Current and past atmospheric heat wave precursors for Sweden: A Machine Learning-based weather regime approach

In this study we assess the occurrence of summertime heatwaves (HW) and their underlaying atmospheric drivers under current and past climate conditions for Sweden using a weather regime classification based on optimized fuzzy rules. It uses daily mean 500hPa geopotential height of ERA5 reanalysis at 0.5° spatial resolution as atmospheric input data for the period of 1940 to 2022. Daily anomalies of 500hPa geopotential height at each grid over the Euro-Atlantic region have been computed as daily deviations from the long-term climatology of 1981-2010. Daily mean temperature from station data over the same 30-year period, distributed in the whole of Sweden, serve as predictand to reflect the variability of local climate. They help to optimize pre-defined fuzzy rules describing individual weather regimes (WRs). A set of twelve temperature-induced WRs are classified as daily timeseries for the years of 1940 to 2022.

HWs are investigated using the Excess Heat Factor (EHF) and the NDQ90 index when daily maximum temperature exceeds the 90^th percentile over the reference period based on ERA5 reanalysis data. The EHF is a measure of heatwave intensity related to human health impacts and consists of two indices describing the deviation of the three-day mean air temperature from the long-term 95^th percentile-based climatology and the short-term anomaly of the previous 30 days. The Chi-square test is used to study the significance of the classified WR along with the co-occurrence of a HW. 

For the case-study of Stockholm, 985 HW events are detected by the NDQ90 index from 1940 to 2022 during May to August. Nearly 83% of detected HWs is found to coincide with the occurrence of four types of anticyclonic WRs. One type of anticyclone explains nearly 40% of the detected summertime HW events. During August, it particularly explains 47% of detected HWs. It is likely because of the strong high-pressure system situated over the North Sea and southern Scandinavia that caused the warmer-than-average temperatures over northern Europe. Similar results are found when using the EHF. In addition, we present how this approach can be extended to investigate the linkages of leading WRs and the occurrence of detected HWs in major European cities.