The Oxford Research Encyclopedia of Climate Science will be available via subscription on April 26. Visit About to learn more, meet the editorial board, or recommend to your librarian.

Show Summary Details

Page of

PRINTED FROM the OXFORD RESEARCH ENCYCLOPEDIA, CLIMATE SCIENCE ( (c) Oxford University Press USA, 2016. All Rights Reserved. Personal use only; commercial use is strictly prohibited (for details see Privacy Policy and Legal Notice).

date: 23 April 2018

Constructing Records of Storminess

Summary and Keywords

Storms are characterized by high wind speeds; often large precipitation amounts in the form of rain, freezing rain, or snow; and thunder and lightning (for thunderstorms). Many different types exist, ranging from tropical cyclones and large storms of the midlatitudes to small polar lows, Medicanes, thunderstorms, or tornadoes. They can lead to extreme weather events like storm surges, flooding, high snow quantities, or bush fires. Storms often pose a threat to human lives and property, agriculture, forestry, wildlife, ships, and offshore and onshore industries. Thus, it is vital to gain knowledge about changes in storm frequency and intensity. Future storm predictions are important, and they depend to a great extent on the evaluation of changes in wind statistics of the past.

To obtain reliable statistics, long and homogeneous time series over at least some decades are needed. However, wind measurements are frequently influenced by changes in the synoptic station, its location or surroundings, instruments, and measurement practices. These factors deteriorate the homogeneity of wind records. Storm indexes derived from measurements of sea-level pressure are less prone to such changes, as pressure does not show very much spatial variability as wind speed does. Long-term historical pressure measurements exist that enable us to deduce changes in storminess for more than the last 140 years. But storm records are not just compiled from measurement data; they also may be inferred from climate model data.

The first numerical weather forecasts were performed in the 1950s. These served as a basis for the development of atmospheric circulation models, which were the first generation of climate models or general-circulation models. Soon afterward, model data was analyzed for storm events and cyclone-tracking algorithms were programmed. Climate models nowadays have reached high resolution and reliability and can be run not just for the past, but also for future emission scenarios which return possible future storm activity.

Keywords: storm, wind, proxy, climate model, tropical cyclone, polar low, Medicane, extreme event


Storms are characterized by a disturbed atmospheric state. They often feature strong winds and gusts, heavy precipitation, hail, thunder, and lightning. There are many different known storm types, from large storm systems like tropical cyclones and extratropical storms of the midlatitudes to small storms like polar lows, Medicanes, thunderstorms, and tornadoes. The sizes and properties of these storms vary tremendously. Among the largest storms are extratropical cyclones, which range between 200 and 2,000 km in diameter. They are low-pressure systems featuring fronts that mostly occur in the midlatitudes. Like all cyclones, they rotate counterclockwise in the Northern Hemisphere and clockwise in the Southern Hemisphere. Unlike tropical cyclones, their main energy source is not the warm ocean, but the temperature contrast between warm air masses in the subtropics and cold ones at the poles. If large pressure gradients lead to high wind speeds with little or no precipitation, we generally call the resulting storm a windstorm.

Tropical cyclones feature diameters between 50 and 2,200 km. Typically, they reach about 600 km, and their spatial extension is generally smaller than those of extratropical storms. Unlike extratropical storms, they do not feature fronts and they have a warm core. More intense tropical storms are given names that vary based on their genesis region. Over the North Atlantic and the eastern and central North Pacific, they are called hurricanes; over the Western North Pacific Basin, they are typhoons; and over the Indian Ocean, around Australia, and the Southern Pacific, they are tropical cyclones. They form in tropical or subtropical regions over warm ocean basins with temperatures of more than about 26°C. Tropical cyclones can originate only at a distance of more than about 5 degrees from the equator because they need the Coriolis force, due to the Earth’s rotation, which is zero at the equator. A small wind shear (varying wind speed and/or direction with altitude) is also needed for tropical cyclogenesis, but too much shear will inhibit the storm’s formation. They are forced by heat from the underlying warm ocean water and moisture transports and quickly dissipate over land where no comparable energy amount is available.

Polar lows (Rasmussen & Turner, 2003) are much smaller than extratropical or tropical cyclones, with diameters of only several hundred kilometers (often less than 300), and they also have a much shorter lifetime (mostly less than 15 h). They are hard to predict because they form very quickly and mostly occur in cold seasons in polar regions, and thus often take place in darkness. Since the late 1970s, polar lows mainly have been detected from satellite imagery that facilitated polar low detection. Nevertheless, due to their quick formation process, they are often detected too late, and since they mostly have high wind speeds, polar lows are extremely dangerous for shipping and offshore activity.

Medicanes are very small storms that only reach about a few tens to a few hundred kilometers in size (Cavicchia & von Storch, 2012). The word Medicane is a combination of the words Mediterranean and hurricane, reflecting the fact that these are hurricane like storms that form over the Mediterranean. There are very few Medicanes each decade, and they mostly happen in fall and early winter (Cavicchia, von Storch, & Gualdi, 2014b). The formation is quite similar to that of tropical cyclones; it is based on a thermodynamic disequilibrium between ocean and atmosphere, with the difference being that the dynamics of Medicanes are controlled by the intrusion of air masses of polar origin in upper atmospheric layers (Emanuel, 2005).

Thunderstorms are only about 25 km in diameter. They are characterized by lightning, thunder, and hail and generally feature strong precipitation and wind speed. Special kinds of thunderstorms are supercells, which contain a deep and persistent rotating updraft or mesocyclone (Lemon & Doswell, 1979). Supercells are the least common type of thunderstorm, but they have the potential to cause severe weather, like large hail, high wind speeds, and the formation of tornadoes. Tornadoes range in size from a few meters up to few kilometers and frequently develop from supercells. They form in rotating columns of air, often in association with funnel clouds, which touch the ground. Due to their often extreme wind speeds, they can cause great damage. They are rated on the Fujita scale, which categorizes tornadoes according to their damage.

Storms may create vast amounts of damage due to their high wind speeds, flooding, bush fires, waves, and storm surges. Thus, they can have a major impact on people (especially those living in coastal areas or working in offshore industries and shipping, agriculture, and forestry), as well as on buildings and property. These large effects lead to a number of questions, such as: Have storms happened more often in recent years? Have they become more severe? And, of the greatest interest, how will storms change in the future?

The successful analysis of these issues for the past may provide an indication for future storm developments. However, to answer these questions, long and homogeneous records of storm activity are essential. Thereby, a period of at least three decades is usually needed to describe climate change effects reliably.

There are direct ways to measure storminess (e.g., to measure wind speed or sea-level pressure). In this article, we use the term storminess to mean “the state of being stormy.” An indirect way to measure storminess is to derive it from sea-level variations or storm-damage data. Changes in storminess imply both changes in storm numbers or frequency and storm strength or intensity. Storm intensity is often defined by near-surface wind speed, usually measured at a height of 10 m. Wind speed is classified according to the Beaufort (Bft) scale, where gales are defined as Bft 8–9 (17.2–24.4 m/s) and storms as weather systems with wind speeds of Bft 10–11 (24.5–32.6 m/s). Finally, hurricane force winds are Bft 12 (>32.7 m/s). Gusts represent short and almost instantaneous maximum wind speed values. In contrast, most measurements use mean values averaged over 10 min.

Historical Wind Measurements

The Greeks were the first peoples interested in making and documenting weather observations. The famous philosopher and scientist Aristotle wrote the first book on weather, Meteorologica, which became the standard reference for the next 2,000 years (Aristotle, 1931). In it, he explained rain, snow, water vapor, clouds, wind, tornadoes, lightning, and hail, as well as many other meteorological events. The (presumably) very first meteorological instrument was a wind vane installed on top of the Tower of Winds in Athens in the 2nd century BC. It was a bronze Triton, a mythological Greek god of the sea, which rotated with the wind and measured its direction.

In the 15th century, the first instruments to measure wind speed were invented in Europe. In Italy, Leon Battista Alberti constructed the first mechanical anemometer in the year 1450 (Jacobson, 2005). (The word anemometer comes from the Greek word for wind, anemos.) Alberti’s instrument featured a swinging disk placed perpendicular to the wind that was spun by the wind. The resulting deflection of the disk measured the wind force (Jacobson, 2005).

