Comparing Knowledge-Driven and Data-Driven Machine Learning Frameworks for Coastal Vulnerability Assessment on the Egyptian North Coast

Authors

• Ali Elgendy, University of Szeged
• Hesham Morgan, Chapman University
• Hesham El-Askary, Chapman UniversityFollow
• László Mucsi, University of Szeged

Document Type

Article

Publication Date

10-19-2026

Abstract

Coastal erosion worsened by climate change and human activities is a major threat to global shorelines. As sea levels rise and storms intensify, this rapid shoreline retreat endangers critical infrastructure and the economic security of coastal populations worldwide. However, traditional knowledge-driven frameworks for assessing these threats often rely on subjective, static rules that fail to capture complex, non-linear environmental interactions. To address this critical gap and demonstrate the value of advanced methods, this study compared the traditional Integrated Valuation of Ecosystem Services and Tradeoffs (InVEST) model against machine learning algorithms, highlighting the potential for more accurate coastal erosion risk assessment. The study analyzed 500 km of the Egyptian Mediterranean coast over 40 years (1985-2025) using satellite imagery. The results mapped severe coastal changes, including 1.35 km of linear shoreline retreat at the Rosetta promontory due to trapped Nile sediment. The comparison showed that the traditional model struggles to identify local high-risk erosion areas, achieving only 33.2% accuracy. In contrast, the model based on the Extreme Gradient Boosting (XGBoost) algorithm, developed using 8946 spatial transects, achieved high accuracy on the blind validation dataset (R2 = 0.86, RMSE = 2.91, MAE = 0.86). By analyzing feature importance, the model showed that coastal erosion risk is mostly driven by ocean depth and human-made coastal defenses, rather than the physical type of the beach. This highly accurate framework addresses a critical planning gap for Egypt’s rapidly growing coastal cities. The transect-level vulnerability map provides a direct input to three existing management procedures: risk-tiered setbacks, vulnerability screening at the EIA scoping stage, and the sequencing of nourishment and nature-based protection, with the InVEST index retained for regional screening and stakeholder communication. Ultimately, this data-driven approach promotes the United Nations Sustainable Development Goals by supporting the design of resilient infrastructure, the protection of coastal cities, and the selection of effective climate action.

Comments

This article was originally published in Ocean & Coastal Management, volume 283 in 2027. https://doi.org/10.1016/10.1016/j.ocecoaman.2026.108411.

Recommended Citation

Elgendy, A., Morgan, H., El-Askary, H., Mucsi, L., 2027. Comparing knowledge-driven and data-driven machine learning frameworks for coastal vulnerability assessment on the Egyptian North Coast. Ocean Coast Manag. 283, 108411. https://doi.org/10.1016/10.1016/j.ocecoaman.2026.108411.

Peer Reviewed

1

Copyright

The authors

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 License.