Show Summary Details

Page of

date: 15 August 2018

# Climate Change and Severe Thunderstorms

## Summary and Keywords

The response of severe thunderstorms to a changing climate is a rapidly growing area of research. Severe thunderstorms are one of the largest contributors to global losses in excess of USD $10 billion per year in terms of property and agriculture, as well as dozens of fatalities. Phenomena associated with severe thunderstorms such as large hail (greater than 2 cm), damaging winds (greater than 90 kmh−1), and tornadoes pose a global threat, and have been documented on every continent except Antarctica. Limitations of observational records for assessing past trends have driven a variety of approaches to not only characterize the past occurrence but provide a baseline against which future projections can be interpreted. These proxy methods have included using environments or conditions favorable to the development of thunderstorms and directly simulating storm updrafts using dynamic downscaling. Both methodologies have demonstrated pronounced changes to the frequency of days producing severe thunderstorms. Major impacts of a strongly warmed climate include a general increase in the length of the season in both the fall and spring associated with increased thermal instability and increased frequency of severe days by the late 21st century. While earlier studies noted changes to vertical wind shear decreasing frequency, recent studies have illustrated that this change appears not to coincide with days which are unstable. Questions remain as to whether the likelihood of storm initiation decreases, whether all storms which now produce severe weather will maintain their physical structure in a warmer world, and how these changes to storm frequency and or intensity may manifest for each of the threats posed by tornadoes, hail, and damaging winds. Expansion of the existing understanding globally is identified as an area of needed future research, together with meaningful consideration of both the influence of climate variability and indirect implications of anthropogenic modification of the physical environment. Few phenomena evoke reactions like the visual appearance of severe thunderstorms; the falling of hail larger than a golf ball, the almost capricious movements of a tornado that strikes one house but not another, or the blinding curtains of wind and rain of a damaging windstorm. The most consistent formal definition for a severe thunderstorm is related to its production of one or more episodes of hail in excess of 2 cm, damaging winds in excess of 90 kmh−1, and tornadoes, which collectively here are referred to as severe convective storms (SCS; Doswell, 2016). These events have a substantial impact, leading to total economic losses exceeding USD$10 billion per year over the United States and €1–2 billion in Europe (Hoeppe, 2016), with loss frequency increasing on both continents.

The occurrence of severe thunderstorms cannot be assumed to be a stationary quantity, and how these severe thunderstorms will respond to a warming and potentially more variable climate is a growing cause for concern and a research interest. Current projections of moderate to high warming scenarios for 2100 from the Coupled Model Intercomparison Project version 5 (CMIP5; Taylor et al., 2012) suggest increasing mean surface temperatures by a range of 4°C–6°C. This rise in the near surface air temperature increases its capacity to hold water, with the relative increases dictated by the Clausius-Clapeyron relationship. Recent theoretical studies leveraging this relationship have suggested that under highly warming scenarios of 4°C the corresponding response in atmospheric moisture content would be an increase of 25%, where sufficient moisture is available (Gettleman et al., 2002; Dai, 2006; Willett et al., 2010; Romps, 2016; Agard & Emanuel, 2017). Increasing the near surface temperature and moisture results in increasing vertical thermodynamic instability. This instability is typically parameterized as convective available potential energy (CAPE), which can be realized either through air being absolutely unstable, or requiring its lifting to a level where it becomes relatively buoyant compared to its surroundings. The importance of increases to moisture cannot be understated, as much of this potential energy is realized by phase changes of water as air rises and cools (Emanuel, 1994). Increasing the available instability leads to larger relative differences between the rising air and its surroundings, and thus the maximum potential updraft speed (wmax) increases owing to its relationship to CAPE ($wmax=2×CAPE$). This increase in updraft velocity is suggested to favor increasingly severe storms, and potentially more severe storm days as instability increases, even on days in the winter and early spring which are typically stable in the present climate (Diffenbaugh et al., 2013; Hoogewind et al., 2017). However, whether such an increase is realized by stronger storms may be highly dependent on the relative warming of the mid-troposphere, as any warming in this layer may offset gains due to near surface instability. Additionally, it is unclear whether updraft velocity would be offset by the commensurately higher concentration of hydrometeors within the storm (Trapp & Hoogewind, 2016). Mid-tropospheric warming would also promote increasing strength of capping inversions, and thus resistance to convection in the form of convective inhibition (CIN), reducing the likelihood of storms initiating.

