Environmental Science Extreme Weather and Climate
Friederike Otto
  • LAST MODIFIED: 24 May 2017
  • DOI: 10.1093/obo/9780199363445-0067


Natural disasters and extreme weather events have been of great societal importance throughout history and often brought everyday life to a catastrophic halt, in a way sometimes comparable to wars and epidemics, only without the lead time. Extreme weather events with large impacts serve as an anchor point of the collective memory of the population in the affected area. Every northern German of the right age remembers the storm surge of 1962 and where they were at the time and has friends or family effected by the event. The “dust bowl” of the 1930s with extensive droughts and heat waves shaped the life of a generation in the United States, and the Sahel droughts in the 1960s and 1970s led to famine and dislocation of population on a massive scale the region arguably never quite recovered from. Hurricane Hyian in 2013 is said to have directly influenced the outcome of the annual Conference of the Parties (COP) United Nation Framework Convention for Climate Change Negotiations in Warsaw, leading to the inclusion of a mechanism to deal with loss and damage from climate-related disasters. Though earthquakes are still fairly unpredictable on short timescales, this is not the case for weather events. Weather forecasts today are so good that we normally know the time and location of the landfall of a hurricane within a 100-mile radius days in advance. Improvements in the prediction of slow-onset events such as droughts (which depend on the rainfall over a large region and whole season) are less striking but have still improved dramatically in the late 20th and early 21st centuries. One of the major reasons for the large increase in the accuracy of weather forecasts is the exponential increase in computing power, which allows scientists to predict and study extreme weather events using complex computer models, simulating possible weather events under certain conditions to understand the statistics of and physical mechanisms behind extreme events. Extreme events are by definition rare and thus impossible to understand from historical records of weather observation alone. Despite the progress on our understanding of and ability to predict extreme weather events, substantial uncertainties remain. Two aspects are of particular importance. Firstly, we know that the climate is changing, having observed almost a one-degree increase in global mean temperature. However, global mean temperature doesn’t kill anyone, extreme weather events do. Their frequency and intensity is changing and will continue to change, but the extent of these changes depends on a host of both global and local factors. Secondly, whether or not a rare weather event leads to extreme impacts depends largely on the vulnerability and exposure of the affected societies. If these are high, even a perfectly forecasted weather event leads to disaster.

General Overviews

The global scientific community brought together under the Intergovernmental Panel on Climate Change (IPCC) has published five assessment reports since its formation in the early 1990s and with our increased understanding of extreme events their trends in frequency and magnitude are included in different chapters in the fifth assessment report (AR5), in both, Stocker, et al. 2014 and its assessment of the physical science basis and Field, et al. 2014 on the impacts of climate change. The two working group assessments particularly report on the most certain findings on observed trends in extreme weather events and their impacts with Field, et al. 2014, providing a succinct overview of key terms such as hazard, vulnerability, and exposure. It is important to put 21st-century understanding of extreme weather events into the context of the general state of knowledge about the climate system and its impacts on society while providing essential background information on methodologies of climate research. Yet the assessment reports do not provide a general overview of the 21st-century state of knowledge on extreme weather events. Given that at least in the short term the largest impacts and highest damages of human-induced climate change will be through extreme weather events, the IPCC saw this shortcoming and published Field, et al. 2012. The SREX report gives an overview of observations and predictions of extreme weather events and their impacts and highlights particularly the social dimension of extreme events. The latter are notably relevant when defining extreme events and trying to identify comprehensive indices (Zhang, et al. 2011). While SREX provides the best overview of the state of the science of extreme events, this subfield of climate science has seen an enormous increase in research activity and has also been recognized by the World Climate Research Programme (WCRP) as a field of science requiring a considered scientific effort and with major breakthroughs expected in the next few years; thus the WCRP asked a team of leading scientists in the field under the Grand Challenge on Extreme Events to coordinate the worldwide research activity to accelerate and enable breakthroughs (WCRP Grand Challenge). (The website does not provide an overview of the impacts of extreme events as the science of understanding these events is only beginning to develop.) A particular development with respect to putting extreme weather events in a climate context is the emerging of the science of extreme event attribution reviewed by the American National Academy of Science (NAS) and dedicated to answering the question whether and to what extent anthropogenic climate change has played a role in recent events. In the following section a brief overview of publications on these impacts is given; however, the rest of the article is focused on extreme weather events from a climatological point of view.

