Mayotte
Classes: 14
Coral
Seagrass/algae
Pavement
Sand
134
Fig. 2. Top: Habitat classification scheme as proposed by Mumby and Harborne (1999) for mapping Caribbean coral reefs. Both geomorphologic and benthic
keys may define a habitat. For instance, three habitats encountered on dense patch reefs are highlighted: a coral zone with presence of massive colonies of
Montastraea sp., an algal zone dominated by the brown algae Lobophora sp., and a bare substrate zone densely covered by gorgonians. Depending on the
precision of the classification scheme, a practitioner can provide details only at geomorphological level, or at benthic levels, or a combination of both. Bottom:
for Glovers Reef, the final classification scheme is a simple five-class scheme, with a forereef (geomorphology) class and four generic benthic classes (brown
algae, seagrass, mixed seagrass, and sand) not related to a specific geomorphologic zone (e.g. sand includes sandy areas in any geomorphological zones).
135
Fig. 3. Example of contrasted habitat classification schemes for two Caribbean sites: Biscayne Bay, FL (USA) and Andros island (Bahamas). Coloured sections
highlight the zones present on each reef. For Biscayne, the study area includes lagoon floor and patch reefs. For Andros, the area of interest includes the oceanic
side of the coral reef system with the forereef, crests, and spur-and-grooves. In both sites, benthic classes are related to a geomorphologic zone, highlighted with
similar color. For instance, for Biscayne, a class of dense seagrass has been defined both for the shallow lagoon floor (blue) and the deep lagoon (green). Sand
is related on the lagoon floor (yellow), without separation between deep and shallow. Diffuse gorgonians are related to diffuse patch reefs (purple). For Andros,
the same bi-component geomorphology-benthos hierarchy occur, e.g. dense gorgonians occur only on high relief escarpment (purple); forereef (green) is split
into four classes: sand (yellow), diffuse gorgonian (light purple), coral Montastraea (dark green), and bare bedrock (grey). Andros and Biscayne both comprise
eight classes of habitats, but with minimum overlap between the two schemes. This illustrates the variety of habitats classification scheme that can occur in the
same region (see also Fig. 2 for Glovers).
136
Fig. 4. Pre-processing of IKONOS images. Top: Example of surface correction aimed at removing wave and glint effects that occur on almost 50% of the 40
images provided by NASA SDP to USF. The level of details available in deep areas is drastically improved after correction. Bottom: An example of false color
composite made with three depth-invariant bottom indices, one from each pair of IKONOS bands 1 3. The backreef appears in pink dividing the outer forereef
from the lagoon. This technique limits misclassification due to depth between shallow dark objects and deep bright objects. Surface correction and depthinvariant indices were not systematically applied, depending on image quality and site characteristics (see Table 3).
137
Table 3
Image processing parameters and classification results
Site
Surface
correction
Depth
correction
IKONOS
bands
Nb control
points
Classes
Overall accuracy
IKONOS (%)
Overall accuracy
Landsat 7 (%)
Addu
Andros
Arue
Biscayne
Boca
Paila
N
N
N
Y
N
N
Y
N
N
Y
4
3
4
3
3
400
150
200
123
150
8
8
8
8
15
66
74
70
84
45
56
N/A
52
56
N/A
Dubai
Glovers
N
N
N
Y
3
3
N/A
150
Heron
165a
Mayotte
1230c
Shiraho
104
7
8
11
5
13
7
5
14
10d
10e
4
74
71
51
77
42
61
78
61
73
68
81
53
N/A
42
71
N/A
N/A
66/61b
N/A
56
50
63
138
Fig. 5. Example of validation of IKONOS classification using in situ large scale transects on Heron Reef. The classifications of three sites (Dubai, Mayotte, and
Heron Reef) have been controlled using this technique. Here, on the south of Heron, the transition between five benthic classes can be compared with in situ
observation where the percent cover of different substrates has been estimated visually in 20 20 m units. Despite the difference in resolution (4 vs. 20 m), we
note the good agreement between in situ data and classification, with accurate definition of transition zones, homogeneous zones, and heterogeneous patchy
zones (data from Joyce, Phinn, Roelfsema, Neil, & Dennison, 2002).
1999), suggesting that IKONOS could be used for monitoring changes in benthic communities for similar sites if a
reference had been available. This is confirmed by Palandro et al. (2003) who combined aerial photographs and one
IKONOS image of Carysfort Reef (Florida) to quantify the
rate of coral loss in the last 20 years.
The general trend in Fig. 7 is a linear decrease of
accuracy with increasing complexity. Accuracies range
from an average of 77% for 4 5 classes to 71% for 7
8 classes, 65% in 9 11 classes, and 53% for more than 13
classes. There is no obvious bias that could be explained
by the skills of the investigators. For instance, in 13 14
classes, both best and worst results are provided by the
same investigator (SA). At 8 10 classes, all results are
very consistent. The variations are the natural consequences of the nature of the site and the way that the images
have been processed (Table 3). We did not compute the
variance of each accuracy because the sampling schemes
for accuracy assessment were not the same and in some
cases the rigorous conditions of application would likely
have been violated (Foody, 2002; Stehman, 1997; Steh-
139
Fig. 6. Examples of classification maps for Arue, Andros, Biscayne, and Glovers. For the three first sites overall accuracy was >70%, while for Glovers the
11-class scheme led to a poor 50% accuracy, prompting the implementation of a more simple five-class scheme (Fig. 2).
140
90
80
70
60
50
40
30
20
2
10
11
12
13
14
15
Roatan
Punaauia
Turks and Caicos
Heron
Andros
Arue
Biscayne
Glovers
Mayotte
Shiraho
Boca Paila
Addu
Dubai
80
70
60
50
40
30
20
2
10
11
12
13
14
15
141
142
4. Conclusion
References
Acknowledgements
Obviously, this study would not have been possible
without the spirit and crew of the NASA Scientific Data
Purchase program at Stennis Space Center under the
successive responsibility of Fritz Pollicelli and Troy
Frisbie. We greatly acknowledge their help in the tasking
requests and their support of our research since 1999.
Andrew Mattee and Michael Satter were our contacts at
Space Imaging. This research was supported by NASA
grants NAG5-10908 to SA and NAG-3446 to FMK and
Kendall Carder. Many individuals helped gather field data
on the various sites. We are indebted to Chris Roelfsema,
Bill Dennison, Fabienne Bourdelin, Claude Payri, Bernard
Thomassin, Michel Pichon, Hajime Kayanne, Saki Harii,
Yoshiyuki Tanaka, Ernesto Arias-Gonzales, Don Hickey,
Nancy Dewitt, Tonya Clayton, David Palandro, Chuanmin
Hu, and the staffs of Heron Island Research Station,
Biscayne National Park, and Service des Peches et de
lEnvironment Marin de Mayotte. We are grateful to GISLAGMAY (Mayotte) and the World Bank, via Andy
143
Extending satellite remote sensing to local scales: land and water resource
monitoring using high-resolution imagery
Kali E. Sawaya*, Leif G. Olmanson, Nathan J. Heinert, Patrick L. Brezonik, Marvin E. Bauer
Department of Forest Resources, University of Minnesota, 1530 N. Cleveland Avenue St. Paul, MN 55108-6112, USA
Received 22 May 2002; received in revised form 19 March 2003; accepted 24 April 2003
Abstract
The potential of high-resolution IKONOS and QuickBird satellite imagery for mapping and analysis of land and water resources at local
scales in Minnesota is assessed in a series of three applications. The applications and accuracies evaluated include: (1) classification of lake
water clarity (r2 = 0.89), (2) mapping of urban impervious surface area (r2 = 0.98), and (3) aquatic vegetation surveys of emergent and
submergent plant groups (80% accuracy). There were several notable findings from these applications. For example, modeling and estimation
approaches developed for Landsat TM data for continuous variables such as lake water clarity and impervious surface area can be applied to
high-resolution satellite data. The rapid delivery of spatial data can be coupled with current GPS and field computer technologies to bring the
imagery into the field for cover type validation. We also found several limitations in working with this data type. For example, shadows can
influence feature classification and their effects need to be evaluated. Nevertheless, high-resolution satellite data has excellent potential to
extend satellite remote sensing beyond what has been possible with aerial photography and Landsat data, and should be of interest to resource
managers as a way to create timely and reliable assessments of land and water resources at a local scale.
D 2003 Elsevier Inc. All rights reserved.
Keywords: IKONOS; Remote sensing; High resolution imagery; Lake clarity; Aquatic vegetation; Impervious surface
1. Introduction
Although high-resolution imagery in the form of aerial
photography has been available for many years, the launch of
the IKONOS-2 by Space Imaging in September 1999 has
signaled a new era in satellite remote sensing. With multispectral digital imagery approaching that of small to medium
scale photography, we are in a new period of applications
development. Our objective has been to evaluate high-resolution satellite imagery in a variety of applications involving
monitoring of land and water resources. We approach each of
three applications at a local scale and address a pertinent
aspect of water quality in Minnesota. Each application
utilizes a single-date scene of IKONOS or QuickBird imagery covering approximately 11 11 km areas.
To demonstrate the potential of the imagery, these
applications each take a closer look at digitally classifying
a particular aspect of monitoring and mapping land and
* Corresponding author. Tel.: +1-612-624-2202; fax: +1-612-6255212.
E-mail address: ksawaya@gis.umn.edu (K. Sawaya).
0034-4257/$ - see front matter D 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2003.04.0006
145
146
147
Fig. 1. Comparison of IKONOS and field measurements of lake water clarity, TSI(SDT).
148
Fig. 2. Comparison of Landsat and IKONOS estimates based on Landsat modeling of lake water clarity, TSI(SDT).
abundance of forest and wetland areas and limited development in the form of parking lots and pavilions for park
visitors. This area has relatively high lake water clarity with
TSI(SDT) of around 50 and SDT of f 2 m. In many of the
residential and commercially developed areas, stormwater
is directed into lakes using them as convenient reservoirs.
The increase in impervious surface area and direct connection to the stormwater system has dramatically changed the
hydrology of many water bodies in Eagan. These changes
Fig. 3. Lake water clarity classification of IKONOS multispectral data overlaid on panchromatic land image. (Imageryn Space Imaging L.P.).
149
investigate whether a similar method is suitable for highresolution IKONOS imagery at a local scale for the City of
Eagan, MN. Eagan is located southwest of St. Paul, MN and
covers 89 km2. In the 1990s, Eagans population growth
increased 33%, reaching 63,557 in 2000. The landscape
consists mostly of single/multi-family residential, commercial/industrial, parks and recreational areas, vacant and
agricultural lands, and numerous small lakes and ponds.
3.2. Methods
We performed several background tests to determine
which spectral transformation provided the strongest relationship to percent impervious surface area. For IKONOS
data, a correlation of 0.90 0.95 suggested that the normalized difference vegetation index (NDVI) provided the best
relationship. Other transformations considered were IR/red
ratio and principal components. The ratio had a correlation
of 0.93 and the second principal component had a correlation of 0.86 with percent impervious. The ratio and the
NDVI had similar results, but we selected NDVI due to its
familiarity. In past research with Landsat TM imagery, we
used tasseled cap greenness (Bauer et al., 2002), but this
was not possible since the coefficients for a tasseled cap
transformation of IKONOS data have not been determined.
IKONOS digital imagery of the study area was acquired
on September 4, 2001. The imagery was clear and cloudfree; therefore, no atmospheric correction was necessary.
