Back to search

KLIMAFORSK-Stort program klima

Machine Ocean – Combining Machine Learning and Earth Observations to Improve Simulations of Turbulent Behavior in the Earth System

Alternative title: Machine Ocean

Awarded: NOK 11.5 mill.

Project Number:

303411

Application Type:

Project Period:

2020 - 2024

Funding received from:

Location:

Partner countries:

Reliable weather and ocean forecasts are fundamental to society, and researchers are constantly working to improve forecast models. The atmosphere and oceans are parts of the complex system we call the Earth System. Globally, this system is relatively simple: solar energy is absorbed on Earth, and an (almost) equal amount of energy is returned to space. Depending on the amount of greenhouse gases in the atmosphere, more or less long-wave radiation is absorbed, which in turn determines the global ground temperature. On a regional scale, the Earth System is more difficult to simulate: energy moves around to compensate for the regional imbalances in incoming and outgoing radiation. This is achieved by winds and currents interacting with each other and with snow and ice. At the center of the interactions is the transfer of momentum - the stress - across the interfaces. This stress is turbulent and impossible to model in an Earth system model (one would need millimeter-scale model resolution). It is therefore necessary to parametrize the stress. This is normally accomplished using semi-empirical bulk formulas to estimate the stress from sparse surface measurements. In Machine Ocean, we have let the data themselves produce a "stress model", via machine learning. Our hypothesis was that the huge increase in data volumes coming from ESA's new Sentinel environmental satellites, as well as new machine learning algorithms that can be trained on and learn from huge datasets, would make it possible to develop a stress model that depends less on empiricism and more on theory and data. This hypothesis turned out to be partially correct: the amount of data from the satellites is still too small for classical machine learning. But we had enough data to be able to confirm that the empirical stress model we compared to cannot be "beaten" by machine learning at this point. This provides extra support for the (empirical) stress model used in most Earth System models today, and that is good news. We have used the Meteorological Institute's storm surge warning model as one of our test beds. Storm surge warnings depend almost entirely on atmospheric pressure and stress transfer, so this is an ideal way to test the impact of improved stress parameterization. At the same time, storm surges are high on the list of dangerous climate changes in Norway, so improving storm surge warnings is very important for society. We have developed methods to correct the mismatch between storm surge models and observations along the coast of Norway with the help of machine learning. Our PhD candidate Paulina Tedesco will defend her PhD in late winter 2024.

The overarching hypothesis for Machine Ocean is that “the present explosion in volume, variety and velocity (VVV) of Earth Observation acquisition (as spearheaded by the Copernicus, in particular the Sentinel 1 mission) combined with new methods to harvest information from big data will allow us to gain further insights into, and significantly reduce the uncertainty in, parameterization of momentum transfer between atmosphere and ocean.” Vertical momentum transfer is one of the most important process in the Earth System, influencing the transfer of carbon, oxygen, heat, freshwater and other quantities between the different spheres, yet possibly the hardest process to measure. The transfer occurs on small horizontal and temporal scales, so it is almost always necessary to parameterize in numerical simulations. The use of machine learning methods to directly predict turbulent stress in Reynolds-averaged Navier–Stokes equations has recently been proposed, and developed for simplified setups, but the field of machine learning applications in fluid dynamics in general and for momentum transfer and larger scale atmosphere models in particular, is in its infancy.

Funding scheme:

KLIMAFORSK-Stort program klima