  • Field, C. B., V. R. Barros, D. J. Dokken, et al., eds. 2014. Climate change 2014: Impacts, adaptation, and vulnerability. Part A: Global and sectoral aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge, UK: Cambridge Univ. Press.

    Save Citation »Export Citation »E-mail Citation »

    IPCC WG2 report on the impacts of climate change provides an overview of impacts, including from extreme weather.

    Find this resource:

  • Field, C. B., V. Barros, T. F. Stocker, et al., eds. 2012. Managing the risks of extreme events and disasters to advance climate change adaptation. A special report of working groups I and II of the Intergovernmental Panel on Climate Change. Cambridge, UK, and New York: Cambridge Univ. Press.

    Save Citation »Export Citation »E-mail Citation »

    Offers a good overview of research on extreme weather events and their impacts.

    Find this resource:

  • NAS National Academies of Sciences, Engineering, and Medicine. 2016. Attribution of extreme weather events in the context of climate change. Washington, DC: National Academies Press.

    Save Citation »Export Citation »E-mail Citation »

    An overview of the science of extreme event attribution that explains the science and highlights the current state of knowledge depending on the type of extreme event (e.g., heat, drought, flood).

    Find this resource:

  • Stocker, T. F., D. Qin, G.-K. Plattner, et al., eds. 2014. Climate change 2013: The physical science basis. Contribution of working group I to the fifth assessment report of the Intergovernmental Panel on Climate Change. Cambridge, UK: Cambridge Univ. Press.

    Save Citation »Export Citation »E-mail Citation »

    IPCC WG1 report on the physical science basis provides an overview in particular on the methodologies used in climate science and extreme event research as well as the latest findings.

    Find this resource:

  • WCRP Grand Challenge.

    Save Citation »Export Citation »E-mail Citation »

    Website from the WCRP grand challenge on extreme events that provides a brief overview of some key aspects of extreme events and a list of relevant papers and reports on the topic.

    Find this resource:

  • Zhang, X., L. Alexander, G. C. Hegerl, et al. 2011. Indices for monitoring changes in extremes based on daily temperature and precipitation data. WIREs Climate Change 2:851–870.

    DOI: 10.1002/wcc.147Save Citation »Export Citation »E-mail Citation »

    A review on indices used to define extreme events.

    Find this resource:

Societal Relevance of Extreme Weather Events

While a hurricane that never makes landfall is still interesting from a meteorological point of view, it is more a question of location, timing, and intensity of the hurricane’s landfall that makes the difference between comparably small effects and extreme impacts. Similarly, while extreme rainfall is what precedes flooding events, it is the layout and management of river catchments, the saturation of the soil, and other factors influencing vulnerability and exposure that causes widespread flooding. Providing illustrative examples of different case studies of extreme events but also other natural disasters such as a cholera epidemic, Field, et al. 2012 underlines the fact that the definition of an extreme event is not at all straightforward and necessarily subjective. Whether or not a heat wave is best defined by, for example, the number of days above a temperature threshold, or the heat stress on the human body over a fixed period, entirely depends on context. Meteorological indices are thus only indicative at best for a local risk assessment. Mitchell and van Aalst 2011 highlights that in some regions there is better information locally available to inform adaptation and resilience, but in many others the uncertainties are extremely high. More research in particular is needed to understand the interplay between extreme weather and other factors defining the risk of extreme impacts, such as population increase and insurance. Pelling 2011 provides an overview and examples of measures that could be taken to manage these risks effectively. What we know about how the frequency and intensity of extreme weather events might change in the future not only depends on local factors but also on the type of event in question. While we can say that on average the number and intensity of heat waves has already increased globally, and in most locations we also see a regional increase in hot days and nights, a comprehensive understanding of future hurricane risks is lacking thus far in the early 21st century. The immediate impacts of extreme weather events on society are large and are quantified in the EM-DAT. There are, however, many remote and indirect ways extreme weather events impact society (impacts on ecosystems, food systems, prices etc.), so that a thorough understanding of the underlying ecosystem or economy is necessary to follow the literature in more depth than Field, et al. 2012. The worst damage from climate change is likely caused by extreme weather events, at least in the short term; hence, extreme weather events play an important role when discussing loss and damage from climate change in a political context as introduced with the UNFCCC or in a legal context of liability (Lord, et al. 2012). While being a comparably specific aspect of the societal impacts of extreme events, the discussion around loss and damage highlights the difficulties in disentangling the hazard from other aspects of risk (Huggel, et al. 2013).


