Geoscience Reference
In-Depth Information
9.6 Conclusions...........................................................................................................................128
9.7 Summary...............................................................................................................................128
Acknowledgments ..........................................................................................................................129
References ......................................................................................................................................129
9.1 INTRODUCTION
There is intense interest among federal agencies, states, and the public to evaluate environmental
conditions on community, watershed, regional, and national scales. Advances in computer technol-
ogy, geographic information systems (GIS), and the use of remotely sensed image data have provided
the first opportunity to assess ecological resource conditions on a number of scales and to determine
cross-scale relationships between landscape composition and pattern, fundamental ecological pro-
cesses, and ecological goods and services. Providing quantifiable information on the thematic and
spatial accuracy of land-cover (LC) data derived from remotely sensed sources is a fundamental step
in achieving goals related to performing large spatial assessments using space-based technologies.
Remotely sensed imagery obtained from Earth-observing satellites now spans three decades,
making possible the mapping of LC across large regions by the classification of satellite images.
However, the accuracy of these derived maps must be known as a condition of the classification.
Theoretically, the best reference data against which to evaluate classifications are those collected
on the ground at or near the time of satellite overpass. However, such data are rarely available for
retrospective multitemporal studies, thus mandating the use of alternative data sources. Accordingly,
the U.S. Environmental Protection Agency (EPA) has established a priority research area for the
development and implementation of methods to document the accuracy of classified LC and land
characteristics databases (Jones et al., 2000).
To meet the ever-growing need to generate reliable LC products from current and historical
satellite remote sensing data, the accuracy of derived products must be assessed using methods that
are both effective and efficient. Therefore, our objective was to demonstrate the viability of utilizing
new high-resolution digital orthophotography along with other airborne data as an effective sub-
stitute when historical ground-sampled data were not available. The achievement of consistent
accuracy assessment results using these diverse sources of reference data would indicate that these
techniques could be more widely applied in retrospective LC studies.
In this study, classification accuracies for four separate LC maps of the San Pedro River watershed
in southeastern Arizona and northeastern Sonora, Mexico (Figure 9.1) were evaluated using historical
aerial photography, digital orthophoto quadrangles, and high-resolution airborne video. Landsat
Multispectral Scanner (MSS) data (60-m pixels) were classified for the years 1973, 1986, and 1992.
Lastly, 1997 Landsat Thematic Mapper (TM) data (30-m pixels) were resampled to 60 m to match
the MSS resolution and classified. All data were analyzed at the Instituto del Medio Ambiente y el
Desarrollo Sustentable del Estado de Sonora (IMADES) in Hermosillo, Mexico. Map accuracy was
assessed by Lockheed-Martin (Las Vegas, Nevada) for 1973 and 1986 and at the University of
Arizona (Tucson, Arizona) for 1992 and 1997. This study incorporated previous accuracy assessment
methods developed for the San Pedro watershed by Skirvin et al. (2000) and Maingi et al. (2002).
9.2 BACKGROUND
9.2.1
Upper San Pedro Watershed Study Area
The study location comprised the upper watershed of the San Pedro River, which originates in
Sonora, Mexico, and flows north into southeastern Arizona. Covering approximately 7600 km
2
(5800 km
in Arizona and 1800 km
in Sonora, Mexico), this area represents the transition between
2
2
