Deep Ocean Learning of Small Scale Turbulence

Turbulent mixing at the sub-meter scale is an essential component of ocean's meridional overturning circulation and its associated global redistribution heat, carbon, nutrients, pollutants, other tracers. Whereas direct turbulence observations in ocean interior are limited to a modest collection field programs, basic information such as temperature, salinity, depth available globally. Here, we show that supervised machine learning algorithms can be trained on existing data develop skillful predictions key properties from T, S, Z, topographic data. This constitutes promising first step toward hybrid physics-artificial intelligence approach parameterization turbulent climate models.