Drivers of the UK summer heatwave of 2018

The boreal summer of 2018 saw high temperature records being challenged or broken across many parts of the northern hemisphere. Dry and sunny weather dominated across the United Kingdom from May through to early August, and the meteorological summer (June, July, August) mean air temperature was the joint highest on record in a national series dating from 1884. Anticyclonic circulation, with a higher chance of a notably dry summer across northern Europe, was predicted well in advance by long-range forecast systems. The summer was exceptional overall, but the peak of the heatwave was not as intense as in some other notable heatwave years. The hot summer was largely a consequence of the atmospheric circulation anomalies and elevated sea-surface temperature in the proximity of the United Kingdom, but these alone are insufficient to fully explain the magnitude of the observed temperature anomalies for the season overall. The warming climate was also a significant factor. The suite of tools now available to modern climate science present an opportunity for us to better quantify the current and future likelihood of such extreme events. It is only when all the evidence is brought together in an integrated way that we can provide a more comprehensive picture. Here, it is shown with established methods, using climate models and observations, that the present-day likelihood of a summer temperature anomaly at or above that of 2018 is approximately 11–12%, which is a factor of 30 higher than estimated for a world without man-made greenhouse gas emissions. The most recent set of probabilistic projections from the UK Climate projections (UKCP18) give a wider but overlapping range of 10–25% for the current climate and a likelihood greater than 50% of summers exceeding 2018 temperatures by the mid-twenty-first century. Summer 2018 was the joint hottest summer season (June, July, August) in the Met Office UK national temperature series dating from 1884 (HadUK-Grid version 1 dataset, Hollis et al., 2019), and it was the warmest on record for England. Summer average temperatures were close to +2.0 degC above the 1981–2010 average for a large swathe of southern and central England and Wales (Figure 1). It was the third warmest summer for daily maximum (behind 1976 and 1995) and daily minimum temperatures (behind 2003 and 2006). UK precipitation was 73% of the average, the 15th driest on record (from 1862), and clear skies were prevalent, with the summer overall recording 124% of average sunshine, making it the fourth sunniest summer (in a series from 1929). The peak in temperatures coincided with a sharp increase in the daily death count for England (Office for National Statistics, 2018), and reported impacts included some extensive and prolonged wildfires as discussed in Sibley (2019, this issue). High temperatures were not confined to the meteorological summer season in 2018. Some exceptionally high temperatures were recorded in the United Kingdom during April 2018, with St James's Park, London reaching 29.1°C on 19 April, and there were two further notable warm spells in May associated with dry and sunny weather both early and late in the month (Figure 2). It was one of the top three warmest Mays on record and the second sunniest. In late June, high pressure located just to the northwest of the United Kingdom resulted in a weak easterly flow across the country. Daily maximum temperatures rose considerably at this time (Figure 2). Daily minimum temperatures were above average but not as notable as the daily maximum. Temperatures peaked on the western side of the United Kingdom, with Glasgow reaching an all-time record high daily maximum temperature for the city of +31.9°C on 28 June, and Porthmadog in Wales reaching +33.0°C on the same day. The Met Office synoptic analysis chart for that day is shown in Figure 3(a). Kornhuber et al. (2019) show that coincident extremes in North America, Western Europe and the Caspian Sea region around this time were linked through a near-stationary ‘wavenumber 7’ Rossby wave circulation pattern across the northern hemisphere, a pattern that is also seen in other notable extreme European heatwave events of recent decades. June was exceptionally dry across central and southern England. Some locations in and around Greater London recorded very little or no rain for 57 days from 30 May to 26 July. Such an extended dry spell is unusual but not unprecedented. Southern England recorded only 7.1mm of rain through June, its second driest June in a series from 1862, behind only 1925. This established dry soils which increased the high maximum temperatures experienced in southeast England during late July (Petch et al., 2019). A second phase of hot weather from 23 to 27 July was associated with high pressure to the east of the United Kingdom over northern Scandinavia as shown in Figure 3(b). This resulted in a flow of tropical continental air reaching the United Kingdom, with the highest temperatures now recorded in the south and east of the country. A maximum of +35.6°C was recorded at Felsham, Suffolk on the 27th, and