Urban change detection: assessing biophysical drivers using machine learning and Google Earth Engine

Area of Science:

• Environmental Science, Urban Planning, Geospatial Analysis
• Remote Sensing and Machine Learning Applications
• Sustainable Development Studies

Background:

• Urban areas undergo rapid transformations due to population growth, economic development, and policy shifts.
• Monitoring urban dynamics and their environmental interactions is vital for sustainable planning.
• The Kaduna River Basin (KRB) in Nigeria faces significant urban transformation challenges.

Purpose of the Study:

• To investigate urban dynamics and their interactions with biophysical conditions in the KRB using machine learning.
• To analyze urban transitions between 1987 and 2020, incorporating environmental variables.
• To identify key drivers of urban change and provide insights for sustainable urban development.

Main Methods:

• Utilized a dataset of 192 points from Google Earth Engine for urban transition analysis (1987-2020).
• Employed machine learning models (Random Forest, SVM, KNN, MARS) with tenfold cross-validation in R.
• Incorporated biophysical variables like population density, precipitation, and surface temperature, with data preprocessing including KNN imputation and One-Hot Encoding.

Main Results:

• Random Forest model achieved 80% overall accuracy in predicting urban change patterns.
• Observed significant land cover changes: 154.81% urban expansion, 95.79% decline in water bodies, 174% vegetation growth.
• Identified population density and water stress index as primary urban change drivers, with climate factors also playing a role.

Conclusions:

• Machine learning provides a robust methodology for monitoring and predicting urban development.
• Findings offer valuable insights into urban transformation processes in the KRB.
• Results support strategies for sustainable urban growth and mitigating adverse environmental impacts.