Considering individual phenomena, for example hail, the warming of mid-tropospheric temperatures likely leads to an increase in the freezing level, which leads to increased melting of smaller hailstones. However, there may be some offset, as stronger updrafts driven by increasing CAPE would favor the growth of larger hailstones, which melt less when falling. Warming the boundary layer and mid-troposphere also has other implications, particularly between polar regions and the midlatitudes, where decreasing meridional thermal gradients result in a reduction in wind speeds, and thus may influence how wind speed and direction change with height. These characteristics are essential to the development of deep-layer vertical wind shear (DLS), which plays an important role in the formation and maintenance of organized severe thunderstorms, such as squall lines and supercells, that produce the majority of observed tornadoes and hailstones. The uncertainty that arises from the balance between these factors, the disconnect in scale between SCS (generally tens of kilometers and lasting for several hours) and the global climate system as encapsulated by climate models (grid scales of 100km or more, and only four analyses daily at 00, 06, 12, and 18UTC), has led to considerable challenges in estimating potential changes. Combining this uncertainty with large degrees of interannual variability and sensitivity to natural variations, temporally and spatially limited observational records, and sensitivity to other anthropogenic modification of the environment has made distinguishing near-term climate change signals difficult to isolate (Trapp et al., 2009).

To provide a concise overview of the impacts of climate change on SCS, it is critical to understand the evolution of this field since the early 1990s and the limitations in data availability and computational power that have forced the field to evolve in the manner that it has. No such review would be complete without a discussion of the limitations of the observations of SCS. Building from this basis, the necessity of using environmental conditions favorable to the development of SCS as a proxy to estimate occurrence is clearer. These two concepts are then used to explore the range of projections globally for a changing climate in terms of both near-term trends in SCS observations and environments and projections of future changes to these environments using climate models. This article will also discuss the latest and most promising evolutions in the field, the application of dynamic downscaling to attempt to directly simulate SCS in future climes, limitations relating to the need to widen these investigations globally, the influence of variability on projected future climates, the need to focus on specific hazards, and indirect influences by anthropogenic sources.

# Evolution of the Field

Thunderstorms are one of the most widely distributed phenomena on the planet within the troposphere, and represent an important mechanism whereby excess heat close to the surface of the planet is distributed vertically through the atmospheric column. This process resolves the development of unstably stratified air (relatively warm near the surface, and cooler aloft) between the lowest portions of the atmosphere and the mid-lower troposphere, common within the midlatitudes. The latent energy released by the condensation of water from rising air constitutes a large contribution to the updraft strength (Emanuel, 1994). While much of this potential is offset by gravity, hydrometeor loading, or pressure perturbations, it is a necessary characteristic to produce convective storms. Yet, as widely spread and frequent as thunderstorms are, only a small subset of these storms produce damaging phenomena, including large hail in excess of 2 cm, damaging winds in excess of 90 km h−1, tornadoes, and excessive precipitation. Such storms are commonly referred to as SCS, though in some countries excessive precipitation is included in this designation. Here, precipitation is excluded, as many non-convective atmospheric processes are capable of producing excessive precipitation (Doswell et al., 1996; Doswell, 2001; Doswell, 2016). In an attempt to isolate storms that pose the greatest risk to the human and built environment, a higher category and subset of SCS is also defined as significant severe thunderstorms (SigSCS; Hales, 1993; Brooks et al., 2003), producing hail in excess of 5 cm, winds in excess of 120 kmh−1, and tornadoes of F2 intensity of greater. For a more comprehensive review of the formative processes and mechanisms leading to SCS and convection, see Trapp (2017).

# Observations and Climatology of Severe Thunderstorms

The limitations of observations have led to the application of radar- or satellite-derived datasets, which are a growing area of SCS climatology (e.g., Cecil & Blankenship, 2012; Cinteneo et al., 2012; Punge & Kunz, 2016; Ni et al., 2017). Such datasets are temporally limited to just over a decade due to changes in quality of both the radar network or satellite sensors, and rely on proxies that relate the observed feature to severe storm occurrence rather than direct observations of SCS. Nonetheless they provide records with fewer spatial limitations for estimating climatology, for example, on a global basis (e.g., Cecil & Blankenship, 2012). When applied with conscious acknowledgement of their limitations, these sources do provide validation of characteristic biases in the surface observational datasets. For example, Soderholm et al. (2017) identified that in the last 18 years, there is little trend in hailstorms in the Brisbane, Australia, region outside of natural variability, a result consistent with the results derived from favorable environmental conditions for SCS (Allen & Karoly, 2014). In the absence of technological changes, it is likely that in the coming decades, these records will be a valuable source of reliable observations for assessing past change or variability of SCS, particularly where observations are sparse or population density is low. However, the currently short temporal length of the record makes these remote sources a challenge to use for the analysis of trends, and hence no further mention of these records will be made here.

