A Pattern-Based Definition of Urban Context Using Remote Sensing and GIS.

Area of Science:

• Remote Sensing
• Urban Geography
• Geospatial Analysis

Background:

• Rapid urbanization and poverty in Sub-Saharan Africa create complex urban environments not defined by a simple urban-rural dichotomy.
• Existing urban mapping methods inadequately capture the nuanced nature of these diverse urban settings.

Purpose of the Study:

• To propose and validate an alternative gradient-based classification scheme for urban mapping in Ghana.
• To develop a nine-class urban context map that captures the transition from unsettled land to compact urban cores.

Main Methods:

• Utilized Landsat Enhanced Thematic Mapper Plus (ETM+) and European Remote Sensing Satellite-2 (ERS-2) synthetic aperture radar (SAR) imagery.
• Employed Spectral Mixture Analysis (SMA) for land cover classification and landscape metrics for fragmentation analysis.
• Integrated SMA-derived measures with SAR texture data using a decision tree classifier to generate the urban context map.

Main Results:

• A nine-class urban context map was generated with an overall classification agreement of 77%.
• User's accuracy was found to be lower than producer's accuracy, indicating reliable classification.
• The classified urban classes showed appropriate differentiation of areas along the urban gradient when compared with Ghana's 2000 Census data.

Conclusions:

• The proposed gradient approach offers a more nuanced understanding of urban environments compared to traditional dichotomies.
• The developed nine-class map effectively represents the spectrum of urban contexts in the study area.
• This method provides a valuable tool for urban planning and management in rapidly developing regions.