The storm intensified as it continued northwards towards the island of Taiwan, becoming a category 4-equivalent Typhoon on the 24th, with maximum (10-minute) sustained winds of 185 km/h. It made a prolonged landfall in northeast Taiwan on the 24th, bringing both heavy rain and high winds that killed 10 people and injured more than 900, while the agricultural sector reported damages of roughly US$50 million (FocusTaiwan, 2024a, FocusTaiwan, 2024b). It subsequently made landfall as a weaker, but still destructive tropical storm on mainland China on July 25. Gaemi brought heavy rainfall to coastal and inland regions, particularly the Hunan province, as it weakened to a tropical depression. Cyclone-based rainfall is uncommon so far inland in China and the heavy precipitation led to flooding and a mudslide that killed 15 people, and another 15 people in neighbouring provinces.35 remained missing a week after the disaster, and 290,000 people were evacuated (CNN, 2024).

The influence of climate change on tropical cyclones is complex compared to other types of extreme weather events. However, attribution studies are increasingly focusing on these destructive events. Rapid attribution studies to date have focused primarily on severe rainfall from such storms. Here, we use several different approaches to investigate the influence of climate change on multiple aspects of Typhoon Gaemi. The study focuses on the three geographic regions that experienced severe impacts – northern Philippines, the island of Taiwan and Hunan province, and analyse whether and to what extent human-induced climate change affected wind speeds and rainfall. To study the conditions that formed and fuelled Gaemi, we also analyse the role of climate change in high sea surface temperatures and potential intensity, a metric combining sea surface temperature, air temperature and air humidity data to predict maximum typhoon wind speeds. The study combines the established World Weather Attribution protocol with a new approach using the Imperial College Storm Model (IRIS) to analyse the role of human-induced climate change in tropical cyclones. 

Main findings

• Typhoon Gaemi brought destructive winds and rainfall to large regions of southeast Asia, including the northern Philippines, Taiwan, and Hunan . At least 90 people were killed, thousands were injured and hundreds of thousands had to leave their homes. The extreme rainfall and high winds triggered landslides, widespread power outages and severe damage to infrastructure and agriculture. 
• In today’s climate, that has already been warmed by 1.2C due to the burning of fossil fuels, weather observations indicate that rainfall events as severe as those brought by Typhoon Gaemi now occur about once every 20 (5 – 30) years in the northern Philippines, about once every 5 (1.5 – 20) years in Taiwan, and about once every 100 (90 – 160) years in Hunan province.
• To determine the role of climate change we combine observations with climate models. In Taiwan and Hunan, the rainfall was about 14% and 9% heavier respectively due to climate change, and in both regions, the rainfall total was made about 60% more likely by climate change. If the world continues to burn fossil fuels, causing global warming to reach 2°C above pre industrial levels, devastating Typhoon rainfall events in both regions will become 30-50% more likely.
• In the northern Philippines, the analysis did not identify a significant trend up to today. Observations indicate that 3-day rainfall events have increased by about 12%, however, there is large uncertainty in these data sets. Climate models suggest both increases and decreases in rainfall in the current climate, but an increase in a future climate with 2°C of warming. 
• The IRIS model was used to investigate Gaemi’s strong winds by analysing category 4-equivalent storms in the Western North Pacific basin, a region that includes the South China and Philippine seas.
• By statistically modelling storms in a 1.2°C cooler climate, this model showed that climate change was responsible for an increase of about 30% in the number of such storms (now 6-7 times per year, up from 5 times), and equivalently that the maximum wind speeds of similar storms are now 3.9 m/s (around 7%) more intense.
• The conditions that formed and fueled Typhoon Gaemi were studied for links to climate change, using potential intensity and sea surface temperatures surrounding the storm track in July 2024. These conditions occur about every second year for potential intensity and about once every 15 years for sea surface temperatures.
• The influence of climate change on potential intensity is highly uncertain, as observations show a very large increase with warming (about a factor of 100 and a potential intensity increase of 6 m/s) that climate models do not capture. Sea surface temperatures as hot as those observed in July 2024 were almost impossible without climate change and have become about 1 degree warmer. If global warming reaches 2°C, sea surface temperatures are projected to be another 0.6°C warmer, and the conditions associated with Typhoon Gaemi will continue to increase in likelihood by a further factor of about 10. 
• Together, these findings indicate that climate change is enhancing conditions conducive to Typhoons, and when they occur the resulting rainfall totals and wind speeds are more intense. This is in line with other scientific findings that tropical cyclones are becoming more intense and wetter under climate change.
• Rural communities with climate sensitive livelihoods (e.g. agriculture), the urban poor residing in the lowest lying land, and those living on exposed hillsides susceptible to landslides were the most affected by the multitude of hazards stemming from the typhoon. 
• The regions affected by Typhoon Gaemi have early warning systems and comprehensive emergency response systems in place for tropical cyclones that help manage impacts. Flood risk associated with extreme rainfall is well-assessed in the affected regions, but existing urban plans and flood control infrastructure are not able to withstand the more extreme floods that are driven by climate change. Unplanned urban development, including in Metro Manila where the population has rapidly increased, is increasing the number of people at risk, especially in lower lying informal areas.

The 2023/24 storm season is the ninth season since the founding of the Western Europe storm naming group. The initiative began in 2015, when the Met Office and Met Éireann, Ireland’s national meteorological service, officially started to identify and name storms that have the potential to cause medium or high impacts, and expanded to include the Royal Netherlands Meteorological Institute (KNMI) in 2019. 

Scientists from the United Kingdom, Ireland, the Netherlands, Sweden and Germany, including scientists from each of the National Meteorological Services in the Western Europe storm naming group, collaborated to assess to what extent human induced climate change and the North Atlantic Oscillation (NAO) influenced the average storm severity, using the wind-based Storm Severity Index (SSI) over a wide region encompassing the United Kingdom and Ireland. The study also investigated the influence of climate change on the average precipitation on stormy days from October 2023 to March 2024, which was one of wettest Oct-Mar periods on record for the UK and the third on record for Ireland, and the wettest over the region south of 54N studied. The study uses peer-reviewed methods to assess changes in storm severity, associated precipitation and precipitation accumulated over the storm season. 

