Support vector machine-based spatiotemporal land use land cover change analysis in a complex urban and rural landscape of Akaki river catchment, a Suburb of Addis Ababa, Ethiopia

Abstract

Intense level of land use land cover (LULC) changes has been observed in Sub-Saharan Africa, particularly in the central highlands of Ethiopia, due to rapid population growth and urbanization process. However, quantifying and identifying the rural-urban landscape changes are challenging. In this study, LULC changes during the years 1984, 1990, 2000, 2010, and 2021 have been analyzed using satellite imageries and Support Vector Machine (SVM) algorithms in a heterogenous rural and urban landscape of the Akaki river catchment, central highlands of Ethiopia. The LULC change drivers were evaluated by applying LULC thematic change analysis combined with key informants’ interviews. Seven LULCs that include: Built-up area (BTA), Cropland (CL), Grassland (GL), Waterbody (WB), Plantation Forest (PF), Woodland (WL), and Bareland (BL) were detected. The result shows that 51.3 % of the catchment area has been transformed into other land uses. BTA increased by 24.7 % while GL and WL reduced by 18.1 % and 5.9 % respectively. Large areas of CL (61 %) and GL (22 %) were changed into an urban landscape. The spatial and non-spatial analysis revealed that the major spatiotemporal LULC change drivers between 1984 and 2005 were land use policy and legislation change and the Eucalyptus tree plantation campaign. Whereas, low-cost housing programs, informal settlers, market opportunity, and real estate development were the main drivers for the LULC changes between 2006 and 2021. The study also found the key informant observation and SVM image classification results are aligned and therefore, we found the SVM-based classifications are suited for such complex rural-urban landscape change and pattern analysis. The outcome of this research can contribute to improving land use policy, its management, and public understanding of the LULC dynamics and its implications.

Keywords

• Spatiotemporal
• Support vector machine
• LULC conversion
• Rural-urban landscape
• LULC drivers