The evolution of instruments to measure wind force and wind speed continued from the 19th century until today. In 1846, the astronomer and physicist John Thomas Romney Robinson invented the first four-cup anemometer (a hemispherical cup anemometer) (Jacobson, 2005) for wind measurements in the United Kingdom, which is the first version of the modern anemometer. The cups rotated horizontally, pushed by the wind, while the number of revolutions in a certain time was counted. It was suitable for measuring average wind speeds in a certain time interval, such as 10 min. To measure wind fluctuations on very short time scales (so-called wind gusts), a Pitot tube may be used, which is usually mounted on a wind vane. It operates on Bernoulli’s equation, which says that total pressure is static pressure plus dynamic pressure. Both total and static pressure are measured; together with air density, the resulting dynamic pressure provides a measure of the wind speed.

There are traditional and nontraditional wind measurement devices. The traditional instruments include wind vanes and cups and Gill and hot-wire anemometers (DeFelice, 1998). Nontraditional wind instruments are laser and sonic anemometers, as well as radar wind profilers. The Gill anemometer (Gill, 1975) uses a freely moving propeller; the wind speed is derived from its rotation. Another propeller anemometer is the vane anemometer, which combines a wind vane and a propeller on the same axis; it measures wind speed and wind direction at the same time. Another concept is adapted by hot-wire anemometers, which relate the heat loss of the wire to the airflow around it. They use electric currents and measure the wind speed around the instrument by a resistance decrease caused by the cooling of the sensor.

Sonic anemometers work with ultrasonic sound waves. They measure the deceleration and acceleration of the waves due to wind speed. They are also suited for measuring wind gusts. Laser anemometers derive wind speed from the backscatter of a laser beam and the according shift in frequency. Radar wind profilers measure vertical wind profiles. An electromagnetic wave is emitted, and the phase shift of the backscattered wave is determined. Thus, the Doppler shift and radial wind field can be derived. Microwave radar sensors measure the reflection of emitted high radio frequency waves over the ocean’s surface from satellites or aircrafts. This technique is applied over the oceans, where it measures the signal’s scattering due to waves, which is then transformed into wind speed and direction. A rough sea state will lead to more backscatter and a strong signal, while a calm sea will cause a weaker signal.

In North America, the first systematic weather observations were introduced in 1644 in Wilmington, Delaware, by Reverend John Campanius Holm (NOAA, 2007). In 1743, the movement of a hurricane and its associated weather patterns were described by Benjamin Franklin, and this was one of the first descriptions of the evolution of such a storm. Among the first weather observers in the United States were presidents such as George Washington and Thomas Jefferson (NOAA, 2007). For instance, Jefferson started to enlist volunteer weather observers in Virginia in 1776. The idea spread, so in 1800, five states adopted the volunteer observing system. This resulted in a weather observer’s network at U.S. Army posts throughout the nation.

At about that time, the Irish admiral Sir Francis Beaufort invented the Beaufort scale, or Beaufort wind force scale. The scale is used to estimate wind speed visually by relating wind speed to observed effects at sea or on land. Such effects could be, for instance, sea conditions ranging from flat sea to huge waves with a lot of foam and spray and reduced visibility. Another example is impacts over land, from calm conditions up to severe structural damage to vegetation and buildings. In 1848, the American scientist Joseph Henry, who was the secretary of the recently created Smithsonian Institution (a group of museums and research centers in the United States), initiated a telegraphic network of 150 volunteer weather observers who would report weather observations across America (NOAA, 2007). The goal of the network was to develop a forecast system for storms that posed a great problem at that time. An official national weather service using weather observations taken at the same time across the country was set up in 1870 (NOAA, 2007; Hughes, 1970).

At that time, the idea grew that vertical atmospheric observations would be necessary. Thus, kite and balloon measurements were introduced in the late 19th and early 20th centuries. These could measure winds at different atmospheric layers. This new, three-dimensional (3D) view of the atmosphere largely improved weather forecasts and later enabled the generation of numerical weather prediction models. In 1943, the first intentional aircraft measurement of a hurricane took place in Texas, which proved that aircraft reconnaissance flights into such forceful storms were possible. Afterward, weather reconnaissance flights have been performed routinely to detect tropical cyclones. Today, the National Weather Service of the National Oceanic and Atmospheric Administration (NOAA) still uses volunteers to observe severe weather. Nowadays, more than 350,000 weather spotters provide reports on severe weather in their communities. Reports of severe storms are very important, as the United States encounters an average of 10,000 severe thunderstorms and more than 1,000 tornadoes each year (NOAA, 2007).

Data Homogeneity

The most straightforward way to construct storm records would be to use wind measurements directly. But as long time series are necessary, many problems have arisen, even though many long-term meteorological records exist. These relate to changes in measurement systems and practices over time. In the past, there were only meteorological observers who subjectively categorized the weather state. The first wind speed measurements used drifting devices like kites and balloons, but later these were replaced by instruments such as anemometers. Not just the instrumentation, but also the frequency of observations and measurements changed over time. A strong storm like a hurricane can show very different features, like much lower wind speeds, if it is just analyzed with meteorological data from the surface like station, ship, or buoy data, in contrast to analyses that also use aircraft or satellite data to provide information at higher levels (Landsea et al., 2004).

The synoptic stations often changed their location, surroundings, or instrumentation. A striking example of the effects of changes in station location is the meteorological station in Hamburg, Germany, where a relocation of the station from the harbor to the airport led to drastically reduced wind speeds, and thus an artificial reduction of storm days (see Weisse & von Storch, 2009, Figure 4.1). On the island of Heligoland in the North Sea, a monitoring station relocation between the main island and a small neighboring island, as well as between different locations on the main island with very contrasting land characteristics (e.g., from a location on land with many surrounding obstacles to a harbor pier with relatively undisturbed conditions), produced self-explanatory changes in wind statistics (Lindenberg, Mengelkamp, & Rosenhagen, 2012). Such station relocations happened quite frequently in the past, and they were not just limited to changes in location (and thus environmental conditions). But they often also implied a change in anemometer height.

A higher building density around the stations also may have an effect, as wind speed is especially sensitive to the surroundings. Stations can be affected at some point by new buildings that shade the wind field, more sealed landscapes that lead to changed heat transports, or the artificial introduction or removal of water sources, which cause fluctuations in humidity in the vicinity. Land-use alterations due to changes in urbanization or vegetation amount caused changes in the wind field; due to an increased roughness length, an “atmospheric stilling” has been observed during the last several decades (McVicar et al., 2012; Vautard, Yiou, Thépaut, & Ciais, 2010). Large effects can occur for synoptic stations, which changed their instruments; this may cause a systematic shift in the long-term wind records. Also, the introduction of new measurement devices like satellites in November 1978, led to higher data quality, but therefore also to data inconsistencies (Weisse & von Storch, 2009). All these changes may lead to changes, trends, or jumps in measurement time series.

Storm Proxies

The potential inhomogeneities in direct wind records hinder the statistical analysis of changes in storminess. But storm records also can be derived from other weather fields besides wind speed. Substitute or proxy time series may be derived from meteorological variables that are less prone to inhomogeneities over longer time scales. Air pressure is an example of such a proxy (Krueger & von Storch, 2012). However, pressure records also are generally vulnerable to errors related to inconsistent observations, to certain outliers or measurement errors caused by issues such as mistranscriptions of old handwritten records, or to station relocations. But pressure has one advantage: it is a rather large-scale variable that is much less affected by changes in instrumentation, station relocation, measurement technique, station surroundings, or other local conditions than is wind speed. Air pressure does not show large spatial variability, which is a typical feature of wind speed. Very long pressure measurement records are available, and some time series date back to the 1750s. These records have been used to derive storm changes for more than 140 years. To derive storm activity from surface pressure, pressure records of single meteorological stations are often applied. The stations need to be chosen adequately so that they provide a realistic description of storms and their changes in the vicinity of the station over time.

Typical storm measures derived from such pressure readings are the number of deep lows below certain thresholds or annual or seasonal frequencies of absolute pressure tendencies (Krueger & von Storch, 2012). Also, lower percentiles of pressure or pressure tendencies are used as storm proxies. In statistics, percentiles divide a distribution of observations or other values into 100 parts of 1% each. A percentile then defines the value below which a certain percentage of data falls. For storms derived from air pressure, often the 5th or 1st percentiles are used, which represent extremely low pressure values (the lowest 5% or 1%, respectively). For wind speed, the 95th or 99th percentiles are good measures to describe storminess, as they give the highest 5% and 1% of the wind speed distribution. But new results based on regional climate model simulations show that such proxies based on pressure readings from single meteorological stations are often only weakly related to storminess (Krueger & von Storch, 2012). The reason for this is that such records detect many kinds of atmospheric disturbances, which are not necessarily directly connected with storms.

There are also many nonmeteorological variables that may serve as storm proxies, as they are directly related to storm properties such as wind speed and direction or duration. Often, such proxies for deriving storms are based on storm impacts. These include losses caused by storms, like damages to forests, industry, and buildings. If storm losses are analyzed in order to derive variations in storm activity, one has to take into account changes in population, building density, insurance, and values over the specified period of time. Other storm-related impacts are sea-level variations such as storm surges or large waves driven by the wind.

