SNIC SUPR
Machine learning for weather forecasting
Dnr:

SNIC 2019/3-611

Type:

SNIC Medium Compute

Principal Investigator:

Gabriele Messori

Affiliation:

Uppsala universitet

Start Date:

2019-12-01

End Date:

2020-12-01

Primary Classification:

10508: Meteorology and Atmospheric Sciences

Allocation

Abstract

(Deep) artificial neural networks have been very successful in many machine learning tasks, especially image recognition, and are applicable to highly non-linear problems. We have successfully used deep neural networks to emulate simple climate models. We want to extend this approach to more complex climate models as well as to renalyses of the Earth’s atmosphere. While this will need more resources, it will allow to test a very pressing research question, namely whether it is possible to make accurate weather forecasts with machine-learning techniques alone. We already used a medium allocation on Kebnekaise for 2018-12-01 to present, which lead to a scientific publication on Geoscientific Model Development (doi: 10.5194/gmd-12-2797-2019). We now want to continue our research in this direction, for which we need another year of computing resources.