daily minimum temperatures were also very high. However, when compared to other extreme summers, the summer of 2018 did not actually have an extended spell of extremely high temperatures. The summer North Atlantic Oscillation (SNAO) represents the principal pattern of year-to-year variability in the summer climate of the north Atlantic and influences summer rainfall and air temperature across northern Europe through its association with the North Atlantic storm track (Folland et al., 2009). The SNAO is defined by the pressure difference between a region east of Shetland and one to the northwest of Greenland. These centres of action differ from the Iceland-Azores winter North Atlantic Oscillation to reflect the more northerly location of the Atlantic storm track in summer. A June-July-August (JJA) SNAO index, as described in Kendon et al. (2019), is shown in Figure 4. Positive SNAO is associated with a north-shifted storm track resulting in reduced rainfall and higher temperatures across the United Kingdom and northern Europe. The 2018 JJA SNAO value is the fourth highest in a series from 1850. The mean sea-level pressure anomaly for summer 2018 in Figure 5 shows the high-pressure anomaly across the north Atlantic and north European region, associated with a north-shifted jet stream (represented by 250 hPa wind speed anomalies in Figure 5), reducing the influence of cyclones and cool, moist Atlantic air on the United Kingdom and northern Europe. An anticyclonic circulation anomaly over northern Europe during the summer was correctly predicted by Met Office long-range forecasts (Dunstone et al., 2019). Ensemble forecasts started at the beginning of November 2017 and May 2018 predicted increased risk of a dry summer across northern Europe, and these were presented, along with other evidence, to UK government contingency planners in May 2018. The predictability of summer rainfall is shown in the left panel of Figure 6. Conversely, there is no significant skill in the north European mean temperature from these predictions, shown in the right panel of Figure 6, with the exception of the long-term trend component.Dunstone et al. (2019) demonstrate, using a set of perturbation experiments, that the pattern of north Atlantic SSTs in May 2018 (and compared to 1976 in Figure 7), sometimes referred to as the north Atlantic tripole pattern, correlates with northern European rainfall through the process of a northward-shifted jet stream responding to the meridional temperature gradient of the north Atlantic sector. They go on to note that, although the North Atlantic SST tripole likely contributed to the observed circulation changes of summers, such as 1976 and 2018, the model produced only a very weak signal, which is often found in predictions of north Atlantic atmospheric variability (Scaife and Smith, 2018). In addition, the world has warmed significantly since the 1970s, and this is also partly reflected in the comparison of 2018 and 1976 in Figure 7. The pattern of SSTs in the north Atlantic are playing out against a warming background climate. Figure 7 also shows that SSTs in the proximity of the United Kingdom were notably warmer than average through summer 2018 in the range +1 to +3 degC above the long-term climatology. July 2018 saw one of the highest summer-month SST anomalies for the region in the HadISST dataset (Rayner et al., 2003). Figure 8 shows the SST and land-surface temperature anomalies, which have a correlation coefficient of +0.76 as both land- and sea-surface temperatures share some response to larger-scale forcings. The elevated SSTs within the UK coastal waters will also have influenced the magnitude and extent of the high temperatures over land. Using the Met Office Unified Model in regional configuration, Petch et al. (2019) have undertaken sensitivity experiments to quantify this impact. Experiments were conducted in which the SSTs within the model domain were reduced by 1.5 or 3.0 degC across the whole domain, running 5-day forecasts over an 11-day period in late July 2018. A 3 degC SST reduction results in an approximate 1 degC reduction in both maximum and minimum temperatures over land in the simulations, as shown in Figure 9, and a 1.5 degC SST reduction has approximately half that effect. The same model was also used by Petch et al. (2019) to demonstrate that soil moisture deficits in southeast England during July, resulting from the extended dry spell from late May, would also have contributed to elevated temperatures as a consequence of the feedback through evaporation and transpiration rates. We expect UK temperatures to be dependent on atmospheric circulation, with warmer summer temperatures, for example, when southerly winds blow across the United Kingdom. We quantify the relative contribution of atmospheric circulation and long-term climate warming on summer temperature using methods outlined in Deser et al. (2016) and O'Reilly et al. (2017). A daily UK mean temperature series was converted to anomalies relative to a 1960–2018 baseline and linearly de-trended (by which a simple linear regression line was used to remove the long-term