Most assessments of extreme weather events focus on temperature and precipitation, as these are the kinds of events we have observed in the long term. However, in a review on updates since the SREX report, Alexander 2016 highlighted that even for these variables, major gaps in data availability and quality exist. In addition to adding more details on what is known about observed changes, the review article provides a brief overview of the history of extreme weather observations. The review is part of Hay, et al. 2016, a special issue on observed and projected changes in weather and climate extremes that gives a brief tour through the IPCC assessment reports and how they report on extreme events. As is shown in Societal Relevance of Extreme Weather Events, it is difficult to define extreme events; depending on specific vulnerabilities, different definitions might be appropriate. At the same time it is desirable to be able to compare findings of different assessments. To achieve this, the ETCCDI came together to identify a list of indices and a common framework to calculate these indices that would be useful in several circumstances. The list incorporates twenty-seven indices, some rather specific, while measures such as TXx (maximum value of daily temperature in a year or month) are now calculated in most studies on extreme events. A comprehensive example of an application of these indices for extreme event predictions is Orlowsky and Seneviratne 2012. What is true for all climate data, from satellite observations to in situ measurements to climate and weather models, also applies for the analysis of extreme weather events. The quality of a data set crucially depends on the location and spatial extent of the event studied. Data sets, modeled and observed, are better in some regions of the world than others with the general rule that better-observed regions are also better represented in climate models. However, this can also be an effect of model tuning and does not necessarily represent a better understanding of processes; models can be “right for the wrong reasons” (Collins, 2007). Therefore, a date set or climate model always needs to be evaluated with respect to the particular class of extreme events of interest before use. Most publicly available data sets and many model simulations are available through the Climate Explorer, a website created and updated at the Royal Netherlands Meteorological Institute (KNMI). The list of data sets presented is by no means exhaustive, and in particular regionally many more data sets are available. The quality is extremely variable, and given the particular importance of homogeneity for the assessment of extreme events, careful quality control is required.


Three main types of climate and weather observations and pseudo-observations exist. All three types have their advantages and disadvantages. The traditional observational data sets are based on in situ measurements from weather stations. A globally available station data set for rainfall and temperature is GHCN-D (Menne, et al. 2012). In regions of the world with a high density of weather stations station data is often the best quality data in particular with respect to precipitation extremes as the high spatial variability in rainfall is represented in this data. Most countries in the world have their own weather station network and the met services of these countries rarely allow this data to be used for free, even for research. To allow an estimation of the frequency and intensity of extreme weather events a data set comprising extreme event indices rather than raw data has been designed, HadEX (Donat, et al. 2013) that provides a good global overview. The resolution is, however, so coarse that it does not allow for analysis in smaller regions. Since the late 1970s satellite measurements are therefore used as an alternative way to measure temperature and estimate rainfall. Funk, et al. 2015 is an example of a data set using satellite measurements and rain gauge data to provide data to analyze precipitation extremes. While having a better spatial coverage than station data satellites do often miss short but intense rainfall events due to the fact that they only measure the instantaneous rainfall a couple of times a day thus they are merged with station data in most cases. Note that these are usually a gridded data set that have smaller (precipitation) extremes and reduced variability due to the averaging over a grid box, in comparison with point data from observing stations. The third category of observational data products is reanalysis data sets. The bases of reanalysis models are numerical weather prediction models, fed with observed data from stations and satellites. After one time step of integration the output data are compared to observed variables and adjusted accordingly. A reanalysis model is a climate model, thus most of the general problems apply here as well (Parker 2016). At least they entail systematic errors from the computational setup that add up to biased assessments in particular in poorly observed regions. Reanalysis data nevertheless is as useful addition to observed data: it is available for the same spatial extent as climate models and thus invaluable to evaluate climate models. And it is particularly important to estimate extreme events that are by definition rare and thus can only be assessed from reanalysis data. Reanalysis is also the only source of data to analyze aspects of the atmospheric circulation. The three most commonly used data sets are ERA-interim (Dee, et al. 2011), NCEP (Kalnay, et al. 1996), and 20CR (Compo, et al. 2011); they differ in what variables are used as input and also in the spatial resolution and length of available data.


