The use of Landsat-3 thermal data to help differentiate land covers

Abstract

Landsat-3 Multispectral Scanner Subsystem (MSS) digital data of the Baltimore, Maryland area taken 24 May 1978 were examined to show the potential usefulness of thermal data (Band 8) to provide better discrimination between agricultural and residential areas, certain types of urban/industrial areas and water, cloud shadows and water, and bare-extractive areas and bright urban cover types. High altitude aircraft imagery taken 3 May 1978, was used for ground truth and training site verification. Two classifications were made per training site; the initial one using Bands 4, 5, and 7 and a second in which the thermal data were included with the visible and near infrared data. This allowed a direct comparison of areas spectrally similar with and without the inclusion of the thermal data. Commission errors determined from selected subsets of the data showed reductions of 95% for the urban/industrial versus water themes, 84% for the residential versus agriculture themes, 64.0% for the bare-extractive versus bright urban themes, and 24% for the cloud shadow versus water themes when the thermal data were included in the signature. The omission errors ranged from less than 10% for the cloud shadow versus water to 47% for the residential versus agriculture themes. This indicates that the thermal data can aid in classifying certain themes providing there is good thermal differentiation.