Extreme and persistent heat has been overwhelming south western parts of the US, Mexico, and the northern countries of Central America. The area has been lying under a large and lingering region of high pressure, known as a heat dome, whereby hot air is trapped close to the ground and further heated under blue skies and sunshine. Whilst heat domes have a well known mechanism for intensifying heatwaves, these past weeks have seen records broken in both daytime and nighttime temperatures in several countries, including Mexico, Guatemala, Honduras and in the south western US.

Heatwaves are among the deadliest types of extreme events. Even though often the death toll is typically underreported especially during or straight after the event, during this hot season Mexico has already reported 125 deaths (SwissInfo). The coinciding ongoing drought is enhancing impacts of the heat even more.

Scientists from Mexico, Panama, the Netherlands, the United Kingdom, the United States, and Sweden collaborated to assess to what extent human-induced climate change altered the likelihood and intensity of the extreme heat in a region that included the US southwest, Mexico, Guatemala, Belize, Honduras and El Salvador. Using peer-reviewed methods, we analyse 5-day maximum daytime and nighttime temperatures in May and June over a large region (see Figure 1) encompassing the region where impacts associated with extreme temperature records were reported.

Main findings 

• The extreme heat in the north and central America has resulted in severe impacts, including more than 125 heat-related deaths in Mexico since March, thousands of cases of heat stroke, and power outages. We likely do not know the full picture of heat-related deaths, since they are usually only confirmed and reported months after the event, if at all. 
• Existing drought conditions have further aggravated the situation by preventing the dispersion of polluting particles, decreasing water availability, and reducing hydropower generation and electricity supply. 
• Observations show that 5-day maximum temperatures in May-June such as recorded this year are expected to occur about every 15 years in today’s climate that has been warmed by 1.2C. However, around the year 2000, when global temperatures were half a degree lower than now, such events were expected to occur only about once every 60 years.
• The night time temperatures over the same 5-day period were also high, but not extreme in today’s climate; there is now a 50% chance per year of similar temperatures occurring. At the turn of the millennium such events would only have been expected to occur with a 13% chance in any given year.
• These return times are estimated for the region as a whole. It is important to highlight that the heat was more rare in the southeastern part of the region, especially for the nighttime temperatures with return periods of up to 1000 years in individual locations. 
• To determine the role of climate change we combine observations with climate models and we conclude that human-induced warming from burning fossil fuels  made the 5-day maximum temperature event about 1.4 degrees hotter and about 35 times more likely. For nighttime temperatures this is about 1.6 degrees hotter and about 200 times more likely.
• These trends will continue with future warming and events like the one observed in 2024 will be very common in a 2C world. 
• Extreme heat warning systems and action plans can help fill important gaps in preparedness across Central America. Heat safety protection laws can be enacted and implemented to protect outdoor workers across all countries. 
• Strengthened grid resilience and water conservation strategies are critical to ensure reliable services during heat events. Improved urban planning, more green spaces, and enhanced infrastructure in informal settlements will also help protect the most vulnerable.

There are at least 17 direct fatalities linked to the fires, more than 150,000 people have been evacuated, and at least 200 structures, including homes, were damaged in the fires (AP News, 2023). The Canadian wildfires have severely impacted air quality locally in Canada, and in the neighbouring United States with Air Quality Index (AQI) values frequently exceeding safe levels in the midwest and northeast USA, and in some cases approaching record levels (e.g. on June 7th AQI reached 341 in New York City, considered hazardous for all residents) (CNBC, 2023). Similarly, in southern Ontario, including the cities of Ottawa and Toronto, air quality reached the “very high risk” level forcing officials to cancel public events and reduce hours for outdoor public services. Schools remained closed for several days in many states, including Nova Scotia, New York, New Jersey, and Connecticut. 

In this study we focus on the fires in eastern Canada, which experienced a particularly unusually active fire season in 2023 and are most directly linked with the very large-scale impacts on air quality. In order to identify the role of human-induced climate change we focus on fire-weather indices rather than on fire regime variables such as area burned, in order to identify whether and to what extent climate change altered fire-prone weather conditions.

To capture the extent and duration of the extreme fire weather across the region, we will use the cumulative daily severity rating (Figure 1). The DSR is a scaled power transformation of the Canadian Fire Weather Index (FWI), and reflects how difficult a fire is to suppress once ignition has occurred; it is commonly used for assessing fire weather on monthly or longer timescales (Van Wagner,, 1987). To capture the peak intensity of the fire season, we also take the annual maximum of the 7-day moving average of the FWI. This index has been used in previous attribution studies (eg. van Oldenborgh et al., 2021), where it has been found to have a good correlation with the area burned. 

Main messages

• Fire weather is one important condition driving wildfires, although changes in vegetation (wildfire fuel), ignition factors, and fire management strategies also contribute to future wildfire risk. 
• In today’s climate, intense fire weather like that observed in May-July 2023 is a moderately extreme event, expected to occur once every 20-25 years. This means in any given year such an event is expected with 4-5% probability. 
• Climate change made the cumulative severity of Québec’s 2023 fire season to the end of July around 50% more intense, and seasons of this severity at least seven times more likely to occur. Peak fire weather (FWI7x) like that experienced this year is at least twice as likely, and the intensity has increased by about 20% due to human-induced climate change. 
• Observed changes are typically larger than in the models. 
• As expected, likelihood and intensity are projected to increase further in a 2°C warmer world. 
• Changes in fire weather are associated with an increase in temperature and decrease in humidity, both of which are driven by human-induced warming; the effect was compounded in 2023 by unusually low precipitation
• The extent, magnitude, and location of concomitant wildfires posed significant challenges for wildfire management which largely focused on disaster response and wildfire containment to limit the impact on lives and infrastructure. 
• The wildfires had disproportionate impacts on indigenous, fly-in, and other remote communities who were particularly vulnerable due to lack of services and barriers to response interventions. 
• The consequences from the wildfires reached far beyond the burned areas with displaced impacts due to air pollution threatening health, mobility, and economic activities of people across North America.
• As fire weather risks increase, changes in fire management strategies and increased resources will be required to meet the increased challenges. 

Several heat deaths have been confirmed in the US, including migrants on the US Mexican border. In Mexico alone over 200 people died due to the heat. Spain, Italy, Greece, Cyprus, Algeria, and China also reported heat deaths, as well as a large increase in hospitalisation due to heat related illnesses. Large parts of the population in Italy and Spain and over 100 million people in Southern US are under heat alerts. In all three regions, demand for power spiked and negatively impacted a number of important crops, including olive oil in Spain and cotton in China. 

Scientists from the World Weather Attribution initiative collaborated to assess to what extent human-induced climate change altered the likelihood and intensity of the extreme July heat in these three regions.

Using published peer-reviewed methods, we analysed how human-induced climate change altered the likelihood and intensity of 1) 18-day average maximum temperatures over the most affected regions in western US, Texas and northern Mexico (fig 1, top). 2) 7-day average maximum temperatures over land in the rectangular box(5W-25E, 36-45N) covering the most affected region (fig1, middle). 3) 14-day average maximum temperatures over the lowlands of China, again covering the most affected region (fig 1, bottom). 

Main findings

• Heatwaves are amongst the deadliest natural hazards with thousands of people dying from heat-related causes each year. However, the full impact of a heatwave is rarely known until weeks or months afterwards, once death certificates are collected, or scientists can analyse excess deaths. Many places lack good record-keeping of heat-related deaths, therefore currently available global mortality figures are likely an underestimate.
• In line with what has been expected from past climate projections and IPCC reports these events are not rare anymore today. North America, Europe and China have experienced heatwaves increasingly frequently over the last years as a result of warming caused by human activities, hence the current heat waves are not rare in today’s climate with an event like the currently expected approximately once every 15 years in the US/Mexico region, once every 10 years in Southern Europe, and once in 5 years for China. 
• Without human induced climate change these heat events would however have been extremely rare. In China it would have been about a 1 in 250 year event while maximum heat like in July 2023 would have been virtually impossible to occur in the US/Mexico region and Southern Europe if humans had not warmed the planet by burning fossil fuels. 
• In all the regions a heatwave of the same likelihood as the one observed today would have been significantly cooler in a world without climate change. Similar to previous studies we found that the heatwaves defined above are 2.5°C warmer in Southern Europe, 2°C warmer in North America and about 1°C in China in today’s climate than they would have been if it was not for human-induced climate change. 
• Unless the world rapidly stops burning fossil fuels, these events will become even more common and the world will experience heatwaves that are even hotter and longer-lasting. A heatwave like the recent ones would occur every 2-5 years in a world that is 2°C warmer than the preindustrial climate.
• Heat action plans are increasingly being implemented across all three regions and there is evidence that they lead to reduced heat-related mortality. Furthermore, cities that have urban planning for extreme heat tend to be cooler and reduce the urban heat island effect. There is an urgent need for an accelerated roll-out of heat action plans in light of increasing vulnerability driven by the intersecting trends of climate change, population ageing, and urbanisation. 

Scientists from Switzerland, India, the Netherlands, France, the United States of America and the United Kingdom, collaborated to assess to what extent human-induced climate change altered the likelihood and intensity of the low soil moisture, both at the surface and the root zones for most crops.