Research on extreme events would be limited without the use of computer models; even data collection relies on reanalysis and interpolation models. Edwards 2001 is a thorough introduction on climate modeling. A spectrum of different types of climate models exists, ranging from simple one-equation energy balance models to numerical weather prediction models and General Circulation Models (GCMs) (Randall 2000). The latter are the only type of models able to represent extreme weather as they simulate atmospheric circulation. Model data from the state-of-the-art general circulation models from modeling centers around the world are available in Taylor, et al. 2012. While the resolution differs, these models are identical to numerical weather forecasting models, and modern modeling systems are developed to allow a seamless interaction between resolutions (Hurrell, et al. 2009). GCMs simulate the climate system but are not a true representation and have biases (Flato, et al. 2013). While they might be perfectly adequate to address some questions about the climate system, others cannot be addressed (Oreskes, et al. 1994; Knutti 2008). For some extreme weather events such as thunderstorms, we know a priori that coarse resolution GCMs are unable to simulate them as they occur on spatial scales of the order of few kilometers. Only regional climate models can be run at such a resolution, and few simulations are affordable. To study extreme events, ensembles of GCMs (a set of model simulations spanning the same time slice) are necessary to analyze the statistics of rare events (Sippel, et al. 2015). There are three types of GCM ensembles: multimodel ensembles consisting of different GCMs as in CMIP5, perturbed parameter ensembles where the parameters representing atmospheric processes on small scales are varied in a single model; and initial condition ensembles where only the initial values of the starting conditions are varied in a single model. While multimodel ensembles generally span a larger range of uncertainties, and thus arguably give a better representation of the possible range, the fact that many models are partially identical gives some models more weight than others. Thus drawing conclusions from a multimodel mean can be misleading (Knutti, et al. 2013). In terms of model biases no type of ensemble is outstanding (Collins, et al. 2010), and model evaluation and bias correction need to be done on a case-by-case basis (Mueller and Seneviratne 2014). Assessments of model biases do not only stem from statistical assessments but also from understanding physical processes driving extreme weather events and whether the models in question reliably represent these processes. Farneti 2017 shows how process understanding and sensitivity studies, in combination with using a hierarchy of models and model intercomparison studies, can greatly improve our ability to model extreme events on decadal time scales.

Understanding Extreme Events in a Climate Context

Extreme weather and climate-related events are by definition rare and always have multiple causes. To understand extreme events thus means to understand the relative importance and interaction of the large- and local-scale processes that characterize extreme events. Held 2005 observes that our ability to simulate the climate system does not correspond to our degree of understanding of it, and this is even more true for extreme weather. Ultimately every extreme event results from a unique combination of external drivers, the land surface conditions and internal climate variability. Large-scale climate conditions such as the El Nino Southern Oscillation (ENSO), however, have a significant influence on the likelihood of a certain type of extreme event to occur in a particular region of the world. In some cases these links are strong (Lyon 2004) while in other cases they are weaker but play a key role in understanding extreme weather (Dong, et al. 2013). Understanding these links allows for better predictions on seasonal as well as longer timescales (Bladé, et al. 2012). Our understanding of the influence of the large- scale circulation can be increased by combining observed records and modeling studies. In particular with the relatively new ability to simulate large ensembles of and conditioning on the large-scale circulation or sea surface temperatures (Mitchell, et al. 2016) a quantification the impact on extreme weather is possible. Whether or not rare meteorological conditions develop in extreme weather events depends to a large degree on the soil conditions. Seneviratne, et al. 2010 gives an overview of the key mechanisms and the potential role of soil moisture in the future. Due to the lack of long-term and large-scale observations of soil moisture many studies on the land surface atmosphere feedback take the form of model-based sensitivity analyses. Given their often large impacts, a better understanding of extreme events and in particular their predictable components is important: scrutinizing a particular event in great detail (Hoerling, et al. 2013) allows for a better understanding of similar events. Overall, our general understanding of extreme weather events and in particular our ability to attribute extreme events to external drivers strongly varies depending on the type of event and is generally higher for directly temperature related events such as heat waves and decreases with decreasing scales of the event and increasing complexity of interacting processes. Hurricanes (Emanuel 2003) are an example of the latter type of extremes, which are of particular importance due to their societal impact.