Table 1. Illustrative Climatologies For Severe Thunderstorm Observations Around the World

Author

Period

Country

Source

Phenomena

Global

Frisby and Sansom (1967)

Various

Obs.

Hail

Fujita (1973)

Various

Obs.

Goliger and Milford (1998)

Various

Obs.

Brooks and Doswell (2001)

Various

Obs.

Zipser et al. (2006)

1998–2004

Satellite

Intense Storms

Cecil and Blankenship (2012)

2003–2010

Satellite

Hail

Ni et al. (2017)

1999–2014

Satellite

Hail

Europe

Dotzek et al. (2009)

Europe

Obs.

Severe

Punge and Kunz (2016) Rev.

Various

Europe

Hail

Antonsecu et al. (2016, 2017)

1800–2014

1950–2015

Europe

Obs.

Groenemeijer et al. (2017)

Europe

Obs.

Severe

Asia

Peterson and Mehta (1981)

1835–1981

Subcontinent

Obs.

Nizamuddin (1993)

India

Obs.

Hail

Bhan et al. (2016)

1975–2010

India

Obs.

Zixiu et al. (1993)

China

Obs.

Severe

Zhang et al. (2008)

1960–2005

China

Obs.

Hail

Yao et al. (2015)

1960–2009

China

Obs.

Banafsheh et al. (2016)

1993–2014

Iran

Obs.

Hail

Jin et al. (2017)

1972–2013

South Korea

Obs.

Hail

North America

Kelly et al. (1985)

1955–1983

U.S.

Obs.

Severe

Grazulis (1993)

1680–1991

U.S.

Obs.

Doswell et al. (2005)

1980–1994

U.S.

Obs.

Hail & Wind

Verbout et al. (2006)

1954–2003

U.S.

Obs.

Cintineo et al. (2012)

2007–2010

U.S.

Hail

Allen and Tippett (2015)

1955–2014

U.S.

Obs.

Hail

King (1997)

1918–1992

Obs.

Smith et al. (1998)

1957–1985

Obs.

Hail

Etkin and Brun (1999)

1950–1981

Obs.

Hail

Etkin et al. (2001)

1980–1997

Obs.

Cheng et al. (2013)

1980–2009

Obs.

South America

Silva Dias (2011)

1960–2008

Brazil

Obs.

Martins et al. (2017)

1991–2012

Brazil

Obs.

Hail

Mezher et al. (2012)

1960–2008

Argentina

Obs.

Hail

Rasmussen et al. (2014)

1998–2013

S. America

Obs., Sat.

Severe

Oceania

McMaster (2001)

1795–1998

Australia

Obs.

Hail

Schuster et al. (2005)

1791–2003

Australia

Obs.

Hail

Niall and Walsh (2005)

1958–2003

Australia

Obs./Ins.

Hail

Allen and Allen (2016)

1795–2014

Australia

Obs.

Soderholm et al. (2017)

1997–2015

Australia

Hail

Africa

S. Africa

Obs.

Hail

Smith et al. (1998)

1962–1982

S. Africa

Obs.

Hail

Note: Readers are directed to these studies and references therein for complete details of the available observations

## Historical Development

Speculation as to the influence of a warming climate on SCS frequency and intensity has persisted since the early 1990s, as there is a propensity for SCS to cause threats to human life and property, and any change to the frequency or intensity of events thus will have significant implications (Gates et al., 1990; Balling & Cerveny, 2003; Botzen & Bouwer, 2016; Hoeppe, 2016). As noted by Brooks (2013), this topic became popular through the mid-late 2000s, as increasingly higher resolution Global Climate Models (GCMs) became capable of simulating the large-scale ingredients that produce severe thunderstorms (e.g., Niall & Walsh, 2005; Diffenbaugh et al., 2008; Diffenbaugh et al., 2013; Hoogewind et al., 2017), but not the storms themselves, a feature that precluded earlier analyses with coarse model data (Gates et al., 1990; Cubasch et al., 2001). This interest has only continued to accelerate since 2013, with an increasing number of studies directly addressing the topic. The field has also expanded to reflect a larger number of regions examining changes to severe thunderstorm frequency in the present climate and future projections thereof, including analyses for Europe, Australia, and Japan in addition to increasingly more detailed analyses over North America.

