Mapping despoiled land cover from Landsat thematic mapper imagery

Abstract

The potential of remotely sensed imagery as a source of information on industrially despoiled land cover was investigated. Emphasis was placed on issues relating to the separability and mapping of despoiled land cover from Landsat TM imagery for the production of estimates of despoiled land cover extent in administratively defined local districts. Spatial filtering was found to enhance the separability of despoiled land in the TM imagery. A land cover classification derived from Landsat TM imagery which had been preprocessed with an appropriate filter, was integrated with local district boundaries within a geographical information system and used to derive estimates of despoiled land cover extent. These initial estimates were adjusted using information on classification accuracy to derive a remotely sensed estimate of despoiled land cover extent in each local district. These were then evaluated against the estimates derived from a manually produced map of despoiled land cover. The results showed a high degree of correspondence between the remotely sensed and manually mapped estimates, with correlation coefficients of up to 0.93 observed (significant at the 99% level of confidence) and illustrate a potential operational role for remote sensing in identifying and monitoring despoiled land within administratively defined local authority areas.