Coral reefs are one of the most biologically diverse and economically valuable ecosystems on the planet. They provide a critical habitat for marine life and serve as a foundation for many human-driven industries. However, in recent decades, coral reef health has declined due to pollution, human interaction, and climate change. By enhancing our understanding of coral reefs, ecologists can devise more effective strategies to mitigate the aforementioned challenges. Current methods for monitoring coral reefs are often limited and invasive. I am developing an autonomous robotic system capable of accurately predicting local ocean currents and safely maneuvering through complex environments under uncertainty. In order to improve our understanding of disturbances (e.g. ocean currents) in an underwater environment, a machine learning diffusion model is used to estimate the flow field around obstacles. After, a safety filter monitors robot trajectories and aids the robot in taking safe actions. Enriching our models with more advanced flow prediction and creating safe trajectories, enables robots to be deployed in fragile ecosystems like coral reefs.