Historic ship logbook data, which in some cases covers several centuries, provides valuable information on air pressure (Küttel et al., 2010), winds, sea state, and temperature. Therefore, they can serve as a proxy for storms. Also, dune sediment may give an indication of storm history (Reimann, Tsukamoto, Harff, Osadczuk, & Frechen, 2011). The rate of sand drift is determined by surface conditions and shear velocity. Thereby, the sand’s grain diameter, the dune’s vegetation and inherent moisture amount, as well as surface roughness, influence the drift rate. All these proxies are generally independent from meteorological observations, and thus they provide valuable information for comparison studies.

Geostrophic Wind Triangles

Proxies to describe storm records need not only be based on single observation stations; they also may be derived from geostrophic wind speed statistics. The geostrophic wind represents wind flowing parallel to lines of constant pressure at a certain height, and it is deduced from air pressure (Dutton, 1986). This derived geostrophic wind speed may be used as a proxy for near-surface wind speed, and thus as a measure for storms. It works best over flat terrain like the open ocean, and for balanced meteorological conditions where the real wind does not deviate much from the geostrophic wind. Balanced conditions refer to the geostrophic balance between the Coriolis force, which is caused by the Earth’s rotation, and the pressure gradient force. Such conditions can be found over the North Atlantic or Baltic Sea, for example. Especially in the midlatitudes, wind speed is generally associated with pressure gradients, and thus with geostrophic wind speed.

But pressure measurements that are available for long time periods are often irregularly scattered over wide areas, so that it is not easy to derive pressure gradients. Some sort of spatial interpolation, therefore, is necessary. The most straightforward approach is to use pressure measurements from three synoptic stations, which form a triangle. Then pressure gradients can be estimated by computing centered differences over the triangle’s area so that the geostrophic wind speed and according statistics can be calculated. In this way, the storm climate for a certain region in the past can be assessed (Schmidt & von Storch, 1993). This method is better suited to measuring over the open ocean than over land areas. This stems from the fact that the geostrophic balance is often less applicable over land, where also other forces like friction have more importance. Also, surface conditions affect wind speed statistics.

Geostrophic wind triangles have been applied successfully over many study domains, such as the German Bight, the North Sea, and the North Atlantic or Europe (Alexandersson, Schmith, Iden, & Tuomenvirta, 1998; Alexandersson, Tuomenvirta, Schmith, & Iden, 2000; Krueger & von Storch, 2011; Matulla, Schöner, Alexandersson, von Storch, & Wang, 2007; Schmith, 1995; Wang, Zwiers, Swail, & Feng, 2009). It has been shown that the technique generally provides a realistic description of the geostrophic wind speed over the triangle. In addition, the statistics of geostrophic wind speed deduced from the triangles are of high quality and superior to the statistics derived from single meteorological stations.

The size of the geostrophic wind triangles influences the resulting wind speed statistics. Smaller triangles with an average side length smaller than 300 km provide geostrophic wind speed statistics that are of higher quality than those for large triangles with average side lengths of more than 800 km (Krueger & von Storch, 2011). These spacious triangles sometimes miss small storms that are connected with large pressure gradients, so the resulting storm climatology is less reliable.

Origins of Climate Models to Study Storminess

At the end of the 19th century, the Norwegian physicist and meteorologist Vilhelm Bjerknes worked at the University of Stockholm on the interaction between fluid dynamics and thermodynamics (Eliassen, 1995). There, he developed a set of conservation equations that basically consist of conservation of mass, momentum, and thermal energy (the so-called primitive equations). These nonlinear differential equations describe atmospheric flow, and today, they are used in most atmospheric and ocean models to simulate large-scale motions. His work led to the development of the first nonelectronic numerical weather prediction method by Lewis Frey Richardson in 1922 (Randall, 2000). The English mathematician, physicist, and meteorologist designed a mathematical model for weather forecast whereby he divided the Earth’s surface and atmosphere into grid cells and then applied the primitive equations to compute the weather 8 h in advance. It took him six weeks to accomplish the task, but still he failed in the end because he made some mistakes, such as not initializing and thus quality-controlling the input data and using a time step that was too long. He then daydreamed that for a successful weather forecast, it would take 64,000 people working on parts of the calculations, coordinated by a leader in the center acting like a conductor. But even with all these humans, Richardson thought that the weather could not be calculated faster than it happened in reality. But at that time, he did not know about the possibilities that digital computers would bring in the future, and that his calculation methods would be used eventually to develop the first general-circulation models.

The American mathematician John von Neumann worked on computer simulations during World War II. He was convinced that weather modeling would lead to an advantage in wartime. Therefore, he brought together several theoretical meteorologists and employed the American meteorologist Jule Gregory Charney to lead the team. In April 1950, they started to simulate a 24-h regional weather forecast for the eastern part of the United States (Phillips, 1995).

To accomplish this task, they linearized the barotropic vorticity equation to treat perturbations on a uniform flow in a narrow west-east channel and expanded Rossby’s frequency formula into a Green’s function that would give a 24-h forecast of the initial flow pattern by simple, weighted, longitudinal integration of the initial distribution of the isobaric height at the barotropic level (Phillips, 1995). Even though the forecast quality was not very high, it proved that the idea of a computerized weather forecast that solved differential equations numerically was feasible. The development of weather forecast models then thrived, and more layers were added. This lead to the first operational numerical weather forecast in Sweden in 1954 at the University of Stockholm, in cooperation with Carl-Gustav Rossby, a famous American meteorologist and oceanographer. Rossby was originally from Sweden but later moved to the United States. His main scientific achievements include the discovery of the jet stream, as well as Rossby waves (Byers, 1960). In 1955, operational weather forecasting was introduced in the United States. Up to that point, numerical weather prediction could not keep up with forecasts analyzed manually, by trained meteorologists. At that time, the forecast models still used many simplifying assumptions before the primitive equations were applied instead, some years later.

In 1956, the American geophysical fluid dynamicist Norman Phillips worked on the atmospheric angular momentum and analyzed whether numerical weather forecast models would be able to simulate the general circulation of the atmosphere (Phillips, 1956). The general-circulation model describes global atmospheric circulation systems, which are basically forced by the temperature contrast between the equator and the poles. He showed that the models were capable to do this, and these experiments were regarded to be the first to use a general-circulation model. This success again led to more model development, which resulted in multilevel, 3D, general-circulation models that used the primitive equations.

In 1965, the first of these models was applied; it was still quite simple, with only a few atmospheric layers and adjustments for certain meteorological processes. Many different modeling groups emerged at that time, and many general-circulation models were built. Many experiments were performed then, like the first simulations that analyzed the effects of a doubling of carbon dioxide and studies that regarded the impact of the deforestation of tropical rainforests (Gates, 2003). Models were improved; they used higher resolutions and a better-simplified description of small-scale or very complex processes that could not be computed directly by the model. These models constituted the first generation of climate models. Simultaneously, ocean models were developed, which inspired the coupling of both atmosphere and dynamical ocean models.

Another direction was taken by regional climate modelers, who used limited-area models nested into global climate models to be able to simulate weather phenomena at higher resolutions. In contrast to global models, regional climate models cover only limited parts of the Earth and use global climate model data as forcing at their lateral and lower boundaries. The regional climate model usually uses less computational resources, so they can operate with much smaller grid distances and can simulate local weather, circulation, or orography, with more detail and temporal resolution.

Today, climate model results (both regional and global) are used regularly to study storms and their changes over time. Tracking algorithms are applied to identify storms and their paths in the gridded model data. Nowadays, climate models can be run at high resolutions due to increasing computing resources, and they have achieved a high quality of output. They are used for studies of the past, reaching back in time for up to thousands or even millions of years in paleoclimate studies, as well as for studies of possible future states that apply future emission scenarios to provide storm assessments for these time spans.

Reconstructing Storm Records with Climate Models

The multiple observations available worldwide are not distributed evenly over the globe, both in space and time. To overcome this problem, data assimilation was introduced. The intention is to achieve the most realistic estimate of atmospheric conditions based on observations and short-range forecasts. Thereby, all observations like meteorological station data; rawinsonde data; aircraft, ship, and buoy measurements; or satellite data are fed into a weather forecast model for process studies, model improvements, or forecast initialization purposes (Reichle, 2008). During the data assimilation, a previous model forecast is compared with current observations to give a new model state that bests fits the observations within a time window. This new model state then becomes the basis for a subsequent new forecast step. The model produces additional variables and information on various atmospheric levels and fills gaps in the data. Such data assimilation is considered to reduce data inhomogeneity by balancing inherent errors in the station records and forecast.