Main findings

• • The 2023/24 storm season, studied here by stormy day wind severity, associated rainfall, and accumulated seasonal rainfall in October-March, has brought deaths, flooding, transport disruptions and power outages, among other impacts, to the UK and Ireland.
  • Successive floods have compounded impacts on the agriculture and housing sectors, leading to cascading impacts on socioeconomic and psychosocial health, and eroding people’s coping capacity, particularly low-income groups. Combined with the cost-of-living crisis, the successive flood events are another layer of disruption at a time when people’s financial resilience is already being tested. 
  • The storm severity index (SSI) was used to define stormy days to study. The SSI considers both the strength of the wind and the area affected. In this analysis we looked at rainfall and wind speed on stormy days identified by the SSI.
  • In today’s climate with 1.2C of warming, stormy days with winds as intense as in the 2023/24 season occur about every 4 years. The associated precipitation is expected to occur about once every 5 years. The seasonal precipitation of the October-March period was more extreme, expected to occur about once every 20 years.
  • Analyses of observations are used to determine whether a trend can be observed in these measures. To determine the role of climate change in these observed changes, we combine observations with climate models.
  • The average precipitation on stormy days are observed to have become approximately 30% more intense, compared to a 1.2C cooler pre-industrial climate. Models agree on the direction of change, combining observations and models indicate that average precipitation on stormy days increased by about 20% due to human induced climate change, or equivalently the 2023/24 level has become about a factor of 10 more likely. 
  • The observed precipitation across Oct-Mar has a strong trend, with a magnitude increase of about 25%. Climate models utilised in this study broadly agree on the direction of the change, and the combination of  observation and model results indicates an increase in magnitude of 6% to 25%, or equivalently the 2023/24 level has become at least a factor of 4 more likely.
  • Models indicate that the trends in average precipitation on stormy days and seasonal precipitation continue into the future, in a climate that is 0.8C warmer than now. Average precipitation on stormy days becomes about another factor of 1.6 times more likely, or 4% more intense, and seasonal precipitation becomes about a factor of 1.5 more likely or 2% more intense.
  • Looking at average SSI on storm days, while some studies using other methods suggest an increase in storminess in a future climate, our analysis has shown a decreasing trend. Our results show that average SSI indices as observed this year became about a factor of 2 less likely. The synthesis of the models also shows a negative trend and, when combined with observations, the results indicate that  a stormy season as observed this year is nowadays a factor of about 1.4 less likely due to human induced climate change. 
  • This highlights the need for ongoing research into how climate change may influence the severity and frequency of windstorms in northern Europe.
  • NAO is a key driver of ‘storminess’ and has been accounted for in this analysis. However, the Oct-Mar 2023/24 averaged NAO was almost neutral.
  • Comprehensive flood risk management is required in the UK and Ireland that encompasses legislative frameworks, strategic planning, and substantial funding. Major UK cities are starting to integrate nature-based solutions into their designs. In Ireland, flood relief projects have been integrating nature-based solutions alongside traditional engineering solutions for over 20 years. Both the UK Met Office and Met Éireann are continuously improving their impact-based weather forecasting mechanisms to facilitate the translation of warning into action, in partnership with other government bodies to ensure their people’s safety. 

To assess to what extent human-induced climate change altered the likelihood and intensity of the heavy precipitation and high wind speeds that caused these impacts researchers from the World Weather Attribution initiative undertook an attribution study on the event.

The focus on two indicators allows us to differentiate the role of climate change in two important characteristics of the event: the 3-day mean precipitation (Rx3day), and the 3-day mean of maximum wind speed magnitude (WSx3day), averaged over the study region where the majority of impacts was observed, defined as a box of 40-50N, 25-45E (Figure 1) and considering only land areas. To account for the climate of the region, with relatively dry and warm summers and wet winters, we study the annual maxima based on July to June cycles.

Main findings 

• Storm Bettina hit the Crimean peninsula in the midst of the active Russia-Ukraine war adding to wide-ranging vulnerabilities across the storm affected areas.
• Storms like Bettina are fairly common in the region at this time of year, which is reflected in the return periods of the event which, in the current climate, are 1 in 3 years for the wind speeds and 1 in 20 years for the associated precipitation (which combines snow and rain).
• Because of human-induced warming, an increasingly larger proportion of precipitation associated with storms like this falls as rain instead of snow, leading to larger flood damages.
• We use observations-based data products and climate models to estimate the role of human-induced climate change in storms like this. The results are very different for rainfall compared to wind speeds. 
• For the precipitation as observed during Storm Bettina, we find that the burning of fossil fuels has increased the likelihood of its occurrence by about a factor of 2. The intensity of an event like this has increased by about 5 percent due to human-induced climate change. 
• Looking at the future, for a climate 2°C warmer than in preindustrial times, models suggest that rainfall intensity and likelihood will increase further. 
• For wind speeds as associated with storm Bettina we find that in observation based products there is a decrease in the likelihood and intensity, while climate models show decreases or increase, leading to no change on average. 
• For a climate 2°C warmer than in preindustrial times, climate models show overall a modest further increase in likelihood and intensity of wind speeds as observed in the 2023 Black Sea event.
• These findings suggest that the decrease we see in wind speeds in the observations are not due to climate change but other drivers e.g. natural variability. Given the model results and the scientific literature, the possibility of an increase in strong winds needs to be taken seriously, even if it cannot be attributed to climate change at this time. 
• Particularly vulnerable groups, notably elderly people, children, and people with disabilities, are more likely to be severely impacted by extreme weather shocks. In particular here, the armed conflict will have exacerbated situations of vulnerability and exposure notably by limiting the ability of communities to respond and recover after the storm and compounding situations of displacement. 
• Disaster response for the storm included flood evacuations, first aid provision, and the set-up of warming points by local governments, Red Cross national societies, and other organisations. 

Using the same methods usually used in WWA’s rapid attribution studies, researchers from Imperial College London and Oxford University found that the extreme rainfall that hit Japan during Typhoon Hagibis in October 2019 was made about 67% more likely by human-caused climate change.

In an additional step the scientists calculated what the monetary consequences of this increase in rainfall are and found that roughly $4 billion of the $10 billion damage in insured losses caused by the rainfall can be attributed to climate change.

There are not many studies attributing economic damages to climate change, so the methodology is less established. Similar studies have calculated the financial damage attributable to climate change in particular extreme events, using the same methodology as in this study, for example a study of Hurricane Harvey, which hit Texas in 2017, found that $67 billion of the damage could be attributed to climate change. This is the first study, for any extreme weather event in Japan, to calculate the damage attributed to climate change, and may underestimate the influence of climate change as observed extreme rainfall in Japan has increased by more than climate models simulate.

The results reflect the growing economic burden Japan – and other countries – already face from climate change and will increasingly experience if emissions are not rapidly eliminated: the “costs of inaction”. The study looks at how much climate change increased the damage from just one extreme event, but as temperatures rise Japan is being hit by a growing number of extreme heat waves and heavy rainfall events.

Key findings:

• The precipitation recorded on 19–20 September 2019 associated with Tropical Storm Imelda was extreme, expected only approx. every 1200 yr at the station with the highest total amount of rainfall. Return times at other stations between East Houston and Beaumont were almost as high. The chances of recording this much precipitation at any of 85 stations along the Gulf Coast is however much higher, at about 1 in 50 years.
• Two standard statistical analyses of the observations show that the probability for such an amount of rain has increased by a factor 2.6 (1.6 to 5.0) since 1900, or equivalently the amount of rainfall in such an event has increased by 18% (11% to 28%) since 1900.
• Taking high-resolution climate models into account, we conclude that two-day extreme precipitation events along the Gulf Coast as intense as observed on 19–20 September 2019 or higher have become 1.6 to 2.6 times more likely due to anthropogenic climate change, or 9% to 17% more intense.
• This study highlights that climate change has clearly led to increased precipitation during extreme events in southeast Texas. Coupled with sea level rise, climate change has resulted in more frequent and intense flooding, especially in coastal areas. This needs to be seen in the context of rapid urban expansion in the area and is characterized by a loss in impervious cover. In part this expansion is driven by population growth. It has resulted in an increase in the number of people and value of property at risk to flooding. An estimated 6.6 million people live in the counties impacted by Imelda.

Introduction

Around 19 September 2019 torrential rainfall from Tropical Storm Imelda caused large-scale flooding in Southeast Texas from Houston to the Louisiana border. Figure 1 shows the highest 1-day precipitation totals from the NOAA calibrated radar (12 UTC 18 September to 12 UTC 19 September), which shows rainfall amounts exceeding 500 mm/day (black) from eastern Houston to Beaumont and the Louisiana border. This is also shown by station data from the GHCN-D v2 dataset. Most stations record precipitation between 8 am and 8 am local time, which corresponds to 13-13 UTC, very close to the time interval of the radar data. The highest daily value recorded at a station is 493.5 mm/day at Beaumont Research Center, Texas. However, given the large amounts of missing data near the maximum defined by the radar data we posit that higher values are likely to be reported at stations that become available with a delay.

