SHAZAM: Self-Supervised Change Monitoring for Hazard Detection and Mapping

2025-03-01

AI Summary

To address key bottlenecks in environmental disaster monitoring—including reliance on labeled data, difficulty modeling seasonal dynamics, and detection latency—this paper proposes a lightweight self-supervised single-temporal change detection framework. Methodologically, it introduces a novel conditional U-Net-based generative seasonal modeling approach, integrated with a theory-driven adaptive seasonal thresholding mechanism; this enables modeling of normal seasonal dynamics and real-time anomaly localization using only a single remote sensing image—requiring no labels, hyperparameter tuning, or domain-specific adaptation. Evaluated across six disaster types (e.g., wildfire, flood, drought) and four public benchmarks, the method achieves F1-score improvements of 0.066–0.234 and substantially enhanced recall. With merely 473K parameters, it produces high-resolution, background-suppressed pixel-level disaster maps, demonstrating strong cross-regional and multi-hazard generalization capability.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Remote Sensing / Geospatial AI
• Data Mining & Knowledge Management: Anomaly/Outlier Detection

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

The increasing frequency of environmental hazards due to climate change underscores the urgent need for effective monitoring systems. Current approaches either rely on expensive labelled datasets, struggle with seasonal variations, or require multiple observations for confirmation (which delays detection). To address these challenges, this work presents SHAZAM - Self-Supervised Change Monitoring for Hazard Detection and Mapping. SHAZAM uses a lightweight conditional UNet to generate expected images of a region of interest (ROI) for any day of the year, allowing for the direct modelling of normal seasonal changes and the ability to distinguish potential hazards. A modified structural similarity measure compares the generated images with actual satellite observations to compute region-level anomaly scores and pixel-level hazard maps. Additionally, a theoretically grounded seasonal threshold eliminates the need for dataset-specific optimisation. Evaluated on four diverse datasets that contain bushfires (wildfires), burned regions, extreme and out-of-season snowfall, floods, droughts, algal blooms, and deforestation, SHAZAM achieved F1 score improvements of between 0.066 and 0.234 over existing methods. This was achieved primarily through more effective hazard detection (higher recall) while using only 473K parameters. SHAZAM demonstrated superior mapping capabilities through higher spatial resolution and improved ability to suppress background features while accentuating both immediate and gradual hazards. SHAZAM has been established as an effective and generalisable solution for hazard detection and mapping across different geographical regions and a diverse range of hazards. The Python code is available at: https://github.com/WiseGamgee/SHAZAM

Problem

• Detects environmental hazards using self-supervised learning.
• Overcomes seasonal variations without labeled datasets.
• Provides high-resolution hazard maps with minimal parameters.

Innovation

• Lightweight conditional UNet for image generation
• Modified structural similarity for anomaly detection
• Theoretically grounded seasonal threshold optimization