Using 270 ground control points, we rectified and registered
the imagery to DOQs covering the same area, with an RMS
error of 1.24 m.
We began our image processing by performing a general
land cover classification using a supervised, maximum
likelihood approach in ERDAS Imagine. Due to spectral
confusion among some water features and asphalt, we hand
Fig. 4. Sample areas delineated over Digital Orthophoto Quadrangles (DOQs) used to calculate varying degrees of impervious surface area for model
calibration and evaluation.
150
The agreement between IKONOS estimates and measurements from the additional sample areas of DOQs was
very high with a linear correlation of 0.98 (Fig. 6). Fig. 7
displays a map of all impervious surfaces in the City of
Eagan as a continuous gray scale of 0 100% impervious
surface area. Note the degree of impervious detail that is
possible to discern in the built environment including bike
trails, vegetation along road medians, and individual
driveways. Resource managers can now recode the 0
100% scale to appropriate indices for their individual
analysis.
The land cover classification step helped to stratify and
improve the accuracy of the final map. Separating water,
vegetation, and bare areas helped prevent mapping several
pervious features as impervious. Without masking, open
water features are erroneously classified as 80 100%
impervious due to their low NDVI. The presence of tree
canopy shadows and bare patches in lawns can also
introduce a small amount of error into the impervious
surface estimations. For future land cover classifications
of high-resolution imagery, we anticipate using objectbased classification methods to minimize some of these
effects.
Shadows from buildings and other tall features affected
the overall classification of impervious surface in highresolution satellite imagery. They tended to be most problematic around the perimeter of forested and urban areas.
Shadows have low NDVI values, thus giving a false
measure of impervious features in certain areas. It is difficult
to ascertain what features are within a shadowed pixel using
digital classification. Therefore, manual image interpretation
Fig. 5. Relationship of percent impervious surface area determined from DOQs and IKONOS NDVI.
151
Fig. 6. Comparison of measured and IKONOS estimates of percent impervious surface area.
Fig. 7. Maps of percent impervious surface area derived from classification of IKONOS NDVI data.
152
County, and West Auburn (55 ha) and Shutz (38 ha) Lakes
in Scott County, MN.
5.2. Methods
We adapted methods typically used for land cover classification (Lillesand et al., 1998) and developed for water
clarity assessments (Olmanson et al., 2002) to do the
evaluation. The aquatic plant classification methods consisted of two procedures: separation of image features into
discrete units and classification of the pixels in each unit.
5.2.1. Satellite image data
For this study, we used an IKONOS image acquired on
September 1, 2001 of the Swan Lake area and a QuickBird
image acquired on July 28, 2002 of the Lake Minnetonka
area. An acquisition window of July 15 through September
15 captures the presence of aquatic vegetation in Minnesota.
The images were of high quality with only minor cloud
cover over the southern portion of the IKONOS image. We
applied a resolution enhancement of the multispectral bands
using the panchromatic band. Atmospheric correction or
normalization of the imagery was not necessary for the
methods used in this study.
5.2.2. Aquatic vegetation reference data
Due to the size of Swan Lake and the abundance of
aquatic plants in the lake, the collection of reference data
would be very difficult without the aid of modern technology. We used Global Positioning System (GPS) technology
and the advanced GPS tracking software in ERDAS Imagine 8.5. We collected field reference data shortly after
acquiring the imagery using a Fujitsu pen computer. While
in the field, we identified different types of aquatic vegetation and located them directly on the IKONOS image using
the field computer. Being able to accurately identify specific
locations on the image while in the field was especially
useful on this large wetland. Having the image available
quickly after its acquisition for use in reference data
collection was also a significant advantage in field sampling
because we could identify unique areas with different
spectral-radiometric responses on the image and target them
for field identification. We targeted emergent vegetation for
the evaluation, but also noted the location of submerged
vegetation appearing at the surface.
For the three lakes south of Lake Minnetonka the
Minnesota Department of Natural Resources collected the
field data. For this study the field data specifically targeted
submerged plants using an echo sounder equipped with a
GPS unit. The hydroacoustic survey collected data pertaining to lake depth and plant depth at thousands of points in
the shallower portions of the lakes where aquatic plants
were present but not at the surface. General point survey
data were collected in areas where the submerged plants
were topped at the surface or where there were floating or
emergent plants.
153
Fig. 8. IKONOS image classification of aquatic vegetation of Swan Lake in Nicollet County, MN (Imageryn Space Imaging L.P.).
154
Submerged plants
Cattail
Sedge and dead sedge
Bulrush
Water lily and FLPW
a
IKONOS (%)
63.6
28.1
2.5
1.6
4.2
58.5
23.7
5.1
6.8
5.9
Table 2
Accuracy of IKONOS image classification of aquatic vegetation for Swan
Lake, MN
Classified
data
Reference data
Cattail Sedge Sedge, Bulrush Lily Submerged Users
dead
accuracy
(%)
Cattail
24
3
Sedge
4
Sedge,
dead
Bulrush
Lily
Submerged
4
Producers
100.0 36.4
Accuracy (%)
88.9
66.7
100.0
2
5
7
1
13
100.0 92.9
13
3
62.5
4
53.8
100.0
92.9
54.2
79.5
6. Conclusions
Through our experiences with these three projects, we
have identified a number of considerations for applications
development using high resolution satellite imagery. First,
we have had the most success applying high-resolution
imagery to local or city scale analyses. When imagery is
acquired over large geographic areas, scene differences can
155
156
and water resource managers who need detailed and accurate monitoring information.
Acknowledgements
We gratefully acknowledge the support of the NASA
Science Data Purchase for providing IKONOS images, the
Upper Great Lakes Regional Earth Science Applications
Center (NASA grant NAG-13-99002), and the University of
Minnesota Agricultural Experiment Station. We also
appreciate the interest and cooperation of the City of Eagan,
Metropolitan Council, Minnesota Pollution Control Agency
and Minnesota Department of Natural Resources. Melanie
Tyler and John D. Madsen, Biological Sciences Department,
Minnesota State University, Mankato, provided the aquatic
vegetation survey field data. We also thank the reviewers
whose comments helped us improve this paper. Finally, we
thank Sarah Finley for her help in compiling and editing this
manuscript.
References
Barbier, E. B., Burgess, J. C., & Folke, C. (1994). Paradise lost? The
ecological economics of biodiversity. London: Earthscan Publications,
267 pp.
Barnes, K. B., Morgan, J. M., & Roberge, M. C. (2000). Impervious
surfaces and the quality of natural and built environments. Baltimore,
MD: Department of Geography and Environmental Planning, Towson
University.
Bauer, M. E., Doyle, J. K., & Heinert, N. J. (2002). Impervious surface
mapping using satellite remote sensing. Proceedings, IEEE Geoscience
and Remote Sensing Symposium, Toronto, Canada, June 24 28, Unpaginated CD ROM, 3 pp.
Carlson, R. E. (1977). A trophic state index for lakes. Limnology and
Oceanography, 22, 361 369.
Civco, D. L., & Hurd, J. D. (1997). Impervious surface mapping for the
state of Connecticut. Proceedings, 1997 ASPRS/ACSM Annual Convention. Seattle, WA, April 7 10 ( pp. 124 135).
Daily, G. C. (Ed.). (1997). Natures services: Societal dependence on natural ecosystems. Washington, DC: Island Press, 392 pp.
Estes, M. G., Gorsevski, V., Russel, C., Quattrochi, D., & Luvall, J. (1999).
The urban heat island phenomenon and potential mitigation strategies.
Proceedings, 1999 National Planning Conference: Approaching the
Millennium, Seattle, WA.
Kloiber, S. M., Brezonik, P. L., & Bauer, M. E. (2002). Application of
Landsat imagery to regional-scale assessments of lake clarity. Water
Research, 36, 4330 4340.
Kloiber, S. M., Brezonik, P. L., Olmanson, L. G., & Bauer, M. E. (2002). A
procedure for regional lake water clarity assessment using Landsat
multispectral data. Remote Sensing of Environment, 82(1), 38 47.
Lillesand, T., Chipman, J., Nagel, D., Reese, H., Bobo, M., & Goldman, R.
(1998). Upper Midwest gap analysis program image processing protocol. Report prepared for the U.S. Geological Survey, Environmental
Management Technical Center, Onalaska, Wisconsin, EMTC 98G001. 25 pp.+Appendixes A C.
Lillesand, T. M., Johnson, W. L., Deuell, R. L., Lindstrom, O. M., &
Meisner, D. E. (1983). Use of Landsat data to predict the trophic state
of Minnesota lakes. Photogrammetric Engineering and Remote Sensing, 49, 219 229.
Mitsch, W. J., & Gosselink, J. G. (1993). Wetlands (2nd ed.). New York:
Van Nostrand Reinhold, 722 pp.
Olmanson, L. G. (1997). Satellite remote sensing of the trophic state conditions of the lakes in the Twin Cities Metropolitan Area. M.S. Paper,
Water Resources Science, University of Minnesota, St. Paul, MN. 28 pp.
Olmanson, L. G., Bauer, M. E., & Brezonik, P. L. (2002). Water quality
monitoring of 10,000 Minnesota Lakes: statewide classification of lake
water clarity using Landsat imagery. Proceedings, 15th William T. Pecora Memorial Remote Sensing Symposium. November 8 15, Denver,
Colorado, unpaginated CD ROM, 5 pp.
Ozesmi, S. L., & Bauer, M. E. (2002). Satellite remote sensing of wetlands.
Wetlands Ecology and Management, 10(5), 381 402.
Ridd, M. K. (1995). Exploring a V-I-S (vegetation-impervious surface-soil)
model for urban ecosystem analysis through remote sensing: comparative anatomy for cities. International Journal of Remote Sensing,
16(12), 2165 2185.
Sleavin, W. J., Civco, D. L., Prisloe, S., & Giannattti, L. (2000). Measuring
impervious surfaces for non-point source pollution modeling. Proceedings, 2000 ASPRS Annual Conference. May 22 26, Washington, DC,
11 pp.
Abstract
This study uses IKONOS imagery to quantify the combined spatial and spectral characteristics of urban reflectance in 14 urban areas
worldwide. IKONOS 1-m panchromatic imagery provides a detailed measure of spatial variations in albedo while IKONOS 4-m
multispectral imagery allows the relative contributions of different materials to the spectrally heterogeneous radiance field to be determined
and their abundance to be mapped. Spatial autocorrelation analyses indicate that the characteristic scale of urban reflectance is consistently
between 10 and 20 m for the cities in this study. Spectral mixture analysis quantifies the relative contributions of the dominant spectral
endmembers to the overall reflectance of the urban mosaic. Spectral mixing spaces defined by the two low-order principal components
account for 96% to 99% of image variance and have a consistent triangular structure spanned by high albedo, low albedo and vegetation
endmembers. Spectral mixing among these endmembers is predominantly linear although some nonlinear mixing is observed along the gray
axis spanning the high and low albedo endmembers. Inversion of a constrained three-component linear mixing model produces stable,
consistent estimates of endmember abundance. RMS errors based on the misfit between observed radiance vectors and modeled radiance
vectors (derived from fraction estimates and image endmembers) are generally less than 3% of the mean of the observed radiance. Agreement
between observed radiance and fraction estimates does not guarantee the accuracy of the areal fraction estimates, but it does indicate that the
three-component linear model provides a consistent and widely applicable physical characterization of urban reflectance. Field validated
fraction estimates have applications in urban vegetation monitoring and pervious surface mapping.