After making landfall on 17 September, Imelda stalled north of Houston on 18 and 19 September. While the storm was downgraded to a Tropical Depression, intense banding contributed to several waves of heavy precipitation across the region creating the horseshoe pattern shown in Figure 1. Furthermore, Imelda’s long residence time and slow forward motion over Southeast Texas generated extremely intense rainfall over relatively small areas, so that the two-day precipitation amounts are even more exceptional in smaller areas. Figure 2 shows the precipitation amounts averaged over the 2-day period 19-20 September 2019. The highest recorded accumulation at any station is 355.3 mm/day, or 710.6 mm/2dy, at Roman Forest 1.9 ENE, Texas.

Observed trends

In the same way as in previous studies on extreme rainfall (in England, France, Chennai, Louisiana and Houston), we fitted a Generalized Extreme Value (GEV) distribution to the NCEI GHCN-D v2 station data that scales with global mean temperature as a proxy for climate change. See Van der Wiel et al. (2017) for a detailed explanation of the procedure, which was also used in the analysis of Hurricane Harvey’s precipitation in Van Oldenborgh et al. (2017). As in the latter publication we use two datasets: 13 stations with at least 80 years of data and at least 1º apart, and 85 stations with at least 30 years of data and at least 0.1º apart. There is one important change in the GHCN-D dataset: an extreme precipitation event in Abbeville on 7–9 August 1940 that had been removed in the version we used for the Harvey analysis has since been reinstated as a correct observation. It has precipitation amounts almost comparable to Harvey and is now visible on the fits. (As an aside, inclusion of this event, plus the data since 2017, does not greatly affect the conclusion of the Harvey analysis in return time, probability ratio or change in intensity, although the uncertainty ranges increase somewhat. The GEV function with positive shape parameter has a large probability for extreme events; that is, the fit is determined more by the many smaller events.)

The GEV fits give a local return time in the current climate, the inverse of which gives the probability of as much rain or more than observed at the most extreme station at a given location (a return time of 50 yr means a probability of 2% every year). For one-day precipitation in the 85-station dataset this is about 550 yr (320 to 750 yr) and for the highest two-day precipitation about 1200 yr (520 to 1700 yr, see Figure 3). The results from the 13-station dataset are similar but with slightly larger uncertainties. The event was therefore more extreme as a two-day event than as a one-day event. Given that the Harvey analysis had shown that multiple-day precipitation caused higher floods, we take as event definition the highest two-day station precipitation.

The regional return time for such a high amount at any of the 85 stations with at least 30 years of data and 0.1º apart that we analysed on the U.S. Gulf Coast 27.5–31 ºN, 85–97.5 ºW). This is much higher as the Gulf Coast is much larger than the area with extreme precipitation, so there are many possible locations. We obtain a return period of 50 yr (20 to 200 yr) for this amount of two-day precipitation in any of these stations on the U.S. Gulf Coast. This appears to be somewhat at odds with the recent observations of three similar flooding events in four years (2016, 2017, and 2019). It requires further analysis, beyond what is possible in this rapid attribution study, to find out whether or not the recent recurrence is more than an unfortunate coincidence.

The fit to the observations shown in Figure 3 also gives the results that such an event has become more common by a factor of roughly 2.6 (1.6 to 5.0) since around 1900, or equivalently that the intensity (i.e. total amount of precipitation) increased by about 18% (11% to 28%).

Modelled effects of climate change

We repeated the same analysis for a 6 member ensemble of 5-yr time slices of the SST-driven climate model EC-Earth 2.3 at T799 (~30km) horizontal resolution (Hazeleger et al., 2012), which was found to describe the three-day extremes we studied in our Harvey analysis well. Our usual procedure is to only use climate model output for the present and past to study the dependence of the event on climate change up to now. However, for this model we do not have enough data to extract that information from the past and present data only. We therefore chose to also use the time slice around 2090. This assumes that the effect of climate change is the same in the past as up to this time, when we describe it as a function of historical and scenario equivalent CO2 concentration. The strong heating in the last time slice gives rise to more intense events, as can be seen in Figure 4.

We defined the event to have the same regional return time as in the observations, 50 yr. The fit gives an increase in probability between 1900 and 2019 of 3.1 (1.9 to 12.3) and a change in intensity of 23% (13% to 31%).

We also analysed the other model of the Louisiana and Harvey analysis that simulated three-day extremes reasonably well, HiFLOR (25km, Murakami et al., 2016). We consider again the spatial maximum, now over the land points in 29–31ºN, 85–95ºW, of the annual maximum of two-day averaged precipitation, Figure 5.

This model gives an increase in probability of a factor 1.6 (1.3 to 2.0), which is equivalent to an increase in intensity of 8% (4% to 12%).

Hazard synthesis

The synthesis plots are shown in Figure 6. For the two observational estimates the natural variability is highly correlated, so the bar labeled ‘observations’ just averages the lower bounds, best fit values and upper bounds. The models are further apart than can be explained by natural variability (red bars). The white extensions denote the model spread necessary to bring the χ²/dof to one (van Oldenborgh et al, in preparation). This has been added to the natural variability to give the ‘models’ summary bar. Finally the observations and models have been averaged using weights (coloured bar) and with equal weights (white box). These almost coincide in this case.

This synthesis shows that the probability of two-day rainfall as intense as the maximum observed during Imelda has increased by a factor 1.5 to 3.1. This is equivalent to an increase in intensity of 9% to 21%.

Vulnerability and exposure

Two years after Harvey, Tropical Storm Imelda again served to highlight the vulnerability of Southeast Texas to severe rainfall. Widespread flooding was reported in Northeast Houston and along the I-10 corridor between Houston and the Golden Triangle (Beaumont-Port Arthur-Orange, largely coinciding with the areas of extreme precipitation of Figures 1 and 2). Several major rivers and bayous in the region exceeded flood stage, including along parts of the San Jacinto, Lower Trinity and Neches Rivers and their tributaries. Flooding also impacted numerous heavily-trafficked roadways catching daytime drivers by surprise. The Harris County Fire Marshal’s Office reported that more than 2,000 water rescues were performed by emergency responders in Harris County, 450 of which involved high water rescues.

In response, Texas’ Governor, Greg Abbott, declared a State of Disaster to provide unlimited state resources to the counties impacted by Imelda. The 13-county disaster declaration area has a population of approximately 6.6 million people. By 21 September, 2,400 flood insurance claims had been filed with the NFIP and the Insurance Council of Texas reports that more than 10,000 vehicles were damaged (80% are estimated to be totaled). It is likely that damages from the event will reach into the billions of dollars. As of 23 September, it was estimated that five deaths had occurred due to Imelda.

Since Harvey, significant steps have been taken to reduce the vulnerability of southeast Texas to coastal flooding. First, several new state laws went into effect on 1 September 2019 including a hazard disclosure law which expands requirements for sellers to disclose flood risk to include past flooding, and whether the home is located in a 500-year floodplain, a flood pool, or near a reservoir. Many homeowners are unaware of their risk and the FEMA flood maps are out of date and do not account for changes in precipitation, or upstream development. (FEMA is set to update its methodology for determining flood risks, which previously only looked at historical information, by 2020.)

