The summer of 2018 witnessed prolonged and severe heatwaves and droughts across northwestern Europe, causing significant socio-economic and environmental impacts. While anthropogenic climate change played a role, natural variability was also a crucial factor. Previous research suggested various drivers for these events, including persistent blocking, subtropical ridge or tropical continental air flow, Rossby Wave-7 patterns, positive summer North Atlantic Oscillation (NAO), Atlantic sea-surface temperature (SST) tripole patterns, and local SST and soil moisture content. Improved seasonal and short-term predictions have been made using SST tripole patterns and local SST/soil moisture content. However, extended-range prediction for the 2018 heatwaves remained largely unstudied. Skillful sub-seasonal or extended-range (2-4 weeks) heatwave prediction is crucial for mitigation and preparation, but predicting heatwave occurrence 3-4 weeks in advance remains challenging. Slow-varying drivers like the Madden-Julian Oscillation, mid-latitude persistent flow regimes, and land-atmosphere interaction could improve sub-seasonal predictability. Reliable identification of heatwave drivers can improve predictability and preparedness. Heatwaves are commonly associated with high-pressure systems and warm air masses, particularly atmospheric blocking for summertime European events. Other persistent regimes, like the subtropical ridge or Atlantic Low, are also linked to Western European heatwaves. Extended-range prediction thus relies on the predictability of atmospheric blocking and persistent weather regimes. The summer NAO, reflecting jet stream and storm track variability, is an important factor, with its positive phase associated with northward jet stream anomalies and warm, dry conditions over northwestern Europe. North Atlantic SST and tropical Atlantic convective anomalies also play a role. Improving dynamical links among heatwave-related circulation regimes and ocean temperature anomalies in forecast models could extend the prediction range.

Numerous studies have explored the 2018 European heatwave and drought, identifying various drivers such as persistent atmospheric blocking (Rösner et al., 2019), subtropical ridges or tropical continental airflow (Sousa et al., 2019), strong Rossby Wave-7 patterns (Kornhuber et al., 2019), positive summer NAO (Drouard et al., 2019), Atlantic SST tripole patterns (McCarthy et al., 2019), and local SST and soil moisture (Petch et al., 2020). Previous work also highlighted the success of incorporating SST tripole patterns and local SST/soil moisture in seasonal and short-term predictions (Dunstone et al., 2019; Petch et al., 2020). However, research on extended-range prediction for these heatwaves was lacking prior to this study. Other studies have focused on the general drivers of heatwaves and their predictability, associating them with high-pressure systems, warm air masses, and atmospheric blocking (Della-Marta et al., 2007; Perkins, 2015; Schaller et al., 2018). The influence of persistent weather regimes over the Atlantic-European sector, like the subtropical ridge or Atlantic Low, on heatwaves has also been noted (Cassou et al., 2005; Sousa et al., 2018). The role of the summer NAO and its relationship with blocking and the Azores High have also been investigated (Folland et al., 2009; Trouet et al., 2018). The influence of North Atlantic SST and tropical Atlantic convective anomalies on European heatwave development has also been established (Duchez et al., 2016).

This study utilized the ERA-Interim reanalysis data from the European Centre for Medium-Range Weather Forecasts (ECMWF) for analysis and forecast verification. Daily 2-m temperature and 500-hPa geopotential height were interpolated onto the S2S grids, and daily anomalies were calculated relative to the 1999–2010 period. NAO and SCA teleconnection indices were obtained from NCEP/CPC. Real-time forecasts from the subseasonal to seasonal (S2S) database, encompassing 11 models with a total of 295 ensemble members, were used. The study focused on the period from early July to mid-August 2018. Three metrics characterized heatwave events: daily 2-m temperature anomaly (aT2), deviation from the 90th percentile value (T2-P90), and excess heat factor (EHF). Circulation indices included a blocking index (BL, using meridional gradients of 500-hPa geopotential height), Atlantic Low (AL) index (center of action approach based on Cassou et al., 2005), Azores High (AZH) index (regional mean of normalized 500-hPa geopotential height anomalies), and North Atlantic Oscillation (NAO) index (meridional gradient method). Bivariate correlation, partial correlation, and multiple linear regression were used for reanalysis data, while quartile statistics were used for S2S forecasts. The S2S forecasts were stratified based on positive/negative values and quartiles of heatwave metrics to examine circulation patterns associated with different heatwave intensities.

The analysis revealed that the persistent blocking regime (BL) was the most significant driver of the 2018 heatwave over Scandinavia (SC), with AL and NAO as secondary contributors (AL positive, NAO negative). The Atlantic Low (AL) was the primary driver for the Western Europe (WE) heatwave. The multi-model ensemble (MME) forecasts captured the evolution of warm spells up to 3 weeks in advance, but heatwave occurrence and significance predictions were less accurate. Successful prediction of warm spells with 1-2 weeks' lead time occurred only when the heatwave reached its mature stage. BL and AL were predictable 2 weeks in advance, contributing to successful warm spell predictions for SC and WE, respectively. While variations in AZH and NAO were captured, their contributions to warm spell predictions remained unclear. Bivariate and partial correlation analyses revealed interdependencies between BL, AL, and AZH; BL was positively correlated with AL and AZH, mitigating the negative relationship between AL and AZH. Multiple linear regression models explained approximately 72% and 57% of the area-averaged temperature variance for SC and WE, respectively, using BL, AL, AZH, and NAO. For SC, BL and AL were most influential, while for WE, AL was dominant. Stratifying forecasts by quartiles of heatwave metrics revealed a linear relationship between BL and heatwave intensity for SC, and between AL and heatwave intensity for WE. The interdependency of BL, AL, and AZH was also captured in the MME forecasts for the first two lead weeks. For longer lead times, the MME forecasts had higher uncertainty.

The findings highlight the importance of atmospheric blocking (BL) and Atlantic Low (AL) regimes in driving the 2018 northwestern European heatwaves. The success in predicting warm spells up to 3 weeks in advance using MME demonstrates the potential of subseasonal-to-seasonal prediction. However, the limitations in predicting heatwave occurrence and significance highlight the need for improved model representation of key processes. The interdependencies between BL, AL, and AZH, captured in both observations and short-lead-time forecasts, suggest that future research should focus on improving model representation of these relationships. The weak contribution of NAO in daily/weekly timescale, contrasting with its significant role in previous studies on a monthly/seasonal timescale (positive correlation between monthly NAO and T2 anomalies but weak negative correlation between daily NAO and T2 anomalies), indicates a complex interplay of timescales. Improvements in representing short-term co-variations between AZH and NAO might improve the lead times for BL and AL prediction and subsequently, the prediction of temperature anomalies.

This study identified BL and AL as the major drivers of the 2018 heatwaves over SC and WE, respectively, demonstrating that the MME forecasts could capture these relationships at short lead times. The relatively poor prediction of heatwave occurrence and significance, even at shorter lead times, highlights the need for improvements in models to better capture the complexities of these events. Future research should focus on improving model representation of the interactions between BL, AL, AZH, and NAO, particularly at shorter timescales, to enhance extended-range heatwave prediction.

The study's reliance on a single extreme heatwave event limits the generalizability of the findings. The use of only 11 S2S models might not fully capture the diversity of models and forecast techniques. The study primarily focused on statistical relationships between circulation patterns and heatwave metrics, with limited exploration of underlying dynamical processes. The analysis mainly focused on northwestern Europe, and the results might not be directly transferable to other regions. Finally, the study's reliance on ERA-Interim reanalysis introduces potential uncertainties related to the accuracy and resolution of the reanalysis data.