D 2003 Published by Elsevier Inc.
Keywords: IKONOS imagery; Urban reflectance; Spectral mixture analysis
1. Introduction
Human settlements occupy a relatively small fraction of
Earths surface area, but their extent, distribution and evolution have enormous impact on environmental and socioeconomic dynamics worldwide. Despite their fundamental
importance, urban areas have not been mapped and characterized with remote sensing to the same extent that other land
cover types have. Optical sensors on operational satellites
provide an efficient means for quantifying past and present
distributions of human settlements as well as their physical
reflectance properties. This is important at two scales. At a
global scale, it is important to understand the physical
characteristics that distinguish developed urban areas from
other types of human modified and undeveloped land surfaces in order to map and monitor the extent and evolution of
urban areas with moderate resolution (20 30 m) optical
* Tel.: +1-845-365-8354; fax: +1-845-365-8179.
E-mail address: small@LDEO.columbia.edu (C. Small).
0034-4257/$ - see front matter D 2003 Published by Elsevier Inc.
doi:10.1016/j.rse.2003.04.008
classes. Ridd (1995) proposed an urban landcover classification scheme based on the distribution of vegetation,
impervious surface and soil (the VIS model) but acknowledged the difficulty of distinguishing between soil and
impervious surfaces with optical sensors. Subsequent studies (e.g., Flanagan & Civco, 2001; Liu & Lathrop, 2002;
Madhavan et al., 2001) have employed a variety of classification methods (Maximum Likelihood, Unsupervised,
Decision Tree) with moderate resolution (30 m) imagery,
but traditional hard classification algorithms are impeded by
the abundance of spectrally mixed pixels. Mixed pixels are
problematic for statistical classification methods because
most algorithms are based on the assumption of spectral
homogeneity at pixel scale within a particular class of land
cover. The classification task is often further complicated by
inconsistencies between the thematic classes sought and the
reflectance properties that can be discriminated with moderate resolution broadband sensors. Urban areas provide
examples of spectrally diverse, scale-dependent thematic
classes containing large numbers of pixels that are spectrally
indistinguishable from other land cover classes. The diversity of land cover types and scales in the urban mosaic
therefore results in relatively high rates of misclassification
between urban and other land cover classes. Combining
spectral, textural and ancillary information can improve
classification accuracy (e.g., Stefanov et al., 2001), but a
physical characterization of reflectance is still necessary to
accommodate the physical processes that influence the
upwelling radiance measured by optical sensors. Several
recent studies have used physical rather than statistical
classifications of urban land cover in individual cities with
some degree of success (e.g., Kressler & Steinnocher, 1996,
2000; Phinn et al., 2002; Rashed, Weeks, Stow, & Fugate,
2002; Small, 2001a; Wu and Murray, 2002). High-resolution aerial photographs reduce the abundance of mixed
pixels (e.g., Akbari, Rose, & Taha, 1999), but film-based
images are less amenable to multispectral image analysis
and classification. Hyperspectral imagery can provide sufficient spatial and spectral resolution to map a wide variety
of urban surfaces (e.g., Herold, Gardner, Hadley, & Roberts,
2002), but relatively few cities have yet been mapped with
imaging spectrometers.
Characterization of urban reflectance is constrained by
the spatial resolution of the sensor. The 30-m resolution of
the Landsat ETM+ sensor and the 20-m resolution of the
Spot HRV sensor are generally not sufficient to discriminate
individual features (e.g., buildings, streets, trees) within the
urban mosaic. As a result, almost all the urban pixels
imaged by these sensors represent a composite radiance
field emanating from several distinct features with different
reflectances within the sensors field of view. This spectral
heterogeneity at scales comparable to the ground instantaneous field of view (GIFOV) of the sensor results in a
preponderance of spectrally mixed pixels. Mixed pixels
violate the cardinal assumption of statistical clustering
algorithms commonly used to classify land cover types.
171
The increased spatial resolution of IKONOS imagery provides an opportunity to image urban areas at scales sufficient to resolve many (but not all) of the individual features
in the urban mosaic. IKONOS orbital platform also makes
it possible to image a wide variety of urban areas worldwide
for a self-consistent analysis and comparison of the reflectance properties of urban land cover. A consistent, physically based description of urban reflectance properties could
help to advance our understanding and simulation of urban
microclimate by providing spatially explicit constraints on
albedo, evapotranspiration and spatial distribution of pervious surface. Characterization of urban reflectance at 4 m
scale would also provide constraints on the spectral endmembers and mixing processes responsible for the mixed
pixels imaged by moderate resolution sensors. If the scaledependent reflectance properties of urban mixed pixels
could be distinguished from the reflectance properties of
other types of land cover it would provide a basis for
mapping urban areas in moderate resolution imagery collected over the past 30 years.
The objective of this study is to explore a physical
characterization of urban reflectance properties in a variety
of urban settings. The characterization will incorporate the
spatial scale and optical reflectance properties of the variety
of land covers contained in the urban mosaics. Spatial
autocorrelation is used to quantify the characteristic scale
lengths of urban reflectance within and among different
cities. Spectral mixture analysis is used to quantify and
compare the reflectance properties of these urban areas.
Spectral mixture analysis provides a physically based representation of composite radiance measurements that allows
surface reflectance to be described as combinations of
spectral endmembers. Representing reflectance as continuous gradations within a spectral mixing space provides a
more flexible, and accurate, description than that resulting
from classification algorithms that assign each pixel membership in one (and only one) of a small number of classes.
It is important to distinguish between identification of
specific target materials and the physical representation of
the radiance image discussed here. Because many different
materials can have indistinguishable reflectance signatures
as measured by IKONOS, it is not generally possible to
identify specific target materials in these unvalidated spectral mixture analyses. The objective here is to determine the
consistency of urban reflectances and to assess the feasibility of using spectral mixture analysis to provide a general
and widely applicable representation of these reflectances.
The key questions addressed by this study are related to the
consistency of urban reflectance properties both within and
among a variety of urban areas worldwide. This study uses a
collection of 14 IKONOS images provided by the NASA
Scientific Data Purchase Program. The acquisition parameters for the NASA-funded IKONOS acquisitions are described in greater detail elsewhere in this volume.
Throughout the analysis it is assumed that the spatial
variations in radiance measured by IKONOS are primarily
172
Fig. 1. Characteristic scale length of urban reflectance. IKONOS panchromatic imagery of Caraz Peru (A) shows spatial variations in visible/near infrared
(VNIR) radiance resulting from the mosaic of roofs, streets, and open spaces. The superimposed grid spacing is 10 m. (B) The normalized two-dimensional
(2D) spatial autocorrelation function (ACF) of a 100 100-m subscene shows the azimuth dependence and quasi-periodic structure resulting from the
dominant scale of building size and street layout. (C) Asymptotic behavior of orthogonal one-dimensional autocorrelation functions extracted from the 2D ACF
shows the consistency of the slope of the central peak. The scale lengths estimated from the 0.1-slope threshold of these two functions are 10 and 12 m and are
consistent with the size of most features in (A). Includes materials n Space Imaging.
173
Fig. 2. Distribution of scale lengths for a variety of urban areas. Each of the 6357 sites corresponds to a 100 100-m subscene of IKONOS panchromatic
imagery. The distribution of scale lengths for the 14 cities indicates that characteristic scales are consistently between 10 and 20 m. Inset histograms show
distributions for different cities.
Fig. 3. False color composites of IKONOS MSI imagery (left) and corresponding principal components (right) for 14 urban areas. Natural color composites of
visible bands (R/G/B = 3/2/1) do not discriminate between vegetation and low albedo surfaces as effectively as false color composites (R/G/B = 3/4/1)
incorporating the NIR band 4. False color composites of the three low-order principal components (R/G/B = PC1/PC2/PC3) minimize visible band correlations
thereby enhancing contrast between surfaces with subtle differences in reflectance. The first principal component corresponds to albedo with bright areas
appearing red in all but one image (blue in She Xian). The second principal component corresponds to high NIR reflectance so vegetation appears green. The
third principal component is orthogonal to the primary mixing plane so pixels with relatively low proportions of PC1 and PC2 and more pronounced nonlinear
mixing outside the primary mixing plane span the range from black to blue. Each image is 1 km2. Includes materials n Space Imaging.
174
3. Spectral mixing
Spectral mixture analysis (SMA) provides a systematic
way to quantify spectrally heterogeneous urban reflectance.
SMA is based on the observation that, in many situations,
radiances from surfaces with different endmember reflectances mix linearly within the IFOV (Nash & Conel, 1974;
Singer, 1981; Johnson, Smith, Taylor-George, & Adams,
1983; Singer & McCord, 1979). This observation has made
possible the development of a systematic methodology for
SMA (Adams et al., 1986, 1989, 1993; Gillespie et al.,
1990; Smith, Ustin, Adams, & Gillespie, 1990; Roberts,
Smith, & Adams, 1993) that has proven successful for a
variety of quantitative applications with multispectral imagery (e.g., Adams et al., 1995; Elmore, Mustard, Manning, &
Lobell, 2000; Pech, Davies, Lamacraft, & Graetz, 1986;
Roberts, Batista, Pereira, Waller, & Nelson, 1998; Smith et
al., 1990). If a limited number of distinct spectral endmem-
175
Fig. 4. Mixing space representation of IKONOS MSI imagery. Normalized eigenvalues (left column) give the variance partition among the principal
components. The two-dimensional scatterplots of the principal components are projections of the three low-order dimensions of a four-dimensional spectral
mixing space. The side view shows the two dimensions associated with the majority (>96%) of the variance while the top and end views incorporate the third
dimension. Variance in the third dimension is primarily associated with nonlinear mixing along the gray axis between the high and low albedo endmembers
and generally accounts for less than 3% of the total variance. The fourth dimension (not shown) contributes less than 1% of the variance but still contains
coherent spectral information. Orthogonal projections of spectral mixing spaces in a variety of urban areas show a similar structure with a narrow triangular
mixing space spanned by high albedo (H), low albedo (L) and vegetation (V) endmembers. Straight edges correspond to linear mixing among endmembers
while convex edges indicate nonlinear mixing. Tapering of the mixing space approaching the Vegetation endmember suggests that a three-component mixing
model is very well constrained for vegetation fractions. The apex seen in the third dimension generally corresponds to a soil (S) endmember.
176
177
Fig. 5. Mixing spaces for nonurban land cover. At 1 km2 scales, other types of land cover do not generally produce the triangular mixing space characteristic of
urban land cover. While these mixing spaces do show cases of linear mixing among well-defined endmembers, nonlinear mixing is evident from the convex
edges of the mixing spaces. Multiple scattering within the open canopy vegetation of the Argentinian pampa produces strong nonlinear mixing. The mixing
space representing the sparsely vegetated Peruvian highland has a well-defined low albedo endmember, but mixing becomes increasingly nonlinear as albedo
increases along dimension 1. The densely vegetated mountains in central China produce a mixing space dominated by illumination differences in which an
exposed soil endmember resides at the bottom of the secondary dimension. The spectrally diverse riparian land cover of the Brazillian Amazon most closely
resembles the triangular mixing space seen in urban areas. Its primary mixing space is dominated by two linear mixing continua and complex topology at the
high albedo apex.