Second, the state legislature created several funding mechanisms to facilitate the construction of flood control infrastructure including the Flood Infrastructure Fund and the Texas Infrastructure Resiliency Fund (TIRF). The Infrastructure Fund was created using nearly $1.7 billion from the state’s savings account, colloquially known as the “rainy day fund,” to be used to build flood control infrastructure in the short term. The Texas Infrastructure Resiliency Fund (TIRF), totaling $1.6 billion, will provide cities, counties and other political subdivisions the opportunity to apply for grants and low- or zero-interest loans that can then be used as the “match” for federally funded projects.

The Texas legislature also passed two bills aimed at improving planning and emergency response statewide: SB6 increases training for emergency managers in the state and provide guidance for disaster response and recovery and SB8 creates a statewide flood plan which will be published  every five years and consist of a consolidated list of flood control infrastructure projects and community initiatives.

Finally, the National Oceanographic and Atmospheric Administration (NOAA) recently updated the intensity-duration-frequency (IDF) curves used for infrastructure design and planning including to delineate flood risks and manage development in floodplains in relation to the National Flood Risk Insurance Program. This update has resulted in an increase in return period rainfall estimates. For example, in Houston, rainfall totals events that were previously classified as 1-100 year events are now 1-25 year events (ibid). (This relative increase is in line with our findings in the attribution of Harvey’s rainfall to climate change, Van Oldenborgh et al, 2017.)

Conclusion

Analyses of the flooding events of the last few years (Van der Wiel et al, 2017; Van Oldenborgh et al, 2017; Risser and Wehner, 2017 and others) have shown that the probability of extreme rainfall on the U.S. Gulf Coast has clearly increased, and climate models indicate that this increase can be connected to changes in climate. Coupled with rapid urbanization characterized by a loss in impervious cover, climate change has resulted in more frequent and intense flooding (Zhang et al. 2018). In this rapid attribution analysis we find the two-day extreme precipitation of Tropical Storm Imelda was also increased substantially by anthropogenic global warming.

Since Harvey (August 2017), significant steps have been taken to reduce risk and increase resilience to floods in southeast Texas, for example by increasing public awareness of flood risks, improving disaster preparedness, response, and recovery, and providing a mechanism for planning and funding for large-scale infrastructure projects. However, as demonstrated by Imelda, extreme weather events will continue to increase in severity as long as climate warming from increasing concentrations of greenhouse gases continues. Further research is needed to evaluate the potential role of climate change in having three of the past four years exhibit events with such high observed rates of extreme precipitation and resulting flooding.

References

Hazeleger et al, 2012. EC-Earth V2.2: description and validation of a new seamless earth system prediction model. Climate Dynamics. 39: 2611-2629, doi:10.1007/s00382-011-1228-5.

Murakami et al, 2016. Seasonal Forecasts of Major Hurricanes and Landfalling Tropical Cyclones using a High-Resolution GFDL Coupled Climate Model. Journal of Climate, 29: 7977–7989, doi:10.1175/JCLI-D-16-0233.1.

M.D. Risser and M.F. Wehner, 2017. Attributable Human-Induced Changes in the Likelihood and Magnitude of the Observed Extreme Precipitation during Hurricane Harvey. Geophysical Research Letters, 44: 12,457–12,464, doi:10.1002/2017GL075888.

G.J. van Oldenborgh et al, 2017. Attribution of extreme rainfall from Hurricane Harvey, August 2017. Environmental Research Letters, 12: 124009, doi:10.1088/1748-9326/aa9ef2

van der Wiel et al, 2017. Rapid attribution of the August 2016 flood-inducing extreme precipitation in south Louisiana to climate change. Hydrology and Earth System Sciences, 21: 897–921, doi:10.5194/hess-21-897-2017

Zhang et al, 2018. Urbanization exacerbated the rainfall and flooding caused by hurricane Harvey in Houston. Nature, 563(7731): 384. doi:10.1038/s41586-018-0676-z

Key findings

• Several major storms pounded Western Europe in January 2018, generating large damages and casualties. The two most impactful ones, Eleanor and Friederike, are analyzed here in the context of climate change.
• Strong winds associated with such storms are observed to have become less frequent in the past four decades.
• Models show a different signal, with no significant change in their frequency until now and a slight increase in the future.
• By analysing a number of climate simulations, we conclude that human-induced climate change has had so far no significant influence on storms like the two studied. However, all simulations indicate that global warming could lead to a marginal increase (0-20%) of the probability of extreme hourly winds until the middle of the century. These trends do not account for the other factors, such as roughness, aerosols, and decadal variability, that contributed to the observed reduction in probability.

Introduction

Storm Friederike led to at least eleven casualties and caused major disruption in the Netherlands and parts of Germany. In advance of Storm Friederike, warnings were issued in both the Netherlands and Germany for severe wind gusts. On January 17 in the Netherlands, a code yellow warning was issued and, subsequently, raised to a code orange. By the morning of January 18, when the worst winds were experienced, a code red was issued for a number of provinces at 09:16 CET. The timing of the strongest winds was around 9:00–11:00 a.m., just after the peak of the morning commute, with many people already on the road and, in some cases, caught unaware by the strong winds. In Germany, the German Meteorological Office (DWD) also issued warnings, asking people to stay indoors, and many schools were closed. In addition to the wind hazard, snow created icy road conditions, and eight people were killed by falling trees or in car accidents caused by dangerous road conditions. Storm Friederike is estimated to have caused around €900 million worth of damage to residential houses and office buildings, and a further €100 million in damage to cars, according to Germany’s Federation of Private Insurers, the GDV. They estimate this was the second most expensive storm to strike Germany in the past 20 years.

In the Netherlands, three people were killed during the storm. For the first time in history, train traffic was completely shut down across the country. Amsterdam Airport Schiphol was closed and more than 300 flights were canceled. Numerous roads were blocked by fallen trees and overturned trucks. Due to their height, trucks were susceptible to being blown off the roads, which caused disruptions and accidents.

The other major storm, Storm Eleanor, led to major disruptions in France during the ski holiday season and is estimated to have cost the insurance and reinsurance industry as much as EUR 1.6 billion. Ski resorts were closed for one or two days in the Alps, with significant economic consequences. Wind gusts of more than 130 km/hr and nearing 150 km/h were reported over several flat regions in France and in Switzerland. Large waves at the Atlantic coasts of Spain and France killed two people. Storm Eleanor was well forecasted by Météo-France, with an orange alert set the day before the event. Over France, according to the severity index developed by Météo-France, Eléanor was the sixth most severe storm since 1995.

More storms than these two were reported during the month. For instance, Storm Carmen, which preceded Storm Eleanor by two days, crossed Southern France with wind gusts exceeding 130 km/h. On January 17, another storm, Fionn, passed over parts of the Mediterranean region and broke wind speed records, including at Cap Corse at the northern tip of Corsica (225 km/h). From a monthly view and in terms of number of events, January 2018 is the most stormy month in France since 1998.

The storm activity was due to a strong westerly flow that persisted throughout the month (as shown in Figure 2, first row) and was enhanced by the jet stream extension eastward of its normal position. The persistence of the flow is also characterized by the frequency of occurrence of the so-called “zonal weather regime” (ZO), as defined by Michelangeli et al. (1995) using cluster analysis on SLP data from the NCAR/NCEP reanalysis. Approximately 45% of the January days were classified in this cluster (Figure 2, remaining panels), which is characterised by mild and wet winter weather. The average frequency of the ZO weather regime is close to 25%. Although not exceptional, this high frequency is significantly higher than normal.

