نوع مقاله : مقاله پژوهشی
موضوعات
عنوان مقاله English
نویسندگان English
Extended Abstract
Introduction
Urban growth, as a consequence of population increase and economic development, is a dynamic and complex process that leads to the expansion of urban areas into surrounding natural lands. This phenomenon results in significant land-use changes and environmental impacts, including habitat destruction, vegetation loss, pollution, and ecological instability. Accurate and continuous monitoring of these changes is crucial for urban planning, resource management, and sustainable development. Remote sensing, particularly the integration of multi-source data and machine learning algorithms, has emerged as an effective tool for monitoring land-use changes and detecting urban built-up areas. The use of optical and radar data together helps overcome challenges such as cloud cover and spectral similarities between urban and non-urban features. Cloud-based platforms such as Google Earth Engine (GEE) provide efficient computational resources to process large datasets for urban monitoring. This study proposes a hybrid approach combining Sentinel-1 radar and Sentinel-2 optical imagery within the GEE platform to map urban built-up areas in Sari City, Iran. By integrating radar and optical data, this approach enhances urban area detection and addresses limitations associated with individual remote sensing datasets.
Materials & Methods
The study utilized Sentinel-1 radar and Sentinel-2 optical imagery, along with spectral indices such as Built-up Extraction Index (BuEI), Normalized Difference Built-up Index (NDBI), Normalized Difference Vegetation Index (NDVI), Green Normalized Difference Vegetation Index (GNDVI), Renormalized Difference Vegetation Index (RDVI), Normalized Difference Water Index (NDWI), and Soil Extraction Index (SoEI), to extract urban features. The Sentinel-1 radar data underwent Principal Component Analysis (PCA) to enhance urban feature visibility, while spectral indices derived from Sentinel-2 imagery facilitated the differentiation between vegetation, water, and built-up areas. A Digital Elevation Model (DEM) was used to extract slope information, which helped in distinguishing urban regions from mountainous areas. To classify the extracted features, three machine learning algorithms were applied: Random Forest (RF), Support Vector Machine (SVM), and Classification and Regression Trees (CART). These algorithms were trained using labeled samples and executed within the GEE environment. Random Forest is an ensemble learning method known for its robustness in handling high-dimensional data and reducing overfitting. SVM is a powerful classification method that finds an optimal decision boundary between classes, while CART is a decision tree-based algorithm that works well for land-cover classification. The classification results were validated using high-resolution Google Earth imagery, and accuracy assessment was performed based on overall accuracy (OA) and the Kappa coefficient (KC) to measure agreement with reference data.
Results & Discussion
The results demonstrated that the Random Forest algorithm achieved the highest accuracy, with an overall accuracy of 95.2% and a Kappa coefficient of 90.4%. The SVM and CART algorithms also performed well, with overall accuracies of 93.8% and 93.3%, respectively. However, RF showed superior performance in identifying small urban patches and reducing classification noise in non-urban areas. The integration of multi-source data and the use of the GEE platform not only improved the accuracy of urban change detection but also enhanced processing speed and reduced the need for advanced computational infrastructure. The combination of Sentinel-1 radar and Sentinel-2 optical imagery effectively improved urban area classification by leveraging the complementary strengths of both data sources. Sentinel-1 radar provided valuable structural information on urban features regardless of weather conditions, while Sentinel-2’s spectral indices improved the separation between built-up areas, vegetation, and water bodies. The use of PCA on Sentinel-1 data further enhanced feature separability by reducing redundancy and highlighting urban structures. Additionally, the DEM-assisted slope analysis contributed to distinguishing urban areas from natural landscapes, particularly in regions with complex topography. The GEE platform played a crucial role in efficiently handling large datasets and performing computationally intensive analyses, making the proposed approach scalable for broader applications. These findings highlight the importance of integrating remote sensing data with advanced machine learning techniques for accurate urban monitoring. The high accuracy achieved by the RF algorithm underscores its suitability for urban classification, particularly in heterogeneous landscapes where built-up areas are interspersed with vegetation and other land-cover types. The study also emphasizes the practical benefits of using cloud-based platforms like GEE, which eliminate the need for local high-performance computing resources and facilitate large-scale environmental monitoring. The ability to rapidly process and analyze satellite imagery is crucial for urban planning, disaster management, and sustainable development initiatives.
Conclusion
This study successfully mapped urban built-up areas in Sari City using a combination of Sentinel-1 and Sentinel-2 imagery within the GEE platform. The Random Forest algorithm outperformed other methods, providing high accuracy in urban area detection. The integration of multi-source data and the use of machine learning algorithms significantly improved the efficiency and accuracy of urban mapping. The findings suggest that the proposed method is a reliable and cost-effective approach for urban monitoring, with potential applications in other regions. Future research could explore the scalability of this method to larger areas and its integration with other data sources for enhanced urban planning and management. The use of GEE eliminates the need for extensive computational resources, making this approach accessible for researchers and policymakers in developing regions. Overall, this study contributes to the growing body of knowledge on urban monitoring and provides a practical framework for sustainable urban development.
کلیدواژهها English
