Locally optimized separability enhancement indices for urban land cover mapping: Exploring thermal environmental consequences of rapid urbanization in Addis Ababa, Ethiopia

• Published in: Remote sensing of environment 2016-03, Vol.175, p.14-31
• Main Authors: Feyisa, Gudina L., Meilby, Henrik, Darrel Jenerette, G., Pauliet, Stephan
• Format: Article
• Language: English
• Subjects: Accuracy; Classification; Climate; Land cover change; Mathematical models; Support vector machines; Surface temperature; Urban areas; Urban heat island; Urbanization; Vegetation
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2015.12.026

Landsat data were used to assess urbanization-induced dynamics in Land use/cover (LULC), surface thermal intensity, and its relationships with urban biophysical composition. The study was undertaken in Addis Ababa city, Ethiopia. Ground-based data and high resolution images were used as reference data in LULC classification. To more accurately quantify landscape patterns and their changes, we applied new locally optimized separability enhancement indices and decision rules (SEI–DR approach) to address commonly observed classification accuracy problems in urban environments. We tested the SEI–DR approach using eight Landsat images acquired between 1985 and 2010. Two approaches were applied to quantify surface heat intensity (SHIn) and to examine its spatial patterns over 25years: thermal gradient analysis and hot spot analysis. A Simultaneous Autoregressive Spatial error model (SARerr) was used to explore relationships between surface temperature and biophysical variables describing urban surfaces. Compared to Maximum Likelihood (ML) and Support Vector Machine (SVM) classification, accuracy improvement achieved through use of the SEI–DR procedure was, respectively, 6% and 5% and the differences were statistically significant (P