Event definitions

Storm Friederike was the result of rapidly developing cyclogenesis and the area with highest wind speeds, located south of the trough center, moved fast from west to east. It crossed the Netherlands and mid-Germany in about half a day. In this analysis, the salient event characteristics will be represented by an indicator defined on the basis of daily maximum wind speed, derived from observations available from the Integrated Surface Database (“Lite” version, ISD-Lite). The database contains hourly global weather data for eight variables. Many of these observations are made at airports. However, many stations only contain three hourly data for the earlier part of the record. Also, when analyzing model output from some of the models contributing to EURO-CORDEX, the daily maximum near-surface wind speed was obtained on the basis of three-hourly wind speeds. For these reasons, we only sampled observations every three hours and the daily maximum wind speed was calculated only if at least four of the eight sampled observations were available.

In Figure 3a, we plot the values of the daily maximum wind observed over Northwestern Europe on the days of the storms. The track of Storm Friederike can be seen in the box [2-15E; 50-53N] where wind speeds are largest. We, therefore, selected as the event indicator the seasonal maximum value of this land area average of daily maximum wind speed (see also Figure 2).

This area contains 68 stations observing wind speed. The area average cannot be exactly calculated using the stations because the distribution of the stations is not even or dense enough, but we take the station average as a reasonable approximation. Using this indicator, Storm Friederike is the eleventh strongest storm in the area since 1 January 1976, with an indicator value of 16.0 ms-1 max daily wind. The 2017-2018 winter season (DJF) becomes the seventh in terms of strongest winter winds over this station set (some seasons had multiple stronger storms). We also considered the daily mean wind for models that did not store higher-frequency data. In terms of that indicator, Friederike was not remarkable with 8.7 m/s-1, as it was a very short duration storm.

For models, the area average is calculated over land grid points, which slightly lowers the indicator value (see comparisons in Table 1 for model evaluation). In order to calculate seasonal return periods, we take the maximum value of the indicator over the winter season (DJF).

The structure of Storm Eleanor was very different. Eleanor was embedded in a large-scale, deep low pressure system. Its strong winds affected a much broader area than Storm Friederike: from Ireland and the U.K. via western France, to Switzerland and the Riviera coast. Its high wind speeds, unusual in the inland part of Western Europe, constituted its most striking aspect. We, therefore, construct the same indicators as for Friederike, which are daily maximum and mean of wind speed, but averaged over a much wider area, from 0 to 10°E and 42°N to 52°N (see Figure 1b and 3b). The value of the indicator is 12.3 m/s for maximum winds and 8.3 m/s for daily mean winds.

Observations, model ensembles and evaluation

For the observational part of the attribution analysis, we used two sources of station data. Unfortunately, the available quantities were slightly different in the different datasets. The analysis is mainly based on the ISD-lite database described above, in which we used the daily maximum of three-hourly instantaneous wind speed. Additional results are based on the Royal Netherlands Meteorological Institute (KNMI) climatological service database, which provides the daily maximum of the hourly averaged wind speed. The KNMI data have also been converted to potential winds, i.e., the wind speed at 10 m assuming a roughness length of 3 cm over land and 2 mm over water, and assuming neutral stability (Weber and Groen, 2009). This corrects to first order for changes in the elevation of the wind anemometer and changes in roughness surrounding the station, which are deduced from the high-frequency variability of the wind (taking into account the response time of the recorder, Weber and Groen 2009). The highest hourly wind of the year series were visually quality controlled. For three series, early data was discarded for obvious inhomogeneities supported by the metadata (Leeuwarden <1990, De Bilt <2002, Lichteiland Goeree <1995). Most series start in 1981, but they are notably more variable and possibly unreliable before circa 1990.

We used four climate model simulation ensembles. The first ensemble is the RACMO regional climate model ensemble downscaling 16 initial-condition realizations of the EC-EARTH 2.3 coupled climate model in the CMIP5 RCP8.5 scenario (Lenderink et al., 2014, Aalbers et al., 2017). The RACMO model uses a 0.11° (12 km) resolution and the daily maximum of near-surface wind speed is analysed. In RACMO, the near-surface wind speed is diagnosed from the model wind and stability vertical profile as the wind speed at 10 m, applying a roughness length of at most 3 cm for land grid points, and a Charnock-type relation for sea grid points. This ensemble was previously used to estimate the change in the odds of wind stagnations in Northwestern Europe (Vautard et al., 2017) and was found to simulate monthly wintertime wind speeds over Western Europe in a satisfactory manner. RACMO simulations are available for the 1950-2100 period. As in previous analyses (see e.g., Philip et al., 2018), we use a 20th century early 30-year period [1951-1980] to estimate odds in the past climate, and the 2001-2030 period to estimate odds in the current climate. We also use two future periods, a period called “near future” [2021-2050] and a period called “mid century” [1941-1970]. As a cross-check, we fitted a time-dependent generalized extreme value (GEV) function to the whole period 1971-2070, as described in van der Wiel et al (2018).

The second model ensemble is the HadGEM3A ensemble (Christidis et al., 2013; Vautard et al., 2018), which includes a set of 15 realizations of atmospheric simulations using observed SSTs (reflecting the actual world) and a set using SSTs where the estimated patterns of anthropogenic heat contribution are removed to reflect the ocean response to a pre-industrial atmospheric composition (as the natural/counterfactual world). The latter runs also use pre-industrial greenhouse gas and aerosol concentrations. Land use and, hence, roughness is put to 1850 values in the HistoricalNat ensemble. For this model, the wind speed daily maximum was not available and the daily mean wind was used instead. No future simulations were available.

The third ensemble is the multi-model EURO-CORDEX ensemble (Jacob et al., 2014), using a 0.11° resolution over Europe. For this ensemble, only 11 simulations were used and bias correction was applied (Vautard et al., in preparation) using the Cumulative Distribution Function transform (CDFt, Vrac et al., 2016). These simulations have been evaluated in the context of the CLIM4ENERGY Copernicus Climate Change Service project. The reference data used for bias correction is the Watch Forcing Data ERA-Interim (WFDEI, Weedon et al., 2014). For wind speed, it is essentially an interpolation of ERA-Interim over a 0.5°×0.5° grid. This data set has a relatively low resolution, so extreme winds are not expected to be accurately represented. This weakness is, therefore, probably propagated to the EURO-CORDEX ensemble. The ensemble is pooled, which is formally possible because the bias correction method corrects data making it homogeneous across the multi-model distribution. However, caution must be taken when interpreting changes using such a pooled ensemble, as changes in the tails of the distribution may be different for each model, leading to potential heterogeneity in extremes for periods different than the reference period.

The fourth ensemble is obtained from simulations using the distributed computing framework known as weather@home (Massey et al. 2015). We used four different large ensembles of December-February wind speeds using the Met Office Hadley Centre for Climate Science and Services regional climate model HadRM3P at 25 km resolution over Europe embedded in the atmosphere-only global circulation model HadAM3P at N96 resolution. The first set of ensembles represents possible winter weather under current climate conditions. This ensemble is called the “all forcings” scenario and includes human-caused climate change. The second set of ensembles represents possible winter weather in a world as it might have been without anthropogenic climate drivers, using different estimates of pre-industrial SST deduced from the CMIP5 ensemble and pre-industrial greenhouse gas and aerosol concentrations. Land-use in both ensembles is identical. This ensemble is called the “natural” or “counterfactual” scenario (Schaller et al., 2016). The third set of ensembles represents a future scenario in which the global mean surface temperature is 1.5°C higher than pre-industrial global temperatures. The fourth scenario is the same as the third, but for 2°C of future global mean temperature. To simulate the third and fourth scenario, we use atmospheric forcings derived from RCP2.6 and 4.5 and sea surface temperatures that match the atmospheric forcing obtained from CMIP5 simulations (Mitchell et al., 2017).

