Forecasting coral reef dynamics using scientific machine learning

Ecological monitoring often does not produce the high-resolution data necessary for forecasting ecosystem states under future environmental conditions. To address this issue, we leveraged scientific machine learning (SciML) techniques to integrate theoretical models with time series data to detect alternative states and predict their long-term dynamics. Our SciML approach utilized two frameworks: neural ordinary differential equations (NODEs) and universal differential equations (UDEs). NODEs use a neural network to approximate a system of differential equations, and UDEs extend this method by incorporating known dynamics. We applied both methods to time series of coral, turf algae, and macroalgae coverage spanning up to 21 years on 25 reefs across the US Virgin Islands (USVI) to answer several research questions. First, which SciML framework provides more accurate estimates of long-term benthic coverage? Second, can UDEs reliably estimate grazing rates on algae? Finally, what are the near-term forecasts for benthic coverage on these reefs? We found that UDEs performed better than NODEs (had lower normalized RMSE) for modeled reefs. In addition, UDEs yielded grazing rates in the range of 0 to 35% of the reef, which is lower than thresholds that would produce coral-dominated reefs. Consequently, near-term forecasts indicated either macroalgae or turf algae would continue to dominate USVI reefs. Through application of SciML, our study enhances understanding of alternative state maintenance and provides valuable tools for ecosystem management and conservation on coral reefs.

Primary Presenter: Zechariah Meunier, Oregon State University (meunierz@oregonstate.edu)

Authors:

Zechariah Meunier, Oregon State University  (meunierz@oregonstate.edu)

Jack Buckner, Oregon State University  (bucknejo@oregonstate.edu)

Nathan Fitzpatrick, University of Hawaii  (nmf68@hawaii.edu)

Emerson Arehart, University of Pennsylvania  (eejjaa@gmail.com)

Ariel Greiner, Pennsylvania State University  (ariel.greiner@mail.utoronto.ca)

Lisa McManus, University of Hawaii  (mcmanusl@hawaii.edu)

James Watson, Oregon State University  (james.watson@oregonstate.edu)