AI Coastal Upwelling Detection in the Mid-Atlantic Bight

Using a combination of AI technology, ocean data, and numerical weather prediction models, we can better predict wind speed and direction for offshore wind energy. Upper ocean temperatures play a critical role in the characteristics of the marine boundary layer, particularly in regions of offshore wind development that experience coastal upwelling. To detect these ocean features and increase the accuracy of these models, we use deep-learning methods. This includes using a convolutional neural network that uses sea surface temperature satellite images from the NASA Geostationary Operational Environmental Satellite 16 (GOES-16). One of the more imperative instruments on the GOES-16 is the ABI: the Advanced Baseline Imager with 16 spectral bands, which has a spatial resolution of 0.5-2.0km with the ability to scan the Western Hemisphere in 15-minute intervals. This is why spatiotemporal models produced by the AIRU-WRF (AI-powered Rutgers University Weather Research & Forecasting) are important to fuse numerical weather predictions with hourly observations. Here, we focus on the Mid-Atlantic Bight (MAB) enclosing the region between the north of Cape Cod, Massachusetts, and the south of Cape Hatteras, North Carolina, bisected by the Hudson Shelf Valley extending from the mouth of the Hudson River Valley. We distinguish the effects of upwelling from other transient features to develop automated detection tools for upwelling and provide those outputs to numerical weather prediction models. With these combined systems we hope to better understand the processes and behaviors of offshore wind and ocean observing methods to harness a potential 22 GW of wind capacity energy.