Farzad et al. (2025) A systematic review and comprehensive evaluation of artificial intelligence approaches for prediction flood susceptibility

Identification

• Journal: Acta Geophysica
• Year: 2025
• Date: 2025-12-16
• Authors: Reza Farzad, Ahmad Sharafati, Yusef Kheyruri
• DOI: 10.1007/s11600-025-01769-1

Research Groups

• Department of Civil Engineering, SR.C., Islamic Azad University, Tehran, Iran
• New Era and Development in Civil Engineering Research Group, Scientific Research Center, Al-Ayen University, Nasiriyah, Thi-Qar, Iraq

Short Summary

This systematic review comprehensively evaluates artificial intelligence (AI) approaches for flood susceptibility prediction, revealing a significant increase in Machine Learning (ML) and Deep Learning (DL) model usage since 2018 and 2020, respectively, with most research originating from Asian countries and relying primarily on satellite data.

Objective

• To systematically review and comprehensively evaluate artificial intelligence (AI) approaches used for predicting flood susceptibility, including prediction parameters, performance metrics, study areas, data sources, and key trends in their advancement.

Study Configuration

• Spatial Scale: Global (systematic review of global research, with a focus on studies from Asian nations like Iran, China, India, and Bangladesh).
• Temporal Scale: Review of articles published from 2004 to 2024, with observed trends in model usage since 2018 (ML) and 2020 (DL).

Methodology and Data

• Models used:
  • Machine Learning (ML) models: Support Vector Machine (SVM), Random Forest (RF), Decision Tree (DT), k-Nearest Neighbor (k-NN).
  • Deep Learning (DL) models: Artificial Neural Network (ANN), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTMs), U-Net, Graph Neural Networks (GNNs).
  • Fuzzy Models (FM): Adaptive Network-Based Fuzzy Inference System (ANFIS).
  • Hybrid Models (HM): Combinations of the above with optimization algorithms (e.g., Particle Swarm Optimization (PSO), Genetic Algorithm (GA), Cuckoo Search Algorithm (CSA), Firefly Algorithm (FA)).
• Data sources: Satellite data (ASTER, SRTM, ALOS, Sentinel-1, Sentinel-2), radar data, direct measurements, rainfall measurements, digital elevation models, flood inventory data.

Main Results

• A total of 180 articles published between 2004 and 2024 were reviewed.
• Machine Learning models constitute the largest share of usage (approximately 44.75%), followed by Deep Learning models (33.15%), Fuzzy models (12.15%), and Hybrid models (9.95%).
• The use of ML models has significantly increased since 2018, while DL models have shown notable growth since 2020.
• The majority of research on flood susceptibility prediction originates from Asian nations, including Iran (23.20%), India and Bangladesh (17.13% combined), and China (16.57%).
• Satellite data, particularly ASTER (44.53%) and SRTM (35.77%), serve as the primary information source for these studies.
• The most frequently used input variables for flood susceptibility mapping include Slope (88%), Curvature (69%), Rainfall (61%), Topographic Wetness Index (TWI) (60%), Elevation (56%), Land Use/Land Cover (52%), and Lithology (41%).
• The Area Under the Receiver Operating Characteristic Curve (AUC-ROC) is the most frequently used performance metric across all model categories due to its high discrimination ability and adaptability in classification tasks.

Contributions

• Provides a comprehensive global review of artificial intelligence approaches for flood susceptibility prediction, evaluating multiple facets including prediction parameters, performance metrics, study areas, and data sources.
• Identifies key trends in the advancement and adoption of Machine Learning and Deep Learning models over the past two decades.
• Highlights effective strategies for flood susceptibility modeling, such as hybrid modeling, data decomposition, model optimization, and ensemble algorithms.
• Reveals geographical hotspots of research (predominantly Asian nations) and dominant data sources (satellite data).
• Serves as a valuable reference for hydrologists in selecting suitable methods or models and identifies existing research gaps, proposing future directions including the exploration of transformer models and the development of ML models for other hydrological parameters.

Funding

No funding.

Citation

@article{Farzad2025systematic,
  author = {Farzad, Reza and Sharafati, Ahmad and Kheyruri, Yusef},
  title = {A systematic review and comprehensive evaluation of artificial intelligence approaches for prediction flood susceptibility},
  journal = {Acta Geophysica},
  year = {2025},
  doi = {10.1007/s11600-025-01769-1},
  url = {https://doi.org/10.1007/s11600-025-01769-1}
}