Assessing the applicability of the V-I-S model to map urban land use in the developing world: Case study of Yogyakarta, Indonesia

The extensive expansion of urban centres over agricultural land in Indonesia is poorly monitored due to the ill-equipped land monitoring system. The objective of this study was to assess the applicability of the vegetation-impervious surface-soil (V-I-S) model in classifying land use in the context of medium-sized and densely populated cities of developing countries. The V-I-S model describes the urban environment as a linear combination of vegetation, impervious surface and soil. As such, it is well suited to relate remotely sensed data, which are closely related to land cover, to biophysical aspects of urbanising environments. A Landsat-5 Thematic Mapper image acquired in June 1994 (dry season) was used to demonstrate the usefulness of the V-I-S model to map urban land use of Yogyakarta, Indonesia. The V-I-S model was applied in a hierarchical classification scheme, in which each level of the hierarchy was designed to segment one or more land use classes considering each component of the V-I-S model. The performance of this approach was evaluated by comparison to a standard maximum likelihood classification. Official land use maps produced in 1994 were used to validate both classifications. The results showed that the V-I-S model applied in the hierarchical classification scheme provides more accurate land use maps (80% overall accuracy) than the maximum likelihood algorithm (53% overall accuracy). The higher accuracy is evidence of the relationships between the V-I-S model and urban land use, based on the physical composition. However, the accuracy of the land use map produced, was not at the level of practical application needed by urban planners. Therefore, further studies need to be conducted to improve the accuracy of land use maps produced using the V-I-S model. (c) 2005 Elsevier Ltd. All rights reserved.