Defining Extreme Events

The majority of publications on extreme events study temperature and heavy precipitation extremes because most observational records consist of temperature and rainfall recordings alone; thus, our understanding and ability to simulate and attribute these types of extreme events is better than ever. This does not mean, however, that these extremes are also the most relevant in terms of impacts and damages; it also does not mean that there are no open questions with respect to these extremes. For example, measuring heat waves is not a straightforward process, and although suggestions are made to find indices of wide-ranging applicability (Perkins and Alexander 2013), ultimately the most useful definition of any extreme weather event will be defined by the vulnerabilities of the society affected. Schär 2015 published a short analysis of future heat-wave risk and not temperature itself but heat stress on the human body is used as the heat-wave measure, thus identifying regions where being outdoors in the summer might be life threatening in the future. Sippel and Otto 2014 shows that defining heat waves in terms of temperature or heat stress leads to different assessments of whether or not the risk of heat waves occurring has changed. Consequently, different definitions of an extreme weather event can lead to substantially different estimates of current trends and future risks. An example of such an event is the multiannual drought in California where anthropogenic climate change has been reported to play an important role by Diffenbaugh, et al. 2015, while Seager, et al. 2015 finds no substantial increase in risk. These examples underline that there is no right or wrong definition of an extreme event but that different definitions are appropriate for different research questions. Hence, it is always essential to be aware of the definition used when collecting scientific evidence from the published literature and assess whether a study really addresses the problem at hand.

Attributing Extreme Events to External Drivers

While the detection of trends in time series of climate variables and their attribution to external drivers has been pursued since the late 20th century (Hasselmann 1997) the attribution of extreme weather events is a new subfield in climate science. The methodology has been introduced in Allen 2003 and implemented for the first time to take account of the European heat wave in 2003 (Stott, et al. 2004). Otto 2016 reviews the short historic development of the field following the Russian heat wave of 2010 and other major extreme events in the first decade of the 21st century (Coumou and Rahmstorf 2012). Since 2012 the Bulletin of the American Meteorological Society (BAMS) publishes and annual special issue on the attribution of extreme events in the previous year (Herring, et al. 2016). In this short timeframe two schools of thought have emerged. Easterling, et al. 2016, referred to as the “risk-based” or “Oxford” method, asks whether and to what extent the overall risk of an extreme event occurring has changed, while the storyline or Boulder approach (Shepherd 2016) aims at understanding the trajectory of the extreme event developing in the presence of an external driver. Stott, et al. 2016 provides an overview of the different methodologies applied following the risk based approach while Trenberth, et al. 2015 argues against applying this approach in light of high-modeling uncertainty but neglecting the fact that most stakeholders are interested in the overall risk. Due to the high interest of the public in the scientific development and the rather heated scientific debate around the two different approaches, the National Academies of Sciences, Engineering and Medicine (NAS) commissioned a report on extreme event attribution that describes the science, highlights the importance of clear communication, and recommends a multimethod approach. Subsequent publications, in particular Herring, et al. 2016, have heeded some of these recommendations and highlight that the Boulder approach, in combination with the risk-based approach, allows for a better understanding of the extreme event that is subsequently being attributed.