The evaluation of the models’ ability to simulate the indicator is made using the ISD-Lite observations, which are available in near-real time. In order to evaluate the capacity of the models to simulate the winds, we extracted wind speed daily maxima at the locations of ISD-Lite stations and averaged these values over all stations in the area. Then, we compared the simulated mean, 95th centile and 99th centile, with the observed equivalent for each model ensemble. For HadGEM3-A and weather@home, as daily maxima were not available, we used daily averages of the wind speeds.

For RACMO, HadGEM3-A and weather@home, model values are pooled together to compute the distribution statistics. For EURO-CORDEX, we calculated both individual model and pooled statistics. Results are presented in Table 1 for the average over all grid points closest to the 68 ISD-lite stations, together with equivalent statistics when the average is made over all land grid points, instead of the positions of the stations. Results show that the models reproduce the indicator with success along the distribution. Comparisons to station data indicate a general underestimation of models within a 10% range. EURO-CORDEX simulations are bias corrected, so the bias is essentially reflecting the WFDEI (ERA-Interim based) bias. The fact that statistics do not differ from one model to the other supports pooling the models’ simulations together in a common distribution. This bias is consistent with models not simulating observational noise due to remaining turbulence. For weather@home, we only have daily values for mean wind speed, so we calculate the maximum mean daily wind speed in a winter season. The simulated values are higher than the observed values for this quantity, especially for the mean, while the 95th and 99th quantile are comparable to observations in particular for the Storm Friederike.

Grid point averages reach lower values than station averages, which is a probable consequence of the higher density of stations near the North Sea coast where winds are stronger which is reflected in the observed area average.

The factor between observation statistics and model statistics for station averages is rather uniform across the distributions, even though observations seem to be heavier tailed than simulations. In order to homogenize attribution results among models and observations and compare return periods with observations, we scaled all simulations by the ratio between 99th centiles of observed station averages and simulated grid-point averages. These bias corrections are a factor 1.13 for RACMO (for both storms), 1.17 (resp. 1.28) for EURO-CORDEX for Friederike (resp. Eleanor), and 1.12 (resp. 1.22) for HadGEM3-A for Friederike (resp. Eleanor).

Storm Friederike

Observations

The winter maximum of the daily maximum of three-hourly maximum wind over the ISD-lite spatial average (over all stations in the box) is shown in Fig. 4a as a function of time (labeled with the year of the second half of winter). The data has been fitted by a GEV distribution in which the position parameter μ and scale parameter σ scale exponentially with time, such that their ratio remains constant. The shape parameter ξ is not time-dependent. We checked whether these assumptions are valid in the models, which have enough data not to need these assumptions. This fit shows a significant decrease (p<0.05 two-sided) in wind speed over 1976–2017, in agreement with earlier analyses (Smits et al, 2005, Vautard et al, 2010). The decrease in intensity of about 12% (95% CI 0–30%) corresponds to a decrease in probability of about a factor of four (95% CI: 1 to 100). Using the global mean temperature as a covariate instead of time gives slightly higher trends. The shape parameter ξ of the GEV is most likely negative, so the distribution has a tail that is thinner than a logarithmic function, so that the ratio of probabilities is more different from one for stronger storms.

The return period of an event like Friederike or worse in the area in which the indicator value reached 16.0 m/s on January 18 in the current climate is of the order of 20 years. In the 1970s, this was roughly five years, so the event has become a fairly rare event due to the decrease in high wind speeds observed during this period.

The result is confirmed in a different dataset from KNMI observations, with most stations showing a clear downward trend over the whole period (1971–2017 for two stations, 1982–2017 for most others, at least 30 years with data) of −15% [−7% to −17%], the same as the ISD-lite data show.. The trends are much less clear when starting in 1990 (using stations with at least 25 years of data). The trends in potential wind are much smaller (around −5%) and not statistically significant, even when pooled over all stations.

RACMO ensemble

The RACMO ensemble is now pooled and the indicator is calculated from the simulations. The indicator is scaled to have the same 99th centile as the observed indicator in the historical period. Indicator statistics are then obtained for three climate periods: 1951-1980, simulating the “past” period, the “current” period taken as 2001-2030, and two future periods assuming the RCP8.5 scenario (2021-2050 and 2041-2060). The observed indicator value for storm Friederike (16.0 m/s) has a present return period of about 13 years (95% CI 10-19 years), which is longer than for the observations. The probability of witnessing higher indicator values is not significantly different in past and current periods (Figure 5). However, the change of probability becomes larger in future periods, with a risk ratio (RR) of about 1.5 [1-2] in the near term (Figure 5b). For this particular case, the increase is also stronger for stronger storms due to an increase in the variability relative to the mean.

Therefore, according to this model’s representation, we do not identify a climate change impact currently, but the increase in probability of storms like Friederike emerges in the coming decades. A fit with a GEV that scales with the global mean temperature of the driving EC-Earth ensemble gives no change (RR between 0.95 and 1.16), overlapping with the 30-year time window analysis (not shown), but this does not include the increase of variability relative to the mean.

HadGEM3A ensemble

The HadGEM3-A ensemble exhibits a significant difference between actual and counterfactual periods, with a current increase of strong daily mean winds in the area struck by Storm Friederike. However, due to the use of the mean wind speed instead of the maximum wind speed, the indicator does not disentangle extreme winds over a short time period from less strong winds over an extended time period. Accordingly, the observed value is not exceptional, due to the fast travelling nature of the extremely high winds in the area: for the value corresponding to Friederike (8.7 m/s), such events occur almost every year in both types of simulations. Note that difference in counterfactual and actual ensembles captures some changes in roughness, but probably not all.

EURO-CORDEX ensemble

In the EURO-CORDEX simulations, the return period corresponding to the scaled indicator (25-40 years) is larger, making it a more extreme event. The shape of the distribution is clearly different from that of the RACMO simulations and that of the observations (compare with Figs. 4 and 5). The RR is generally not significantly different from one (Figure 7b), despite a rather systematic increase. Such increase becomes marginally significant in the middle of the century with RR values in the range 1 to 3 for lower wind thresholds. Again, this indicates a tendency for more storms like Friederike in the future with a signal emergence not yet achieved. The GEV with smoothed EC-Earth global mean temperature as covariate confirms this conclusion, with an increase in probability of 1.0 to 1.2 (p~0.1), this corresponds to the assumption the percentage increase is a constant 0.0 to 1.4% per degree global warming over the whole range of Fig. 7a. This assumption holds well over the four 30-year time periods considered before.

Weather@Home

For weather@home, using the suboptimal definition of maximum of daily mean wind to define Storm Friederike (8.7 m/s), we find no significant change in the likelihood of storms like Friederike to occur (Figure 8). In contrast to the EURO-CORDEX assessment this also holds for rarer events (not shown).

Table 2: Event return periods and risk ratios summarized for all model ensembles and for Storm Friederike. Risk ratios (RRs) are calculated with respect to a past or counterfactual period.

Storm Eleanor

Observations

The same analysis on the Eleanor index as in section 4.1 gives a more significant downward trend for this storm (p<0.01 two-sided), with a decrease of about 20% (3–35%) (Figure 9). This corresponds to an increase in return period of a factor 8 (1.5–100). The return time also is about 20 years in the current climate according to this fit.