Main findings

• Heat and low rainfall in West-Central Europe had far reaching impacts on a variety of sectors including human health, energy, agriculture, and municipal water supply. It was exacerbated by e.g. poor water infrastructure and leakages, and it came at a time when food and energy prices were already high resulting in compounding social and economic impacts.
• In this study, we particularly focus on the dry soils which caused severe economic and ecological impacts across the Northern Hemisphere (excluding the tropical regions) and were particularly severe in West-Central Europe. We therefore focus on these two regions, North-Hemisphere extratropics and West-Central Europe, to analyse the agricultural and ecological drought from June to August 2022.
• Observation-driven land surface models show that very low summer surface and root-zone soil moisture, such as observed in 2022, happens about once in 20 years in today’s climate in both regions.
• While the magnitude of historical trends vary between different observation-based soil moisture products, all agree that the dry conditions observed in 2022 over both regions would have been less likely to occur at the beginning of the 20th century.
• To determine the role of climate change in these observed changes, we combine the observation-based datasets with climate models and conclude that human-induced climate change increased the likelihood of the observed soil moisture drought events. The change in likelihood is larger in the observation-based data compared to the models.
• We also assessed the role of climate change in temperature and rainfall in these regions and found that the strong increase in high temperatures is the main reason for the increased drought.
• Combining all lines of evidence we find for West-Central Europe that human-induced climate change made the 2022 root zone soil moisture drought about 3-4 times more likely,  and the surface soil moisture drought about 5-6 times more likely.
• For the Northern Hemisphere extratropics, human-induced climate change made the observed soil moisture drought much more likely, by a factor of at least 20 for the root zone soil moisture and at least 5 for the surface soil moisture, but as is usually the case with hard to observe quantities, the exact numbers are uncertain.
• The models analysed also show that soil moisture drought will continue to increase with additional global warming, which is consistent with projected long-term trends in climate models as reported e.g., in the IPCC AR6.

Scientists from the US, Canada, the UK, the Netherlands, France, Germany and Switzerland collaborated to assess to what extent human-induced climate change made this heatwave hotter and more likely.

Using published peer-reviewed methods, we analysed how human-induced climate change affected the maximum temperatures in the region where most people have been affected by the heat (45–52 ºN, 119–123 ºW) including the cities of Seattle, Portland, and Vancouver (with well over 9 million people in their combined metropolitan areas).

Main findings

• Based on observations and modeling, the occurrence of a heatwave with maximum daily temperatures (TXx) as observed in the area 45–52 ºN, 119–123 ºW, was virtually impossible without human-caused climate change.
• The observed temperatures were so extreme that they lie far outside the range of historically observed temperatures. This makes it hard to quantify with confidence how rare the event was. In the most realistic statistical analysis the event is estimated to be about a 1 in 1000 year event in today’s climate.
• There are two possible sources of this extreme jump in peak temperatures. The first is that this is a very low probability event, even in the current climate which already includes about 1.2°C of global warming — the statistical equivalent of really bad luck, albeit aggravated by climate change. The second option is that nonlinear interactions in the climate have substantially increased the probability of such extreme heat, much beyond the gradual increase in heat extremes that has been observed up to now. We need to investigate the second possibility further, although we note the climate models do not show it. All numbers below assume that the heatwave was a very low probability event that was not caused by new nonlinearities.
• With this assumption and combining the results from the analysis of climate models and weather observations, an event, defined as daily maximum temperatures (TXx) in the heatwave region, as rare as 1 in a 1000 years would have been at least 150 times rarer without human-induced climate change.
• Also, this heatwave was about 2°C  hotter than it would have been if it had occurred at the beginning of the industrial revolution (when global mean temperatures were 1.2°C cooler than today).
• Looking into the future, in a world with 2°C of global warming (0.8°C warmer than today which at current emission levels would be reached as early as the 2040s), this event would have been another degree hotter. An event like this – currently estimated to occur only once every 1000 years, would occur roughly every 5 to 10 years in that future world with 2°C of global warming.

In summary, an event such as the Pacific Northwest 2021 heatwave is still rare or extremely rare in today’s climate, yet would be virtually impossible without human-caused climate change. As warming continues, it will become a lot less rare.

Our results provide a strong warning: our rapidly warming climate is bringing us into uncharted territory that has significant consequences for health, well-being, and livelihoods. Adaptation and mitigation are urgently needed to prepare societies for a very different future. Adaptation measures need to be much more ambitious and take account of the rising risk of heatwaves around the world, including surprises such as this unexpected extreme. Deaths from extreme heat can be dramatically reduced with adequate preparedness action. Heat action plans that incorporate heatwave early warning systems can strengthen the resilience of cities and people. In addition, longer-term plans are needed to modify our built environments to be more adequate for the hotter climate that we already experience today and the additional warming that we expect in future.  In addition, greenhouse gas mitigation goals should take into account the increasing risks associated with unprecedented climate conditions if warming would be allowed to continue

Background information

The heatwave considered in this study is linked to a slow-moving strong high pressure system, sometimes called Omega-blocking or “heat dome”, which brings descending and thus warm and dry air, as well clear skies, further heating the near-surface air. This high pressure system also reached record levels in terms of its strength, measured as the “thickness” of the lower part of the atmosphere, the so-called troposphere. The pressure values observed in the very strong blocking anticyclone are comparable to those observed in other parts of the world in recent heatwaves. The “Omega” blocking pattern is typically associated with heatwaves in this region. While the pressure system was record-breaking in its values, it was far less unusual compared to climatology than the associated extreme temperatures. Recent research suggests that climate change increases the chances for such stagnant high pressure systems in summer through weakening of the summer jet stream. As of yet, it is unclear if, and to what extent, such long-term dynamical changes play a role in this event.

An important feature of this extreme heatwave is that it occurred following a very dry spring over the Western U.S., so the absence of evaporative cooling could be an important factor in the exceptional temperatures observed. However, the northern part of the region impacted by this heatwave experienced wet anomalies in the weeks and months preceding the heat.  Anticyclonic subsidence, and downslope winds were also present, and probably acted as additional heating factors. Overall, it is difficult at this stage to assess the extent to which these factors either in isolation or combined provide a good explanation of why the observed temperatures were so much higher than anything ever recorded in this part of the world. Hence, more research is needed to understand the processes as well as potential influence of human-caused climate change on them.

Here we provide a first estimate of the role of climate change on the extreme temperatures measured in the Pacific Northwest. We analyse the maximum daily temperatures as these are relevant to the impact of the event. While the minimum temperatures are also important for health impacts, we used only one index to keep the assessment straightforward. In this rapid study, we do not analyse the impact that human-caused climate change may have on specific aspects leading to the observed synoptic situation. We ask whether and to what extent human-caused climate change altered the likelihood and intensity of the analysed event. Specifically we analyse (1) how the likelihood of the yearly maximum temperature to be as high or higher than observed in June 2021 has changed and (2) how much less severe a heatwave with the same return period would have been in a world without human-caused climate change. It is important to highlight that, because the temperature records of June 2021 were very far outside all historical observations, determining the likelihood of this event in today’s climate is highly uncertain. All numbers shown assume that the heatwave was a very low probability event (about 1 in 1000 years) that was not caused by new nonlinearities. As in previous analyses, we only give a lower bound of the estimate of the influence of climate change on the change in probability of the event as the best estimate and upper bound are very ill-defined for extreme heat.

Based on this first rapid analysis, we cannot say whether this was a so-called “freak” event (with a return time on the order of 1 in 1000 years or more) that largely occurred by chance, or whether our changing climate altered conditions conducive to heatwaves in the Pacific Northwest, which would imply that “bad luck” played a smaller role and this type of event would be more frequent in our current climate.

In either case, the future will be characterized by more frequent, more severe, and longer heatwaves, highlighting the importance of significantly reducing our greenhouse gas emissions to reduce the amount of additional warming.

The latest heat-related death numbers are alarming, yet they are likely a severe undercount and the real toll will only become clear after mortality statistics are reviewed for the role of heat in exacerbating underlying conditions.

References

[1] World Meterological Organization: June ends with exceptional heat

[2] Global News: Scenes of destruction after wildfire destroys village of Lytton, B.C.

[3] Vancouver Sun: Heat wave linked to massive spike in sudden deaths across Lower Mainland

[4] Oregon Health Authority: Summer 2021 Oregon ESSENCE Hazard Report

[5] CNN: Historic Northwest heat wave linked to dozens of deaths and hundreds of emergency room visits

[6] New York Times: Air-Conditioning Was Once Taboo in Seattle. Not Anymore.

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

• The observational analysis — looking at similar events in the past — will be much harder. There is a gradient in hurricane activity along the coast and there are mountains with strongly varying rainfall amounts. Hurricanes are also rare here so we have to combine different areas along the coast to ascertain whether they fit the distribution of other extreme rain events (like happened in the Gulf) or behave differently than the rest.
• We have only a single model ready with the required resolution for this region and a solid analysis requires multiple large ensembles of high-resolution models.
• Lack of manpower in the WWA consortium. We have no stable funding at the moment and cannot do this analysis “on the side” together with the ongoing analysis of the Kerala floods and the normal work. We would strongly welcome partners that can support our work and enable our impactful analyses.

One analysis of the forecast strength of Florence has already been made public by Reed et al. They use a different technique from our regular WWA analyses, which answers a slightly different question. Whereas we compute the change in probability of a class of events, such as all events with extreme rainfall above a certain limit, their study investigates only this specific event (more precisely, one forecast of the event). The results of such a study not only depend on the general effect of climate change on precipitation of hurricanes, but also on the details of this particular hurricane and even the forecast model. More analyses are needed to assess the robustness of this quick analysis, although the basic result that global warming increases the precipitation is a very robust one supported by observations and modelling studies (e.g. Liu et al, 2018).

Note also that in addition to direct impacts of wind and rain, Florence is expected to cause a large storm surge. Such storm surges are of course affected by ongoing sea level rise, but this is not yet part of our attribution assessment.

We do not find any evidence for an intensification of these types of cold waves due to the Arctic warming faster than the mid-latitudes. On the contrary, they seem to be warming faster than the winter mean as the Arctic air coming south is less cold now.

Key points

• The two-week cold outbreak in the northeastern U.S. and southeastern Canada was highly unusual in the current climate; however, it is not unprecedented.
• Cold waves like this have decreased in intensity and frequency over the last century, but still occur.
• Temperatures like these are now about fifteen times rarer. This is equivalent to cold waves being about 4ºF (2ºC) warmer than they used to be.

Introduction

Over recent years (2014, 2016 and now), there have been a number of extreme cold events in North America, although in different regions and of different durations. Low temperatures have been observed over much of northeastern North America during the two weeks between December 26, 2017 through January 8, 2018 (Fig. 1, top), and these temperatures are unusually cold in a large region centered over the Great Lakes extending to the eastern seaboard (Fig. 1, bottom). In 2014, cold air was concentrated in the Great Lakes area (van Oldenborgh et al, 2015). December 2016 also saw a strong cold spell in North America, but over more western regions (see our analysis of that event). We here analyse this year’s cold using the same methodology as we applied to those other recent cold events.

Comparison with recent observations

First, to help us define formally this year’s event, we compare this year’s temperatures in the general area of the cold wave to what is used as the “normal” period of 1981-2010. We compare the coldest two weeks in the 2017/18 winter so far with the coldest two weeks in previous winters in North America. We chose this comparison, rather than using temperatures on the exact same dates in the past, because a cold wave can occur any time in winter and affects people similarly, regardless of the exact timing. We find that the coldest two weeks this winter have been 7ºF to 11ºF (4ºC to 6 ºC) colder than the coldest two weeks have been, on average, during the past “normal” period of 1981–2010 over most of the area affected by the cold wave (Fig. 2).

Event definition

The area with coldest two-week daily mean averaged temperatures is approximately 40ºN–50ºN, 65ºW–95ºW. This is indicated on the map in Figure 3. Note that only a few Canadian stations have up-to-date data in the GHCN-D v2 dataset. We use the spatial average over land points in this area as an index for this cold wave. This index is thus optimised to express these large anomalies. A time series of daily values of this temperature index since mid-October shows a clear and expected decrease through the end of 2017 — signaling the arrival of winter — but the temperatures over December 26, 2017 through January 7, 2018 are much colder than expected for this time of year, with deviations below the 1981–2010 mean close to 25ºF (14°C) (Fig. 3). The last dip was on Saturday, January 6, after which this cold wave came to an end.

Comparison with historical observations

To put this in a longer historical perspective and compute actual odds of such an event, we used the Berkeley Earth System Temperature daily temperature analyses 1880–2013, extended by ERA-interim up to 2016/17 and ECMWF analyses and forecasts for the last few months. Over the overlap period, this data agrees well after a small bias correction in the mean. The time series of the temperature of the coldest two-week period of the year, averaged over the area of the 2017–2018 cold wave, is shown in Figure 4. Note that although we have chosen the area to give the largest cold anomaly in 2017/18, there have been many similar or colder two-week periods in that region in earlier periods, notably in 1960–1994 and before 1920, but none since 1993/94. If not unprecedented in absolute values, this year’s cold stands out for occurring so early in the season. Only in 1886/87 there was an earlier colder two-week period. In 1887/88 a similarly cold period started one day later than this year.

Return time

To quantify the effect of warming on the coldest two-week period of the winter in the area, we fitted the series to a Generalised Extreme Value (GEV) distribution whose mean location is allowed to shift proportionally to the smoothed global mean temperature time series (Fig. 5). That factor is estimated to be about two, indicating that the temperature of the coldest two-week period has increased about two times faster than the global mean temperature rise. This has been explained as a consequence of stronger warming trends in the Arctic, where the cold air originates  (Screen, 2014; van Oldenborgh et al, 2015). The fit also shows that these two weeks are very unusual in the current climate, with a return time of very roughly 250 years in this region.

Change in probability

The significant dependence of the GEV location on the values of global mean temperature allows us to ask what the return time of such an event was in the climate of a century ago, i.e., for values of the global mean temperature at the beginning of the 20th century. Our estimates indicate that a cold wave like this occurred on average every seventeen years. This means that cold waves like this have become approximately a factor fifteen less frequent due to global warming, as shown in Figure 5.

Trends over North America

In fact, the temperature of two-week cold waves has increased over all of North America, as expected in a warming world (Fig. 6). We see no evidence of an increased intensity of these cold waves, but rather a faster warming than the mean winter temperature in northern North America. The same was shown for one-day cold extremes in previous articles (van Oldenborgh, et al, 2015), both in observations and climate models.

Conclusions

We conclude that this was an exceptional two-week cold wave in the area in the current climate. Cold outbreaks like this are getting warmer (less frequent) due to global warming, but cold waves still occur somewhere in North America almost every winter.

References

Barnes, E. A. and J. Screen. (2015) The impact of Arctic warming on the midlatitude jet-stream: Can it? Has it? Will it? WIREs Climate Change, 6: 277–286. doi: 10.1002/wcc.337

Van Oldenborgh, G.J., R. Haarsma, H. De Vries, en M.R. Allen. (2015) Cold extremes in North America vs. mild weather in Europe: the winter of 2013–14 in the context of a warming world. Bulletin of the American Meteorological Society, 96: 707–714, doi: 10.1175/BAMS-D-14-00036.1

Screen, J.A. (2014) Arctic amplification decreases temperature variance in northern mid- to high-latitudes. Nature Climate Change, 4: 577–582. doi: 10.1038/nclimate2268

We also analysed extreme rainfall in the Houston area in three ensembles of 25 km resolution models. The first also shows 2 × CC scaling, the second 1 × CC scaling and the third did not have a realistic representation of extreme rainfall on the Gulf Coast. Extrapolating these results to the 2017 event, we conclude that global warming made the precipitation about 15% (8%–19%) more intense, or equivalently made such an event three (1.5–5) times more likely. This analysis makes clear that extreme rainfall events along the Gulf Coast are on the rise. And while fortifying Houston to fully withstand the impact of an event as extreme as Hurricane Harvey may not be economically feasible, it is critical that information regarding the increasing risk of extreme rainfall events in general should be part of the discussion about future improvements to Houston’s flood protection system.

Introduction

Hurricane Harvey formed as a tropical storm over the Atlantic Ocean on August 17, 2017 and crossed into the Caribbean Sea the next day. It weakened to a tropical depression as it crossed the Yucatan Peninsula, but attained hurricane strength over the Gulf of Mexico on August 24, rapidly intensifying to reach Category 4 strength just before making landfall on the Texas coast 50 km east of Corpus Christi on August 25, causing severe wind damage in coastal towns. Harvey moved slowly inland, remaining nearly stationary about 100 km inland for four days before moving back into the Gulf and making a second landfall in Louisiana on August 30. While Hurricane Harvey was a significant hurricane in terms of its size and wind speed, ultimately, the storm will be remembered for the extreme flooding it caused in Houston and surrounding areas.

Between August 25–30, unprecedented rainfall totals for a tropical cyclone (TC) in the contiguous United States were recorded. The station Cedar Bayou at FM1942 , about 40 km west of Houston, had observed an accumulated 1318 mm (51.89^”) by Thursday August 31, 10 AM CDT. Widespread flooding necessitated more than 120 000 rescues, exceeding the capacity of formal emergency response organisations and requiring the assistance of volunteers with access to boats or large vehicles. Over 80 deaths have been attributed to Harvey, mostly as a result of drowning, and financial analysts estimate it to be among the costliest natural disaster in US history (NOAA NCEI 2017). It is estimated that flooding ultimately impacted more than 100 000 homes, of which nearly 80 000 are estimated to have been flooded to a depth of at least 0.46 m (18^”) and 23 000 to at least 1.5 m (5 ft) (FEMA 2017).

In the immediate aftermath of the event, the question was raised as to what extent the impacts of Hurricane Harvey were intensified due to anthropogenic climate change. In this paper, we analyze the rainfall associated with Hurricane Harvey, contextualising it with relevant flood risk factors, to answer this question.

The observed and expected response of TCs to greenhouse gas-induced climate change has been the subject of intense research. Globally, there is an expectation that increasing greenhouse gases will lead to a decrease or no change in the overall number of TCs, but that the maximum wind speed and precipitation of the strongest storms should increase (Hesselbjerg Christensen et al 2013). However, there is low confidence in region-specific projections. For the Atlantic basin, there is considerable spread in the expected change in TC frequency resulting from CO₂ increases, even considering only the strongest storms (Knutson et al 2013). Furthermore, changes in observing practices limit confidence in century-scale trends in Atlantic hurricane frequency (Vecchi and Knutson, 2011). That is, at this stage, there is no clear scientific evidence to support the notion that the existence of Harvey was made more likely by global warming.

