RGB color indices as proxy for symbiont cell density and chlorophyll content during coral bleaching
Ferrara, E. F.; Bauer, L.; Puntin, G.; Bautz, F. R.; Celayir, S.; Do, M.-S.; Eck, F. L.; Heider, M. C.; Wissel, P. M.-C.; Arnold, A.; Wilke, T.; Reichert, J.; Ziegler, M.
Show abstract
Coral bleaching, the breakdown of the symbiosis between the coral host and endosymbiotic microalgae, is the main cause of widespread coral reef degradation. Current methods for assessing coral health based on visual appearance, such as the use of color reference cards, are limited by subjective human color perception and low resolution. Digital photography with RGB (Red, Green, Blue) color channel analyses offers a fast, non-invasive, and standardized alternative to estimate physiological parameters. However, the link between coral color and physiological parameters during bleaching may vary depending on the type of stressor. While such approaches are extensively used in plant studies, their application in estimating Symbiodiniaceae cell density and chlorophyll content in corals requires further attention. In this study, we analyzed the correlation between Symbiodiniaceae cell density and chlorophyll content across three coral species (Acropora muricata, Pocillopora verrucosa, and Stylophora pistillata) with 19 color indices derived from the RGB channels currently established as predictors of chlorophyll content in plants. Corals were exposed to three bleaching conditions (acute short-term and chronic long-term heat stress and menthol bleaching) to identify the best color indices for assessing coral health through image analysis. We found that the Red index had the strongest linear correlation with symbiont cell density and chlorophyll content across species (R2 up to 0.97), so that relative changes in this color index can be directly interpreted as corresponding changes in tissue parameters. To train a model that predicts symbiont densities of a distinct sample set using the Red index, we found 10 to 12 samples to be sufficient to achieve an accuracy of > 95 % of the models trained on the full datasets. This research contributes to improved image analysis as a reliable and non-invasive tool for monitoring, by providing guidelines for a systematic use of RGB data to interpret coral health.
Matching journals
The top 2 journals account for 50% of the predicted probability mass.