For longer time scales, so-called reanalysis data sets were invented. These assimilate historical weather data on climatological time scales ranging from several decades to centuries with a single procedure, which does not change over time. Therefore, they use the same numerical weather forecast model and compute the weather states in the past in the same way throughout the period of interest. Reanalyses aim to achieve increased data homogeneity by quality control mechanisms to account for factors such as instrument changes, spatial and temporal interpolation due to an uneven distribution of data, and an inherent smoothing of measurement errors. They provide data sets with many variables on equal grids, atmospheric layers, and time intervals.

But reanalyses have some drawbacks, which are mostly related to the better data quality achieved in more recent times. In the middle of the 19th century, only very limited meteorological station data or ship measurements were available. Therefore, data coverage was sparse, especially over open oceans. Station density increased over the next decades, as did the instrumentation. The introduction of satellites at the end of the 1970s provided much better data coverage and quality compared to the relatively few data that was available before. But therefore, the addition of satellite data also led to inhomogeneities in long-term time series. However, due to the nonchanging assimilation system and quality control, these are presumably much smaller in reanalyses than in raw measurement data.

Today, there are reanalyses that cover more than 100–140 years in the past. Many reanalyses avoid the problems related to the data quality and coverage increase due to the introduction of satellite data in the late 1970s, beginning in 1979 (e.g., ERA-Interim, 1979–present, T255 (0.7 degrees, about 80 km), 60 vertical levels: Simmons, Uppala, Dee, & Kobayashi (2007)); the National Aeronautics and Space Administration Modern Era Retrospective-analysis for Research and Applications (NASA MERRA, 1979–present, 0.5 degrees (about 55 km), 72 vertical levels: Rienecker et al. (2011); and the NCEP Coupled Forecast System Reanalysis (NCEP CFSR, 1979–2010, 0.5 degrees (about 55 km), 64 vertical levels: Saha et al. (2006, 2010)). Reanalyses models span grid distances from more than 200 km down to less than about 50 km.

Due to the inhomogeneities in measurement time series, relatively low resolution of the data sets, and reduced reliability of derived storm statistics, an alternative method evolved: dynamical downscaling. This method uses atmospheric global climate model data to drive a regional climate model. The global model uses greenhouse gas emissions that include carbon dioxide, methane, sulfur, black carbon, and nitrogen oxide, among others, as well as volcano aerosols and solar forcing data as estimated for the simulation period of interest. For present-day conditions, the global model often uses observed sea surface temperatures and sea ice coverage to give a realistic representation. Alternatively, reanalyses are used as input to simulate the climate of the past.

For such hindcast studies of the past, a method was invented to make sure that the regional climate model will not change its large-scale weather patterns essentially in comparison to the forcing global model or reanalysis input. Only those weather phenomena of regional type, which can be well resolved by the regional model, are altered. This technique is called spectral nudging (von Storch, Langenberg, & Feser, 2000; Waldron, Paegle, & Horel, 1996). It reduces the regional model’s freedom at the global spatial scale, while the model is given full freedom to develop its own regional weather. In addition, spectral nudging decreases the internal model variability and dependency on initial conditions (Weisse & Feser, 2003). This leads to larger reproducibility, while the results are, on average, closer to observations.

The application of spectral nudging means that the regional model is changed after each completed model time step by adding a spectral nudging term. Therefore, the regional climate model results are transferred into spectral space via a Fourier transformation and the large spatial scales are selected by their wavenumbers (von Storch et al., 2000). Similarly, the large scales of the global reanalysis or global climate model input data are selected. The nudging term depends on the difference between the regional and global models, and thus they can be either positive or negative. By adding this term to the low wave numbers of the regional climate model, it becomes closer to the large-scale input field or is nudged toward it. Afterward, the regional model results are transferred back to physical space, again via a Fourier transformation. These results of the regional climate model then show the desired properties of high agreement of global forcing data and the regional model at large scales and new small features, which ideally show an added value, at regional scales (Feser, 2006).

Climate models are used not just to simulate past states, but also to simulate future scenarios. For that purpose, climate models apply an estimation of future greenhouse gas emissions and nowadays of solar forcing as well (Matthes et al., 2017) which depend on the specific scenario chosen. Climate models are basically similar to numerical weather prediction models and use the same variables, but normally different parameterizations. Both global and regional climate model and reanalysis data may be used to reconstruct past storm climate or to simulate future storminess (e.g., Feser & von Storch, 2008; Pinto et al., 2007; Schubert-Frisius, Feser, von Storch, & Rast, 2017; Zahn, Von Storch, & Bakan, 2008; Zappa, Shaffrey, Hodges, Sansom, & Stephenson, 2013).

The climate models compute a multitude of meteorological variables, of which wind speed, wind gusts, and air pressure are suitable for deriving storm statistics, including seasonal or yearly extreme values, storm numbers, intensity, and duration. Also, the number of storm days per year or some other measure that exceeds a certain predefined threshold to represent storm activity can be defined. Another well-established method is storm tracking (Neu et al., 2013). These algorithms detect elements such as maxima in vorticity, which describes the rotation of the air, or minima in surface pressure, and then adequately connect them to form storm tracks. Thereby, the cyclone’s development, path, and most probable movement must be considered. Then these storm tracks and derived storm statistics can contribute to form cyclone climatologies.

Past North Atlantic Winter Storminess

Most studies that describe changes in storms for long time scales like the past 100 years or more do not rely on direct wind measurements, but rather on wind proxies (Feser et al., 2015). This is due to a lack of long wind measurement time series, as well as potential inhomogeneities in the data. Long pressure records exist, so these are often used to compute geostrophic wind speed, which allows a derivation of storm statistics. Climate model simulations were performed that cover a span ranging from the last 100 years back to more than 1,000 years ago (e.g., Brönnimann et al., 2012; Fischer-Bruns, von Storch, González-Rouco, & Zorita, 2005; Xia, von Storch, & Feser, 2013; Xia, von Storch, Feser, & Wu, 2016).

In addition, reanalysis data sets can look many decades back in time. Most of these studies show wide variability in decadal time scales for the number of storms over the North Atlantic for the last 100–150 years. According to geostrophic wind speed studies (Alexandersson et al., 1998, 2000) in the early 1880s, high storm numbers were apparent, and then a decrease in storm numbers followed until the early 1960s (Figure 1). The subsequent increase peaked in the mid-1990s, and a drop followed. Since the studies based on reanalysis data and climate models for the past are mostly limited to the last three to six decades, almost all of them return an increase in storm numbers over the northern North Atlantic from about the 1970s until the mid-1990s, which forms part of the decadal variability in the longer studies (e.g., Leckebusch, Renggli, & Ulbrich, 2008; Raible, Della-Marta, Schwierz, Wernli, & Blender, 2008; Schneidereit, Blender, Fraedrich, & Lunkeit, 2007; Simmonds & Keay, 2002; Trigo, 2006; Wang, Swail, & Zwiers, 2006; Weisse, von Storch, & Feser, 2005).

Constructing Records of StorminessClick to view larger

Figure 1. Storm index for northwestern Europe (i.e., British Isles, North Sea, and Norwegian Sea) for the years 1881–2004, based on geostrophic wind speed percentiles according to the methodology of geostrophic wind triangles described in Alexandersson et al. (1998). Blue circles represent extreme values (95th-percentiles) of standardized geostrophic wind speed anomalies averaged over 10 sets of station triangles. The gray curve shows filtered data of low frequency. The blue lines show linear trends of extreme wind speeds (95th-percentiles), for the entire time period, and for a reanalysis period (the ERA-40 period, 1957–2001).

Source: Feser et al. (2015).

However, in the southern North Atlantic, south of about 60 degrees north, negative storm number tendencies can be seen for reanalysis-based studies (Raible et al., 2008; Trigo, 2006; Wang et al., 2006; Weisse et al., 2005). If model simulations that cover the 20th century and going as far back as the last 1,000 years and long-term storm proxy records are analyzed, just decadal variability, or rather a decrease in storm numbers, is found (Fischer-Bruns et al., 2005; Xia et al., 2013). This result applies to the British Isles, the North Sea, and the northeast Atlantic. Studies give inconsistent results for the Baltic Sea and central Europe.

