Interpretable Dual-Stream Learning for Local Wind Hazard Prediction in Vulnerable Communities

2025-05-20

AI Summary

Current wind disaster prediction systems over-rely on meteorological variables while neglecting community-level vulnerability, hindering fine-grained risk assessment and resilience planning in vulnerable regions. To address this gap—particularly in data-sparse, infrastructure-deficient Native American communities across the U.S. Great Plains—this paper proposes the first interpretable dual-stream learning framework that jointly leverages structured numerical weather data and unstructured disaster-related textual narratives. The model employs RoBERTa to encode event descriptions and random forests to process meteorological features, integrated via late fusion and augmented with gradient-based sensitivity analysis for transparent decision-making. Ablation studies and sensitivity analyses demonstrate that our approach significantly outperforms conventional baselines, enabling block-level risk prediction with both high accuracy and strong interpretability. This advances practical utility for emergency response coordination and community-scale resilience planning.

Technology Category

• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models
• Machine Learning: Structured Learning
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Wind hazards such as tornadoes and straight-line winds frequently affect vulnerable communities in the Great Plains of the United States, where limited infrastructure and sparse data coverage hinder effective emergency response. Existing forecasting systems focus primarily on meteorological elements and often fail to capture community-specific vulnerabilities, limiting their utility for localized risk assessment and resilience planning. To address this gap, we propose an interpretable dual-stream learning framework that integrates structured numerical weather data with unstructured textual event narratives. Our architecture combines a Random Forest and RoBERTa-based transformer through a late fusion mechanism, enabling robust and context-aware wind hazard prediction. The system is tailored for underserved tribal communities and supports block-level risk assessment. Experimental results show significant performance gains over traditional baselines. Furthermore, gradient-based sensitivity and ablation studies provide insight into the model's decision-making process, enhancing transparency and operational trust. The findings demonstrate both predictive effectiveness and practical value in supporting emergency preparedness and advancing community resilience.

Problem

• Predict local wind hazards in vulnerable communities with sparse data
• Integrate weather data and event narratives for better risk assessment
• Provide interpretable predictions for emergency preparedness and community resilience

Innovation

• Interpretable dual-stream learning framework
• Combines Random Forest and RoBERTa transformer
• Late fusion for structured and unstructured data