From a limited understanding as recently as 2007, the field has evolved remarkably to reflect a diverse and wide body of research, though a large number of questions remain unasked and unanswered. A challenge of the interdisciplinary nature of this area of research is the tendency for researchers to be focused in their training predominantly either on the mesoscale (<400km, <1 day) or on the synoptic to climate scale (>400km and on periods of days to decades). Understanding of the interaction between the climate system and the mesoscale is dependent on understanding the energy cascade, where planetary to regional climate signals that vary from periods of weeks to years modulate the positioning and frequency of jet streams and baroclinic systems that create conditions favorable to the development of SCS, and the upscale propagation of energy into the synoptic and climate scales as convectively generated energy derived from the latent heat conversion from water vapor to condensate is released. The evolution of this field has occurred as a response to bringing the statistical and large-scale process perspective of researchers on the synoptic and climate scales together with a more forecast- and process-based understanding common for the mesoscale.

Perhaps the first mention of the potential impact of the climate system on SCS originates with the work of Gates et al. (1990), who described the potential to identify the impacts of climate change on SCS using changes in favorable environmental conditions. A subsequent study by Griffiths et al. (1993) proposed four methods to assess this influence based on the limitations of the understanding of SCS and the extremely limited climate models of the day:

1. (i) Correlating SCS climatology with general circulation indices such as El Niño-Southern Oscillation (ENSO) in the present and projecting changes in the general circulation features for the future.

2. (ii) Use of climate models to assess the present and future frequency of environments favorable to the occurrence of SCS such as CAPE and DLS (based on the suggestions of Gates et al., 1990).

3. (iii) Comparison of the present and future frequencies of synoptic patterns which produce SCS and their associated environments.

4. (iv) Direct simulation of SCS processes and storms over a limited region using nested higher-resolution model runs. These simulations draw lateral boundary conditions from climate models, akin to those used for precipitation studies (Giorgi, 1990), though requiring the further step of resolving of storm scale processes.

Despite this forward-thinking vision, climate models of the time were not capable of representing the environmental conditions associated with SCS. Thus even to produce useful environmental (ii) or synoptic conditions (iii), extremely coarse resolution of models required the downscaling of the GCM data using regional nests, a computationally expensive process (Griffiths et al., 1993; Ye et al., 1998; McMaster, 1999). Concurrently, direct simulation was only a new concept for numerical weather prediction in an operational context, and no resources were available to even approach the 4 km convective scale needed to accomplish (iv).

This led to correlations with general circulation indices (i), for example average minimum nighttime temperature being used to infer changes to hail frequency based for France (Dessens, 1995), which yielded an estimate of 40% change in hailstorm frequency for every degree of warming. Moving beyond elementary correlations, the answer to this challenge came by applying the concept of drawing one-dimensional thermodynamic profiles from climate model data (McMaster, 1999), similar to proximity soundings, where an atmospheric profile is sampled from a nearby rawinsonde station in both space and time to a severe weather event (e.g., Rasmussen & Blanchard, 1998). The technique, applied in conjunction with model or reanalysis data of sufficient spatial resolution, provided a meaningful description of the local atmospheric state relevant to SCS (e.g., Ye et al., 1998; McMaster, 1999; Brooks et al., 2003; Niall & Walsh, 2005). This evolution was necessary in order to provide the tools to make the assessment of favorable SCS environments described in (ii) by Griffiths et al. (1993) possible.

## Environmental Proxies and Favorable Conditions

To understand the basis for environmental proxies representing favorable conditions for SCS, concepts drawn from mesoscale forecasting are needed. At its origin, surface observations for analyzing the mesoscale (Fujita, 1963) and balloon-launched soundings have been used operationally to characterize conditions favorable to the development of severe thunderstorms. Applying the concepts of proximity soundings to large numbers of SCS observations has regularly been used to produce forecasting guidelines for favorable environmental conditions (e.g., Rasmussen & Blanchard, 1998; Rasmussen, 2003; Craven et al., 2002). To produce statistically robust predictands, a sufficiently large sample of SCS with proximal in space and time environmental conditions is needed to identify characteristic conditions favorable to a given phenomenon, the ingredients-based approach (Doswell et al., 1996). Analyzing these parameters, or combining multiple meteorological covariates to limit the favorable parameter space, allows production of a parameter exceedance to approximate incomplete report data (Brown & Murphy, 1996). In a similar vein, the relationship between synoptic patterns and SCS have been popular in Europe (e.g., Kapsch et al., 2012), but their application over the United States has been limited, as these factors are more relevant in the spring, while land surface and mesoscale features tend to dominate in the summer months (Doswell & Bosart, 2001).