178
179
Fig. 6. Scale dependence of variance partition. Normalized eigenvalue distributions for larger urban areas (left) indicate that the second dimension of the mixing
space accounts for much of the variance (20 50%) in urban reflectance at spatial scales of several kilometers. Eigenvalue distributions of smaller (1 km2) city
center areas show a greater fraction of variance associated with the primary dimension spanned by high and low albedo endmembers. The difference is due to
the greater abundance of vegetated area within the larger images than within the 1-km2 city centers.
180
Fig. 7. Spectral endmembers. Radiance vectors corresponding to the apexes of the triangular mixing spaces show a remarkable consistency of form, although
amplitude is variable. Variations in amplitude result from actual differences in endmember reflectance as well as differences in illumination and atmospheric
conditions. The slope of the low albedo endmembers (thin curves) is consistent with wavelength-dependent atmospheric scattering effects. IKONOS DNs were
converted to radiance using the calibration coefficients provided by Space Imaging and NASA (Zanoni et al, Pagnutti et al, this volume).
differences in solar zenith, collection geometry, atmospheric turbidity, it is surprising that the amplitude differences
among the endmembers are not greater.
fH e 11 fV e12 fL e13 r1
181
182
Fig. 9. Maximum endmember fractions and RMS error distributions. (A) The cumulative distributions of the largest endmember fraction of each pixel span the
full range of permissible maximum fractions. Cities with significant areas of unmodeled soil (Caraz, Juarez, Phoenix, Pune, and Salvador) have more than 10%
area with maximum fractions less than 0.33. The three-component model does not account for the soil endmember and cannot replicate the spectra of these
pixels. (B) The RMS of the differences between the modeled and observed radiance vectors is normalized by mean pixel radiance to account for the magnitude
of the misfit relative to the amplitude of the observed radiance vector. Distributions of radiance-normalized RMS error are consistently low with error
amplitudes less than 3% of the mean radiance of the corresponding pixel. The three urban areas with higher errors (Bangkok, Caraz, Salvador) correspond to
examples with an unmodeled fourth endmember in the third dimension.
Fig. 8. Endmember fraction estimates and linear mixing model errors. Inversion of a constrained, three-component linear mixture model results in three
endmember fraction estimates for each pixel in the image. Endmember fraction images range from 0 (black) to 1 (white). The RMS error for each pixel is
calculated for the difference between the observed radiance and the forward implementation of the linear model using the endmember vectors and the
endmember fraction estimates. RMS error images range from 0% to 6% and are displayed with a 2% linear stretch to emphasize specific features not well fit by
the three-component linear model. Most urban areas have a strong low albedo component corresponding to shadow and the low surface reflectance of many
building materials.
183
5. Potential applications
Representation of multispectral radiances, or surface
reflectances, as spectral endmember abundances provides
a simple way to describe reflectance properties imaged by
the IKONOS sensor. The strength of the spectral mixture
model is its ability to represent a wide variety of surface
reflectance types as simple combinations of endmember
abundances. The inversion of the linear mixing model is
computationally trivial compared to statistical classifications
like maximum likelihood. Spectral mixture analysis is
preferable to hard classification for many physical science applications because it accommodates the fundamental
physical process responsible for the preponderance of mixed
pixels observed in almost all multispectral imagery. Representing a high-dimensional urban mixing space with only
three endmembers is obviously a gross simplification, but it
does provide a more flexible representation than hard
classifications which attempt to represent a continuum of
reflectance characteristics with exclusive membership in one
of a limited number of idealized classes. Endmember
fraction distributions can also be hardened into a finite
number of thematic classes if necessary (Adams et al., 1995;
Roberts et al., 1998). Spectral mixture models have the
added advantage of producing output in physically intuitive
units (fractional area). Relative abundances of high and low
albedo surface and vegetation can be converted to physical
quantities like albedo, pervious surface area and leaf area
index more easily than radiance measurements can and
more accurately than thematic classes can.
Vegetation fraction estimates derived from IKONOS
imagery provide a tool to monitor urban vegetation health
and abundance. An example of intraurban variations in
vegetation fraction is shown for Chicago in Fig. 10. The
spatial distribution and abundance of vegetation has a direct
impact on the urban environment by modulating solar
energy flux and evapotranspiration. Urban vegetation may
also have an impact on urban air quality as leaves sequester
surface reactant pollutants and particulates (Abdollahi &
Ning, 2000). Mapping spatial distribution and abundance of
urban vegetation is therefore important for modeling urban
air quality. IKONOS 4-m GIFOV approaches the scale of
crown diameter for many tree species. Quantitative assessment of vegetation fraction at this scale could be useful for
monitoring defoliation and blight in large urban parks.
Vegetation distribution can also be used as a proxy for
pervious surface when modeling urban hydrology (Small,
2002b).
Examination of spectral mixing spaces may also prove
useful for feature extraction tasks. In spite of its limited
spectral resolution, IKONOS 11-bit radiometry is capable
of distinguishing subtle differences in surface reflectance
that would be indistinguishable with 8-bit imagery. IKONOS combination of high spatial resolution and bit depth
results in greater clustering within the mixing space. The
Phoenix mixing space shown in Fig. 4 provides an example
184
Fig. 10. Example applications of spectral mixture analysis of urban IKONOS imagery. Vegetation fraction estimates have sufficient spatial resolution to map
intraurban vegetation abundance at street scale in Chicago. Warmer colors show higher vegetation abundance with fractions greater than 50% saturated red.
Localization of features with consistent reflectance properties within the mixing space can facilitate extraction of features like road networks (green) in
Pasadena. Includes materials n Space Imaging.
of this clustering (Fig. 4). Targets with distinct and consistent reflectance will appear as distinct clusters in IKONOS
mixing space whereas their reflectance would be mixed
with, or indistinguishable from, adjacent features in the
mixing space of a sensor with lower spatial resolution or
fewer resolvable brightness levels. An example of a fine
scale feature extraction from mixing space clustering is
shown in Fig. 10. The paved road surfaces in this image
of Pasadena have sufficiently consistent and distinct reflec-
6. Conclusions
Spectral mixture analysis provides a physically based
approach to quantify the optical reflectance properties of the
urban mosaic. The spectral mixing space concept accom-
modates the inevitable existence of mixed pixels and provides insight into the variety of distinct and gradational
reflectance patterns present in IKONOS imagery. The mixing space representation also reveals the presence of spectral
endmembers and the extent to which mixing among the
endmembers is linear. When mixing is predominantly linear,
it is possible to define linear mixture models that can be
inverted to yield endmember fraction estimates and measures of misfit to observed data.
Spatial autocorrelation of IKONOS panchromatic imagery provides statistical estimates of the spatial scale of urban
reflectance variations. Two-dimensional autocorrelation
functions consistently show a well-defined, axisymmetric
peak with a width corresponding to the spatial scale of the
most areally abundant features. The width of the peak
therefore provides an estimate of the characteristic spatial
scale of the high contrast reflectance patterns corresponding
to individual features within the urban mosaic (roads,
buildings, trees, etc.). The distribution of length scale
estimates from 6357 sites in 14 urban areas indicates that
the majority of sites have characteristic length scales between 10 and 20 m. This explains why urban areas are
characterized by spectral heterogeneity when imaged by
moderate resolution (20 30 m) sensors. This also suggests
that a significant fraction of IKONOS 4-m pixels will be
spectrally heterogeneous in urban imagery. This is supported by the maximum fraction distributions in Fig. 9.
Principal component transformation of urban IKONOS
imagery allows reflectance patterns to be interpreted in the
context of a spectral mixing space. Eigenvalue distributions
of the imagery used in this study indicate that the majority
(55% to 95%) of image variance corresponds to albedo
variations represented in the first principal component (PC).
A significant, but lesser, fraction of image variance is
associated with vegetation. In the urban areas investigated
here, almost all (>96%) image variance is associated with
the first two PCs. The remaining two PCs contain useful
information about distinct reflectances not discriminated in
the first two PCs and may be useful for isolating additional
endmembers. The spectral mixing space defined by the first
three principal components consistently takes the form of a
triangular scatterplot with linear or concave edges indicating
that mixing is predominantly linear. The spectral endmembers residing at the apexes of the mixing space correspond
to high albedo, low albedo and vegetation endmembers. The
mixing spaces of some of the urban areas used in this study
also revealed the presence of a fourth endmember, usually
corresponding to soil.
Inversion of a simple three-component linear mixture
model produces stable, consistent estimates of endmember
abundance fractions for each pixel in the image. RMS
misfits between observed radiance vectors and modeled
radiance vectors based on fraction estimates are generally
less than 3% of the mean of the observed radiance vector.
Most of the urban areas used in this investigation are
dominated by the low albedo endmember as a result of
185
shadowing and low reflectance building materials. Abundance of vegetation and high albedo features varies considerably within and among cities. Endmember abundance
maps derived from IKONOS can be used to constrain spatial
variations in solar energy flux and evapotranspiration as
well as to map spatial distributions of vegetation and
pervious surfaces.
Acknowledgements
The IKONOS data used in this study were provided by
Space Imaging through the NASA Scientific Data Purchase
program. Includes materials n Space Imaging. This research
would not have been possible without the support of the
Columbia Earth Institute, CIESIN and the NASA SocioEconomic Data and Applications Center (SEDAC).
References
Abdollahi, K. K., & Ning, Z. H. (2000). Urban vegetation and their relative
ability in intercepting particle pollution (PM2.5). Third Symposium on
the Urban Environment, Davis, CA, 1.15.
Adams, J. B. (1986). Spectral mixture modeling: A new analysis of rock
and soil types at the Viking Lander 1 site. Journal of Geophysical
Research, 91, 8089 8122.
Adams, J. B., Sabol, D. E., Kapos, V., Filho, R. A., Roberts, D. A., Smith,
M. O., & Gillespie, A. R. (1995). Classification of multispectral images
based on fractions of endmembers: Application to land cover change in
the Brazilian Amazon. Remote Sensing of Environment, 52, 137 154.
Adams, J. B., Smith, M. O., & Gillespie, A. R. (1993). Imaging spectroscopy: Interpretation based on spectral mixture analysis. In C. M. Englert, & P. Englert (Eds.), Remote Geochemical Analysis: Elemental and
Mineralogical Composition ( pp. 145 166). New York: Cambridge
Univ. Press.
Akbari, H., Rose, L. R., & Taha, H. (1999). Characterizing the Fabric of the
Urban Environment: A Case Study of Sacramento, CA. Report No.
LBNL-44688, Lawrence Berkeley National Laboratory, Berkeley, CA.
Boardman, J. W. (1993a). Automating spectral unmixing of AVIRIS data
using convex geometry concepts. Proceedings Third JPL Airborne
Earth Science Workshop, Pasadena, CA.
Boardman, J. W. (1989b). Inversion of imaging spectrometry data using
singular value decomposition. IGARSS89 12th Canadian Symposium
on Remote Sensing ( pp. 2069 2072).
Elmore, A. J., Mustard, J. F., Manning, S. J., & Lobell, D. B. (2000).
Quantifying vegetation change in semiarid environments: Precision
and accuracy of spectral mixture analysis and the normalized difference
vegetation index. Remote Sensing of Environment, 73, 87 102.
Flanagan, M., & Civco, D. L. (2001). Subpixel impervious surface mapping. Proc. 2001 ASPRS Annual Convention, St. Louis, MO (13 pp).
Getis, A. (1994). Spatial dependence and heterogeneity and proximal databases. In S. Fotheringham, & P. Rogerson (Eds.), Spatial Analysis and
GIS ( pp. 105 120). London: Taylor and Francis.
Gillespie, A. R., Smith, M. O., Adams, J. B., Willis, S. C., Fischer, A. F., &
Sabol, D. E. (1990). Interpretation of residual images: Spectral mixture
analysis of AVIRIS images, Owens Valley, California. Proceedings of
the 2nd Airborne Visible/Infrared Imaging Spectrometer (AVIRIS)
Workshop, Pasadena, CA ( pp. 243 270).
Green, A. A., Berman, M., Switzer, P., & Craig, M. D. (1988). A transformation for ordering mutispectral data in terms of image quality with
implications for noise removal. IEEE Transactions on Geoscience and
Remote Sensing, 26(1), 65 74.
186
Change identification using multitemporal spectral mixture analysis: Applications in Eastern Amazonia. In C. Elvidge, & R. Lunetta (Eds.),
Remote Sensing Change Detection: Environmental Monitoring Applications and Methods ( pp. 137 161). Ann Arbor: Ann Arbor Press.
Roberts, D. A., Smith, M. O., & Adams, J. B. (1993). Green vegetation,
nonphotosynthetic vegetation and soils in AVIRIS data. Remote Sensing
of Environment, 44, 255 269.
Ryan, R., Baldridge, B., Schowedgerdt, R., Choi, T., Helder, D., & Blonski,
S. (2003). IKONOS spatial resolution and image interpretability. Remote Sensing of Environment, 88, 37 51. (doi:10.1016/S00344257(03)00230-X)
Singer, R. B., & McCord, T. B. (1979). Large scale mixing of bright and
dark surface materials and implications for analysis of spectral reflectance. Journal of Geophysical Research, 1835 1848.
Settle, J. J., & Drake, N. A. (1993). Linear mixing and the estimation of
ground cover proportions. International Journal of Remote Sensing,
14(6), 1159 1177.
Small, C. (2001a). Estimation of urban vegetation abundance by spectral
mixture analysis. International Journal of Remote Sensing, 22(7),
1305 1334.
Small, C. (2001b). Multiresolution analysis of urban reflectance. Proceedings of the IEEE Workshop on Remote Sensing and Data Fusion of
Urban Areas, 8 9 Nov. 2001, Rome, Italy.
Small, C. (2001c). Spectral dimensionality of urban radiance. Proceedings
of 10th JPL Airborne Earth Science Workshop, NASA Jet Propulsion
Laboratory, Pasadena, CA.
Small, C. (2002a). A global analysis of urban reflectance. International
Journal of Remote Sensing (In press).
Small, C. (2002b). Reflectance properties of pervious and impervious surfaces. Proceedings of the ASPRS Pecora Land Remote Sensing Conference, Denver CO.
Smith, M. O., Johnson, P. E., & Adams, J. B. (1985). Quantitative determination of mineral types and abundances from reflectance spectra
using principal component analysis. Journal of Geophysical Research,
90, 792 804.
Smith, M. O., Ustin, S. L., Adams, J. B., & Gillespie, A. R. (1990).
Vegetation in deserts: I. A measure of abundance from multispectral
images. Remote Sensing of Environment, 31, 1 26.
Stefanov, W. L., Ramsey, M. S., & Christensen, P. R. (2001). Monitoring
urban land cover change: An expert system approach to land cover
classification of semiarid to arid urban centers. Remote Sensing of
Environment, 77(2), 173 185.
Wu, C. S., & Murray, A. T. (2003). Estimating impervious surface distribution by spectral mixture analysis. Remote Sensing of Environment,
84(4), 493 505.
Wulder, M., & Boots, B. (1998). Local spatial autocorrelation characteristics of remotely sensed imagery assessed with the Getis statistic.
International Journal of Remote Sensing, 19(11), 2223 2231.
Zanoni, V., Stanley, T., Ryan, R., Pagnutti, M., Baldridge, B., Roylance, S.,
Snyder, G. L., & Lee, G. (2003). The joint agency commercial imagery
evaluation (JACIE) team: Overview and IKONOS joint characterization
approach. Remote Sensing of Environment, 88, 17 22. (doi:10.1016/
S0034-4257(03)00226-8)
Use of IKONOS and Landsat for malaria control in the Republic of Korea
Penny M. Masuoka a,*, David M. Claborn b, Richard G. Andre c, Joseph Nigro c,
Scott W. Gordon d, Terry A. Klein e, Hung-Chol Kim e
a
Uniformed Services University of the Health Sciences, Goddard Space Flight Center, Code 920, Greenbelt, MD 20771, USA
b
Navy Disease Vector Ecology and Control Center, Naval Air Station, Box 43, Jacksonville, FL 32212-0043, USA
c
Department of Preventive Medicine and Biometrics, Uniformed Services University of the Health Sciences, Bethesda, MD 20814-4799, USA
d
U.S. Army Center for Health Promotion and Preventive Medicine, 1312 Cobb Street SW, Fort McPherson, GA 30330, USA
e
Preventive Services Directorate, 18th Medical Command, Unit 15281, APO AP 96205-0054, USA
Received 26 April 2002; received in revised form 17 February 2003; accepted 24 April 2003
Abstract
Malaria reemerged in the Republic of Korea (ROK) in 1993. While limited numbers of U.S. soldiers in high-risk areas use chloroquine/
primaquine chemoprophylaxis to prevent malaria, control of mosquito larvae through larviciding also can be used to reduce the risk of
malaria transmission. In order to estimate the cost of larviciding, accurate estimates of the spatial extent of mosquito larval habitats are
necessary. The purpose of this study was to determine whether an accurate estimate of the area covered by mosquito larval habitats can be
obtained using Landsat 7 Enhanced Thematic Mapper+ (ETM+) and/or IKONOS data for the Korean test site.
To estimate the area covered by larval habitats near Camp Greaves [Paekyeon-Ri, near Tongil-Chon (village)] in the ROK, an IKONOS
and a Landsat 7 ETM+ image were classified using a parallelepiped classification. In a comparison with rice paddy field sites, 24 (92%) of
the sites were classified correctly on the IKONOS image and 17 (65%) were classified correctly on the Landsat image. Comparing the
classifications on a pixel-by-pixel basis, the agreement between the two classifications was 79%. Part of the disagreement was due to the
difference in resolution of the two images. In spite of local differences, the two classifications produced similar area estimates.
Although either Landsat or IKONOS could be used in Korea for a reasonable estimate of habitat area, only IKONOS can resolve small
irrigation ponds. While ponds represent a small portion of the total larval habitat, they are an important source for mosquito breeding during
the late rice-growing season in the ROK since they contain higher larval densities. High-resolution imagery, such as IKONOS, would be
necessary for planning and implementing treatment of these smaller habitats.
D 2003 Elsevier Inc. All rights reserved.
Keywords: IKONOS; Landsat; Malaria control
1. Introduction
After being absent from the Republic of Korea (ROK)
since the 1970s, Plasmodium vivax malaria reemerged in
1993 with the occurrence of two cases (Chai et al., 1994).
The number of cases has grown almost every year since,
resulting in 1642 cases in 1997 (Feighner, Pak, Novakoski,
Kesley, & Strickman, 1998) and peaking in 2000 with 4142
cases (Korea National Institute of Health). The focus of
malaria has been just south of the Demilitarized Zone
(DMZ) in Kyonggi and Kangwon Provinces. The primary
188
Table 1
Characteristics of IKONOS and Landsat images
Image
Band
Wavelength
(Am)
IKONOS
(1) Blue
(2) Green
(3) Red
(4) Near-IR
Panchromatic
0.45 0.52
0.52 0.60
0.63 0.69
0.76 0.90
0.45 0.90
4
4
4
4
1
Landsat ETM +
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
0.45 0.52
0.53 0.61
0.63 0.69
0.75 0.90
1.55 1.75
10.4 12.5
2.09 2.35
0.52 0.90
30
30
30
30
30
60
30
15
Image
IKONOS
Landsat ETM +
11 11
185 185
Blue-green
Green
Red
Near-IR
Mid-IR
Thermal
Mid-IR
Panchromatic
Resolution
(m)
2. Methods
2.1. Field work
Field work for this project was performed from June
through September 2000, and concentrated on two military
bases near the DMZ: Camp Greaves and Camp Casey.
Camp Greaves [Paekyeon-Ri, near Tongil-Chon (village)]
is located in a rural area just south of the DMZ. Camp Casey
(Tongducheon) is approximately 35 miles east of Camp
Greaves in a more populated area with less agriculture.
Standard larval survey techniques using a plastic dipper
in all types of standing fresh water were conducted at both
sites. Seven potential types of larval habitats were identified
and sampled: (1) rice fields, (2) streamside pools, (3)
irrigation ponds, (4) irrigation ditches, (5) drainage ditches,
(6) swamps, and (7) rivers. Only two small swamps occurred in the study area and were not found to be important
larval habitats. The Imjin River parallels the rear boundary
of Camp Greaves and could not be sampled due to military
security measures. However, a nearby upriver site was
sampled and was negative for mosquito larvae.
During larval sampling, water quality testing was conducted to determine the potential effect of minerals on
mosquito larvae populations. Samples were tested for nitrate
and phosphate concentrations, total dissolved solids, and pH.
Statistical analysis of these data showed that none of these
factors was predictive for the presence or absence of Anopheles larvae (Claborn, Hshieh et al., 2002). These results
suggest that most standing or slowly moving fresh water is
suitable as a larval habitat for A. sinensis in this area.
Each field site was located by using a Garmin III Global
Positioning System (GPS) unit (Garmin International,
Olathe, KS, http://www.garmin.com/). Since GPS readings
could not be taken in the center of the rice fields due to
potential crop damage, four readings were taken at the
corners of the field and then the points were averaged to
get an accurate estimate of the center of the field. GPS
points were plotted on topographic maps in the field using
ArcView version 3.2 (Environmental Research Systems
Institute, Redlands, CA, http://www.esri.com/). After the
completion of the fieldwork, points also were displayed
on the IKONOS image. Almost all of the 93 points collected
were located within 5 m of the actual ground points. Four
points had a greater locational error, possibly due to
electromagnetic interference from nearby transmission lines
(Earth Observation Magazine (EOM) Archives, 2002, http://
www.eomonline.com/Common/Archives/Oct95/gps.htm,
April 15, 2002) or from deliberate interference from local
189
Fig. 1. IKONOS image with buffer zone around Camp Greaves and larval sampling sites. (Includes material from Space Imagingn.)
190
the Camp Greaves and the Camp Casey sites. The IKONOS
image lies entirely within the Landsat image but only covers
the Camp Greaves site. The images were georeferenced to a
UTM projection with a WGS-84 datum. A visual comparison of the images with each other and with georeferenced
topographic maps showed that the geometric correction of
the images was good.
PCI remote sensing software (PCI Geomatics, Richmond
Hill, Ontario, Canada, http://www.pcigeomatics.com) was
used to perform supervised classifications on the IKONOS
and Landsat images. Training sites for the classification
were selected at locations where researchers had sampled
Fig. 2. Comparison of IKONOS false color composite and classification with Landsat. On classification images, the river class is shown in yellow, rice fields in
green, and ponds in red. Each image is 5.5 km in width. (Includes material from Space Imagingn.)
3. Results
3.1. Classification
Figs. 2 and 3 show the result of the classifications of the
Landsat and IKONOS images. Based on a visual comparison between the classification and the original image,
191
Fig. 3. An enlarged subset of the images from Fig. 2 comparing an IKONOS false color composite and classification with Landsat. On classification images,
the river class is shown in yellow, rice fields in green, and ponds in red. Image width is 1.9 km. (Includes material from Space Imagingn.)
