Patch-level colorimetric quantification of coral bleaching for marine pollution monitoring using standardized...

Coral bleaching is widely used as a visible ecological indicator of marine environmental degradation, reflecting cumulative stress from sedimentation, nutrient enrichment, chemical contamination, and localized thermal anomalies. Reliable pollution monitoring therefore requires quantitative and standardized bleaching metrics that can be consistently compared across sites and time. In this study, we present an automated framework for patch-level colorimetric quantification of coral bleaching using standardized CoralWatch references, enabling reproducible and objective assessment under variable underwater conditions. The system integrates a dehazing network (RAUNE-Net) for optical restoration, foundation models such as Grounding DINO and SAM2 for chart localization and segmentation, geometric rectification, and ΔE-based color analysis in CIELAB space for quantitative bleaching inference. Implemented as a PySide6 desktop application, AutoCoralMatch provides an intuitive graphical interface supporting batch image processing, AI-assisted annotation, and standardized bleaching report generation. Field evaluations conducted in both controlled aquaria and Arabian Gulf reef sites demonstrate high fidelity in patch extraction, classification, and bleaching quantification, exhibiting strong correlation with expert assessments. By explicitly linking patch-level color changes with pollution-associated stressors such as turbidity, nutrient enrichment, and coastal sediment plumes, the system enables consistent and quantitative documentation of bleaching patterns in areas affected by marine pollution. The platform supports integration of visual bleaching indicators into environmental modeling, impact assessment, and long-term coastal monitoring programs. Through its transparent design and open data release, AutoCoralMatch establishes a methodological foundation for scalable, data-driven decision support in marine ecosystem management.