ABSTRACT
Many phenomena encountered in the Earth sciences exhibit challenging aspects such as teleconnections, cross-talk, weak signals, non-linear dynamics, phase changes, and chaos. AI and deep learning are being explored as tools to tackle these challenges. Deep learning is strong at finding correlations in large datasets, even subtle ones ranging over long distances in space or time, and is therefore useful in dealing with teleconnections. Using ML to correct cross-talk between multiple components has been demonstrated in engineering settings, and may carry over to Earth science contexts. Use cases involving weak signals and non-linearity have been shown to be addressable by deep learning, and chaotic dynamics have also seen their predictability extended by neural network techniques.
