Enhanced Detection of Geomorphic Changes in the Khor Al-Sabiyeh Coastal Environment Using SAR, Optical Imagery, and Machine Learning
Ahmad E. Aldousari 1*, Temitope D. Timothy Oyedotun 2*, Saud Reyadh AlKhaled 3, Helene Burningham 4

1 Department of Geography, Assistant Professor, Social Sciences College, Kuwait University, Kuwait City, Kuwait
2 Department of Geography, Professor, Faculty of Earth and Environmental Sciences (FEES), University of Guyana, P O Box 10 1110, Turkeyen Campus, Guyana
3 Department of Architecture, Assistant Professor , College of Architecture, Kuwait University, Kuwait City, Kuwait
4 Coastal and Estuarine Research Unit, Professor, UCL Department of Geography,
Gower Street, London, WC1E 6BT, UK
* dr.dousari[at]ku.edu.kw (Corresponding author)
* : temitope.oyedotun[at]uog.edu.gy (Corresponding author)

Abstract

Coastal wetlands are among the most ecologically significant landscapes, yet they are highly susceptible to geomorphic changes driven by both natural processes and anthropogenic pressures. This study presents an integrated remote sensing and machine learning approach to enhance the detection of geomorphic changes in Khor Al-Sabiyeh, a critical coastal wetland in Kuwait. By utilising the complementary strengths of Synthetic Aperture Radar (SAR) and optical imagery (Landsat-8 and Sentinel-2), we developed an analytical framework that overcomes the limitations of conventional monitoring methods. Using Google Earth Engine (GEE) for data preprocessing, we generated cloud-free annual composites for 2020 and computed a suite of spectral indices, including NDVI, NDMI, MNDWI, GCVI, SR, and custom band ratios, which provided detailed insights into geomorphic dynamics, water bodies, and land surface conditions. This study used supervised machine learning classifiers, particularly Random Forest, to detect and classify geomorphic transformations with high accuracy. The model was validated using cross-validation techniques and statistical metrics, such as overall accuracy and the Kappa coefficient, confirming its reliability and robustness. The results revealed distinct spatial patterns of erosion, accretion, and land cover changes, which have direct implications for environmental planning in the region. The results from this research show the potential of integrating SAR, optical datasets, and machine learning for a timely and accurate assessment of landscape changes in fragile coastal systems. The methodological framework adopted in this study is transferable and scalable, offering valuable applications for similar systems globally. This approach supports evidence-based environmental governance and enhances resilience in the face of climate change and human-induced alterations.

Keywords: Coastal Wetlands- Geomorphic Change Detection- Khor Al-Sabiyeh Kuwait- Machine Learning- Optical Imagery- Remote Sensing Integration- Synthetic Aperture Radar (SAR).

Topic: Topic B: Applications of Remote Sensing