Applying these proximity soundings to model output, as suggested by both Gates et al. (1990) and Griffiths et al. (1993), was uncommon until 2003, when Brooks et al. (2003) realized the potential of model-analyzed states of the historical atmosphere to be used for the climatology of severe thunderstorms where the observed frequency was unknown. The linchpin to this leap forward was the relating of rawinsonde to reanalysis-derived proximity soundings and the illustration that thermodynamic and kinematic parameters were reasonably well represented by model data (McMaster, 1999; Lee, 2002; Niall & Walsh, 2005), despite many of the earlier reanalyses having biases in environmental moisture (Trenberth & Guillemot, 1998). Subsequent analysis has shown the utility of reanalyses with later-generation, higher-resolution products (Niall & Walsh, 2005; Allen et al., 2011; Allen & Karoly, 2014; Gensini et al., 2014a; Pucik et al., 2015). Generally, reanalyses have been shown to be better at representing kinematic parameters, with the performance for CAPE varying regionally and limited by sensitivity to the convective parcel. These factors have led to the mixed-layer average parcel being favored, though problems remain with the grid scale of the reanalysis and the rendition of boundary layer moisture, particularly over continental interiors and warm coastal features (Lannici & Warner, 1991; Allen & Karoly, 2014; Pucik et al., 2015). Another limitation of model and reanalysis data is the poor handling of the sharp changes in vertical temperature associated with the interface between the environmental mixed layer and boundary layer, a region of the vertical profile which leads to the formation of convective inhibition (Lanicci & Warner, 1991; Allen & Karoly, 2014; Gensini et al., 2014a).

Brooks et al. (2003) used proximity soundings from the NCEP/NCAR reanalysis (Kalnay et al., 1996) for an observed set of SCS, SigSCS, and tornadoes to describe discriminants for conditions favorable to these events. This illustrated a methodology that allowed departure from the flawed observational records and allowed for exploration of proxy SCS occurrence for regions with fewer observations. Applying this methodology to global data from the same reanalysis, this established the first long-term climatology of environments favorable to the development of severe thunderstorms and tornadoes not only over the United States, but also globally (Figure 1).

Click to view larger

Figure 1. Global distribution of mean annual Brooks tornado environments 1958–1999.

Adopted from Tippett et al. (2015), their Figure 5, based on the original equation from Brooks et al. (2003) utilizing data from NCEP/NCAR reanalysis.

These results provide an important baseline in our understanding of the frequency of severe thunderstorms in a global context, and until the recent work of Cecil and Blankenship (2012) and subsequent studies using remote sensing data has been difficult to verify due to observational sparsity. The environment discriminants of Brooks et al. (2003) were later refined on independent data from Australia using linear discriminant analysis by Allen et al. (2011), which suggested that a meteorological covariate combining both DLS (in this case 0–6km AGL, S06), and 50mb or 100mb Mixed-Layer CAPE (MLCAPE) would be the most capable of distinguishing between observed SCS and SigSCS, and defining the edge of the environmental phase space favorable to SigSCS such that:

$Display mathematics$

The S06 term has been used in various different weights in subsequent studies (e.g., Marsh et al., 2007; Trapp et al., 2007a; Gensini & Ashley, 2011; Allen & Karoly, 2014; Seeley & Romps, 2015; Viceto et al., 2017; Púčik et al., 2017). However, from a physical standpoint the phase space relationship shows a stronger relationship between shear and severity, which reflects the increasing likelihood of SigSCS with increasing shear and resulting storm organization (e.g., for hail; Dennis & Kumjian, 2017), and thus the S06 term is more sensibly weighted (Brooks, 2013; Blamey et al., 2017). Here, we refer to this myriad of products between CAPE and S06 as SEV environments. Other parameters commonly applied include 2–4 km lapse rates of temperature and convective inhibition (CIN). Analyses of environmental frequency and rarity can also be made using extreme value theory (Heaton et al., 2011; Gilleland et al., 2013; Mannshardt & Gilleland, 2013) and the product of CAPE and S06. Finally, while proxies can provide a useful indication, there are no specific thresholds that can give an ideal discrimination between SCS and non-SCS favorable environments (Doswell & Schultz, 2006). Further limitations to this approach are the non-conditionality of environments, in that even if an environment is favorable, there is no guarantee that a storm will form within the environment. Part of this problem lies in the incapability of reanalyses to resolve the finer-scale initiation processes, and draws in part from insufficient vertical resolution that produces issues with the measurements of convective inhibition (Brooks et al., 2003; Allen & Karoly, 2014; Gensini et al., 2014a). Convective parameterizations are applied in order to overcome the inability of the model’s resolution to realistically resolve the convective and microphysical processes within clouds (Houze, 1997), but often result in premature depreciation of CAPE, or in some cases its total elimination (Marsh et al., 2007; Allen et al., 2014a). Despite these limitations, reanalyses have been widely applied to estimate the climatological frequency of SCS (Table 2).

