Document: ALI-3-89-1

New land cover dataset for supporting regional and national scale landscape analyses in the conterminous U.S.

GALLANT, A.L.*

USGS EROS Data Center, Sioux Falls, SD 57198 USA ¹

Abstract:
A new land cover classification for the conterminous U.S. has been produced by the U.S. Geological Survey's EROS Data Center. This digital classification was designed to support large-area landscape analyses and summaries, and was derived by mosaicking and analyzing some 860 scenes of Landsat Thematic Mapper data representing leaf-on and leaf-off conditions, and ancillary data layers depicting characteristics of terrain, population, soils, hydrology, transportation, and, where available, wetlands and vegetation. The classification contains 21 cover types and reflects conditions centered on 1992. Although several classification approaches were explored during the project, we found that data pre-processing procedures had the greatest effect on the quality of the results. Thus, we are currently developing methods to improve scene selection, reduce data distortion during image mosaicking, and partition the data into mapping zones that decrease within-dataset spectral and ecological variance. An accuracy assessment of the classification has, to date, been completed for the eastern half of the country, and associated analyses are being used to determine the type and spatial distribution of classification errors made in different parts of the country in order to focus future efforts on overcoming these classification challenges. Additionally, we are aiming for a more flexible database structure for upcoming land cover products by using the current data to explore the feasibility of mapping continuous land cover variables and discrete variables that cross-reference land cover types with land uses. These changes in the database will allow users to partition or aggregate land cover characteristics to best suit specific applications.

Keywords: land cover classification, Landsat Thematic Mapper

This abstract is being presented at: 1:30 PM in session:
Oral Session #64: Remote Sensing.