Atmospheric oscillations and large-scale weather patterns are important for storm generation and development. A very prominent circulation is the North Atlantic Oscillation (NAO), which is defined as the pressure difference fluctuation at sea level between the Icelandic Low and the Azores High. The difference is expressed as an index called the NAO index, which is high for large pressure differences and low for small differences. In winter, a high NAO index often appears, which goes along with more westerly winds and a higher frequency of low-pressure systems. Their frontal systems, in combination with large temperature and pressure differences, can lead to storm generation, so storm seasons are generally more pronounced for high NAO indexes. The NAO thus has a large impact on storm generation over the North Atlantic, which explains a large part of pressure variability over the North Atlantic. With a low NAO index, westerly winds occur less often and low-pressure systems are shifted toward the Mediterranean Sea, which leads to a southward relocation of storm activity over Europe. But although a high NAO index usually leads to higher numbers and more intense storms, severe storms also may occur with a low NAO index.

The NAO index for winter shows merely large interannual and multidecadal variations from the 1820s until today (Figure 2). It shows a period of high index values beginning in the 19th century, and then a decrease followed from about 1905 until about the mid-1960s, a subsequent peak occurred in the 1990s, and a drop took place after about 1990. It shows some similarities to the storm frequency time series for the North Atlantic derived from geostrophic wind speed triangles (Figure 1) and explains approximately 30% of changes in winter storm numbers, but the correlation is not stable over time. The connection between the NAO index and storm activity has been rather weak in the past, but in more recent decades, the correlation is increasing.

Constructing Records of StorminessClick to view larger

Figure 2. Updated NAO index after Jones, Jonsson, and Wheeler (1997) for boreal winter (December–March) 1823/1824 to 2012/2013, renormalized with respect to the full time period. The black line shows an 11-year running mean. This is a statistical measure that first disaggregates a time series of values into different subsets of a certain length, and then computes a number of averages over these subsets.

Source: Feser et al. (2015).

Over the North Atlantic, there is a region where storm paths are forced by the predominant wind systems, the so-called storm track. Here, extratropical cyclones travel eastward along the polar jet stream. This storm track both influences and is influenced by atmospheric circulation, large-scale weather phenomena, temperature, land-sea contrast, and other obstacles close to the ground. Many studies of past changes (Harnik & Chang, 2003; Hickey, 2003; Schiesser, Pfister, & Bader, 1997; Schneidereit et al., 2007; Sickmoeller, Blender, & Fraedrich, 2000; Trigo, 2006; Wang et al., 2006) found a poleward shift of the North Atlantic storm track.

Future European Winter Storm Perspectives

For the future, there is much uncertainty about how and where storms will develop, if there will be any changes in their numbers, or if they will become more intense. To account for possible evolutions of demography, societies, economy, and technology, the Intergovernmental Panel on Climate Change (IPCC) set up a number of emission scenarios. An updated version of these scenarios was published in the IPCC Special Report on Emissions Scenarios (IPCC, 2000), which describe possible future changes of human, technical, and economical factors and their high uncertainties and derive associated greenhouse gas emissions from them. These anthropogenic emissions include the greenhouse gases carbon dioxide, methane, nitrous oxide, carbon monoxide, hydrofluorocarbons, perfluorocarbons, sulfur hexafluoride, hydrochlorofluorocarbons, chlorofluorocarbons, sulfur dioxide, and nitrogen oxide. These atmospheric emissions then can be fed into global or regional climate models to assess future climate conditions such as changes in storm activity.

Due to the high uncertainty of future societal or economic changes, for instance, it is favorable to use a range of emission scenarios so that most of these often very different developments can be covered. For the 5th Assessment Report of the Intergovernmental Panel on Climate Change (IPCC, 2014), four new emission scenarios were developed. These so-called Representative Concentration Pathways (RCPs) together span a range of year 2100 radiative forcing values from 2.6 to 8.5 W/m2 (van Vuuren et al., 2011).

Both global and regional climate models have been used to study future changes in storm activity. Most of them compare a present-day or preindustrial time span to scenario simulations of about 50 to several hundred years, mostly ending in the year 2100. These studies do not give a univocal answer about whether winter storms will become more frequent in the future over the North Atlantic and northwestern Europe. Overall, about the same number of articles present increases as decreases. Decreases were mainly seen in the high-latitude North Atlantic (Bengtsson & Hodges, 2006; Carnell & Senior, 1998; Harvey, Shaffrey, Woollings, Zappa, & Hodges, 2012; Lambert, 2004; Lambert & Fyfe, 2006; Leckebusch & Ulbrich, 2004), and increases in the midlatitude North Atlantic and the North Sea (e.g., Beniston et al., 2007; de Winter, Sterl, & Ruessink, 2013; Harvey et al., 2012; Ulbrich & Christoph, 1999; Ulbrich et al., 2008), while the results were very heterogeneous, with comparable numbers of studies showing a decrease as an increase for the Baltic Sea and central Europe (e.g., Andersen, Kaas, & May, 2001; Geng & Sugi, 2003; Pinto, Karremann, Born, Della-Marta, & Klawa, 2012; Pinto, Neuhaus, Leckebusch, Reyers, & Kerschgens, 2010). But almost all the studies agree that storms will become more intense over the North Atlantic and Europe for changed climatic conditions (Feser et al., 2015). This was shown, for instance, for the British Isles, the North Sea, and northwestern Europe. Many studies also presented an increase in track density of intense cyclones over the North Atlantic, particularly close to Europe (Pinto, Zacharias, Fink, Leckebusch, & Ulbrich, 2009). A tripolar pattern was described for wind intensity changes, showing a decrease over the Norwegian Sea, the Mediterranean Sea, and the subtropical central Atlantic, and an increase over central Europe (Zappa et al., 2013).

In future scenario simulations, a possible shift of the North Atlantic storm track often depends on the model generation used. The storm track is shown to move poleward in older-model scenario simulations (e.g., Bengtsson & Hodges, 2006; Bengtsson et al., 2009; Fischer-Bruns et al., 2005; Hall et al., 1994; Knippertz et al., 2000; Ulbrich & Christoph, 1999; Yin, 2005). Analyses based on more recent model generations, though, present an eastward extension of the storm track from the eastern North Atlantic toward Europe (e.g., Ulbrich et al., 2008; Woollings, Gregory, Pinto, Reyers, & Brayshaw, 2012; Zappa et al., 2013). This result was confirmed with many studies using different climate models and storm detection algorithms (Feser et al., 2015).

Changes in Tropical Cyclone Activity

Tropical storms show large variability in their numbers, as well as in their strength in the past (Knutson et al., 2010). For the last decades, though, no significant trend was described in storm numbers over the various ocean basins. Only over the North Atlantic did several studies detect century-scale increases in hurricane frequency. But these are subject to many uncertainties, such as low measurement density in earlier times and changes in observing techniques. Therefore, it is still uncertain if these changes in tropical storm numbers, as well as variability in cyclone intensity or rainfall, can be attributed to anthropogenic climate change. In addition, tropical cyclogenesis, duration, track location, and associated storm surge flooding have not been found to exceed natural variability.

A different result is given for future scenario simulations with both global and regional climate models. These show a more univocal picture, and all of them return an increase in tropical cyclone intensity. This increase is projected to range between 2% and 11% until the end of this century (IPCC, 2012). But there are larger changes for the individual ocean basins; the range is between ‒1% and 9%. A decrease or no change in global tropical cyclone numbers is given under global warming conditions. The decrease ranges between 6% and 34% according to the individual studies (Knutson et al., 2010), and the result was more robust in the Southern Hemisphere than the Northern Hemisphere. Model studies of higher resolution with grid distances of 60 km and less generally project an increase in the number of stronger tropical cyclones. Also, for rainfall of tropical cyclones, an increase is projected for future climate conditions. The change in most studies was about 20% within 100 km of the storm center (IPCC, 2012).

Changes in Mesoscale Storms

Not much is known about changes in mesoscale storms like polar lows or Medicanes, both in the past and for possible future states. This is because very few long-term studies exist on this topic. The term mesoscale generally describes spatial scales between very few and several hundred kilometers. In a regional hindcast study of the period of time between January 1948 and February 2006 for polar lows over the North Atlantic, it was found that their number shows large year-to-year variability, but no systematic change (Zahn & von Storch, 2008). For the same region, global scenario simulations until the end of the 21st century were downscaled with a regional climate model to analyze projected future changes in polar lows. The results described significantly lower numbers of these small storms for various emission scenarios (Zahn & von Storch, 2010). As a reason for this decrease, midtroposphere temperatures were identified, which warmed faster than sea-surface temperatures. This leads to greater atmospheric stability, and thus inhibition of polar low formation or intensification. In addition, the main genesis region of North Atlantic polar lows shifted north for elevated greenhouse gas concentrations. This is caused by melting sea ice, which provides more ice-free regions farther poleward, which are essential for polar low formation.