RACMO ensemble

Storm Eleanor is now investigated through the wind daily maximum indicator with an average over the large region as defined above. Due to the southern boundary that is excluding a small band of the large region, we used for this model a boundary at 43.5N instead of 42N. This makes the indicator return value for stations slightly lower than when calculated over the full region (11.9 m/s instead of 12.3 m/s). The corresponding RACMO return period is in the range 3 to 5 years. The climate change is not significant for the current period and marginally significant for future periods, as for Storm Friederike (Figure 10). The estimated RR is 1.1 and slightly higher for future periods. Interestingly, for stronger storms, the RR increases. The same results hold for a GEV fit with covariate of all data in 1971-2070, with a RR significantly different from one (95% CI 1.0 to 1.2, not shown).

HadGEM3A ensemble

For HadGEM3-A, using the daily mean wind, we find no climate change signal in the estimation of the probabilities of high winds of any magnitude, but for the very extreme winds, we find marginally significant changes in the direction of more frequent high winds under current conditions than under natural conditions (Figure 11). The estimated return period for the indicator value corresponding to Eleanor, which does not fall in the extreme tail, lies also between three and five years. A risk ratio close to one is estimated.

EURO-CORDEX ensemble

Using the EURO-CORDEX ensemble, the return period of the large-scale Storm Eleanor, characterized by the chosen indicator, is estimated to about 7-10 years. A climate change signal is absent in the simulations when comparing 1971-2000 and 2001-2030 periods. For the indicator value, the RR is in the range [0.5-1]. Only for later periods and for larger indicator values, a marginally significant increase in the RR in the range [1-2] can be seen (Figure 12). A GEV with a modelled global mean temperature (from EC-Earth) as covariate also gives a non-significant increase with a RR between 0.99 and 1.15 (95% CI).

Weather@Home

For weather@home, using the maximum of daily mean wind to define Storm Eleanor (8.3 m/s), we find no significant change in the likelihood of storms like Eleanor to occur. In contrast to the EURO-CORDEX assessment, this also holds for rarer events where the weather@home model shows a non-significant decrease in high wind speeds.

Table 3: Event return periods and risk ratios summarized for all model ensembles and for Storm Eleanor.

Synthesis and conclusions

Western European countries have been struck by high-impact wind storms during the month of January 2018. The link between storms like Eleanor (on 3/1/2018) and Friederike (on 18/1/2018) and climate change have been studied through this attribution analysis involving several simulation ensembles and observations from tens of weather stations.

From an analysis of two sets of observations, we conclude that near-surface storms in the areas of the two storms have a decreasing trend in wind speed and, hence, frequency trend over the past 40 years (see Figure 13), consistent with previous observation-based studies on storminess in these areas (Smits et al., 2005; Soubeyroux et al., 2017) and with global land wind stilling (Vautard et al., 2010; McVicar et al., 2012). This trend was shown to be close to zero over the Netherlands area when using the potential wind, showing a strong influence of roughness changes there (see also Wever, 2012). Other processes, such as aerosols increase, could also induce a wind decrease (Bichet et al., 2012), and decadal-scale long-term variability has been shown to have a significant role as well (e.g., Matulla et al, 2008).

We next turn to the model results. Due to the differing experiments that we used, the risk ratios have been computed over different intervals. To compare those we need to convert them to a common interval. We do this by assuming the risk ratio is an exponential function of some indicator of global warming f(yr):

RR(y₁,y_end) = RR(y₂,yr_end) RR(yr₁,yr₂) = RR(yr₂,yr_end) exp[f(yr₁)-f(yr₂)]

In the following, we use the RCP4.5 CO₂ concentration for f(yr), as the global mean temperature has no observations in the future and the projections depend on the model.

In contrast to the observations, global and regional climate models do not simulate such a decrease over the past decades. Instead, simulations of the daily maximum of 3-hourly instantaneous wind, of the same spatial and temporal characteristics of these storms and, hence, the observational analysis, indicate increases in probability between 1975 and 2055, corresponding to increases in wind speed for this return time. These are not all significantly different from one, but model consensus and future trends support the presence of such positive tendency. The change is small though: a risk ratio of 1.5 for Friederike with an uncertainty range of 1 to 2, corresponding with an increase in intensity of the wind of only about 5% (0% to 10%). For Eleanor, the numbers are even smaller: an increase in RR of about 1.25 (1.0 to 1.6) or an increase in intensity of 2% (0% to 5%).

The changes in daily mean wind are smaller still and indistinguishable from no change. However, as these do not correspond directly to the impact of these storms, we do not take them into account in the synthesis.

The climate model simulations do not always include changes in aerosols and either have no roughness changes (regional models) or capture these only partially (HadGEM3-A). This explains at least partially the conflict with the observed trends, as the potential wind results for the Netherlands showed that roughness plays a large role in the observed decrease in storminess. By contrast, these model ensembles mainly reflect changes due to greenhouse gases.

We conclude that storms like Friederike and Eleanor have not become significantly more or less frequent due to climate change, but our model results indicate that global warming due to greenhouse gases could make storms like them somewhat more frequent in the future, with a frequency increase up to at most a factor of two, or equivalently a few percent higher wind speeds. However, the observations show a clear decline in high wind speeds, in accordance with earlier studies. This is equivalent to declining probabilities of these kind of storms, but our analysis and previous studies find explanations for these changes in factors other than greenhouse gases. The increase in roughness potentially explains a major part of this decrease, and does not exclude other factors, such as decadal variability and aerosol effects. Until a quantitative attribution of past observed decreases is established, and with that an understanding of the interplay between greenhouse gas forcing and those other factors, and scenarios for them, the confidence on future evolutions of wind storms will remain low, based on simulations reflecting mainly the effects of greenhouse gas increases.

Acknowledgements

This work was supported by the EUPHEME project, which is part of ERA4CS, an ERA-NET initiated by JPI Climate and co-funded by the European Union (Grant #690462). It was also supported by the French Ministry for an Ecological and Solidary Transition through national convention on climate services. We would like to thank all volunteers who have donated their computing time to weather@home.

References

Aalbers, E.E., G. Lenderink, E. van Meijgaard, B.J.J.M. van den Hurk (2017). Local-scale changes in mean and heavy precipitation in Western Europe: climate change or internal variability? Climate Dynamics, 50 (11–12): 4745–4766. doi: 10.1007/s00382-017-3901-9

Bichet, A., Wild, M., Folini, D., & Schär, C. (2012). Causes for decadal variations of wind speed over land: Sensitivity studies with a global climate model. Geophysical Research Letters, 39(11). doi: 10.1029/2012GL051685

Christidis, N., Stott, P. A., Scaife, A. A., Arribas, A., Jones, G. S., Copsey, D., … & Tennant, W. J. (2013). A new HadGEM3-A-based system for attribution of weather-and climate-related extreme events. Journal of Climate, 26(9): 2756-2783. doi: 10.1175/JCLI-D-12-00169.1

Lenderink, G., B.J.J.M. van den Hurk, A.M.G. Klein Tank, G.J. van Oldenborgh, E. van Meijgaard, H. de Vries and J.J. Beersma (2014). Preparing local climate change scenarios for the Netherlands using resampling of climate model output. Environmental Research Letters, 9(11): 115008. doi: 10.1088/1748-9326/9/11/115008

Matulla, C., Schöner, W., Alexandersson, H., von Storch, H. (2008). European storminess: late nineteenth century to present. Climate Dynamics, 31: 125–130. doi: 10.1007/s00382-007-0333-y