Predicting Extreme Events on Seasonal and Decadal Timescales

The skill of weather forecasting on timescales of up to a week has constantly increased since the mid-20th century thanks to a better understanding of process, computing power, and improved observations. Predictions on seasonal timescales have been developed and recently increased noticeably in usefulness with the introduction of probabilistic forecasts. NASEM in particular allows for the skilled prediction of extreme events on these timescales; prior, large-scale, and long-term assessments of frequency and magnitude of extreme events could not be made. With increasing global mean temperatures, experts expect increased frequency and magnitude of some types of extremes, which is due to the larger water-holding capacity of a warmer atmosphere. In a milestone publication on the prediction of future rainfall, Allen and Ingram 2002 found, however, that the increase is lower than projected in earlier publications on basic thermodynamic assumptions. Regardless of the exact rate, an increase in mean precipitation and temperature leads to an increase in distribution and thus to an increase in heat and rainfall extremes. Fischer and Knutti 2015 quantified the expected increase in heat and rainfall extremes. These findings, however, are only based on the global average, and regional trends can weaken or gain strength. The Held and Soden 2006 so-called rich-get-richer paradigm describes a general trend of wet areas getting more rainfall while dry areas get drier; however, Allan 2014 questions whether this applies to the entire globe, or particularly to land, where it matters more for people. Whether or not this paradigm holds, local soil-moisture conditions and regional circulation processes need to be taken into account to make place-specific predictions. Herring, et al. 2016 (see Attributing Extreme Events to External Drivers) provides examples of extreme events that change in unexpected ways. In addition to long-term predictions necessary to build resilient societies, the ability to predict extreme weather a few months in advance is crucial to make decisions about, for example, which crop to plant, or whether to prepare for droughts or floods. These so-called seasonal predictions are particularly important in parts of the world with high interannual variability in the climate. These areas are often identical to regions where the weather is partly dominated by large-scale teleconnection systems such as the El Niño Southern Oscillation (ENSO). The presence of a strong El Niño event in turn then also leads to high predictability of temperature and precipitation extreme in these areas. Weisheimer and Palmer 2014 overviews the reliability of temperature and precipitation extremes in different parts of the world, finding in general that dry extremes are more predictable than wet ones, and seasonal predictions of temperature extremes are on average more skillful. In regions with high day-to-day variability in the weather, such as western Europe, seasonal predictions are largely unreliable unless the preceding season was extremely wet or dry, which increases or decreases the likelihood for heat extremes in the summer (Quesada, et al. 2012).

Importance and Implications of Understanding, Attributing, and Predicting Extreme Weather

Extreme weather events around the world have been associated with poor harvests, water shortages, and forced migration in communities struck by floods, droughts, and hurricanes (UNSDIR). Media attention around such events has always been high, but in a changing climate with clear impacts on extreme weather events discussed in Predicting Extreme Events on Seasonal and Decadal Timescales and Attributing Extreme Events to External Drivers speculation about the role of climate change in driving the event as well as the impacts added a new component in the media assessment. Concurrently, national governments such as the UK Department for International Development (DFID) as well as the UN in the Sendai framework for disaster risk reduction are demanding a stronger scientific evidence base for their decision making around climate change, resilience, and disaster risk reduction. Clearly, extreme events are not caused by climate change, but they are expected to change in their frequency and magnitude. Yet understanding its influence on specific events is more complex and needs to be done on a case-by-case basis. While it is now possible to assess whether climate change has altered the probability of occurrence of specific extreme events (see Attributing Extreme Events to External Drivers), scientific studies have been limited to a small number of disasters, and in some cases the scientific community does not yet have the understanding and adequate tools to investigate the role of human influence today and in the future, especially in data-poor regions. However, confusion in the understanding of changes in extreme events is problematic, potentially influencing public perceptions of climate change (Taylor, et al. 2014) and introducing misunderstanding in policy debates. Policy designed to address the impacts of climate change requires an understanding of what the impacts of climate change have been and might be in the future and how these impacts translate to human lives and livelihoods. In order to achieve this, there is a need for better communication between scientists and policymakers about existing scientific evidence, and for more scientific research that addresses both the attribution of weather events to anthropogenic emissions and the attribution of the impacts of these weather events to be linked to attributing those to external drivers.

back to top