For the North Pacific, quite similar results were found, which showed larger interannual variability but no obvious long-term change of polar lows over the last six decades (Chen & von Storch, 2013). For future emission scenario studies until the end of the 21st century, again a decrease in the number of polar lows emerged (Chen, von Storch, Zeng, & Du, 2014). Thereby, the decrease was greater for scenarios with a stronger warming signal. A higher vertical stability due to fast-rising atmospheric temperatures in connection with slower-rising sea surface temperatures was again responsible for the reduction in storm numbers. As for the North Atlantic, a shift of polar low genesis regions to the north was found. There have been very few studies on polar lows in other ocean basins.

Over the Mediterranean, storms with tropical features may form from time to time. Some of these, known as Medicanes, reach wind speeds of up to 119–153 km/h. Hindcast studies of the period between 1948 and 2011 show year-to-year variability in the storm’s frequency, but no detectable trend (Cavicchia, von Storch, & Gualdi, 2014a). There are two hotspots of Medicane formation, in the Western and central Mediterranean Sea, reaching a maximum level during the cold season. For future scenarios, Medicanes are projected to become less in number, but their intensity may increase moderately (Cavicchia et al., 2014b).

Significant Questions That Remain, are Being Pursued, or Should be Pursued

Many open questions remain on the important topic of how storms have changed and how they will change in the future. Extratropical storms in the North Atlantic have been studied extensively, and some agreement can be found for different studies and regions considered in the past. But for future scenario simulations, the results on the change of North Atlantic storm frequency are inconclusive, while the general agreement is that they will become more intensive under climate change conditions. Another important question is whether the North Atlantic storm track will change its position. Newer model generations project an eastward extension of the storm track into Europe, while older model generations rather return a poleward shift.

Tropical cyclones have been studied very intensively over different ocean basins, but many open questions still exist, including the intensity of tropical cyclones and associated rainfall patterns. Several best track data sets produced by different weather services in several countries worldwide, which use available observation and satellite data, show discrepancies over the northwestern Pacific (Barcikowska, Feser, & von Storch, 2012). This stems from the fact that they use different algorithms, and presumably in some cases additional data, to estimate tropical cyclone intensities. This is most obvious for intense storms, which have the largest impact on populations and their environment, and more research is needed for this topic.

Polar lows have been studied over the North Atlantic and in a very limited number of studies over the North Pacific, but many open areas of research remain for different regions of the world. Here, one should mention in particular the Southern Hemisphere. This area lacks observation data of high quality, especially for the presatellite era, which makes the analysis of all kinds of storms difficult. Since Medicanes happen only very rarely, and since they were hard to detect in the past, a definite answer about whether they did become more or less frequent between 1948 and 2011 cannot be provided. More studies are needed in order to analyze possible future changes in their genesis regions, tracks, or intensity.

A large area of research is the attribution of climate change. Here, one may look at the influence of aerosols on storms for elevated greenhouse gas concentrations. But most of all, long and homogeneous measurement time series are needed to derive storm statistics for the past, so that an indication may be deduced for possible future states. These long time series are available as reanalysis data sets, but they all suffer from increasing measurement density over time or the inclusion of new available observations like satellite data. Some very long-term reanalyses over more than 140 years exist, but even though some of them only incorporated very few meteorological variables, all have used an increasing station density over time. A great effort to pursue would be to rerun these long-term reanalyses with the same station density, and ideally the same data quality over time, so that more reliable storm statistics could result.

Another way to extend long time series further back in time would be to use statistical methods like nonlinear statistical analog-upscaling (Schenk & Zorita, 2012) to reconstruct wind fields and storms. Thereby, measurement data from very few stations are combined with regional climate model simulations to gain high-resolution weather data over long time periods. Another practice is the application of proxies to derive long-term storm records. Currently, some historical handwritten meteorological measurement data are digitized; these then can serve as storm proxies like geostrophic wind triangles (Wagner, Tinz, & von Storch, 2016). Also, studying ship log data (Küttel et al., 2010) as a proxy for storms or using coastal dunes (Bierstedt, Hünicke, Zorita, & Ludwig, 2017) to deduce data on past wind field variations are promising approaches that currently are being pursued.


Alexandersson, H., Schmith, T., Iden, K., & Tuomenvirta, H. (1998). Long-term variations of the storm climate over NW Europe. Global Atmosphere-Ocean System, 6, 97–120.Find this resource:

Alexandersson, H., Tuomenvirta, H., Schmith, T., & Iden, K. (2000). Trends of storms in NW Europe derived from an updated pressure data set. Climate Research, 14, 71–73.Find this resource:

Andersen, U. J., Kaas, E., & May, W. (2001). Changes in the storm climate in the North Atlantic/European region as simulated by GCM timeslice experiments at high resolution. Retrieved from this resource:

Aristotle. (1931). Works of Aristotle. Vol. III, Meteorology (Trans. E. W. Webster). Oxford: Clarendon Press.Find this resource:

Barcikowska, M., Feser, F., & von Storch, H. (2012). Usability of best track data in climate statistics in the western North Pacific. Monthly Weather Review, 140, 2818–2830.Find this resource:

Bengtsson, L., & Hodges, K. I. (2006). Storm tracks and climate change. Journal of Climate, 19, 3518–3543.Find this resource:

Bengtsson, L., Hodges, K. I., & Keenlyside, N. (2009). Will extratropical storms intensify in a warmer climate? Journal of Climate, 22, 2276–2301.Find this resource:

Beniston, M., Stephenson, D. B., Christensen, O. B., Ferro, C. A. T., Frei, C., Goyette, S., . . . Woth, K. (2007). Future extreme events in European climate: An exploration of regional climate model projections. Climatic Change, 81, 71–95.Find this resource:

Bierstedt, S. E., Hünicke, B., Zorita, E., & Ludwig, J. (2017). A wind proxy based on migrating dunes at the Baltic Coast: Statistical analysis of the link between wind conditions and sand movement. Earth System Dynamics, 8, 639–652.Find this resource:

Brönnimann, S., Martius, O., von Waldow, H., Welker, C., Luterbacher, J., Compo, G. P., . . . Usbeck, T. (2012). Extreme winds at northern mid-latitudes since 1871. Meteorologische Zeitschrift, 21, 13–27.Find this resource:

Byers, H. B. (1960). “Carl-Gustaf Arvid Rossby,” 1898–1957. Biographical Memoirs 34. Washington, DC: National Academy of Sciences.Find this resource:

Carnell, R. E., & Senior, C. A. (1998). Changes in mid-latitude variability due to increasing greenhouse gases and sulphate aerosols. Climate Dynamics, 14, 369–383.Find this resource:

Cavicchia, L., & von Storch, H. (2012). The simulation of medicanes in a high-resolution regional climate model. Climate Dynamics, 39, 2273–2290.Find this resource:

Cavicchia, L., von Storch, H., & Gualdi, S. (2014a). A long-term climatology of medicanes. Climate Dynamics, 43, 1183–1195.Find this resource:

Cavicchia, L., von Storch, H., & Gualdi, S. (2014b). Mediterranean tropical-like cyclones in present and future climate. Journal of Climate, 27, 7493–7501.Find this resource:

Chen, F., & von Storch, H. (2013). Trends and variability of North Pacific polar lows. Advances in Meteorology, ID-ID 170387.Find this resource:

Chen, F., von Storch, H., Zeng, L., & Du, Y. (2014). Polar low genesis over the North Pacific under different global warming scenarios. Climate Dynamics, 43(12), 3449–3456.Find this resource:

de Winter, R. C., Sterl, A., & Ruessink, B. G. (2013). Wind extremes in the North Sea Basin under climate change: An ensemble study of 12 CMIP5 GCMs. Journal of Geophysical Research, 118, 1601–1612.Find this resource:

DeFelice, T. P. (1998). An introduction to meteorological instrumentation and measurement. Upper Saddle River, NJ: Prentice Hall.Find this resource:

Dutton, J. A. (1986). The ceaseless wind: An introduction to the theory of atmospheric motion. New York: Dover Publications.Find this resource:

Eliassen, A. (1995). Jacob Aall Bonnevie Bjerknes Biographical Memoirs V.68. Washington, DC: National Academy of Sciences.Find this resource:

Emanuel, K. (2005). Genesis and maintenance of Mediterranean hurricanes. Advances in Geosciences, 2, 217–220.Find this resource:

Feser, F. (2006). Enhanced detectability of added value in limited area model results separated into different spatial scales. Monthly Weather Review, 134, 2180–2190.Find this resource:

