Fine spatial resolution satellite sensor imagery for per-field land cover classification

Aplin, Paul (1999) Fine spatial resolution satellite sensor imagery for per-field land cover classification. University of Southampton, Doctoral Thesis.

Record type: Thesis (Doctoral)

Abstract

Land cover information is a useful aid to our understanding and management of the environment. Commonly, the only means of obtaining affordable land cover information at the national scale is via remote sensing. National land cover mapping has been relatively inaccurate and uninformative due to the limited detail provided by coarse spatial resolution satellite sensor imagery. Currently, the finest spatial resolution multispectral satellite sensor imagery in common use is 20 m. Within the next few years, however, finer spatial resolution satellite sensor imagery will become available widely, enabling considerably smaller land cover features to be identified. Notably, three satellite sensors are due for launch within the next year (IKONOS 2, QuickBird, OrbView-3), capable of generating 4 m spatial resolution multispectral imagery.

The potential of fine spatial resolution satellite sensor imagery for national land cover and land use mapping was demonstrated. Airborne sensor imagery was used as a substitute for forthcoming satellite sensor imagery and two local scale studies were used to draw general conclusions of relevance to the remote sensing of land cover at the national scale. Per-pixel classification, the traditional method of mapping from remotely sensed imagery, was relatively inaccurate due to within-field variation. Per-field classification, achieved by integrating remotely sensed imagery with cartographic vector data, was considerably more accurate due to the removal of within-field variation. Fine spatial resolution satellite sensor imagery was of particular value where detailed information was required, such as in urban areas or to generate within-field texture-based measures of discrimination.

Generally, land cover was classified more accurately than land use, since the former was identifiable directly from remotely sensed imagery but the latter was not. Overall, it was demonstrated that per-field classification of forthcoming fine spatial resolution satellite sensor imagery will enable more accurate and informative national land cover mapping than current coarse spatial resolution satellite sensor imagery, thereby increasing our ability to understand and manage the environment.

This record has no associated files available for download.