192
Table 2
Confusion matrices of land cover at larval sample sites compared to land
cover determined by classification of the satellite imagery
Classified data
Confusion matrixIKONOS
Rice
24
Pond
0
Ditch
0
Swamp
0
Unknown
2
Total
26
Pond
Ditch
Swamp
Total
0
0
0
0
3
3
4
0
0
0
9
13
0
0
0
0
1
1
28
0
0
0
15
43
8
0
0
0
5
13
1
0
0
0
0
1
28
0
0
0
15
43
The same sample site locations were used for both matrices.
Fig. 5. Image showing pixels (in black) that were classified as habitat on
Landsat but as nonhabitat on IKONOS. Image covers the same area as
Fig. 2 and is 5.5 km wide.
193
Table 4
Cost comparison of chemoprophylaxis to larviciding for control of malaria
within two military bases and a surrounding 1-km buffer zone
Camp Greaves
Camp Casey
430.4 ha of habitat
Larvicide treatment = US$40,263.13
Chemoprophylaxis cost for 760
persons = US$28,522.80
122.5 ha of habitat
Larvicide treatment = US$11,450.37
Chemoprophylaxis cost for 8000
people = US$330,264.00
Fig. 6. Image showing pixels (in black) that were classified as habitat on
IKONOS but as nonhabitat on Landsat. Image covers the same area as
Fig. 2 and is 5.5 km wide.
Image
IKONOS
Landsat
Rice fields
Ponds
River
Nonhabitat
4,198,151
48,709
1,465,431
6,789,022
4,304,250
None
1,604,925
6,502,500
Nonhabitat includes all areas other than rivers, rice fields, and ponds.
4. Conclusions
We have found that similar land cover area estimates of
mosquito larval habitat can be obtained from IKONOS and
Landsat 7 ETM+ imagery. For estimating the costs of
larviciding and other types of planning, which need only
rough estimates of the major habitat areas defined, Landsat
would be adequate. On a local level, the IKONOS image
allowed a better classification of rice fields, accurately
identifying 92% of the field sites versus 65% for the Landsat. Although we were unable to classify the three very
small ponds sampled in the field, many larger ponds were
visible on the IKONOS image. The use of IKONOS has the
advantage of being able to portray and classify land cover
features such as ponds and rice fields that are less than
30 30 m in size. Although ponds represent a relatively
small portion of the total habitat area, they are an important
breeding habitat for mosquitoes in Korea since they contain
higher larval densities than the rice fields late in the growing
season (Claborn, Hshieh et al., 2002). For areas where small
features represent the majority of the habitat, high-resolution
imagery would be necessary. In addition, high-resolution
imagery would be much more useful in planning sampling
collections and larviciding tasks than Landsat imagery.
The two imagery types could be used together in planning and implementing a malaria control program. Landsat,
in conjunction with information on the location and number
194
Acknowledgements
The work described in this paper was partially supported
by NASA grant NAG5-8532. The authors would like to
thank the following people for their support: Locke Stuart,
Nancy Maynard, Donald Roberts, Rita Aissi-Wespi, McKinley Rainey, William Herman, Kenneth McPherson, and
Alex Ornstein. The soldiers of the 702 Preventive Medicine
Section, 2nd Infantry Division, 18th Medical Command,
provided assistance in the field work phase of this research.
Several anonymous reviewers and the editors of this issue
provided helpful comments and suggestions.
References
Arvidson, T., Gasch, J., & Godward, S. N. (2001). Landsat 7s long term
acquisition planan innovative approach to building a global archive.
Special issue on Landsat 7. Remote Sensing of Environment, 78(1 2),
13 26.
Beck, L. R., Rodriguez, M. H., Dister, S. W., Rodriguez, A. D., Rejmankova, E., Ulloa, A., Meza, R. A., Roberts, D. R., Paris, M. A., Spanner,
R. K., Washino, C., Hacker, L. J., & Letgers, J. F. (1994). Remote
sensing as a landscape epidemiological tool to identify villages at high
risk for malaria transmission. American Journal of Tropical Medicine
and Hygiene, 51(3), 271 280.
Chai, I. H., Lim, G. I., Yoon, S. N., Oh, W. I., Kim, S. J., & Chai, J. Y.
(1994). Occurrence of tertian malaria in a male patient who has never
been abroad. Korean Journal of Parasitology, 32, 195 200.
Claborn, D. M., Hshieh, P. B., Roberts, D. R., Klein, T. A., Zeichner, B. C.,
The Woods Hole Research Center, PO Box 296, Woods Hole, MA 02543-0296, USA
Department of Geography, University of Maryland, College Park, MD 20742-8225, USA
Received 9 August 2002; received in revised form 4 June 2003; accepted 4 July 2003
Abstract
High-resolution imagery from the IKONOS satellite may be useful for many resource management applications. We assessed the utility of
IKONOS imagery for applications in the mid-Atlantic region, including mapping of tree cover, impervious surface areas, and riparian buffer
zone variables in relation to stream health ratings. We focused on a 1313-km2 area in central Maryland using precision-georeferenced
IKONOS products. We found the IKONOS imagery to be a valuable resource for these applications, and were able to achieve map accuracies
comparable to manual aerial photo interpretation. We were also able to use derived data sets for consistent assessments over areas that would
be difficult to accomplish with traditional photographic mapping methods. For example, we found that a stream health rating of excellent
required no more than 6% impervious cover in the watershed, and at least 65% tree cover in the riparian zone. A rating of good required less
than 10% impervious and 60% tree cover. A number of issues associated with application of the IKONOS data arose, however, including
logistics of image acquisition related to phenological and atmospheric conditions, shadowing within canopies and between scene elements,
and limited spectral discrimination of cover types. Cost per unit area was also a nontrivial consideration for the image data products we used,
but allowed us to provide valuable derived products to agencies in support of their planning and regulatory decision-making processes. We
report on both the capabilities and limitations of IKONOS imagery for these varied applications.
D 2003 Elsevier Inc. All rights reserved.
Keywords: IKONOS imagery; Resource management; Mid-Atlantic region
1. Introduction
Information on land cover has become an integral part of
the environmental and developmental planning process. It
has aided in the advancement of more effective land use
planning, habitat assessments, and hydrological applications. Historically, land cover and land use information
was obtained by a combination of field measurements and
aerial photo interpretation. This approach typically required
intensive interpretation by expert analysts, and cross validation methods to ensure that analyst interpretations were
consistent. Recently, satellite imagery has become available
at spatial resolution nearly comparable to aerial photographs, with the added advantage of digital multispectral
* Corresponding author. The Woods Hole Research Center, PO Box
296, Woods Hole, MA 02543-0296, USA. Tel.: +1-508-540-9900; fax: +1508-540-9700.
E-mail address: sgoetz@whrc.org (S.J. Goetz).
0034-4257/$ - see front matter D 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2003.07.010
196
197
Fig. 1. Study area map of Montgomery County, MD, within the mid-Atlantic region and Chesapeake Bay watershed.
1
2
3
4
Pan
445 516
506 595
632 698
757 853
450 900
Blue
Green
Red
Near-infrared
Panchromatic
198
Fig. 2. Location and acquisition dates of IKONOS image panels of the study area. Note the cloud cover remaining in Panel 7 despite a 14-month acquisition
window. (Includes material from Space Imagingn.)
3. Methods
3.1. IKONOS image preprocessing
The four multispectral IKONOS bands were radiometrically corrected to at-sensor reflectance utilizing the methods
Table 2
IKONOS acquisition metadata
Panel
Date
Local
time
Sun
elevation
Sun
azimuth
Collection
elevation
Collection
azimuth
0
1
3
4
5
7
5/1/2001
5/1/2001
5/23/2001
4/6/2000
6/26/2000
4/20/2000
10:59
11:00
11:02
10:35
10:53
10:45
62.08
62.16
67.36
51.28
66.98
57.25
143.46
143.80
139.09
139.97
126.80
140.47
78.25
88.49
85.83
76.46
79.06
77.35
24.83
95.46
276.78
119.51
301.54
349.31
199
200
Table 3
Comparison of county area classified as tree or nontree using supervised
and unsupervised approaches
Classification type
Type
Percent of area
Unsupervised
Tree
Nontree
Tree
Nontree
44.6
55.4
49.7
50.3
Supervised
Overall
accuracy
j statistic
Z statistic
Supervised
Unsupervised
0.86
0.83
0.72
0.67
15.7
13.3
4. Results
4.1. Tree cover
The extent of the tree cover classifications derived using
the supervised and unsupervised approaches is summarized
in Table 3. The unsupervised classification provided an
estimate of f 5% (67 km2) more forest than the supervised
classification. Accuracy statistics generated for the initial
201
supervised and unsupervised classifications (Table 4) indicate that both classifications were significantly better than
random, but the differences between the classified maps,
while not statistically significant ( p = 0.07), were substantial enough to convince us to take a different approach (i.e.,
decision trees trained with a large sample of reference
data).
The tree cover map derived from the decision tree
approach (Fig. 3) was spatially detailed and boundaries
between scenes are virtually nonexistent, despite the range
of dates in the IKONOS acquisitions in both years and
seasons. This was due, in large part, to the accuracy of the
decision tree algorithm and the abundance of training data.
Trees occupied nearly 38% of the county (Table 5). Areas
obscured by cloud or cloud shadow, which could not be
correctly classified, comprised 7% of the total county area.
The original forest classification done ca. 1992, and
subsequently manipulated for use as training data for our
mapping, depicted just 28% of the county as forested
(36,849 ha). The primary reason for the difference was the
depiction of a forest/nonforest classification in the natural
Fig. 3. IKONOS tree cover map of Montgomery County, MD. Note the forest cover data (from natural features GIS coverage; Fig. 4) used in the areas obscured
by clouds.
202
Table 5
Classification of county using decision tree classifier
Classified type
Area (ha)
Percent of county
Tree
Nontree
Cloud
Total
49,522
72,268
9478
131,268
37.7
55.1
7.2
100
Fig. 4. IKONOS image detail with classification results. (a) IKONOS imagery; (b) 1992 natural features coverage of forest areas; (c) tree cover derived from
imagery using a decision tree classifier; and (d) IKONOS image with natural features coverage and derived tree cover maps overlaid. Note the small patches of
trees in (c) compared to the forest cover in (b) and the substantial amount of tree cover missing from the forest area coverage (d). (Includes material from Space
Imagingn.)
Tree
Nontree
Validation
sample
(natural
features)
Classified
(IKONOS)
Number
correct
Producers
accuracy
Users
accuracy
212,703
287,297
217,224
282,577
208,351
278,236
98.0
96.9
95.9
98.5
203
Fig. 5. Impervious surface areas (yellow) for a portion of Montgomery County derived from IKONOS imagery (right) compared to planimetric coverage
derived from aerial photographs (left), showing changes in residential development since the area was initially mapped (ca. 1993). IKONOS tree cover (green)
is also shown, with the IKONOS panchromatic image as a backdrop. (Includes material from Space Imagingn.)