Feser, F., Barcikowska, M., Krueger, O., Schenk, F., Weisse, R., & Xia, L. (2015). Storminess over the North Atlantic and Northwestern Europe—A review. Quarterly Journal of the Royal Meteorological Society, 141, 350–382.Find this resource:

Feser, F., & von Storch, H. (2008). A dynamical downscaling case study for typhoons in SE Asia using a regional climate model. Monthly Weather Review, 136(5), 1806–1815.Find this resource:

Fischer-Bruns, I., von Storch, H., González-Rouco, J. F., & Zorita, E. (2005). Modelling the variability of midlatitude storm activity on decadal to century time scales. Climate Dynamics, 25, 461–476.Find this resource:

Gates, W. L. (2003). Ein kurzer Überblick über die Geschichte der Klimamodellierung. Promet, 29(1–4), 3–5.Find this resource:

Geng, Q., & Sugi, M. (2003). Possible change of extratropical cyclone activity due to enhanced greenhouse gases and sulfate aerosols—study with a high-resolution AGCM. Journal of Climate, 16, 2262–2274.Find this resource:

Gill, G. C. (1975). Development and use of the Gill UVW anemometer. Boundary-Layer Meteorology, 8, 475–495.Find this resource:

Hall, N. M. J., Hoskins, B. J., Valdes. P. J., & Senior, C. A. (1994). Storm tracks in a high-resolution GCM with doubled carbon dioxide. Quarterly Journal of the Royal Meteorological Society, 120, 1209–1230.Find this resource:

Harnik, L., & Chang, E. K. M. (2003). Storm track variations as seen in radiosonde observations and reanalysis data. Journal of Climate, 16, 480–495.Find this resource:

Harvey, B. J., Shaffrey, L. C., Woollings, T. J., Zappa, G., & Hodges, K. I. (2012). How large are projected 21st century storm track changes? Geophysical Research Letters, 39.Find this resource:

Hickey, K. R. (2003). The storminess record from Armagh Observatory, Northern Ireland, 1796–1999. Weather, 58(1), 28–35.Find this resource:

Hughes, P. (1970). A century of weather service. New York: Gordon and Breach Science Publishers.Find this resource:

Intergovernmental Panel on Climate Change (IPCC). (2000). Special report on emissions scenarios: A special report of Working Group III of the Intergovernmental Panel on Climate Change. Cambridge, UK, and New York: Cambridge University Press.Find this resource:

Intergovernmental Panel on Climate Change (IPCC). (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. C. B. Field, V. Barros, T. F. Stocker, D. Qin, D. J. Dokken, K. L. Ebi, . . . P. M. Midgley (Eds.). Cambridge, UK, and New York: Cambridge University Press.Find this resource:

Intergovernmental Panel on Climate Change (IPCC). (2014). Climate change 2014: Synthesis report. Contribution of Working Groups I, II, and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change. Core Writing Team, R. K. Pachauri, & L. A. Meyer (Eds.). Geneva, Switzerland: Intergovernmental Panel on Climate Change (IPCC). Available online.Find this resource:

Jacobson, M. Z. (2005). Fundamentals of atmospheric modeling (2nd ed.). New York: Cambridge University Press.Find this resource:

Jones, P. D., Jonsson, T., & Wheeler, D. (1997). Extension to the North Atlantic oscillation using early instrumental pressure observations from Gibraltar and south-west Iceland. International Journal of Climatology, 17, 1433–1450.Find this resource:

Knippertz, P., Ulbrich, U., & Speth, P. (2000). Changing cyclones and surface wind speeds over the North Atlantic and Europe in a transient GHG experiment. Climate Research, 15, 109–122.Find this resource:

Knutson, T. R., Mcbride, J. I., Chan, J., Emanuel, K., Holland, G., Landsea, C., . . . Sugi, M. (2010). Tropical cyclones and climate change. Nature Geoscience, 3, 157–163.Find this resource:

Krueger, O., & von Storch, H. (2011). Evaluation of an air pressure–based proxy for storm activity. Journal of Climate, 24, 2612–2619.Find this resource:

Krueger, O., & von Storch, H. (2012). The informational value of pressure-based single-station proxies for storm activity. Journal of Atmospheric Oceanic Technology, 29, 569–580.Find this resource:

Küttel, M., Xoplaki, E., Gallego, D., Luterbacher, J., García-Herrera, R., Allan, R., . . . Wanner, H. (2010). The importance of ship log data: Reconstructing North Atlantic, European, and Mediterranean sea level pressure fields back to 1750. Climate Dynamics, 34, 1115–1128.Find this resource:

Lambert, S. J. (2004). Changes in winter cyclone frequencies and strengths in transient enhanced greenhouse warming simulations using two coupled climate models. Atmosphere-Ocean, 42(3), 173–181.Find this resource:

Lambert, S. J., & Fyfe, J. C. (2006). Changes in winter cyclone frequencies and strengths simulated in enhanced greenhouse warming experiments: Results from the models participating in the IPCC diagnostic exercise. Climate Dynamics, 26, 713–728.Find this resource:

Landsea, C. . . . Anderson, C., Charles, N., Clark, G., Dunion, J., Fernández-Patagás, F., Hungerford, P., . . ., Zimmer, M. (2004). The Atlantic Hurricane Database Reanalysis Project: Documentation for 1851–1910 Alterations and Additions to the HURDAT Database. In Richard J. Murnane & Kam-biu Liu (Eds.), Hurricanes and typhoons: Past, present, and future (pp. 177–221). New York: Columbia University Press.Find this resource:

Leckebusch, G. C., Renggli, D., & Ulbrich, U. (2008). Development and application of an objective storm severity measure for the Northeast Atlantic region. Meteorologische Zeitschrift, 17, 575–587.Find this resource:

Leckebusch, G. C., & Ulbrich, U. (2004). On the relationship between cyclones and extreme windstorm events over Europe under climate change. Global and Planetary Change, 44, 181–193.Find this resource:

Lemon, L. R., & Doswell, C. A., III. (1979). Severe thunderstorm evolution and mesocyclone structure as related to tornadogenesis. Monthly Weather Review, 107, 1184–1197.Find this resource:

Lindenberg, J., Mengelkamp, H.-T., & Rosenhagen, G. (2012). Representativity of near surface wind measurements from coastal stations at the German Bight. Meteorologische Zeitschrift, 21, 099–106.Find this resource:

Matthes, K., Funke, B., Anderson, M. E., Barnard, L., Beer, J., Charbonneau, P., . . . Versick, S. (2017). Solar forcing for CMIP6 (v3.1). Geoscientific Model Development, 10, 2247–2302.Find this resource:

Matulla, C., Schöner, W., Alexandersson, H., von Storch, H., & Wang, X. (2007). European storminess: Late nineteenth century to present. Climate Dynamics, 31, 125–130.Find this resource:

McVicar, T. R., Roderick, M. L., Donohue, R. J., Li, L. T., Van Niel, T. G., Thomas, A., . . . Dinpashoh, Y. (2012). Global review and synthesis of trends in observed terrestrial near-surface wind speeds: Implications for evaporation. Journal of Hydrology, 416–417, 182–205.Find this resource:

National Oceanic and Atmospheric Administration (NOAA). (2007). A history of observing the weather. Retrieved from

Neu, U., Akperov, M. G., Bellenbaum, N., Benestad, R., Blender, R., Caballero, R., . . . Wernli, H. (2013). IMILAST: A community effort to intercompare extratropical cyclone detection and tracking algorithms. Bulletin of the American Meteorological Society, 94, 529–547.Find this resource:

Phillips, N. A. (1956). The general circulation of the atmosphere: A numerical experiment. Quarterly Journal of the Royal Meteorological Society, 82, 123–164.Find this resource:

Phillips, N. A. (1995). Jule Gregory Charney Biographical Memoirs V. 66. Washington, DC: National Academy of Sciences.Find this resource:

Pinto, J. G., Karremann, M. K., Born, K., Della-Marta, P. M., & Klawa, M. (2012). Loss potentials associated with European windstorms under future climate conditions. Climate Research, 54, 1–20.Find this resource:

Pinto, J. G., Neuhaus, C. P., Leckebusch, G. C., Reyers, M., & Kerschgens, M. (2010). Estimation of wind storm impacts over Western Germany under future climate conditions using a statistical–dynamical downscaling approach. Tellus A, 62, 188–201.Find this resource:

Pinto, J. G., Ulbrich, U., Leckebusch, G. C., Spangehl, T., Reyers, M., & Zacharias, S. (2007). Changes in storm track and cyclone activity in three SRES ensemble experiments with the ECHAM5/MPI-OM1 GCM. Climate Dynamics, 29, 195–210.Find this resource:

