Forecasting large-scale circulation regimes using deformable convolutional neural networks and global spatiotemporal climate data

TY - JOUR

T1 - Forecasting large-scale circulation regimes using deformable convolutional neural networks and global spatiotemporal climate data

AU - Nielsen, Andreas Holm

AU - Iosifidis, Alexandros

AU - Karstoft, Henrik

N1 - © 2022. The Author(s).

PY - 2022/5

Y1 - 2022/5

N2 - Classifying the state of the atmosphere into a finite number of large-scale circulation regimes is a popular way of investigating teleconnections, the predictability of severe weather events, and climate change. Here, we investigate a supervised machine learning approach based on deformable convolutional neural networks (deCNNs) and transfer learning to forecast the North Atlantic-European weather regimes during extended boreal winter for 1 to 15 days into the future. We apply state-of-the-art interpretation techniques from the machine learning literature to attribute particular regions of interest or potential teleconnections relevant for any given weather cluster prediction or regime transition. We demonstrate superior forecasting performance relative to several classical meteorological benchmarks, as well as logistic regression and random forests. Due to its wider field of view, we also observe deCNN achieving considerably better performance than regular convolutional neural networks at lead times beyond 5-6 days. Finally, we find transfer learning to be of paramount importance, similar to previous data-driven atmospheric forecasting studies.

AB - Classifying the state of the atmosphere into a finite number of large-scale circulation regimes is a popular way of investigating teleconnections, the predictability of severe weather events, and climate change. Here, we investigate a supervised machine learning approach based on deformable convolutional neural networks (deCNNs) and transfer learning to forecast the North Atlantic-European weather regimes during extended boreal winter for 1 to 15 days into the future. We apply state-of-the-art interpretation techniques from the machine learning literature to attribute particular regions of interest or potential teleconnections relevant for any given weather cluster prediction or regime transition. We demonstrate superior forecasting performance relative to several classical meteorological benchmarks, as well as logistic regression and random forests. Due to its wider field of view, we also observe deCNN achieving considerably better performance than regular convolutional neural networks at lead times beyond 5-6 days. Finally, we find transfer learning to be of paramount importance, similar to previous data-driven atmospheric forecasting studies.

KW - Atmosphere

KW - Forecasting

KW - Machine Learning

KW - Neural Networks, Computer

KW - Weather

U2 - 10.1038/s41598-022-12167-8

DO - 10.1038/s41598-022-12167-8

M3 - Journal article

C2 - 35589754

SN - 2045-2322

VL - 12

JO - Scientific Reports

JF - Scientific Reports

IS - 1

M1 - 8395

ER -