Table 1 Summary of existing studies.

Ge¹¹

Disaster prediction knowledge graph utilizing remote sensing, geographic info, and expert knowledge

Multi-source spatio-temporal information acquisition; efficiency proven in forest fire and landslide studies

Does not consider computational efficiency or scalability in larger-scale disaster situations

Akbarian et al.¹²

Decision Support System (DSS) for flood disaster management, including PCA, KMeans Clustering, and neural networks

Positive correlation between rainfall and flood disasters, good for real-time flood prediction in varying climates

Feasible only where credible data is available; applicability in third-world countries limited

Jiang et al.¹⁴

Artificial neural network for landslide risk prediction, based on rainfall intensity levels

High accuracy in predicting landslide risks, particularly under varying rainfall intensities

Primarily focuses on rainfall, not accounting for other geographical or human factors

Raza et al.¹⁶

User-centric communication model for disaster-stricken areas, based on ad hoc clustering

Maximizes throughput and helps restore communication during disasters using social media data

Not tested in large-scale real-world disasters or real-time data conditions

Luo et al.¹³

Multi-sensor information fusion algorithm, combining Kalman joint filters, BP neural networks, and SVMs

Improved prediction accuracy and fault tolerance, real-time decision-making

Real-time data fusion challenges not fully addressed for diverse disaster types

Keum, Han, and Kim¹⁷

Flood prediction model based on hydraulic theory, machine learning (PNN classifier), and LHS

Real-time flood mapping with high accuracy (85%) and fast runtime (1 min 12 s)

Model scalability and applicability to various flood types and cases not fully explored

Kumar et al.¹⁸

Machine learning techniques (SVM, fine tree classifiers) for flood forecasting

Improves flood forecasting accuracy, reduces computational costs, enhances prediction abilities

Real-time data input could further enhance predictive abilities for disaster response

Orimoloye¹⁹

Machine learning for drought forecasting, using regression-based algorithms

Effective for mapping drought and forecasting its effects on agriculture

Requires more detailed studies for sub-Saharan Africa, particularly for climatic and anthropogenic characteristics

Jiang et al.²⁰

Machine learning models (Random Forest, MLP) for waterlogging disaster prediction

Useful for predicting waterlogging and related disaster losses, accounting for varying rainfall conditions

Can be improved with more specific geographic and temporal data for more accurate locality-based estimates

Cao²¹

AI-driven disaster resilience review, particularly for emergency crises and ECDs (including COVID-19)

AI helps in proactive disaster management and enhances system integration and global scalability for disaster response

High implementation costs and need for specialized skills, limiting widespread adoption