were obtained using the three alternative methods. Aerial photography provided reference data to
assess the 1973 and 1986 LC maps with overall classification accuracies of 70% (1973) and 67%
(1986). Assignments of class labels to sample points on 1992 reference DOQQs were verified by
comparison with higher-resolution airborne video data, with overall 1992 map classification accu-
racy of 75%. Accuracy assessment of the 1997 products used contemporaneous airborne color
video data and resulted in an overall map accuracy of 72%. There was no evidence of positive bias
in accuracy resulting from use of homogeneous vs. heterogeneous pixel contexts in sampling the
The use of historical aerial photography, high-resolution DOQQs, and airborne videography as
a proxy for actual ground sampling for satellite image classification accuracy has merit. Selection
of a reference data set for this study depended on the date of image acquisition. For example, prior
to 1992, historical aerial photographs were the only data available. DOQQs covered the period
since initiation of the high-resolution NAPP in 1992, and high-resolution airborne videography
provided a cost-effective means of acquiring many reference sample points near the time of image
acquisition. Problems that were difficult to avoid included inadequate sampling of rare classes and
reconciling cover changes between acquisition dates of aerial photography or DOQQs and satellite
image data. Other issues, including the need for consistent geometric rectification and criteria for
mutually exclusive and reproducible LC class descriptions, need special attention when satellite
image classification and subsequent LC map accuracy assessment are performed by different teams.
The U.S. Environmental Protection Agency, Office of Research and Development provided fund-
ing for this work. The authors wish to thank participants from U.S. EPA, Lockheed Martin Environ-
mental Services, Instituto del Medio Ambiente y el Desarrollo Sustentable del Estado de Sonora
(IMADES), and the Arizona Remote Sensing Center at the University of Arizona for their assistance.
Asner, G.P., C.A. Wessman, C.A. Bateson, and J.L. Privette, Impact of tissue, canopy, and landscape factors
on the hyperspectral reflectance variability of arid ecosystems,
Remote Sens. Environ.
, 74, 69-84, 2000.
University of Arizona Press, Tucson, 1991.
Bahre, C.J., Human impacts on the grasslands of southeastern Arizona, in
A Legacy of Change
The Desert Grassland
M.P. and T.R. Van Devender, Eds., The University of Arizona Press, Tucson, 1995.
Brown, D.E., C.H. Lowe, and C.P. Pase, A digitized classification system for the biotic communities of North
America, with community (series) and association examples for the Southwest,
, 14 (Suppl. 1), 1-16, 1979.
Congalton, R., A review of assessing the accuracy of classifications of remotely sensed data,
37, 35-46, 1991.
Congalton, R.G. and K. Green,
Assessing the Accuracy of Remotely Sensed Data: Principles and Practices
CRC Press, Boca Raton, FL, 1999.
Congalton, R.G., R.G. Oderwald, and R.A. Mead, Assessing Landsat classification accuracy using discrete
multivariate statistical techniques,
49, 1671-1678, 1983.
Drake, S.E., Climate-Correlative Modeling of Phytogeography at the Watershed Scale, Ph.D. dissertation,
University of Arizona, Tucson, 2000.
Drake, S.E., Visual interpretation of vegetation classes from airborne videography: an evaluation of observer
proficiency with minimal training,
Photogram. Eng. Remote Sens.,
, 62, 969-978, 1996.
Easterling, D.R., T.R. Karl, E.H. Mason, P.Y. Hughes, D.P. Bowman, R.C. Daniels, and T.A. Boden, United
States Historical Climatology Network (U.S. HCN) Monthly Temperature and Precipitation Data,
Revision 3, Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, Oak Ridge,
Photogram. Eng. Remote Sens.