204
Table 7
Accuracy assessment of impervious surfaces map
Validation
Classified Number Producers Users
sample
(IKONOS) correct accuracy accuracy
(planimetric)
Impervious
108,730
Nonimpervious 31,636
110,466
29,900
98,513
19,683
90.6
62.2
89.2
65.8
Fig. 6. Maps of small watershed aggregated values for (a) stream health rankings, (b) impervious surface cover (%), (c) tree cover (%), and (d) riparian buffer
tree cover (%). (b) (d) were calculated using the IKONOS-derived maps of these land cover variables.
amounts of impervious surface area, tree cover, and riparian tree buffers. Variance ratios ( F statistics) were 42.1,
16.3, and 15.0, respectively, all significant at p < 0.001
(N = 245). Streams with excellent ratings had significantly
lower impervious cover and higher tree cover, both within
the watershed and riparian buffers, than streams rated
good, fair, or poor. This trend continued progressively
through each rating category, with increasing average
impervious area and decreasing tree cover as stream health
decreased. All pairwise comparisons confirmed statistically
significant differences ( p < 0.001) between stream rating
categories. There were no substantial differences between
those streams with preliminary assessments and those with
completed assessments, except in the case of those rated
excellent (n = 32) versus preliminary excellent (n = 6), but
this was likely a result of the small sample size of the
latter.
The goodness of fit test of Hosmer and Lemeshow
(1989) failed to reject the null hypothesis of no difference
between the observed and model-predicted values of stream
health; thus, the multivariate model was an adequate predictor of stream health rating. There is no varianceexplained statistic in logistic regression when the dependent
variable is categorical, but reduction in residual deviance
from the null model was highly significant (v2 = 33.1,
p < 0.001). There was, however, multicollinearity among
variables in the regression. Across watersheds, the extent of
tree cover both within and outside of the 100-ft buffer was
positively correlated with each other (r = 0.71), so the
properties of the 100-ft buffer were not independent of
the surrounding land cover. Impervious area and total tree
cover were inversely correlated (r = 0.55), as were impervious area and tree buffers (r = 0.43), but these relationships were reduced in agricultural areas, which had both
low tree and impervious cover. As implied by the ANOVA,
the individual independent variables in the logistic regression were each statistically significant, and stepwise inclusion tests revealed that impervious cover was the primary
predictive variable, followed by tree cover within the
buffers, and then total tree cover. Residual deviance was
significantly reduced when riparian buffers were added to a
model based on impervious surfaces alone, but subsequent
addition of total watershed tree cover did not further
contribute to model significance. The results suggest that
guidelines for excellent stream health rating would be no
more than 6% impervious with at least 65% forested
Table 8
Small watershed sample size and average statistics by stream health rating
category
Stream health
rating
Area
(km2)
Impervious
(%)
Tree
cover (%)
Buffered
(%)
Excellent
Good
Fair
Poor
38
81
76
50
272
658
451
356
3.6
4.9
13.9
19.5
50.6
44.6
37.0
29.6
76.8
71.3
63.2
56.3
205
5. Discussion
5.1. Mapping applications
Land cover classification in the study region using
IKONOS imagery with traditional supervised and unsupervised approaches produced substantially different maps
and, for example, tree cover estimates. There were several
factors that impacted the classifications, including the
timing of data acquisition (thus phenological stage),
associated viewing and atmospheric conditions, the quality of training data (in the case of supervised classification), and insufficient discrimination resulting from
limited IKONOS spectral bands. Some of the more
pronounced classification errors were between wet agricultural fields, dense forests, and shadows. For example,
spring small grains (wheat, barley, and rye) and hay
(alfalfa and grasses) had emerged in agricultural fields,
which produced a similar spectral response as deciduous
trees in partial leaf flush. In other areas, the visibility of
the forest floor through the tree canopy produced spectral
responses similar to residential areas. Shadows were also
an especially important consideration with the use of this
high spatial resolution imagery, particularly shadowing
within the forest canopy from adjacent trees. Some of
these issues can be reduced through the use of multitemporal imagery, but this would be logistically difficult
206
207
Acknowledgements
This work was funded by NASAs Earth Science
Applications Division (grants NAG1302010 and NAG1399011). We acknowledge Montgomery County MNCPPC (Mary Dolan and Nazir Baig) and Montgomery
DEP (Cameron Wiegand) for data sets and collaborative
assistance. IKONOS image data were acquired by Space
Imaging and provided by the NASA Scientific Data
Purchase program.
6. Conclusions
Utilizing high spatial resolution imagery can be beneficial to many different resource management applications.
Imagery like IKONOS can aid in the development of a wide
range of mapping and spatial modeling applications, but a
number of issues must be considered with these relatively
new and unique data sets. The issues include (i) programmatic considerations of the timing of acquisitions to capture
features of interest (e.g., leaf-off for impervious versus leafon for tree cover), as well as timely data acquisition (ours
took 18 months to acquire); (ii) technical considerations,
such as compensation for the effects of object shadowing
resulting from improved resolution of individual scene
elements, and limited spectral resolution and range; and
(iii) economic factors, such as cost per square kilometer
relative to DOQs, and feasibility of consistent repeat
acquisitions. For example, we noted an inability of IKONOS to adequately spectrally discriminate some land cover
types due to high spatial variability within scene elements
resulting from variable illumination and viewing conditions.
Spectral variability within scene objects also contributed to
reduced class type discrimination between more generalized
land cover types (e.g., confusion between deciduous forest
and some agricultural crops).
New sets of interpretation strategies need to be developed to maximize the information obtained from IKONOS,
while minimizing the problematic issues specific to high
spatial resolution imagery. Our exploration of several such
approaches relevant to resource mapping applications suggests the great practical utility of IKONOS imagery, particularly for impervious surface, tree cover, and riparian
buffers, all of which are related to stream health. Our results
provided very specific guidelines for predicting stream
health ratings, which allows for targeted and adaptive
References
Birk, R.J., Stanley, T., & Snyder, G.L. (2003). U.S. Government commercial
data purchase programs for use in government-sponsored science and
applications activities. Remote Sensing of Environment, 88, 3 16.
(doi:10.1016/S0034-4257(03)00227-X)
Blonski, S. (2001). IKONOS-based simulations of Landsat 7 VNIR data:
Comparison with actual, coincident images. In Proceedings of the 2001
High Spatial Resolution Commercial Imagery Workshop, Greenbelt,
MD, USA, March 19 21 (sponsored by NASA/NIMA/USGS Joint
Agency Commercial Imagery Evaluation Team) (CD-ROM).
Breiman, L., Freidman, J., Olshend, R., & Stone, C. (1984). Classification
and regression trees. Monterey, CA: Wadsworth.
Cooper, J. R., Gilliam, J. W., Daniels, R. D., & Robarge, W. P. (1987).
Riparian areas as filters for agricultural sediment. Soil Science Society of
America Journal, 51, 416 420.
Correll, D. L. (1997). Buffer zones and water quality protection: General
principles. In N. E. Haycock, T. P. Burt, K. W. T. Goulding, & G. Pinay
(Eds.), Buffer zones: their processes and potential in water protection
( pp. 7 20). Harpenden, UK: Quest Environmental.
Dial, G.F., Bowen, H., Gerlach, B., Grodecki, J., & Oleszczuk, R. (2003).
IKONOS satellite, sensor, imagery, and products. Remote Sensing of
Environment, 88, 23 36. (doi:10.1016/S0034-4257(03)00229-3)
Fisher, J., & Goetz, S. J. (2001). Considerations in the use of high spatial
resolution imagery: An applications research assessment. In American
Society for Photogrammetry and Remote Sensing (ASPRS) Conference
Proceedings, St. Louis, MO (p. 8), Available from ASPRS http://
www.asprs.org and at http://www.geog.umd.edu/resac.
Friedl, M. A., & Brodley, C. E. (1997). Decision tree classification of land
cover from remotely sensed data. Remote Sensing of Environment, 61(3),
399 409.
Friedl, M., Brodley, C., & Strahler, A. (1999). Maximizing land cover
classification accuracies produced by decision trees at continental to
global scales. IEEE Transactions on Geoscience and Remote Sensing,
37(2), 969 981.
Goetz, S. J. (1997). Multi-sensor analysis of NDVI, surface temperature
and biophysical variables at a mixed grassland site. International Journal of Remote Sensing, 18(1), 71 94.
Goward, S.N, Davis, P.E., Fleming, D., Miller, L., & Townshend, J.R.G.
(2003). Empirical comparison of Landsat 7 and IKONOS multispectral
208
Department of Geography, 2181 LeFrak Hall, University of Maryland, College Park, MD 20242, USA
Earth Science Applications Division, NASA Stennis Space Center, Stennis Space Center, MS 39529, USA
c
Remote Sensing Directorate, Lockheed Martin Space Operations-Stennis Programs, John C. Stennis Space Center, MS 39529 USA
b
Received 7 February 2003; received in revised form 28 April 2003; accepted 10 June 2003
Abstract
Over the last decade, NASA and other federal agencies have been increasingly encouraged to explore commercial sources of land remote
sensing data rather than pursuing government-funded sources of these measurements (prior to U.S. Commercial Executive, 2003. U.S.
Commercial Remote Sensing Policy. White House, Washington, DC). The science and applications users have been skeptical of this move to
commercial sources of observations both because of uncertainty over the capabilities of the private sector and the unsatisfactory outcomes of
previous efforts to commercialize satellite-acquired land remote sensing observations. In an effort to more fully explore the potential of
commercial remotely sensed land data sources, the NASA Earth Science Enterprise (ESE) implemented an experimental Scientific Data
Purchase (SDP) that solicited bids from the private sector to meet ESE-user data needs. The SDP activity spanned nearly 5 years and supplied
many US and international researchers with sources of land remote sensing observations that had not been previously available. The images
from the Space Imaging IKONOS system provided a particularly good match to the current ESE missions such as Terra and Landsat 7 and
therefore serve as focal point in this analysis.
Throughout the SDP process, there have been many lessons learned concerning interactions between US industry, government agencies
and the science user community. The specifics of the Space Imaging IKONOS experience under the NASA SDP are most valuable with
respect to possible future uses of commercial vendors to supply NASA ESE user needs for space-acquired land observations. Areas where
valuable lessons were learned included the technical, scientific, proprietary and management aspects of the interactions. As this activity has
evolved, user confidence in the technical and scientific qualities of the IKONOS measurements has increased substantially. There are still
areas where further progress could be achieved, with respect to proprietary and management aspects of scientific commercial data buys.
To date, the NASA scientific and applications users who have examined the IKONOS imagery have found the data to be of high quality,
providing substantial value to their specific pursuits. They have found that the novel attributes of IKONOS, particularly in the spatiotemporal
domain have introduced new analysis challenges not previously experienced with EOS sensors such as Landsat and MODIS. The technical
qualities of the observations have been substantially improved during the SDP activity as a result of independent validation and verification
by the Joint Agency Commercial Imagery Evaluation (JACIE) of the IKONOS observations.
The experience gained from the Space Imaging IKONOS SDP activity, suggests that US private sector is technically capable of meeting
the needs of NASA ESE science and application users. The future success of such interactions between industry, government and users
appears to be far more dependent on the organizational and legal aspects of such arrangements than technical capabilities of the data
providers.
D 2003 Elsevier Inc. All rights reserved.
1. Introduction
The NASA Scientific Data Purchase (SDP) experiment
to acquire scientific-quality satellite-acquired remotely
* Corresponding author. Tel.: +1-301-405-2770; fax: +1-301-3149299.
E-mail address: sgoward@umd.edu (S.N. Goward).
