Monitoring Urban Expansion and Land Use/Land Cover Changes in Banadir, Somalia Using Google Earth Engine (GEE)

Abstract

Land Use Land Cover (LULC) changes in the Banadir region are rapidly changing because of the increasing interaction of human activates with the environment as the population increases. However, there is no published evidence on this phenomenon. This study used multi-temporal Landsat images (1989, 2003, and 2018) to extract and evaluate LULC changes in Banadir Somalia. Based on Google Earth Engine (GEE) platform, two Machine Learning (ML) algorithms were used and compared, using supervised classification approaches via Support Vector Machine (SVM) and the Random Forest (RF) classifiers. This study shows that the average classification accuracy of SVM is showing the highest overall accuracy of 96.7% (Kappa = 0.946), which is 1.7% higher than that achieved by RF. The analysis of LULC changes in Banadir for the last three decades revealed that the urban area increased from 43.1 km2 in 1989 to 87.7 km2 in 2018, which means that the urban expanded or increased by two folds. Vegetation land cover dropped down from 276.5 km2 in 1989 to 96.42 km2 in 2018, which is a two-thirds decrease in vegetation in the last 30 years whereas bare surface area increased from 93.74 km2 in 1989 to 229.05 km2 in 2018, which is 1.96% increase per year. This study concludes that Banadir has experienced a dramatic increase in the urban and bare surface and a decrease in vegetation for the past 30 years.