Pinto, J. G., Zacharias, S., Fink, A. H., Leckebusch, G. C., & Ulbrich, U. (2009). Factors contributing to the development of extreme North Atlantic cyclones and their relationship with the NAO. Climate Dynamics, 32, 711–737.Find this resource:

Raible, C., Della-Marta, P. M., Schwierz, C., Wernli, H., & Blender, R. (2008). Northern Hemisphere extratropical cyclones: A comparison of detection and tracking methods and different reanalyses. Monthly Weather Review, 136, 880–897.Find this resource:

Randall, D. A. (2000). General circulation model development: Past, present, and future. London: Academic Press.Find this resource:

Rasmussen, E. A., & Turner, J. (2003). Polar lows—Mesoscale weather systems in the polar regions. New York: Cambridge University Press.Find this resource:

Reichle, R. H. (2008). Data assimilation methods in the earth sciences. Advances in Water Resources, 31, 1411–1418.Find this resource:

Reimann, T., Tsukamoto, S., Harff, J., Osadczuk, K., & Frechen, M. (2011). Reconstruction of Holocene coastal foredune progradation using luminescence dating—An example from the Świna barrier (southern Baltic Sea, NW Poland). Geomorphology, 132, 1–16.Find this resource:

Rienecker, M. M., Suarez, M. J., Gelaro, R., Todling, R., Bacmeister, J., Liu, E., . . . Woollen, J. (2011). MERRA: NASA’s modern-era retrospective analysis for research and applications. Journal of Climate, 24, 3624–3648.Find this resource:

Saha, S., Moorthi, S., Pan, H.-L., Wu, X., Wang, J., Nadiga, S., . . . Goldberg, M. (2010). The NCEP Climate Forecast System reanalyses. Bulletin of the American Meteorological Society, 91, 1015–1057.Find this resource:

Saha, S., Nadiga, S., Thiaw, C., Wang, J., Wang, W., Zhang, Q., . . . Xie, P. (2006). The NCEP Climate Forecast System. Journal of Climate, 19, 3483–3517.Find this resource:

Schenk, F., & Zorita, E. (2012). Reconstruction of high resolution atmospheric fields for Northern Europe using analog-upscaling. Climate of the Past, 8, 1681–1703.Find this resource:

Schiesser, H. H., Pfister, C., & Bader, J. (1997). Winter storms in Switzerland North of the Alps 1864/1865–1993/1994. Theoretical and Applied Climatology, 58, 1–19.Find this resource:

Schmidt, H., & von Storch, H. (1993). German Bight storms analysed. Nature, 365(6449), 791.Find this resource:

Schmith, T. (1995). Occurrence of severe winds in Denmark during the past 100 years (pp. 83–86). Presented at the Proceedings of the Sixth International Meeting on Statistical Climatology, June 19–23, 1993, Galway, Ireland, All-Ireland Statistics Committee and Cosponsors, All-Ireland Statistics Committee and Cosponsors. Retrieved from this resource:

Schneidereit, A., Blender, R., Fraedrich, K., & Lunkeit, F. (2007). Icelandic climate and North Atlantic cyclones in ERA-40 reanalyses. Meteorologische Zeitschrift, 16, 17–23.Find this resource:

Schubert-Frisius, M., Feser, F., von Storch, H., & Rast, S. (2017). Optimal spectral nudging for global dynamic downscaling. Monthly Weather Review, 145, 909–927.Find this resource:

Sickmoeller, M., Blender R., & Fraedrich, K. (2000). Observed winter cyclone tracks in the northern hemisphere in re-analysed ECMWF data. Quarterly Journal of the Royal Meteorological Society, 126, 591–620.Find this resource:

Simmonds, I., & Keay, K. (2002). Surface fluxes of momentum and mechanical energy over the North Pacific and North Atlantic Oceans. Meteorology and Atmospheric Physics, 80, 1–18.Find this resource:

Simmons, A., Uppala, S., Dee, D., & Kobayashi, S. (2007). ERA-Interim: New ECMWF reanalysis products from 1989 onwards. ECMWF Newsletter, No. 110. 25–35.Find this resource:

Trigo, I. F. (2006). Climatology and interannual variability of storm-tracks in the Euro-Atlantic sector: a comparison between ERA-40 and NCEP/NCAR reanalyses. Climate Dynamics, 26, 127–143.Find this resource:

Ulbrich, U., & Christoph, M. (1999). A shift of the NAO and increasing storm track activity over Europe due to anthropogenic greenhouse gas forcing. Climate Dynamics, 15, 551–559.Find this resource:

Ulbrich, U., Pinto, J. G., Kupfer, H., Leckebusch, G. C., Spangehl, T., & Reyers, M. (2008). Changing Northern Hemisphere storm tracks in an ensemble of IPCC climate change simulations. Journal of Climate, 21, 1669–1679.Find this resource:

van Vuuren, D. P., Edmonds, J., Kainuma, M., Riahi, K., Thomson, A., Hibbard, K., . . . Rose, S. K. (2011). The representative concentration pathways: An overview. Climatic Change, 109(5), 5–31.Find this resource:

Vautard, R., Cattiaux, J., Yiou, P., Thépaut, J.-N., & Ciais, P. (2010). Northern Hemisphere atmospheric stilling partly attributed to an increase in surface roughness. Nature Geoscience, 3, 756–761.Find this resource:

von Storch, H., Langenberg, H., & Feser, F. (2000). A spectral nudging technique for dynamical downscaling purposes. Monthly Weather Review, 128, 3664–3673.Find this resource:

Wagner, D., Tinz, B., & von Storch, H. (2016). Signal stations: Newly digitized historical climate data of the German Bight and the southern Baltic Sea coasts. Journal of Atmospheric Ocean Technology, 33, 2735–2741.Find this resource:

Waldron, K. M., Paegle, J., & Horel, J. D. (1996). Sensitivity of a spectrally filtered and nudged limited-area model to outer model options. Monthly Weather Review, 124, 529–547.Find this resource:

Wang, X., Zwiers, F., Swail, V., & Feng, Y. (2009). Trends and variability of storminess in the Northeast Atlantic region, 1874–2007. Climate Dynamics, 33(7–8), 1179–1195.Find this resource:

Wang, X. L., Swail, V. R., & Zwiers, F. W. (2006). Climatology and changes of extratropical cyclone activity: comparison of ERA40 with NCEP-NCAR reanalysis for 1958–2001. Journal of Climate, 19, 3145–3166.Find this resource:

Weisse, R., & Feser, F. (2003). Evaluation of a method to reduce uncertainty in wind hindcasts performed with regional atmosphere models. Coastal Engineering, 48, 211–225.Find this resource:

Weisse, R., & von Storch, H. (2009). Marine climate and climate change. Storms, wind waves, and storm surges. Heidelberg, Germany: Springer Praxis.Find this resource:

Weisse, R., von Storch, H., & Feser, F. (2005). Northeast Atlantic and North Sea storminess as simulated by a regional climate model during 1958–2001 and comparison with observations. Journal of Climate, 18, 465–479.Find this resource:

Woollings, T., Gregory, J. M., Pinto, J. G., Reyers, M., & Brayshaw, D. J. (2012). Response of the North Atlantic storm track to climate change shaped by ocean–atmosphere coupling. Nature Geoscience, 5, 313–317.Find this resource:

Xia, L., von Storch, H., & Feser, F. (2013). Quasi-stationarity of centennial Northern Hemisphere midlatitude winter storm tracks. Climate Dynamics, 41, 901–916.Find this resource:

Xia, L., von Storch, H., Feser, F., & Wu, J. (2016). A study of quasi-millennial extratropical winter cyclone activity over the Southern Hemisphere. Climate Dynamics, 47, 2121–2138.Find this resource:

Yin, J. H. (2005). A consistent poleward shift of the storm tracks in simulations of 21st century climate. Geophysical Research Letters, 32(18), L18701.Find this resource:

Zahn, M., & von Storch, H. (2008). A long-term climatology of North Atlantic polar lows. Geophysical Research Letters, 35.Find this resource:

Zahn, M., & von Storch, H. (2010). Decreased frequency of North Atlantic polar lows associated with future climate warming. Nature, 467, 309–312.Find this resource:

Zahn, M., von Storch, H., & Bakan, S. (2008). Climate mode simulation of North Atlantic polar lows in a limited area model. Tellus Series A—Dynamic Meteorology and Oceanography, 60(4), 620–631.Find this resource:

Zappa, G., Shaffrey, L. C., Hodges, K. I., Sansom, P. G., & Stephenson, D. B. (2013). A multi-model assessment of future projections of North Atlantic and European extratropical cyclones in the CMIP5 climate models. Journal of Climate, 26, 5846–5862Find this resource: