Hybrid methods in flood inundation modeling: a systematic review.

Area of Science:

• Environmental Science
• Hydrology
• Computer Science

Background:

• Flooding is a growing threat exacerbated by climate change, necessitating effective flood inundation models for mitigation.
• Traditional process-based models face computational limitations for real-time applications.
• Machine learning models offer efficiency but struggle with data dependency and interpretability.

Purpose of the Study:

• To systematically review state-of-the-art hybrid flood models.
• To define hybridization and classify its techniques.
• To propose a benchmarking framework for model evaluation.

Main Methods:

• Systematic literature review of hybrid flood models.
• Classification of hybridization strategies (input, structure, processing).
• Categorization of flood model evaluation metrics (accuracy, speed).

Main Results:

• Hybridization enhances flood models' physics awareness, real-time applicability, and adaptability.
• A novel classification of hybridization techniques is proposed.
• A comprehensive benchmarking framework for comparing flood models is presented.

Conclusions:

• Hybrid flood models offer a promising solution to overcome limitations of standalone approaches.
• Standardized evaluation and benchmarking are crucial for advancing flood inundation modeling.
• Physics-informed machine learning represents a key future direction for flood modeling research.