Image processing and AI techniques for climate change detection using remote sensing: a comprehensive review.

Area of Science:

• Remote Sensing
• Artificial Intelligence
• Climate Science

Background:

• Climate change drives complex transformations across Earth systems, necessitating advanced monitoring.
• Traditional methods struggle with heterogeneous, multiscale changes and large satellite data archives.

Purpose of the Study:

• To review image processing and AI techniques for climate change detection using Earth observation data.
• To provide a unified taxonomy linking methods to applications and a decision framework for practitioners.

Main Methods:

• Synthesis of classical change detection, machine learning, deep learning (Siamese, segmentation networks), spatio-temporal models, and multi-sensor fusion.
• Analysis of performance metrics (OA, IoU, F1-score) and challenges (data quality, generalization, interpretability).

Main Results:

• Deep learning methods show higher accuracy in complex environments, outperforming classical approaches.
• Classical methods are effective for large-scale, data-scarce applications.

Conclusions:

• Significant gaps exist in model generalization, labeled dataset availability, and multi-sensor time-series integration.
• Future directions include foundation models, standardized benchmarks, and interoperable decision-support systems.