However, the impacts of Harvey may have been influenced by global warming; studies consistently indicate that greenhouse gas-induced warming should lead to increases in the total and maximum rainfall by TCs (Knutson et al 2010, Scoccimarro et al 2014, Villarini et al 2014). In general, the maximum moisture content of air increases with 6%–8.5% per degree warming, according to the Clausius-Clapeyron (CC) relationship (Clapeyron 1834, Clausius 1850, Held and Soden 2006, O’Gorman 2015). If relative humidity stays the same, which is the norm near oceans, the actual amount of water vapour in the air increases by the same amount. Studies exploring the response of TC rainfall to greenhouse warming find rates of increase at least as large as CC-scaling, with various studies indicating increases that follow or exceed CC-scaling (e.g. Knutson et al 2010, 2013, Scoccimarro et al 2014). The hypothesis underlying higher scaling is that the extra heat of condensation gives extra energy to drive the circulation in a well-organised system. The moisture flux is thus enhanced twice: not only with higher moisture content, but also with higher velocities. This could result in up to two times CC-scaling, as was found on smaller and shorter time scales by Lenderink et al (2017). Another contribution may be a possible increase in the probability that a hurricane stalls over the coast either by a systematic trend in persistency of high pressure events globally (Mann et al 2017) or simply a local tendency in mean circulation.

Van der Wiel et al (2017) showed that the high tail of the distribution of extreme precipitation on the US Gulf Coast can be described well by a generalised extreme value distribution (GEV), in spite of the many different mechanisms that cause these high precipitation events there (Schumacher and Johnson 2006). It was found that the change in intensity was compatible with 2 × CC scaling. The distribution of the most extreme events was also found to be simulated reasonably well by the GFDL HiFLOR model with a 25 km atmospheric resolution, and to a lesser extent by the 50 km resolution FLOR-FA model. These models simulated 1 × CC scaling for the most extreme events, as is observed for global one-day precipitation extremes (Westra et al 2013). However, regionally and for longer time-scale events, the scaling may well be different. For instance, in Boulder, Colorado, lower scaling was found for five-day precipitation extremes (Hoerling et al 2013, Eden et al 2016).

We follow the same methodology as van der Wiel et al (2017) to attribute the extreme precipitation from Hurricane Harvey to anthropogenic climate change and refer to that paper for an extensive discussion of the methods and assumptions. In addition, we are including two supplementary observational datasets and two additional high-resolution models. This analysis does not consider the relatively low-resolution (50 km) FLOR-FA model.

The analysis herein focuses on extreme precipitation as the primary cause of flooding. We do not consider the backwater effects of elevated water levels due to storm surge or relative sea level rise in Galveston Bay on the ability of the system to drain, but previous studies have suggested that this may be an important factor in determining the intensity and extent of coastal flooding (Torres et al 2015, Sebastian et al 2017). In addition, we acknowledge that other anthropogenic factors have contributed to increased flood risk in Houston, specifically urban development, which has led to floodplain encroachment, increased impervious cover, reduced overland and channel roughness, decreased storage capacity (Brody et al 2008, 2011), and resource withdrawal that has led to subsidence of up to 3 m (10 ft). Moreover, the operation of emergency flood control structures, such as the Addicks and Barker Reservoirs west of downtown Houston, further exacerbated flooding in some areas of the city. While the impact of urban development and subsidence on flooding during Hurricane Harvey is not directly addressed in this paper, we provide suggestions for future research regarding this event and its impacts.

Event definition

The time scale of the event definition is set by the main impact: flooding in the city of Houston. Hydrographs show that most rivers and bayous in Houston crested on Monday August 28 or Tuesday August 29 local time, see figures 1(a) and (b). Note that Buffalo Bayou is an exception, as it was affected by the controlled release of waters from the Addicks and Barker Reservoirs upstream (figure 1(c)). The 10 minute meteorological station ‘Houston’ also shows that the three days up to Monday August 28 produced the highest amounts of precipitation. We therefore take the three-day average for Saturday August 26 to Monday August 28 to be the most relevant time scale. This implies the rainfall accumulations we consider are lower than the all-storm totals mentioned in the introduction.

The highest official gauge recording over the three-day period reported immediately after the event was at the William P. Hobby International Airport (Hobby Airport), WMO 72244, with 824.7 mm 3dy⁻¹ (32.47^” 3dy⁻¹), corresponding to 274.9 mm dy⁻¹ (10.82^” day⁻¹) on August 26–28. At airports, this is usually 0–24 UTC. Later updates of the GHCN-D v2 dataset added to stations with higher precipitation: Houston NWSO (USC00414333) with 999.2 on August 26–28 and Baytown (USC00410586) with 1043.4 mm 3dy⁻¹ (41.07^” 3dy⁻¹), corresponding to 347.8 mm dy⁻¹ (13.69^” dy⁻¹) on August 27–29, presumably 8–8 local time. We use the latter value as the highest observed point value for this event.

Amateur stations give even higher amounts, but were not used in this study. It should be noted that often-quoted ’72 hr sums’ use higher resolution data (hourly or 10 minute) and choose the beginning arbitrarily. This precludes comparison with historical observations that are usually daily with a fixed observation time.

Van der Wiel et al (2017) showed that extreme precipitation could be considered homogeneous along the Gulf Coast, using the land area 29°N–31°N, 85°W–95°W. As the most-affected areas are just west of this box, we extended it to Corpus Christi, 27.5°N–31°N, 85°W–97.5°W (figures 2(c) and (d)). The extreme western part of this area has slightly lower extreme precipitation, but this does not affect the analysis, especially not the spatial maximum that is used for the models. Note that extreme precipitation in this area is not only due to tropical storms, but also to a variety of other mechanisms such as the cut-off low studied in van der Wiel et al (2017). Extreme precipitation events occur throughout the year, with only a modest increase in the hurricane season (June–November).

Data and methods

Observational data

For the observational analysis, we primarily use the two rain gauge-based datasets that were also used in van der Wiel et al (2017). The first is the GHCN-D v2 collection of rain gauge data extended with GTS data. There are 312 stations in the box covering 1878–2017. The density of stations is much higher than the decorrelation length of three-day precipitation. To obtain less-dependent datasets, we analyse the station data in two ways: all stations with 30 or more years of data and at least 0.1° apart (85 stations), and a subset with 80 years or more of data and at least 1° apart (13 stations). The subsetting was done in the order of station IDs. The first subset is slightly stricter than in the previous study (van der Wiel et al 2017) to ensure the validity of the moving spatial blocks technique that we use in the bootstrap to obtain uncertainty estimates in the presence of spatial dependencies. We finally note that the rain gauges have varying observation times.

For comparison against the model data, we use the CPC Unified Precipitation Analysis, a 25 km gridded dataset based on rain gauges. This dataset has a maximum three-day precipitation of 251.1 mm dy⁻¹ or 614.2 mm 3dy⁻¹ (9.89^” day⁻¹ or 24.18^” 3dy⁻¹), see figure 3(b). This is only slightly lower than the highest point value available at the time in spite of the area-averaging, showing the large spatial extent of the extreme precipitation. However, the observations at Baytown and Houston NWSO, and a comparison with the rain radar data (figure 3(c)), show that this dataset likely underestimated the rainfall. However, as it is the best estimate we have with a long time series, we use it anyway, keeping this underestimation in mind.

At this resolution, the grid points are insufficiently independent to do an extreme value analysis, so for these we averaged the grid boxes into a 50 km grid. At this resolution the highest 2017 value is 186.7 mm dy⁻¹ or 560.2 mm 3dy⁻¹ (7.35^” dy⁻¹ or 22.05^” 3dy⁻¹).

To check these datasets, we also considered the NASA GPM/IMERG data and the NOAA calibrated rain radar fields. The NASA dataset is 0–24 UTC and the NOAA dataset 12–12 UTC. In figures 3(c) and (d) we show the highest three-day precipitation in 2017 up to September 30 (only for August 25–30 for NOAA) fields. The NOAA calibrated rain radar dataset has somewhat higher point maxima, up to 385.5 mm dy⁻¹ or 1156.5 mm 3dy⁻¹ (15.18^” dy⁻¹ or 45.53^” 3dy⁻¹), as it can catch local maxima that usually are not caught by official rain gauges. However, as it has no long historical record, we could not use it for the quantitative analysis. The NASA GPM/IMERG datasets show lower maximum amounts than any other dataset, 203.3 mm dy⁻¹ (8.00^” dy⁻¹) at 10 km resolution.

EC-Earth

We used the output of atmosphere-only EC-Earth 2.3 (Hazeleger et al 2010) experiments at T799 (~25 km) described in detail in Haarsma et al (2013). There are four experiments of six ensemble members each: pre-industrial 1850–1854, present-day 2002–2006, near future 2030–2034 and end-of-century 2094–2098 (120 years total). The present-day experiment uses prescribed daily observationally-based reconstructions of sea surface temperatures (SST) at 0.25° resolution. The other experiments use the 2002–2006 observed SSTs transformed to the appropriate epoch by subtracting or adding the mean SST change of the ECHAM5/MPI-OM model used in the ESSENCE project with SRES A1B forcings (Sterl et al 2008) to the 2002–2006 conditions. Other boundary conditions, such as atmospheric greenhouse gas concentrations, were taken from the RCP4.5 scenario. This model has been shown to represent Atlantic hurricanes well in Haarsma et al (2013) and Baatsen et al (2015).

GFDL HiFLOR

As a second model, GFDL HiFLOR is used. HiFLOR is a fully coupled global climate model, based on GFDLs CM2.1 and CM2.5 (Delworth et al 2006, 2012). It uses a 25 km horizontal atmospheric and land grid, coupled to a coarser ocean and sea ice model. A detailed model description is provided in Murakami et al (2015) and van der Wiel et al (2016). Four static forcing experiments (1860, 1940, 1990, 2015) of variable length are available. The first 20 years of each integration are removed to allow for fast, near-surface ocean spin-up, and the remaining years concatenated to form a dataset that covers 565 years. Full details of the model experiments used are provided in van der Wiel et al(2017). The models are biased towards smaller precipitation extremes in the Gulf Coast region, though due to the relatively high resolution, this bias is significantly lower than in coupled models of average CMIP5 resolution (van der Wiel et al 2016).

