MIAMI — Scientists have created an AI model that predicts moderate heat stress, a key precursor to coral bleaching, up to six weeks in advance at locations along Florida’s coral reefs. Predictions are usually accurate within a week.
This study presents a site-specific explainable machine learning framework to support coral scientists and restoration professionals in developing local reef management and emergency response plans.
“This model allows coral scientists and resource managers to be informed in advance whether heat stress is likely to occur during a given season and, more importantly, the week in which heat stress is most likely to begin.” The lead author is Marybeth Arcodia of the University of Miami Rosenstiel School of Marine, Atmospheric, and Earth Sciences. Arcodia holds a dual appointment within the Department of Atmospheric Sciences and the Frost Institute for Data Science and Computing. “Through explainable AI, we can also identify the environmental factors driving predictions in each reef area.”
“Our model identifies potential factors influencing heat stress at specific reef locations,” said co-author and postdoctoral researcher Richard Karp. Associate Professor at Rosenstiel School Oceanic and Atmospheric Research Collaborative Research Institute, “This information gives managers the opportunity to identify trigger points for emergency action plans and can support planning and response decisions.”
The research team combined atmospheric science, coral ecology, and data science to build a predictive tool tailored for Florida’s coral reefs.
Localized AI predictions on practical timescales
The research team used the XGBoost machine learning model to predict the onset of moderate coral heat stress at three reef sites using environmental data from 1985 to 2024. Inputs include cumulative and instantaneous heat stress indicators, sea surface temperature anomalies, air temperature, wind, solar radiation, loop current and El Niño conditions indicators, leveraging NOAA Coral Reef Watch and other public datasets.
“Our prediction system produced good predictions up to six weeks in advance and was accurate in most cases within about a week after heat stress actually began.” Carp. “We also outperformed two benchmark approaches, a multiple logistic regression model and a frequency-based method, in predicting whether heat stress will occur and identifying when heat stress begins.”
Researchers also applied Explainable AI techniques using SHAP. This is a way to show which environmental factors most strongly influence each forecast and understand how the drivers of heat stress vary depending on reef location and forecast lead time.
Although surface temperature consistently ranks as one of the most important predictors, other key environmental factors vary by site and lead time, highlighting the value of local predictions.
“These insights are delivered on a timescale where management action is still possible,” Karp added. “These can help prioritize monitoring, inform when to initiate emergency actions, and guide where resources are most effectively targeted.”
Actionable predictions that support proactive coral reef conservation
Coral reefs in Florida and the Caribbean are experiencing increasingly frequent severe heat stress and bleaching events, including the record marine heat wave in 2023, increasing the need for site-level early warning tools.
The authors emphasize that the new AI framework aims to: Complement rather than replace existing operational systems NOAA coral reef monitoring, etc. Local seasonal timing signals that signal the onset of heat stress.
the study, “An Explainable Machine Learning Prediction System for Early Warning of Heat Stress on Florida Coral Reefs” has been published open access. environmental research communication December 16, 2025.
The study took two years to complete from conceptualization to publication. Funding was provided by the U.S. Department of Energy’s Office of Biological and Environmental Research’s Regional and Global Model Analysis Program Area and a NOAA grant as part of PCMDI. #NA19OAR4590151 and #NA24OARX431C0022, and by a grant from the NOAA Coral Reef Conservation Program. #31476 and #31640.
Authors include Marybeth C. Arcodia of the University of Miami Rosenstiel School of Oceanic, Atmospheric, and Earth Sciences and the University of Miami Frost Data Science Institute, Richard Karp of the Rosenstiel Collaborative Institute for Oceanic and Atmospheric Research, and Elizabeth A. Barnes of the University of Colorado Department of Atmospheric Sciences.
/Open to the public. This material from the original organization/author may be of a contemporary nature and has been edited for clarity, style, and length. Mirage.News does not take any institutional position or position, and all views, positions, and conclusions expressed herein are solely those of the authors. Read the full text here.
