Automated detection and classification of synoptic-scale fronts from atmospheric data grids

Abstract

Abstract. Automatic determination of fronts from atmospheric data is an important task for weather prediction as well as for research of synoptic-scale phenomena. In this paper we introduce a deep neural network to detect and classify fronts from multi-level ERA5 reanalysis data. Model training and prediction is evaluated using two different regions covering Europe and North America with data from two weather services. We apply label deformation within our loss function, which removes the need for skeleton operations or other complicated post-processing steps as used in other work, to create the final output. We obtain good prediction scores with a critical success index higher than 66.9 % and an object detection rate of more than 77.3 %. Frontal climatologies of our network are highly correlated (greater than 77.2 %) to climatologies created from weather service data. Comparison with a well-established baseline method based on thermodynamic criteria shows a better performance of our network classification. Evaluated cross sections further show that the surface front data of the weather services as well as our network classification are physically plausible. Finally, we investigate the link between fronts and extreme precipitation events to showcase possible applications of the proposed method. This demonstrates the usefulness of our new method for scientific investigations.