The model data of HiFLOR are identical to the data that were used in van der Wiel et al (2017) because neither extensions to these experiments nor data over an extended region are publicly available at this time. However, the statistics along the coast are fairly constant (figure 2(d)), so the only problem arising when using a smaller box is that the probability for an event to occur anywhere on the coast is underestimated. In the observational gridded dataset, the return time for an event to occur in the larger box is about 25% lower than a similar event in the smaller box. However, the estimate of the change in return period or the changes in intensity due to anthropogenic climate change is not impacted by the size of the box after this bias correction. We have therefore chosen to include HiFLOR in this analysis, using the available model data.

Weather@home

The third model is the regional climate model (RCM) HadRM3P with ~25 km horizontal resolution and 19 vertical levels (Massey et al 2015, Guillod et al 2017). The domain includes Central America, the United States, and the entire Gulf of Mexico region. Most modelled hurricanes affecting the Gulf Coast originate inside this domain. It is nested in and driven by the global HadAM3P model at N96 resolution. Three experiments are used in our analysis. (1) A 30 year climatology (1986–2015) with 30 simulations per year (= 900 members; ACTCLIM). (2) An actual ensemble with 1000 simulations for the August–October 2017 period (ACTUAL). (3) A natural or counterfactual ensemble with 1500 simulations for the August–October 2017 period (NATURAL). A smaller ensemble of ~100 ACTUAL and NATURAL simulations is available for the February–July 2017 period. Lower boundary conditions are prescribed using OSTIA SSTs for ACTCLIM, and GloSea5 forecast SSTs for ACTUAL and NATURAL. ΔSSTs from CMIP5 historical and natural runs are used to obtain the naturalised SSTs (see Haustein et al (2016) for details). Forcings are prescribed according to the CMIP5 protocol (Meinshausen et al 2011).

Methods

Estimates of return periods and changes therein and changes in intensity are obtained with fits to GEVs that scale with the smoothed global mean temperature (GMST), inspired by the Clausius–Clapeyron relation. We investigated the alternative of using of SST averaged over the Gulf of Mexico, which would be more directly related to local atmospheric moisture content. The results this gives are very similar to those obtained using GMST, albeit with larger uncertainties due to the stronger noise in the local SST (from weather, local forcings and observational problems). Again, full details and the underlying assumptions are given in van der Wiel et al (2017).

Spatial dependencies are accounted for by a spatial moving block analysis as recommended by Efron and Tibshirani (1998) for temporal dependencies. This technique was also used in Eden et al (2016) and van der Wiel et al (2017). The quality of the fits is checked by comparing the fit and observations for the current climate (red in subsequent plots) and a previous climate (blue in subsequent plots). We always quote two-sided 95% confidence intervals, which are estimated by a non-parametric bootstrap procedure.

Return periods and trends in observations

GHCN-D rain gauges

As mentioned before, for 2017, we take the observed value at Baytown (USC00410586) with 1043.4 mm 3dy⁻¹, corresponding to 347.8 mm dy⁻¹, over August 27–29. This is also the highest value observed in the box over all years. The 2017 event under study here is not included in the fits, although, of course, the 2016 Louisiana event is included.

In the larger station set of 85 stations (5193 station years, about 800 degrees of freedom), the return period of the 2017 three-day maximum in the current climate is more than 9000 yr (97.5% CI), figures 4(a), (c) and (e). The large extrapolation required makes the uncertainties large. This is a factor of four (2.1–5.0) larger than it was in the climate of 1900, corresponding to a 19% (12%–22%) increase in intensity.

This is confirmed by the fit to the much smaller dataset of 13 stations with at least 80 years of data and minimum distance 1° that are almost independent (1310 station years, about 1000 degrees of freedom), figures 4

(b), (d) and (f). The GEV fit gives a return period of larger than 9000 yr in the current climate (97.5% CI), a factor of four (2.0–11) larger than in the climate of around 1900. The increase in intensity is also compatible: 18% (11%–25%).

We finally considered the spatial maximum in the box (not shown). For this, we use the 13-station dataset, which has a roughly constant number of stations between 1910 and 2010. A GEV fit over this period gives the return period for an event like this happening anywhere along the Gulf Coast between Corpus Christi, TX, and Apalachicola, FL. This is about once in 800 yr, with a 97.5% upper bound of less than once every 100 yr (i.e. less than a 1% probability every year). The risk ratio (RR), or the change in probability, has larger uncertainties in this measure than when using all stations. We find a value between 0.5 and 11.

CPC gridded analysis

As mentioned in section 3.1, we averaged the CPC gridded analysis to 0.5° to reduce the spatial dependencies. On this scale, the 2017 event is the highest in the dataset of 85 grid points (237.8 mm dy⁻¹ or 9.36^” dy⁻¹), of which only five are independent. We find a return period for this much rain in a 50 km grid box of about 10 000 yr (2200–30 000 yr). Note that this likely an underestimate, as the rain radar gave higher area averages. The probability in the current climate is a factor 5.4 (1.6–10) larger than in the climate of 1950, corresponding to an 18% (7%–24%) increase in intensity (figures 5(a) and (c)).

If we consider the annual and spatial maximum in the Gulf Coast region in the 0.25° dataset (251.1 mm dy⁻¹ or 9.89^” dy⁻¹), this also shows an increase, albeit, again, with larger uncertainties (figures 5(b) and (d)). The return period of an extreme event like the one observed in 2017 or worse anywhere in the box is around 230 yr (60–5000 yr). The RR is very uncertain over this short time period: 3.2 (0.6–80) since 1950, corresponding to an increase in intensity of very roughly 12% (−5% to 29%).

We conclude from these two observational datasets that the probability of observed intense rainfall has increased by a factor of roughly four, very likely more than two, corresponding with an increase in intensity of about 20%, very likely more than 12%.

This was an extremely rare event given the past observations, even taking the trend into account. The return periods for a point observation as large as observed or larger is more than 9000 yr (97.5% CI). The chances of observing an event like this anywhere on the Gulf Coast are higher, but still low: less than 1 in 100 years (<1% yr⁻¹, 97.5% CI) for a point observation.

We show in the supplementary material available at stacks.iop.org/ERL/12/124009/mmedia that the probability in 2017 is not increased due to natural variability.

Attribution to anthropogenic factors

To investigate the source of the observed increase, we have to use climate models in which we can study a counterfactual climate without anthropogenic influences, such as the emissions of greenhouse gases. As in van der Wiel et al (2017), we consider the spatial maximum over 27.5°N–31°N, 85°W–97.5°W of the annual maximum of three-day averaged precipitation. For HiFLOR, we could only use the somewhat smaller box 29°N–31°N, 85°W–95°W. This implies that there are fewer locations to reach the maximum precipitation. We correct the model output for this difference together with the bias correction.

Model evaluation

We investigated whether the models were fit for purpose in two ways: by comparing the tail of the extreme precipitation and the seasonal cycle of somewhat less extreme precipitation to the observations. The details are shown in the supplementary material. Based mainly on the shape of the tail, we decided to use the EC-Earth and GFDL models, but not the HadRM3P model.

EC-Earth

The extreme precipitation is fitted to a GEV that depends on the RCP4.5 equivalent CO₂concentration, not the global mean temperature (figure 6(a)). This was done because there are no GMST observations for the two future periods and the CMIP5 RCP4.5 multi-model global mean temperature is not well-defined before 1860, as not all models simulate that period. The equivalent CO₂ concentration and multi-model global mean temperature are correlated at r = 0.996.

The GEV fit shows that the spatial maximum of the annual maximum of three-day precipitation along the Gulf Coast is simulated reasonably well (figure 6(c)), although due to the high scale parameter σ (steeper slope), the return period is lower than in the observations (figure 5(d)). The fit assumes that the scaling up to 2100 is the same as up to 2017. It gives an increase in intensity from of 1861–2017 of 17% (10%–23%), corresponding to an increase in probability of a factor 2.2 (1.5–4.1).

HiFLOR

As mentioned in the supplementary material, the GEV fits provide a multiplicative bias correction of 80% to correct for the modelled extremes that are smaller than observed and the difference in box sizes (see section 3.3). The model bias over the same box is about 40%. The return period is compatible with the observed one, as expected after bias correction. The increase in probability (RR) is 1.6 with a 95% CI of 1.3–2.3. This corresponds to an increase in intensity of 8% (4%–11%, figures 6(b) and (d)).

Synthesis

Figure 7 summarises the results obtained above. The top three results show the changes for local extremes, i.e. for an event to occur at a specific location. The next four show the changes in the spatial maximum, i.e. for an event to occur anywhere along the US Gulf Coast. The latter fits involve only the most extreme events, giving O(10) times fewer degrees of freedom. The uncertainties due to natural variability are therefore larger, but the two models that passed the evaluation can only reproduce these most extreme events reliably. In the observations, the regional extremes are smoothly connected to less extreme events, so we can use the changes in the local extremes to deduce the changes in the regional extremes. We therefore compute the last line as the unweighted average of the three estimates for changes in local extremes from observational datasets and the two estimates from models for changes in regional extremes.

It should be noted that changes at the return period of the extreme rainfall from Hurricane Harvey in 2017, estimated at more than 9000 yr from the station data, can only be deduced from a GEV fit to the various datasets by assuming the properties do not change in the extrapolation from the more common extreme events, as none of the models has enough data to sample this probability directly.

