Underwater synthetic image generation for multi-temporal monitoring of crustose coralline algae

Abstract. Monitoring crustose coralline algae (CCA) is crucial for understanding the health and evolution of marine ecosystems. In particular, CCA-dominated assemblages serve as key indicators of ecological change, especially in the context of global climate pressures and ocean acidification. Developing robust tools to detect subtle and spatially distributed variations in their appearance is fundamental for early warning systems and practical conservation actions. However, acquiring multi-temporal underwater images with reliable ground truth is complex due to environmental variability and logistical challenges. This paper proposes a framework integrating underwater photogrammetry with Neural Radiance Fields (NeRF) to create synthetic images for multi-temporal ecological monitoring. The authors investigate the creation of synthetic views from consistent camera positions across various monitoring periods to enhance the consistency and interpretability of image-based evaluations over time. More specifically, we have focused on generating synthesized images with the same camera pose across different epochs to assess expected environmental changes, enabling repeatable, ground-truth-validated tests of image-based monitoring techniques. A quantitative image comparison framework was developed, incorporating several metrics, such as Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and perceptual color difference (ΔE) within spatially constrained masks in specific regions of interest (ROIs). Results demonstrate the ability to detect chromatic alterations over time, with SSIM values indicating consistency and ΔE metrics revealing widespread perceptual changes. The methodology provides a robust and reproducible foundation for evaluating monitoring protocols and supporting ecological interpretation.