0034-4257/$ - see front matter D 2003 Elsevier Inc. All rights reserved.
doi:10.1016/j.rse.2003.06.007
210
Fig. 1. The spectral band passes of IKONOS versus Landsat 7. Note the differences particularly in the near infrared. This leads to a divergence in spectral
measurements from the two systems (see Goward et al., this issue for further details).
Table 1
Spectral band passes for IKONOS and Lansat-7 visible and near-infrared
bands
Spectral band
Blue
Green
Red
NIR
445 516
506 595
632 698
757 853
450 515
525 605
630 690
750 900
nm
nm
nm
nm
nm
nm
nm
nm
moving from local to global-scale observatories (Woodcock & Strahler, 1987) (Fig. 2).
The radiometric precision of the IKONOS sensor, with its
11-bit analog to digital data conversion, is a factor of 8
improvement compared to the 8-bit radiometry for
Landsat 7s ETM+, providing a comparable improvement
in radiometric measurement precision that IKONOS
supplies spatially. Potentially, this combination of radiometric precision and spatial detail could provide the high
quality means needed to validate the lower resolution
measurements from Landsat and other systems.
The orbital equatorial crossing time is also nearly the
same as Landsat, Terra, EO-1 and SPOT, although the
IKONOS systems nadir, repeat frequency is much lower
at ~ 140 days, a necessary trade-off for the increased
spatial resolution of the IKONOS system. With platform
agility it can revisit most locations with 3 days within
F 30j from nadir (Dial et al., 2003).
As a result, IKONOS supplies an important step in
multiscale land remote sensing, one that is most closely
associated with field measurements and ground truth.
The similarities in spectral measurements and orbit that
exist between IKONOS and other systems, combined with
the increased radiometric precision and spatial resolution of
the IKONOS satellite, make it a valuable tool in comparing
and evaluating higher and lower resolution systems. Thus,
the IKONOS images offer significant potential in providing
the validation and local assessments needed to support
regional and global scale studies of the earth system, using
Landsat, SPOT, ASTER, MODIS, AVHRR and VEGETATION as the primary observatories.
211
Fig. 2. Comparison of IKONOS (left); acquisition date: 04/02/2001, image upper left corner latitude/longitude: 45.45/12.38, and Landsat 7 (right); acquisition
date: 08/26/2001, image upper left corner latitude/longitude: 47.0/11.12 observations for the city of Venice, Italy. The differing spatial resolution of the two
sensors is clear. The trade off is that Landsat systematically monitors all land areas of the earth seasonally whereas the IKONOS system can only sample small
portions of the earths land areas each year.
212
4. Lessons learned
From 1999 to present, the NASA SDP procured approximately 2400 IKONOS data products (550 GB), costing
over $11 million, for NASAs Earth Science Enterprise
researchers. These data sets have been used to support a
wide range of earth science applications, as noted by the
papers published in this special issue.
There is no question that NASA-funded researchers have
enjoyed the access to these novel observations that has been
provided under this experimental NASA Scientific Data
Purchase. Analysis of these observations has posed some
new challenges, but they have also provided new insights
about land cover properties. Many of the early concerns that
the NASA ESE community of users had about the measurements have been alleviated through working with the actual
data.
An observatory as novel as IKONOS takes some time to
adjust to both in terms of getting used to the new perspective provided as well as becoming comfortable with the
technical characteristics of the observations. As stated in the
JACIE paper (Zanoni et al., 2003), it took a large team of
engineers and scientists from US government laboratories
and universities over 3 years to complete a detailed assessment of metric properties of the imagery supplied by Space
Imaging. The extent of the IKONOS validation is consistent
with prior validation experiences with government-funded
ESE science missions. It takes time to develop full confidence in the output of new observatories. Having a team of
scientists and engineers working on such system validation
is vital to mission scientific success (Goward & Masek,
2001; Justice & Townshend, 1994, 2002).
4.1. Technical assessment
Overall, NASA and affiliated researchers, in conjunction
with Space Imaging engineers and technicians, believe that
the IKONOS system supplies high quality, high spatial
resolution multispectral imagery that is suitable to support
the research goals and requirements of NASA-supported
Earth scientists. Validation and independent scientific assessment of the IKONOS observations extended over 4
years. A scientifically comprehensive validation takes considerable time and effort. After 4 years, there are few
technical questions related to sensor performance and Space
Imaging-applied, post-acquisition processing that remain to
be addressed. Technical areas that have been evaluated and
required detailed attention include:
213
214
Fig. 3. (a) IKONOS imagery from Antarctica; acquisition date: 10/05/2000; image upper left corner latitude/longitude: 65.25/ 60.89. (b) IKONOS image
of Kerguelen Island in the Indian Ocean; acquisition date: 09/24/2002; image upper left corner latitude/longitude:
49.09/70.54. These observations
demonstrate the value of a space-based high spatial resolution observatory. Acquisition of such detailed imagery from aircraft would have been quite difficult
and very expensive. A space-based system such as IKONOS provides ready access to all areas of the earths land areas, an important attribute for an
observation system that is used to study global changes.
215
Fig. 4. Seasonal changes recorded between adjacent IKONOS scenes acquired from an area in Maryland just north of Washington, DC; one acquired on 04/07/
2000, image upper left corner latitude/longitude: 39.36/ 77.16, and one acquired on 04/06/2000, image upper left corner latitude/longitude: 39.33/ 77.02.
Note that at this time of the year, vegetation foliage are rapidly growing so that the scene from the early date shows much less green foliage than the later date.
Fig. 5. Impact of haze and cirrus clouds on observation quality for two images acquired of Congo, Africa. Note the significant loss of contrast in the image on
the right, acquired on 03/26/2001, image upper left corner latitude/longitude: 1.19/16.01, due to haze and clouds (the image without the haze and cirrus clouds,
on the left, was acquired on 10/23/2000).
216
217
218
Acknowledgements
This assessment of the NASA Scientific Data Purchase
experience with the Space Imaging IKONOS observations
was carried out with support from the Scientific Data
Purchase administered by the civil servant and contractor
staff at the NASA Stennis Space Center. Drs. Goward and
Townshend received support from NASA Grant NAG
1398001 from NASA Stennis to pursue this study. All of
the participants in the JACIE program as well as contributors to the JACIE Workshops dedicated to IKONOS
assessment substantially contributed to the understanding
noted here. It is our hope that we have properly represented
all participants perspectives in this regard.
References
Andrefouet, S., Kramer, P., Torres-Pulliza, D., Joyce, K. E., Hochberg, E. J.,
Garza-Perez, R., Mumby, P. J., Riegl, B., Yamano, H., White, W. H.,
Zubia, M., Brock, J. C., Phinn, S. R., & Muller-Karger, F. E. (2003).
Multi-sites evaluation of IKONOS data for classification of tropical coral
reef environments. Remote Sensing of Environment, 88, 127 142.
(doi:10.1016/j.rse.2004.04.005)
Arvidson, T., Gasch, J., & Goward, S. N. (2001). Landsat 7s long term
acquisition planan innovative approach to building a global archive,
special issue on Landsat 7. Remote Sensing of Environment, 78(1 2),
13 26.
Birk, R. J., Stanley, T., & Snyder, G. L. (2003). U.S. government commercial
data purchase programs for use in government-sponsored science and
applications activities. Remote Sensing of Environment. (doi:10.1016/
j.rse.2003.07.007)
Bowen, H. S. (2002). Absolute radiometric calibration of the IKONOS
sensor using radiometrically characterized stellar sources. Pecora
15/Land Satellite Information IV/ISPRS Commission I/FIFOS 2002,
pp. CD-ROM. Denver, CO: American Society of Photogrammetry and
Remote Sensing.
Dial, G., Bowen, H., Gerlach, F., Grodecki, J., & Oleszczuk, R. (2003).
IKONOS satellite, imagery, and products. Remote Sensing of Environment, 88, 23 36. (doi:10.1016/j.rse.2003.08.014)
Goetz, S. J., Wright, R., Smith, A. J., Zinecker, E., & Shaub, E. (2003).
IKONOS imagery for resource management: tree cover, impervious
surfaces and riparian buffer analyses in the Mid-Atlantic region. Remote
Sensing of Environment, 88, 194 207. (doi:10.1016/j.rse.2003.07.010)
Goward, S. N., Davis, P. E., Fleming, D., Miller, L., & Townshend, J. R. G.
(2003). Empirical comparison of Landsat 7 and IKONOS multispectral
219
Pace, S., Frelinger, D., Lachman, B., Brooks, A., & Gabriele, M.
(2000). The earth below: Purchasing science data and the role of
public private partnerships. Washington, DC: Rand Science and
Policy Institute.
Pace, S., Sponberg, B., & Maccauley, M. (1999). Data policy issues and
barriers to using commercial resources for mission to planet earth.
Washington, DC: Rand Science and Policy Institute.
Pagnutti, M., Ryan, R., Kelly, M., Holekamp, K., Zanoni, V., Thome, K., &
Schiller, S. (2003). Radiometric characterization of IKONOS multispectral imagery. Remote Sensing of Environment, 88, 52 67. (doi:10.1016/
j.rse.2003.07.008)
Ryan, R., Baldridge, B., Schowedgerdt, R., Choi, T., Helder, D., & Blonski,
S. (2003). IKONOS spatial resolution and image interpretability. Remote Sensing of Environment, 88, 37 51. (doi:10.1016/j.rse.
2003.07.006)
Sawaya, K., Olmanson, L., Heinert, N., Brezonik, P., & Bauer, M. (2003).
Extending satellite remote sensing to local scales: Land and water resource monitoring using high-resolution imagery. Remote Sensing of
Environment, 88, 143 155. (doi:10.1016/j.rse.2003.04.006)
Seelan, S. K., Laguette, S., Casady, G. M., & Seielstad, G. A. (2003).
Remote sensing applications for precision agriculture: A learning community approach. Remote Sensing of Environment, 88, 156 168.
(doi:10.1016/j.rse.2003.04.007)
Small, C. (2003). High resolution spectral mixture analysis of urban reflectance. Remote Sensing of Environment. (doi:10.1016/j.rse.
2003.04.008)
Teillet, P. M., & Holben, B. N. (1993). Toward operational radiometric
calibration of NOAA AVHRR imagery in visible and near infrared channels. Canadian Journal of Remote Sensing (submitted for publication).
Thenkabail, P. S., Stucky, N., Griscom, B. W., S.Ashton, M., Diels, J.,
Meer, B. V. D., & Enclona, E. (2003). Biomass estimations and carbon stock calculations in the oil palm plantations of African derived
savannas using IKONOS data. Remote Sensing of Environment (in
review).
Townshend, J. R. G., Huang, C., Kalluri, S. N. V., Defries, R. S., Liang, S.,
& Yang, K. (2000). Beware of per-pixel characterization of land cover.
International Journal of Remote Sensing, 21, 839 843.
Townshend, J. R. G., & Justice, C. O. (2002). Towards operational monitoring of terrestrial systems by moderate resolution imaging. Remote
Sensing of Environment, 83, 351 359.
Woodcock, C. E., & Strahler, A. H. (1987). The factor of scale in remote
sensing. Remote Sensing of Environment, 21, 311 322.
Zanoni, V., Stanley, T., Ryan, R., Pagnutti, M., Baldridge, B., Roylance, S.,
Snyder, G. L., & Lee, G. (2003). The joint agency commercial imagery
evaluation (JACIE) team: Overview and IKONOS joint characterization
approach. Remote Sensing of Environment, 88, 17 22. (doi:10.1016/
j.rse.2003.07.005)