END-TO-END PHYSICS-INFORMED REPRESENTATION LEARNING FOR SATELLITE OCEAN REMOTE SENSING DATA: APPLICATIONS TO SATELLITE ALTIMETRY AND SEA SURFACE CURRENTS

In non-equilibrium statistical mechanics, which describes forced-dissipative systems such as air–sea interaction, there is no universal probability density function (pdf). Some such systems have recently been demonstrated to exhibit a symmetry called a fluctuation theorem (FT), which strongly constrains the shape of the pdf. Using satellite data, the mechanical power input to the ocean by air–sea interaction following or not a FT is questioned. A FT is found to apply over specific ocean regions.

Data from the ARKTIKA-2018 expedition and new satellite data help us to follow rapid changes in the upper layer of the Laptev and East Siberian seas (LS, ESS) in summer 2018. With satellite-derived surface temperature, an improved SMOS salinity, and wind, we study how the fresh river water is mixed with cold sea water and ice-melted water at small time and spatial scales. The wind pushes fresh water northward and northeastward, close to and under the ice, forcing it into the deep Arctic Ocean.

With present-day techniques, ocean surface currents are poorly known near the Equator and globally for spatial scales under 200 km and timescales under 30 d. Wide-swath radar Doppler measurements are an alternative technique. Such direct surface current measurements are, however, affected by platform motions and waves. These contributions are analyzed in data collected during the DRIFT4SKIM airborne and in situ experiment, demonstrating the possibility of measuring currents from space globally.

Learning physics from data using a deep neural network is a challenge that requires an appropriate but unknown network architecture. The package introduced here helps to design an architecture by translating known physical equations into a network, which the experimenter completes to capture unknown physical processes. A test bed is introduced to illustrate how this learning allows us to focus on truly unknown physical processes in the hope of making better use of data and digital resources.

FluxEngine is an open-source software toolbox designed to allow for the easy and accurate calculation of air–sea gas fluxes. This article describes new functionality and capabilities, which include the ability to calculate fluxes for nitrous oxide and methane, optimisation for running FluxEngine on a stand-alone desktop computer, and extensive new features to support the in situ measurement community. Four research case studies are used to demonstrate these new features.

The Sea surface KInematics Multiscale (SKIM) monitoring mission is a proposal for a future satellite that is designed to measure ocean currents and waves. Using a Doppler radar, the accurate measurement of currents requires the removal of the mean velocity due to ocean wave motions. This paper describes the main processing steps needed to produce currents and wave data from the radar measurements. With this technique, SKIM can provide unprecedented coverage and resolution, over the global ocean.