Determination of Land Use from LANDSAT Imagery: Applications to Hydrologic Modeling.
Abstract
An operational procedure for determining land use from LANDSAT imagery was applied to five watersheds: Crow Creek, Walnut Creek, Rowlett Creek, Pennypack Creek, and Castro Valley (University of California, Davis (USD) staff performing the work on the latter three watersheds). Based on this work the following conclusions can be made: At the grid cell level LANDSAT land use was in error approximately one-third of the time. By aggregating land use over the entire watershed, LANDSAT's average misclassification of land use reduces to 2-8% for the major land use categories. The commercial classification was less accurate at the grid cell level than the UCD procedure classification; errors at the watershed level were nearly the same for both. Evaluated in terms of the difference in discharge frequency curves derived using the same hydrologic model but different land use (LANDSAT and conventional), the LANDSAT derived land use found to be completely adequate. The number and type of land use categories derived from LANDSAT data were sufficient to be able to apply two standard hydrologic modeling techniques; Snyder's unit hydrograph and percent imperviousness and the SCS curve number method. The UCD procedure works. It is a complete, self-contained package of computer programs and manual operations that permit a user to identify land use from LANDSAT digital data without requiring the use of expensive interactive image processing equipment. LANDSAT land use can be directly incorporated into a grid cell data bank, thus providing and automated environment for applying the LANDSAT classification in routing hydrologic investigations. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1979
- Accession Number
- ADA102190