McVicar, T. R., Roderick, M. L., Donohue, R. J., Li, L. T., Van Niel, T. G., Thomas, A., Grieser, J., Jhajharia, D., Himri, Y., Mahowald, N.M., Mescherskaya, A.V.,Kruger, A.C., Rehman, S. and Dinpashoh, Y. (2012). Global review and synthesis of trends in observed terrestrial near-surface wind speeds: Implications for evaporation. Journal of Hydrology, 416: 182-205.  doi: 10.1016/j.jhydrol.2011.10.024

Michelangeli, P.A., Vautard, R., Legras, B. (1995). Weather regimes: recurrence and quasi-stationarity. Journal of Atmospheric Science, 52: 1237-1256. doi: 10.1175/1520-0469(1995)052<1237:WRRAQS>2.0.CO;2

Philip, S., S. F. Kew, G. J. van Oldenborgh, E. Aalbers, R. Vautard, F. Otto, K. Haustein, F. Habets, R. Singh and H. Cullen (2018). Validation of a rapid attribution of the May/June 2016 flood-inducing precipitation in France to climate change. Climate Dynamics, in preparation. doi: 10.1175/JHM-D-18-0074.1

Schaller, N., A.L. Kay, R. Lamb, N.R. Massey, G.-J. van Oldenborgh, F.E.L. Otto, S.N. Sparrow, R. Vautard, P. Yiou, A. Bowery, S.M. Crooks, C. Huntingford, W. Ingram, R. Jones, T. Legg, J. Miller, J. Skeggs, D. Wallom, S. Wilson & M.R. Allen. (2015). Human influence on climate in the 2014 Southern England winter floods and their impacts. Nature climate change, 6, pages 627–634 (2016). doi:10.1038/nclimate2927

Smits, A. A. K. T., Klein Tank, A. M. G., & Können, G. P. (2005). Trends in storminess over the Netherlands, 1962–2002. International Journal of Climatology, 25(10), 1331-1344. doi:10.1002/joc.1195

Soubeyroux, J.-M., Dosnon, F., Richon, J., Schneider, M., and P. Lassègues, (2017). Caractérisation à haute résolution spatiale des tempêtes historiques en métropole application à la tempête Zeus du 6 mars 2017. La Climatologie, 14.

Vautard, R., J. Cattiaux, P. Yiou, J.-N Thepaut, and P. Ciais. (2010). Northern Hemisphere atmospheric stilling partly attributed to an increase in surface roughness. Nature Geoscience, 3: 756–761. doi: 10.1038/ngeo979

Vautard, R., A. Colette, E. van Meijgaard, F. Meleux, G. J. van Oldenborgh, F. Otto, I. Tobin, P. Yiou. (2017). Attribution of wintertime anticyclonic stagnation contributing to air pollution in Western Europe. Bulletin of American Meteorological Society, in Explaining extreme events of 2016 from a climate perspective, special supplement, S70-S75. doi: 10.1175/BAMS-D-17-0113.1

Vautard, R., N. Christidis, A. Ciavarella, C. Alvarez-Castro, O. Bellprat, B. Christiansen, I. Colfescu, T. Cowan, F. Doblas-Reyes, J. Eden, M. Hauser, G. Hegerl, N. Hempelmann, K. Klehmet, F. Lott, C. Nangini, R. Orth, S. Radanovics, S. I. Seneviratne, G. J. van Oldenborgh, P. Stott, S. Tett, L. Wilcox, P. Yiou (2017). Evaluation of the HadGEM3-A simulations in view of climate and weather event human influence attribution in Europe. Climate Dynamics, submitted. doi: 10.1007/s00382-018-4183-6

Vrac, M., Noël, T. and R. Vautard. (2016). Bias correction of precipitation through Singularity Stochastic Removal: Because occurrences matter. Journal of Geophysical Research, 121(10): 5237-5258. doi: 10.1002/2015JD024511

Weber, N. and G. Groen. (2009). Improving potential wind for extreme wind statistics, KNMI Scientific Report WR 2009-02.

Wever, N. (2012). Quantifying trends in surface roughness and the effect on surface wind speed observations. Journal of Geophysical Research: Atmospheres: 117(D11). doi: 10.1029/2011JD017118

The excessive nature of this record rainfall event, which led to flooding of more than 5,000 homes and businesses and left over 60,000 people without power, has led many to question whether climate change played a role, especially since there have been several large floods over the last decades.
Recent advances in the science of extreme event attribution now make it possible for scientists to, using peer-reviewed methods, rapidly provide an objective, quantitative initial estimate of the relative contribution of global warming to specific classes of extreme weather events. As a result, these analyses provide estimates of the return-time period of the event both today and in the past — before there was a strong human influence on the climate system. The ratio of these is a measure of the extent to which climate change affected the likelihood of the event. Overall, extreme event attribution can provide valuable information to decision-makers faced with tough questions about changing risks and to underpin adaptation strategies at a more local level.

To assess the potential link between the U.K.’s record rainfall and man-made greenhouse gases in the atmosphere, we conducted independent assessments using three peer-reviewed approaches. These approaches involve statistical analyses of the historical temperature record, the trend in a global coupled climate model, and the results of thousands of simulations of possible weather with a regional climate model. Applying multiple methods provides scientists with a means to assess confidence in the results.

Based on these three approaches – all of which are in agreement – the team found that global warming increased the likelihood of the heavy precipitation associated with a storm like Desmond. The increase is small but robust. It should be noted that a positive attribution for an extreme rainfall event like Desmond is still somewhat rare. Evidence of this can be found in a summary of the events analysed as part of the annual BAMS Special Issue on Explaining Extreme Events from a Climate Perspective (pdf, 5,4 MB). Whereas the vast majority of heat events studied found a climate change signal, less than half of the papers looking at extreme rainfall events found a human influence.

By comparing recent extreme events with the historical record and climate model simulations, the team found that an event like this is now roughly 40% more likely due to climate change than it was in the past, with an uncertainty range of 5% to 80%. It is important to note that this analysis only considers externally driven changes in precipitation. It does not take into account other factors that influenced the flooding. While our analysis provides evidence that climate change has aggravated the flood hazard in this part of the world, risk is also determined by trends in exposure and vulnerability. As events like this become more common in the U.K., it will be important to discuss both changing risks associated global warming and the overall adequacy of flood defences.

Data from NASA’s Integrated Multi-satellitE Retrievals for GPM (IMERG) were used to estimate rainfall for the period from November 30 to December 7, 2015. This analysis found that some rainfall near the Irish Sea measured over 392 mm (~15.4 inches) during this period. As much as 304 mm (~12 inches) of rain were reported to have fallen in only 24 hours.
 Credits: SSAI/NASA/Hal Pierce.

Scientific parameters

Variable:
Precipitation

Event definition:
24 hour maximum precipitation

Domain:
54º–57ºN, 6ºW–2ºE

Results were checked against local station data. Over the large area, the ECMWF analysis gives an average precipitation of 36.4 mm on 5 December (0–0 UTC)

Observational data:
ECMWF 24-hour precipitation forecast; ECA & D

Models used:
Global coupled model: EC-Earth 2.3 model (16 runs, 1861–2015) 100 km resolution

Regional model:
weather@home HadRM3P at 50km resolution over Europe, driven by HadAM3P & OSTIA SSTs

Results

Overall
5% to 80% more frequent

Method #1 (KNMI)
Observations: 30%–180%

Method #2 (KNMI)
Global Coupled Model: 10%-80%

Method #3 (Oxford)
Regional Large-ensemble: 5%-50%

This analysis is available from Hydrology and Earth System Sciences (HESS) (pdf, 806 KB).