Combining Deep Learning and Multi-Source GIS Methods to Analyze Urban and Greening Changes.

Area of Science:

• Environmental Science
• Urban Planning
• Remote Sensing Technology

Background:

• Urban development often leads to a decline in green cover, impacting ecosystem services and societal well-being.
• Existing studies lack comprehensive spatiotemporal analysis of urban growth and greening dynamics using advanced technologies.
• Innovative remote sensing (RS) and deep learning (DL) approaches are needed to accurately quantify these environmental changes.

Purpose of the Study:

• To develop and validate an innovative methodology for analyzing urban and greening changes over time.
• To integrate deep learning (DL) for precise classification and segmentation of built-up and vegetation areas.
• To assess the spatiotemporal dynamics of urban expansion and green cover loss using remote sensing data.

Main Methods:

• A U-Net deep learning model was trained and validated for image classification and segmentation.
• Satellite and aerial imagery from 2000 to 2020 were analyzed for an urban area in Matera, Italy.
• Geographic Information System (GIS) techniques were integrated with DL outputs for spatiotemporal analysis.

Main Results:

• The U-Net model demonstrated high accuracy in classifying urban and greening features.
• A significant increase in built-up area density (8.28%) was observed between 2000 and 2020.
• A notable decline in vegetation cover density (5.13%) was recorded during the same period.

Conclusions:

• The proposed methodology effectively quantifies urban and greening spatiotemporal development using innovative RS technologies.
• The findings highlight the rapid transformation of urban landscapes and the associated loss of green cover.
• This approach supports informed decision-making for sustainable urban planning and environmental management.