warming trend shown in Figure 10). For each day of summer 2018, the closest mean sea-level pressure (MSLP) analogues were found for the same calendar month, but from previous years, for a region bounded by 47.5°N, 5°E and 62.5°N, 12.5°W. Each day's MSLP field in 2018 was reconstructed as a weighted combination of its analogue days, and the day's temperature anomaly in 2018 was estimated by applying the same weights to the temperature anomalies of the analogue days. The circulation-based reconstruction is therefore a temperature estimate based only on temperature from other days in the period 1960–2017 that have an analogous circulation pattern, with no knowledge of 2018 temperatures. In order to increase the daily auto-correlation of the reconstruction, to better match observed behaviour, half the mean of the previous 30 days’ reconstructed temperature anomaly was also added to each day, referred to as a ‘persistence’ adjustment. Elevated SSTs in the coastal waters as described previously would be an expected contributor to this persistence. The resulting reconstructed series is shown in Figure 10. The correlation coefficient between the reconstruction and the de-trended observed series is +0.67. This demonstrates that a significant proportion of the annual variability in summer mean temperature anomalies can be explained with the atmospheric circulation and a persistence term. The reconstruction gives a temperature anomaly of +0.84 degC relative to 1960–2018, which is comparable in magnitude to the linear trend component of +0.7 degC shown in Figure 11. This approach therefore suggests that approximately half of the summer mean temperature anomalies of 2018 may be explained by the atmospheric circulation regimes and an additional, similar-sized contribution from the warming trend. Using the UNSEEN method (Thompson et al., 2017), the Met Office decadal prediction system can be used to estimate the present-day likelihood of summer mean temperatures at or above those observed in 2018. The large ensemble of simulations provides a synthetic dataset of virtual observations that enables an assessment of the likelihood of seeing extremes such as 2018. Furthermore, it allows the identification of events that are even more extreme than those previously observed but that are both plausible and possible in the current climate. This has been carried out with a 59-year hindcast set of two 40-member ensembles, one initialised in November and the other in May, for the period 1960–2018. This provides a set of 4720 samples of summer (June to August) conditions. In order to make use of the maximum amount of data – and information about extremes – from both model and observations, a simple linear trend correction was applied to the model and observations. In this way, the full length of the data can be viewed as representative of the present-day climate. One effect of removing the trend is to modify the magnitude of hot events from the observational record such that 1976 becomes the warmest year of the period (Figure 10), with 2018 in seventh place, indicating that a 1976-type event in the current climate would be more extreme than what was experienced in 2018. In order to determine the applicability of the UNSEEN method to UK summer temperatures, the model ensemble was first assessed against observations as described in Thompson et al. (2017). Figure 12 summarises the likelihood of summers, with mean temperatures exceeding those of 2018. Within this ensemble, there is an approximately 11% chance of exceeding summer 2018 in the current climate and an approximately 1% chance of exceeding it by 1 degC or more (indicated by the dashed line in Figure 12). Over the same period (1960–2018), the de-trended observations have seven summers equal to or exceeding 2018 in a 59-year series, approximately 12% of years, so the observational record and the large model ensemble provide a very similar estimate of the likelihood. A suite of 20 models that contributed to the Coupled Model Inter-comparison Project (CMIP5) have been used, along with the methodology of Christidis et al. (2018) and Christidis et al. (2013), to estimate the change in extreme events when comparing the present-day climate (referred to as ALL) with the likely distribution from a counterfactual ‘natural’ world (referred to as NAT) without the influence of human activity (Stott et al., 2016). A total of 56 simulations of the historical climate and 66 simulations of the natural climate are available. These model simulations are distinct from those used in UNSEEN. These are used to address the following question: ’How have anthropogenic forcings changed the likelihood of summers in the UK being warmer than 2018?’ Summer mean air temperature from observations (HadUK-Grid) and the model ensembles re-gridded to match the observational coverage are shown in Figure 13 relative to a 1901–1931 baseline. The ALL and NAT ensembles overlay each other through the early twentieth century but start to diverge by the late twentieth and early twenty-first century, reflecting an estimated increased likelihood of a 2018 summer in the ALL scenario compared to NAT. The ensemble suggests that anomalies comparable to 2018 are relatively rare for the past climate but become relatively common by the mid-twenty-first century and would be below average by the end of the century. The period 2011–2025 (i.e. centred on 2018) is used to construct return period estimates from the ALL and NAT distributions using extreme value analyses with a generalised Pareto distribution (Christidis et al., 2013). Within the NAT ensemble, the likelihood of a 2018 summer is estimated as 0.4% (range 0.1–0.7%) and in the ALL ensemble as 12.3% (range 10.8–13.8%). This represents an estimated increase in the likelihood of a 2018 summer, due to anthropogenic climate change, by a factor of 30. The same model ensemble suggests that the likelihood increases to 53% (range 50–57%) by the 2050s. The probabilistic projections component of the UKCP18 project (Lowe et al., 2018) provides future climate change projections and an assessment of key known uncertainties (Murphy et al., 2018). The methodology combines results from a 57-member perturbed physics ensemble (PPE) of the Met Office climate model HadCM3, along with multi-model projections from the CMIP5 earth system models, which are combined within a Bayesian statistical framework. The method is more fully described in Sexton et al. (2012) and Harris et al. (2013). The UKCP18 projects a greater chance of hotter, drier summers in the future, with overall warming by the end of century ranging from +0.5 to +5.7 degC relative to a 1981–2000 baseline. Warming trends are projected to be greater in summer than winter, exaggerating the seasonal cycle, and there is a north/south contrast with greater increases in maximum summer temperatures over the south of the United Kingdom, which already experiences the highest summer temperatures in the United Kingdom. The probabilistic simulations have also been used to quantify the likelihood of exceeding the 2018 UK summer temperatures, providing an alternative approach to both UNSEEN and event attribution. Figure 14 shows the cumulative probability distributions for a 1981–2000 baseline period (green), estimates for a 2018 climate (blue) and a projection for 2050 using the Representative Concentration Pathway (RCP) 4.5 scenario for global emissions of greenhouse gases and land use change (purple, Thomson et al., 2011). The UKCP18 gives a probability of approximately 20% for the 2018 event occurring in the current climate but with a range (not shown) of 10–25%. This is wider than for the attribution and UNSEEN analysis, but the methods all overlap. UKCP18 can also estimate the probability of exceeding the 2018 warming level by 2050, which ranges from 52 to 66%, dependent on the choice of emission scenario. In the high emissions case, it may exceed 90% by the end of the twenty-first century. The summer of 2018 in the United Kingdom was the joint hottest on record for mean temperature and regularly experienced spells of daily maximum temperatures in the low to mid-30s Celsius. A positive phase of the SNAO with a northward-shifted jet stream resulted in anticyclonic conditions dominating across western Europe and the United Kingdom. However, the synoptic situation in the summer did not result in an extended spell of southerly or southeasterly air masses over the United Kingdom that would tend to result in an extended spell of exceptionally high temperature, such as that experienced in the summers of 1976, 1995, 2003, 2006 and 2019. Recent research by Dunstone et al. (2019) has shown that a tripole pattern in north Atlantic SSTs in late spring contributed to long-range predictability of the signal for high-pressure dominance and low rainfall across northern Europe. Most of the anomalies in the UK mean temperature of summer 2018 can be explained based on circulation, accounting for elevated SSTs in the coastal waters of the United Kingdom, as well as other feedbacks such as extremely low soil moisture following an extended dry spell through June and July. However, these factors alone are insufficient to fully explain the observed temperature anomaly, which also has an important contribution from the underlying warming trend in the UK climate. To further quantify the role of our changing climate, multiple approaches using a range of climate model ensembles have been used to estimate the present-day likelihood of summer temperatures exceeding those of 2018. A range of climate models and observations are in broad agreement that the present-day likelihood of summers warmer than 2018 is approximately 11–12%. Anthropogenic climate change has significantly increased the risk of a 2018 summer temperature anomaly, and formal attribution experiments estimate this to be a factor of 30 higher than for a world without human greenhouse gas emissions. The latest set of UK climate projections (UKCP18) estimate that a 2018-like summer could be more common than not by the mid-twenty-first century. This work was supported by the Met Office Hadley Centre Climate Programme funded by BEIS and Defra. We thank two anonymous reviewers for constructive feedback that improved the manuscript.