Unmixing With Spectral Variability for Marine Mucilage Analysis

The marine mucilage outbreak of Spring 2021 in the Sea of Marmara is one of the worst occurrences of mucilage in terms of the area affected and the duration of the outbreak. Unmixing, for the spectral and spatial analysis of this outbreak, avoids the need for manually prepared ground truth maps or carefully selected thresholds, while enabling distinguishing not only mucilage from water, but potentially also accumulated mucilage from diffused or dispersed mucilage. The previous unmixing-based approaches used to address the mucilage phenomenon utilized the linear mixture model (LMM). However, the assumption of linearity, which does not address spectral variability or non-linear interactions, is tenuous for this liquid mixture. This work addresses spectral variability for unmixing-based marine mucilage analysis for the first time. Extended linear mixture model (ELMM) is used to take spectral variations of endmembers in each pixel into account, and experimental results validate the proposed approach with respect to previous unmixing-based approaches.