Both the observations and EC-Earth show an increase in intensity ΔI of around 16% per degree global warming (figure 7(a)). As this part of the Gulf of Mexico and the Gulf Coast have warmed by about the same amount as the global mean, this is equivalent to two times Clausius–Clapeyron scaling. In contrast, the HiFLOR model only shows an increase of about 8%, which is more compatible with 1 × CC scaling. The spread in results is not compatible within natural variability, χ²/dof ≈ 4 for the two datasets and two model results, so there are systematic differences between the models that must be taken into account.

The same picture appears for the RRs (figure 7(b)); again, the EC-Earth results are in line with the observations (in spite of the too high variability). Extreme rainfall on the Gulf Coast in HiFLOR is less sensitive to global warming than EC-Earth in this region.

We conclude that precipitation extremes on the US Gulf Coast have increased due to global warming. The increase is higher in the observations (12 < ΔI < 22%) and one of the two models (10 < ΔI < 23%) than in the other model (4 < ΔI < 11%). Both should include possible stalling effects. If the station observations are homogeneous enough, this points to 2 × CC scaling. However, there may be inhomogeneities in the observations (well after 1948) that could cause an overestimation of the trend, which would imply somewhat lower scaling. The unweighted average of the local change in intensity in three relatively independent observational datasets and the regional change in the two models gives an increase of 15% with an uncertainty range 8%–19%.

These increases are equivalent to an increase in probability of at least a factor two in the observations, somewhat less in the models. An unweighted average on a logarithmic scale gives a most likely increase of a factor of 3, with uncertainty range of 1.5–5.

Extrapolating these trends to the future, we expect another similar increase if global warming is limited to 2 °C above pre-industrial levels. However, under a ‘business-as-usual’ scenario in which the world continues to rely primarily on fossil fuels, the intensity of extreme rainfall events on the Gulf Coast would increase by about 50% as the world warms another four degrees. This corresponds to an increase in probability of a factor of roughly 10 (both numbers have large uncertainties). This is in rough agreement with the increase of a factor of 18 that Emanuel (2017) finds relative to 1981–2000 for large-area averaged precipitation (their value for 2017 is larger than ours mainly due to the linear interpolation rather than the exponential one we use).

Vulnerability and exposure

While this study focuses on the rainfall hazard, the primary impact was flooding. Drivers that exacerbate or reduce impacts in storms of this scale are a complex combination of an extreme natural hazard, long-term planning decisions and short term disaster preparedness and response decisions. These are discussed in detail in the supplementary material. Although the extreme rainfall levels from Hurricane Harvey are extremely rare, additional factors, such as rapid population growth, urban growth policies, and ageing water management infrastructure further exacerbate the ultimate impacts of this storm. Recent flood events resulting from storms such as Tropical Storm Allison (2001), Hurricane Ike (2008), Memorial Day (2015), and Tax Day (2016) further illustrate the importance of managing exposure and vulnerability when reducing the level of flood impacts in Houston.

Conclusions

As described in the introduction and methods, this study has focused primarily on the changes in extreme rainfall in the US Gulf Coast region, applied to the rainfall during Hurricane Harvey that caused record flooding in Houston. The first result is that the three-day rainfall sums that were responsible for most of the flooding were extremely rare, with a return time for station observations of more than 9000 yr (97.5% CI) in the current climate, taking the trend into account. This would have caused flooding in any city.

The second result is that we find strong evidence that global warming over the last century, primarily caused by anthropogenic greenhouse gas emissions, has increased the intensity of the three-day rainfall extremes on the Gulf Coast, or equivalently, has increased the probability of a given rainfall event. Due to the rarity of the observed rainfall in Houston, applying the increase observed for less extreme events to this event requires a (reasonable) extrapolation. We find that the intensity of rainfall increased by 15% (8%–19%) and the probability of this much rain or more by a factor of three (1.5–5). This increase in rainfall intensity contributed to the flooding observed in Houston and surrounding areas. Less extreme rainfall events have also resulted in impactful flooding in Houston in previous years, and their return times are also decreasing.

We also acknowledge that several other factors have likely also contributed to increased flood risk in Houston over the past century and should be further explored. First, given Houston’s proximity to the coast, relative sea level rise has likely contributed to increases in flood risk in the region. While NOAA estimates that sea level rise near Galveston Island is increasing by approximately 6.47 mm yr⁻¹, further analysis is required to disentangle the relative contributions of anthropogenic climate change and regional subsidence to increases in sea levels in Galveston Bay and their effects on flood risk in Houston. Second, the effects of regional changes in land use and land cover on flooding during Hurricane Harvey, as well as the long-term performance and operation of flood adaptation measures employed in Houston, are of interest and importance to local governing bodies. Finally, additional research is required to determine to what extent Harvey’s storm surge contributed to compound flooding in coastal watersheds, and whether compound flooding during tropical cyclones should be considered in the design of coastal flood defences.

While these questions were not answered in this study, our results provide one of the big pieces of the puzzle showing that although rainfall during Hurricane Harvey was exceptional, the trend in extreme rainfall needs to be taken into account when considering upgrades to flood infrastructure in Houston and surrounding areas since additional global warming will continue to increase the risk of extreme precipitation events further.

References

Baatsen, M., Haarsma, R.J., Van Delden, A.J. and de Vries, H. (2015) Severe autumn storms in future western Europe with a warmer Atlantic Ocean. Climate Dynamics, 45: 949–64. doi: 10.1007/s00382-014-2329-8

Brody, S.D., Zahran, S., Highfield, W.E., Grover, H. and Vedlitz, A. (2008) Identifying the impact of the built environment on flood damage in Texas. Disasters, 32: 1–8. doi: 10.1111/j.1467-7717.2007.01024.x

Brody, S.D., Blessing, R., Sebastian, A. and Bedient, P. (2011) Examining the impact of land use/land cover characteristics on flood losses. Journal of Environmental Planning and Management, 57: 1252–65. doi: 10.1080/09640568.2013.802228

Clapeyron, E. (1834) Me ́moire sur la puissance motrice de la chaleur. Journal de l’Ecole Royale Polytechnique, XIV: 153–90.

Clausius, R. (1850) Ueber die bewegende Kraft der Wa ̈rme und die Gesetze, welche sich daraus fu ̈r die Wa ̈rmelehre selbst ableiten lassen. Annalen der Physik, 155: 368–97. doi: 10.1002/andp.18501550306

Delworth, T.L. et al. (2006) GFDL’s CM2 global coupled climate models. Part I: formulation and simulation characteristics. Journal of Climate, 19: 643–74. doi: 10.1175/JCLI3629.1

Delworth, T.L. et al. (2012) Simulated climate and climate change in the GFDL CM2.5 high-resolution coupled climate model. Journal of Climate, 25: 2755–81. doi: 10.1175/JCLI-D-11-00316.1

Eden, J.M., Wolter, K., Otto, F.E.L. and van Oldenborgh, G.J. (2016) Multi-method attribution analysis of extreme precipitation in Boulder, Colorado. Environmental Research Letters, 11: 124009. doi: 10.1088/1748-9326/11/12/124009

Efron, B. and Tibshirani, R.J. (1998) An Introduction to the Bootstrap. New York: Chapman and Hall. ISBN 9780412042317

Emanuel, K. (2017) Assessing the present and future probability of Hurricane Harvey’s rainfall. Proceedings of the National Academy of Sciences, 114: 12681–84. doi: 10.1073/pnas.1716222114

FEMA, 2017. Historic disaster response to Hurricane Harvey in Texas. HQ-17-133.

Guillod, B.P., Jones, R.G., Bowery, A., Haustein, K., Massey, N.R., Mitchell, D.M., Otto, F.E.L., Sparrow, S.N., Uhe, P., Wallom, D.C.H., Wilson, S. and Allen, M.R. (2017) weather@home 2: validation of an improved global – regional climate modelling system. Geoscientific Model Development, 10: 1849-1872. doi: 10.5194/gmd-10-1849-2017

Haarsma, R.J., Hazeleger, W., Severijns, C., de Vries, H., Sterl, A., Bintanja, R., van Oldenborgh, G.J. and van den Brink, H.W. (2013) More hurricanes to hit western Europe due to global warming Geophysical Research Letters, 40: 1783–88. doi: 10.1002/grl.50360

Haustein, K., Otto, F.E.L., Uhe, P., Schaller, N., Allen, M.R., Hermanson, L., Christidis, N., McLean, P. and Cullen, H. (2016) Real-time extreme weather event attribution with forecast seasonal SSTs. Environmental Research Letters, 11(6):  doi: 10.1088/1748–9326/11/6/064 006

Hazeleger, W., C. Severijns, T. Semmler, S. Ştefănescu, S. Yang, X. Wang, K. Wyser, E. Dutra, J.M. Baldasano, R. Bintanja, P. Bougeault, R. Caballero, A.M. Ekman, J.H. Christensen, B. van den Hurk, P. Jimenez, C. Jones, P. Kållberg, T. Koenigk, R. McGrath, P. Miranda, T. Van Noije, T. Palmer, J.A. Parodi, T. Schmith, F. Selten, T. Storelvmo, A. Sterl, H. Tapamo, M. Vancoppenolle, P. Viterbo, and U. Willén. (2010) EC-Earth: A seamless Earth-system prediction approach in action. Bulletin of the American Meteorological Society, 91: 1357–1363. doi: 10.1175/2010BAMS2877.1

Held, I.M. and Soden, B.J. (2006) Robust responses of the hydrological cycle to global warming Journal of Climate, 19: 5686–99. doi: 10.1175/JCLI3990.1

