Predicting Ocean States at the North Atlantis II Seamounts with a Deep Learning Model

Located in the western North Atlantic, the subsurface temperature and salinity structure at the Atlantis II Seamounts is sensitive to fluctuations in the Gulf Stream, and historical ocean structure profiles at this location show a bimodal distribution, making seasonal climatologies a poor predictor of ocean states. In addition, historical CMIP6 ensemble means show dynamic changes in temperature and salinity between 1980-2007 (~0.5 °C and ~0.1 PSU, respectively), with no corresponding change in sea surface height anomalies, suggestive of decadal changes independent of gulf stream position. Both sea surface height and sea surface temperature and mixed layer depth are key variables to predict accurate density gradient profiles using statistics based methods. We use a deep learning state profiles using Improved Synthetic Ocean Profile (ISOP) model (specifically a convolutional neural network- based model) to predict sea surface height anomalies and sea surface temperature for a 1 month to 2 year time interval. The training dataset consists of the global 1850-2007 CMIP6 sea surface salinity, temperature, height, and mixed layer depth that capture climate modes of variability that impact the Gulf Stream. Multiple interpretation techniques, such as saliency maps, backward optimization, and layer-wise relevance propagation, connect the deep learning output to relevant physical features and climate modes of variability such as the North Atlantic Oscillation and climate change.