The United States Environmental Protection Agency (USEPA) Environmental Monitoring and Assessment Program Landscapes (EMAP-L) group has been charged with assessing the status and trends in the condition of the nation's landscapes. EMAP-L is proposing to use Landsat MultiSpectral Scanning (MSS) and Thematic Mapper (TM) satellite imagery to calculate a series of landscape indicator metrics that relate to landscape condition. This paper will address the sources of error that may result from calculation of landscape metrics from derived data, accumulation of error through the landscape analysis process, and the identification of procedures that may reduce error throughout the entire process of calculating landscape indicators. Several sources of error confound the ability to produce estimates of landscape indicator values with known confidence. Sources of error include the acquisition instrument, georeferencing of the data to known ground coordinates, reduction and aggregation of spectral data, clustering algorithms, labeling and classification of spectral clusters, merging of disparate data sets, and sensitivities and limitations of landscape statistics. Although research has been conducted for a few of these areas, little is known about how each of these contribute to error in calculation of landscape metrics. More importantly, little is known about how error accumulates as data are processed to produce landscape statistics. Knowledge of the sources of error and error propagation will be critical in establishing data quality objectives for landscape indicators. Furthermore, it is critical to identify procedures and methods that will help reduce error associated with each individual source, as well as accumulation of error.
David Totman
Lockheed Environmental Systems & Technologies Co.
980 Kelly Johnson Dr.
Las Vegas, NV 89119
Telephone: 702-897-3329
Fax: 702-897-3312
Email: dave@sun390.las.epa.gov