Christensen, J.H., K. Krishna Kumar, E. Aldrian, S.-I. An, I.F.A. Cavalcanti, M. de Castro, W. Dong, P. Goswami, A. Hall, J.K. Kanyanga, A. Kitoh, J. Kossin, N.-C. Lau, J. Renwick, D.B. Stephenson, S.-P. Xie and T. Zhou, 2013: Climate Phenomena and their Relevance for Future Regional Climate Change. In: Climate Change 2013: The Physical Sci- ence Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

Hoerling, M.P., Wolter, K., Perlwitz, J., Quan, X., Escheid, J., Wang, H., Schubert, S., Diaz, H.F. and Dole, R.M. (2013) Northeast Colorado extreme rains interpreted in a climate change context. Bulletin of the American Meteorological Society. 95: S15–18. doi: 10.1175/1520-0477-95.9.S1.1

Knutson, T.R., Mcbride, J.I., Chan, J., Emanuel, K., Holland, G., Landsea, C.W., Held, I.M., Kossin, J.P., Srivastava, A.K. and Sugi, M. (2010) Tropical cyclones and climate change. Nature Geoscience, 3: 157–63. doi: 10.1038/ngeo779

Knutson, T.R., Sirutis, J.J., Vecchi, G.A., Garner, S., Zhao, M., Kim, H-S., Bender, M., Tuleya, R.E., Held, I.M. and Villarini, G. (2013) Dynamical downscaling projections of twenty-first-century Atlantic hurricane activity: CMIP3 and CMIP5 model-based scenarios. Journal of Climate, 26: 6591–6617. doi: 10.1175/JCLI-D-12-00539.1

Lenderink, G., Barbero, R., Loriaux, J.M. and Fowler, H.J. (2017) Super-Clausius–Clapeyron scaling of extreme hourly convective precipitation and its relation to large-scale atmospheric conditions. Journal of Climate, 30: 6037–52. doi: 10.1175/JCLI-D-16-0808.1

Mann, M.E., Rahmstorf, S., Kornhuber, K., Steinman, B.A., Miller, S.K. and Coumou, D. (2017) Influence of anthropogenic climate change on planetary wave resonance and extreme weather events. Scientific Reports, 7: 45242. doi: 10.1038/srep45242

Massey, N., Jones, R., Otto, F. E. L., Aina, T., Wilson, S., Murphy, J. M., Hassell, D., Yamazaki, Y. H. and Allen, M. R. (2015) weather@home— development and validation of a very large ensemble modelling system for probabilistic event attribution. Quarterly Journal of the Royal Meteorological Society, 141: 1528–1545. doi: 10.1002/qj.2455

Meinshausen, M., Smith, S.J., Calvin, K., Daniel, S., Kainuma, M.L.T., Lamarque, J-F., Matsumoto, K., Montzka, S.A., Raper, S.C.B., Riahi, K., Thomson, A., Velders, G.J.M. and van Vuuren, D.P.P. (2011) The RCP greenhouse gas concentrations and their extensions from 1765–2300. Climatic Change, 109: 213. doi: 10.1007/s10584-011-0156-z

Murakami, H. and Vechi, G.A. (2015) Simulation and prediction of category 4 and 5 hurricanes in the high-resolution GFDL HiFLOR coupled climate model. Journal of Climate, 28: 9058–79. doi: 10.1175/JCLI-D-15-0216.1

NOAA NCEI, 2017. Billion-dollar weather and climate disasters: overview

O’Gorman, P.A. (2015) Precipitation extremes under climate change. Current Climate Change Reports, 1: 49–59. doi: 10.1007/s40641-015-0009-3

Schumacher, R.S. and Johnson, R.H. (2006) Characteristics of US extreme rain events during 1999–2003. Weather and Forecasting, 21: 69–85. doi: 10.1175/WAF900.1

Scoccimarro, E., Gualdi, S., Villarini, G., Vecchi, G.A., Zhao, M., Walsh, K. and Navarra, A. (2014) Intense precipitation events associated with landfalling tropical cyclones in response to a warmer climate and increased CO2 Journal of Climate, 27: 4642–54. doi: 10.1175/JCLI-D-14-00065.1

Sebastian, A., Dupuits, E.J.C. and Morales-Na ́poles, O. (2017) Applying a Bayesian network based on Gaussian copulas to model the hydraulic boundary conditions for hurricane flood risk analysis in a coastal watershed. Coastal Engineering, 125: 42–50. doi: 10.1016/j.coastaleng.2017.03.008

Sterl, A., Severijns, C., Dijkstra, H., Hazeleger, W., van Oldenborgh, G.J., van den Broeke, M., Burgers, G., van den Hurk, B., van Leeuwen, P.J. and van Velthoven, P. (2008) When can we expect extremely high surface temperatures? Geophysical Research Letters, 35: L14703. doi: 10.1029/2008GL034071

Torres, J., Bass, B., Irza, N., Fang, Z., Proft, J., Dawson, C., Kiani, M. and Bedient, P. (2015) Characterizing the interactions of hurricane storm surge and rainfall-runoff for the Houston-Galveston region. Coastal Engineering, 106: 7–19. doi: 10.1016/j.coastaleng.2015.09.004

van der Wiel, K., Kapnick, S.B., Vecchi, G.A., Cooke, W.F., Delworth, T.L., Jia, L., Murakami, H., Underwood, S. and Zeng, F. (2016) The resolution dependence of contiguous US precipitation extremes in response to CO2 forcing. Journal of Climate, 29: 7991–8012. doi: 10.1175/JCLI-D-16-0307.1

van der Wiel, K., Kapnick, S.B., van Oldenborgh, G.J., Whan, K., Philip, S., Vecchi, G.A., Singh, R.K., Arrighi, J. and Cullen, H. (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

Vecchi, G.A. and Knutson, T.R. (2011) Estimating annual numbers of Atlantic hurricanes missing from the HURDAT database (1878–1965) using ship track density. Journal of Climate, 24: 1736–46. doi: 10.1175/2010JCLI3810.1

Villarini, G., Lavers, D.A., Scoccimarro, E., Zhao, M., Wehner, M.F., Vecchi, G.A., Knutson, T.R. and Reed, K.A. (2014) Sensitivity of tropical cyclone rainfall to idealized global-scale forcings. Journal of Climate, 27: 4622–41. doi: 10.1175/JCLI-D-13-00780.1

Westra, S., Alexander, L.V. and Zwiers, F.W. (2012) Global increasing trends in annual maximum daily precipitation. Journal of Climate, 26: 3904 –18. doi: 10.1175/JCLI-D-12-00502.1

Averaging the monthly mean temperatures over the CONUS, 2017 was found to have the second warmest February on record, behind that of 1954, with a temperature of 4.9 °C (41 °F; Figure 2, based on preliminary data for 2017 from the ECMWF analysis). This was 3.1°C (5.6°F) warmer than the 1981–2010 normal. Averaged over the area east of the Rocky Mountains (approximated by 105° W) it was the warmest February on record. In contrast, in California February temperatures were only just above normal in 2017 after three very warm years. Farther north, in Washington State, February temperatures were 2.5 ºC (4.5 ºF) degrees below normal at −0.8 °C (30.6 ºF).

February CONUS temperatures exhibit a warming tendency over the last 120 years (Figure 2). Two questions tied to this apparent warming are: i) is it distinct from changes one would expect from natural weather fluctuations (“detection”)? and ii) if so, can the causes behind the trend be ascertained (“attribution”)? A combined analysis of the historical observations and climate models can help answer these questions.

The analysis of observations provides us with an estimate of how rare the event was, and how its likelihood has changed since the 1900s, a time for which we have good temperature observations but global warming had only just started. The causes of the observed changes can be assessed with climate model experiments with which one can isolate the effects of external factors known to influence the climate: human-caused greenhouse gas and aerosol emissions, and also solar variability and volcanic eruptions.

Trend in the observations

We use the NOAA/NCEI series of CONUS temperatures derived from many thousands of weather stations (Lawrimore et al, 2011). Earlier research has shown that this series is not affected materially by urbanization and siting (Menne et al, 2010, Williams et al 20212); using only rural, well-sited stations actually gives slightly higher trends than the full database. Similar results are found if the analysis is repeated using other surface temperature data (e.g., NASA-GISTEMP, Berkeley Earth and CRU datasets). These temperature anomalies occur mainly in the lowest kilometer of the atmosphere, with satellite observations of the lower troposphere showing anomalies that are about half the magnitude of the near-surface observations. Besides this well-known difference with surface observations, the satellite record is also too short to robustly determine trends in February CONUS temperatures or to assess changes since 1900. We have therefore not included it in this analysis.

The upper tail of February CONUS temperatures is described well by a Normal distribution. We assume that it varies to first order proportional to the global mean temperature, GMST (smoothed), but allow a slower or quicker pace (this method was first used in van Oldenborgh, 2007 and is described fully in van Oldenborgh et al, 2015). The remaining variability is due to the random weather, with virtually no low-frequency variability left over. El Niño – Southern Oscillation (ENSO), the Pacific Decadal Oscillation (PDO) and  Atlantic Decadal Variability (AMV) do not influence the CONUS temperature in February, simplifying the analysis.

This fit, excluding 2017, shows that the CONUS temperature in February has risen a bit faster than the GMST, with a 95% uncertainty range between 0.8 and 2.7 times the global mean temperature. The return period of the 2017 temperature is about 12 years, with an uncertainty range of 5 to 40 years. This means that the chance of getting a temperature this high or higher is around 8% each year, with 95% certainty within the range 2% to 30%. Thus, in the current climate this was not really an exceptional event any more.

