Title: ML-Down
DNr: Berzelius-2023-355
Project Type: LiU Berzelius
Principal Investigator: Kristofer Krus <>
Affiliation: SMHI
Duration: 2024-01-01 – 2024-07-01
Classification: 10501


The goal of this project is to develop our deep learning-based methods for statistical downscaling of climate data further. We have already created a model that can downscale ERA-5, which is global reanalysis data on a 0.25° resolution grid, to CERRA, which is regional reanalysis data on a 5.5 km resolution grid. We further plan to develop the method so that we can use a network that has been trained on input data from one data source (for example ERA-5), to also be able to handle input data from other data sources well, on which it has not been trained. To achieve this, we plan to use bias adjustment. If we can successfully do this, we expect that we can substitute this form of statistical downscaling for some of the dynamic downscaling that we do today; the latter basically requires a lot of extra simulation, and is much more computationally expensive than the former. We therefore expect to be able to reduce the amount of computation that we need to do by a significant amount.