Exploring Google Earth Engine, Machine Learning, and GIS for Land Use Land Cover Change Detection in the Federal Capital Territory, Abuja, between 2014 and 2023

Abstract

This study aims to visualize various land use land cover (LULC) classes, estimate the net change in LULC types between the years 2014 and 2023, and use transition mapping to track LULC transitions within waterbody, vegetation, bareland, and buildup to better understand how land use types change from one another. Google Earth Engine (GEE), Random Forest (RF), and Geographic Information System (GIS) were used for LULC classification and change detection. The dataset employed for this study is Landsat-8 with 30 m resolution, and four land cover classes, which include waterbody, vegetation, bareland, and buildup, were classified for this study using RF. The training samples for each class were divided into two groups: 60% were used for LULC classification, and 40% were used for accuracy assessment. The overall accuracy (OA) for 2014 and 2023 was 91% and 92%, while the Koppa coefficient (KC) for 2014 and 2023 was 0.88% and 0.89% giving validation to the quality and reliability of this research. The result showed that between 2014 and 2023, waterbody increased from 27.56 km2 to 36.67 km2, vegetation cover decreased from 3,153.73 km2 to 1,283.62 km2, bareland increased from 3,660.74 km2 to 4,845.30 km2, and buildup area increased from 541.21 km2 to 1,217.69 km2. The transition results show that vegetation cover is being rapidly replaced by bareland and buildup. The results of this study are crucial for directing land use planning in Abuja and other developing regions of the world to enhance sustainable urban expansion. To help the government and policymakers support sustainable development, further research should be conducted to deepen the understanding of urban growth and its consequences.

Keywords

• Google Earth Engine
• Random Forest
• GIS
• Land use Land cover
• Abuja