A deep learning based classification of atmospheric circulation types over Europe: projection of future changes in a CMIP6 large ensemble

Abstract

High- and low pressure systems of the large-scale atmospheric circulation in the mid-latitudes drive European weather and climate. Potential future changes in the occurrence of circulation types are highly relevant for society. Classifying the highly dynamic atmospheric circulation into discrete classes of circulation types helps to categorize the linkages between atmospheric forcing and surface conditions (e.g. extreme events). Previous studies have revealed a high internal variability of projected changes of circulation types. Dealing with this high internal variability requires the employment of a single-model initial-condition large ensemble (SMILE) and an automated classification method, which can be applied to large climate data sets. One of the most established classifications in Europe are the 29 subjective circulation types called Grosswetterlagen by Hess & Brezowsky (HB circulation types). We developed, in the first analysis of its kind, an automated version of this subjective classification using deep learning. Our classifier reaches an overall accuracy of 41.1% on the test sets of nested cross-validation. It outperforms the state-of-the-art automatization of the HB circulation types in 20 of the 29 classes. We apply the deep learning classifier to the SMHI-LENS, a SMILE of the Coupled Model Intercomparison Project phase 6, composed of 50 members of the EC-Earth3 model under the SSP37.0 scenario. For the analysis of future frequency changes of the 29 circulation types, we use the signal-to-noise ratio to discriminate the climate change signal from the noise of internal variability. Using a 5%-significance level, we find significant frequency changes in 69% of the circulation types when comparing the future (2071–2100) to a reference period (1991–2020).

Future changes of circulation types and their effects on surface air temperature and precipitation in the SMHI large ensemble

“…During the review process of this study, an article from Mittermeier et al (2022) was published that analyzed, among other questions, CT frequency changes in the SMHI-LENS SSP370 scenario over a large Europe-North Atlantic domain. Their results are consistent with our finding that the frequency of some CTs changes significantly under future climate change.…”

“…Details, however, differ from their as well as from other studies that analyze future circulation changes. For example, Huguenin et al (2020) find no change in any summer and winter CT frequencies towards the end of the twenty-first century, but their focus is on Central Europe and, probably more important in this context, they use a different classification method, namely the Grosswettertypes (GWT) classification method which was also applied in Mittermeier et al (2022). GWT groups the large-scale pressure fields mainly based on their predominant flow direction and basically does not take into account the pressure gradient magnitude which limits its efficacy in separating effects of CTs on other variables such as temperature and precipitation (Hansen and Belušić 2021).…”

Automated Classification of Atmospheric Circulation Types for Compound Flood Risk Assessment: CMIP6 Model Analysis Utilising a Deep Learning Ensemble

“…They used their trained neural networks ensemble to analyse future changes in a single-model initial-condition large ensemble of the global climate model EC-Earth3 under the Shared Socio-Economic Pathways (Riahi et al, 2017) 3-7.0 scenario. Mittermeier et al (2022) demonstrated the enhanced capabilities of neural networks in classifying Großwetterlagen. However, the occurrence rate of Cyclonic Westerly was strongly underestimated by their networks.…”

“…Two studies have previously focused on the automatic classification of Großwetterlagen, namely James (2007) and Mittermeier et al (2022). James (2007) proposed an objective classification based on composite plots, whereby a circulation type was assigned to each day based on the composite with the highest correlation after temporal smoothing.…”

“…Liu et al (2016) utilised a deep convolutional neural network to detect extreme weathers, such as tropical cyclones, in climate datasets. Mittermeier et al (2022) subsequently modified this neural network architecture and used it for the classification of Großwetterlagen, achieving higher accuracy than the composite method by James (2007). They used their trained neural networks ensemble to analyse future changes in a single-model initial-condition large ensemble of the global climate model EC-Earth3 under the Shared Socio-Economic Pathways (Riahi et al, 2017) 3-7.0 scenario.…”