Around 1900 the return period would have been about 160 years, i.e., a probability of 0.6%, with an uncertainty from 0.05% to 1.4% each year. Thus, on the basis of the observed trend we find that the probability has increased by roughly a factor of 13. We have 95% confidence that the increase in likelihood of events like February 2017 is at least a factor of four. However, it is important to note that this increase isn’t all attributable to climate change. For this attribution we rely on climate models.

A similar fit can also give the return period of the 1954 temperature. At that time it was a very unusual situation with a return period of roughly 1 in 200 year (about 0.5% chance). That also means, however, that the odds of finding such an event in 120 years of data are about 50-50, just like throwing a 6 in three throws of dice. The weather maps also confirm that February 1954 had much more unusual weather than February 2017, while the circulation for this year is not as unusual, adding support to the fact that the heating trend has made high temperatures like last month more commonplace than they were in the 1950s.

The same analysis for Washington State, starting in 1900, gives a return period of a February temperature this cold of about once every 30 years, 3% each year. Due to a strong trend, two times the global mean, around 1900 it would have been a very normal year occurring every four years or so.

Attribution

The next question is what caused the heating trend in the February CONUS observations over the last century. This question cannot be directly assessed using the observational record, with which we can only assess correlation and not causality,  and must be assessed using climate models, in which the various external “forcings” (such as changes in insolation, volcanoes, greenhouse gas concentrations, etc.) can be controlled and their impact understood. We use a collection of coordinated experiments from climate models run at centers across the world under the “5th Coupled Model Intercomparison Project”, or CMIP5, which were assessed in the  IPCC Fifth Assessment Report, the data for which are publicly available. In addition to CMIP5 experiments, two ensembles of atmosphere-only models are used: the UK Met Office HadGEM3-A model at N219 (60km) and the very large ensemble of Weather@Home runs of HadAM3P simulations at N96. Fifteen of the CMIP5 models and the two SST-forced models have been run at least three times with and without human-caused emissions of greenhouse gases and aerosols, allowing us to isolate the effect of the influence of greenhouse gas changes on the likelihood of February warm spells within the climate models.

For the CMIP5 analysis, we first checked which of the 15 models, with the required simulations for the analysis, have a CONUS temperature distribution that is compatible with the observed one, after a constant bias correction, similar to Lewis and Karoly 2013; King et al. 2015. Two models did not pass this test and were removed from the analysis. For the remaining models we counted how many times a temperature anomaly of +3.5 ºC (+6.3ºF) above 1961-1990 was exceeded. This threshold includes the 1954 and 2017 events. We did this three times: in 2007–2027 in the RCP8.5 scenario to estimate the return interval in the climate representative of 2017, in 1901–2005 from the the natural world model experiments (i.e., those without human-caused greenhouse gas and aerosol emissions; NAT), and in a future world around 2050 (2040-2060) assuming the RCP8.5 scenario, which represents continued intensive use of fossil fuels.

Historical and projected increases in greenhouse gases lead to increased incidence of warm CONUS Februaries, and decreased incidence of cold CONUS Februaries. The CMIP5 climate model simulations (Figure 4) show an 18-year-return period (95% confidence interval between 12-37 years) for Februaries over the CONUS as warm or warmer than 2017, similar to the observational estimate. Without the historical human-induced increases in greenhouse gases, the “best estimate” of the return period in the models is around 60 years (95% CI between 27-184 years), or a tripling of its probability — the lower bound on the 95% range of the probability change is 1.8x now from past greenhouse gas increases. The model estimate of the greenhouse-driven increase in warm February probability is within the uncertainty range of the observed increase in probability, though  on the lower range of the observational estimates. Around 2050 temperatures like this are projected to be completely normal, occurring approximately every three years on average (95% CI between 2.6-4.5 years).

Of note is that even in the 2050 ensemble, the probability of “cold” winters over the CONUS remains: natural fluctuations in weather and climate are expected to continue generating cool conditions, though at a reduced rate. That is, long-term warming and increased incidence of warm events does not obviate cool events – it just reduces their probability.

The 15 HadGEM3-A historical runs 1960–2015 (Christidis et al, 2013) reproduce the variability and trend of the February CONUS temperatures well, though the model has a systematic 2.3°C cold error in its temperatures, which is adjusted in the analysis,. The simulations with only natural forcings (solar variability and volcanic eruptions) show no trend, showing that the whole trend in the historical simulations from 1960, about 2.1 times the global mean temperature (consistent with the relation estimated in observations), is due to  emissions. The return time of the temperature observed in February 2017 is about 11 years, in agreement with the observed value. In 1900 this would have been around 200 years. The probability of a temperature like the one observed in 2017 or higher has thus increased by a factor of about 20 in this model, with a lower bound of 13 (95% CI, one-sided).

We analyzed 4898 Weather@Home simulations of February 2017 in current climate conditions forced by seasonal forecast SSTs (Haustein et al, 2016) compared to 3710 possible February temperatures with the anthropogenic signal from atmosphere and SSTs removed. The difference between these two large ensembles reveals a smaller but significant increase in the likelihood of warm Februaries occurring. What is now a 1 in 12 year event would have been a 1 in 22 year event in the world that might have been (risk ratio 1.9 with 95% confidence larger than 1.6) while what would have been a 1 in 160 year event  in the world that might have been would now be a 1 in 62 year event (risk ratio 2.6 with 95% confidence larger than 1.8).

These modeling and observational results are not very surprising given the broad observational, theoretical and modeling background on the connection between increasing greenhouse gases, global and regional temperature changes, and temperature extremes. The character of mechanisms behind the observed rise of the global mean temperature has been studied extensively in the scientific literature, and has been assessed in the Fifth IPCC Assessment Report (AR5, Chapter 10, Bindoff et al, 2013), leading to the conclusion that ‘it is extremely likely that human influence has been the dominant cause of the observed warming since the mid-20th century’. For most of the world, local enhanced likelihood of  warm spells can already be related to this global change, including warming over the contiguous United States in winter.

Synthesis

Figure 6 summarises all the results on the change in probability from the observed trend (blue) and the three model ensembles (red). All results are compatible with a common trend plus natural variability (χ²/dof = 0.9). We therefore computed a weighted average of all results, this is shown in purple: the probability has increased by a factor more than three.

Conclusions

Results from global climate models are inconsistent with the hypothesis that the increase in odds of warm Februaries was caused by natural forcing agents such as solar activity, which has declined since the 1960s, and volcanic eruptions. Meanwhile the model results indicate that past historical increases in greenhouse gases have raised the odds of warm Februaries in the CONUS considerably. The observed trend is compatible with the effects of human-induced emissions of greenhouse gases. Since past and projected future greenhouse gas increases will continue to raise the temperatures, the frequency of winter months like February 2017 should be expected to increase over the coming decades.

Overall, we find that the chances of seeing a February as warm as the one experienced across the Lower 48 has increased more than threefold because of human-caused climate change. The record-warm February of 1954 was, at the time, a very rare event (probability about 0.5% per year) but similar events should now be expected every few years.

References

Bindoff, N.L., P.A. Stott, K.M. AchutaRao, M.R. Allen, N. Gillett, D. Gutzler, K. Hansingo, G. Hegerl, Y. Hu, S. Jain, I.I. Mokhov, J. Overland, J. Perlwitz, R. Sebbari and X. Zhang, 2013. Detection and attribution of climate change: from global to regional supplementary material (pdf, 10.1 MB). In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)].

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

Haustein, K., F. E. L. Otto, P. Uhe, N. Schaller, M. R. Allen, L. Hermanson. N. Christidis, P. McLean and H. Cullen. (2016) Real-time extreme weather event attribution with forecast seasonal SSTs. Environmental Research Letters, 11: 064006. doi: 10.1088/1748-9326/11/6/064006

King, A., G. J. van Oldenborgh, D. J. Karoly, S. C. Lewis and H. Cullen. (2015) Attribution of the record high Central England temperature of 2014 to anthropogenic influences. Environmental Research Letters, 10: 054002. doi: 10.1088/1748-9326/10/5/054002

Lawrimore, J. H., M. J. Menne, B. E. Gleason, C. N. Williams, D. B. Wuertz, R. S. Vose, and J. Rennie. (2011) An overview of the global historical climatology network monthly mean temperature data set, version 3. Journal of Geophysical Research, 116, D19121. doi: 10.1029/2011JD016187.

Lewis, S. C., and D. J. Karoly. (2013) Anthropogenic contributions to Australia’s record summer temperatures of 2013, Geophysical Research Letters, 40: 3705–3709, doi: 10.1002/grl.50673

Menne, M. J., C. N. Williams Jr., and M. A. Palecki. (2010) On the reliability of the U.S. surface temperature record. Journal of Geophysical Research, 115, D11108. doi: 10.1029/2009JD013094

van Oldenborgh, G. J. (2007) How unusual was autumn 2006 in Europe? Climate of the Past, 3: 659-668, doi: 10.5194/cp-3-659-2007

van Oldenborgh, G. J., R. Haarsma, H. de Vries and M. R. Allen. (2015) Cold extremes in North America vs. mild weather in Europe: the winter of 2013–14 in the context of a warming world. Bulletin of the American Meteorological Society, 96: 707–714. doi: 10.1175/BAMS-D-14-00036.1

Williams, C. N., M. J. Menne, and P. W. Thorne. (2012) Benchmarking the performance of pairwise homogenization of surface temperatures in the United States. Journal of Geophysical Research, 117, D05116. doi: 10.1029/2011JD016761