Whether environmental discriminants are applicable worldwide is also a relevant question (e.g., Dessens, 1995; Brooks et al., 2003), as many such relationships are developed using the high-density observations of the United States, and thus only sample a subset of favorable environments. For example, a large portion of the Australian continent lies closer to the equator than any location in the United States. In contrast, Europe is considerably northwardly displaced relative to the United States, suggesting the potential for out-of-sample environments in both locations, whereas South Africa is more similar to Australia. Nonetheless, independent environment relationships for both Australia and Europe (Brooks, 2009; Allen et al., 2011; Allen & Karoly, 2014), and subsequent larger proximity studies (Brooks, 2009; Pucik et al., 2015; Blamey et al., 2017) suggest that similar environmental parameters occur on a global basis, and the frequency of occurrence is the main source of variation for similar types of environment. For example, Brooks (2009) noted similar environmental spaces for SCS and tornadoes for Europe and the United States between both continents, but showed that the conditional probability in Europe had a higher likelihood of producing severe tornadoes, which was speculated to be a result of forcing. These similarities do not exclude the possibility that favorable environments of different parameter spaces do not exist in other parts of the world (e.g., cool season or low CAPE tornadoes; Hanstrum et al., 2002; Sherburn et al., 2016; or environments as yet unknown).

Table 2. Examples of Studies Using Environmental Proxies for Climatology to Estimate Frequency Through Time

Author

Period

Source

Region

Phenomena

Brooks et al. (2003)

1997–1999

NCEP/NCAR

US & Global

Severe/Tor

Niall and Walsh (2005)

1958–2003

NCEP/NCAR

Australia

Hail

Romero et al. (2007)

1970–2000

ERA-40

Europe

Sev/Tor

Riemann-Campe et al. (2011)

1958–2001

ERA-40

Global

Brooks (2009)

1958–1999

NCEP/NCAR

Europe & US

Sev/Tor

Gensini and Ashley (2011)

1980–2009

NARR

US

Severe

Hand and Cappelluti (2011)

2004–2008

UK Met GCM

Global

Hail

Tippett et al. (2012)

1979–2011

NARR

US

Mannshardt and Gilleland (2013)

1958–99

NCEP/NCAR

US

Severe

Gilleland et al. (2013)

1958–99

NCEP/NCAR

US

Severe

Allen and Karoly (2014)

1979–2011

ERA-Interim

Australia

Severe

Tippett et al. (2014)

1979–2012

NARR

US

Allen et al. (2015a)

1979–2012

NARR

US

Hail

Cheng et al. (2015)

1980–2009

NARR

North America

Mohr et al. (2015)

1971–2000

COSMO

Europe

Hail

Cheng et al. (2016)

1980–2009

NARR

North America

Blamey et al. (2017)

1979–2010

CFSR

South Africa

Hail

Westermayer et al. (2017)

2008–2013

ERA-Interim

Europe

H/W/T

Púčik et al. (2017)

1981–2000

ERA-Interim

Europe

Severe

## Applications of Environmental Proxies and Recent Developments

As climate models have evolved and the length of reanalysis records has increased, assessing long-term trends, sources of variability, and implications of warming has become more plausible. Where the field has shown most promise is in the capability to simulate SCS environments on the storm scale using high-resolution dynamic downscaling, with large-scale model data used as boundary conditions for running a model with mesoscale resolution (Griffiths et al., 1993). This concept had been applied to other fields, but arguably made the most sense for severe thunderstorms (e.g., Trapp et al., 2007b; Leslie et al., 2008). Taking this step to simulate storms and infer their severity, instead of just the proximity environments, has undergone considerable leaps since the field evolved (Trapp et al., 2007b; Gensini & Mote, 2015; Trapp & Hoogewind, 2016; Hoogewind et al., 2017). This approach provides an improvement on favorable environments and thus a more realistic climatology, as environmental conditions are necessary but not sufficient to produce SCS, leading to environment-only methodologies highlighting an excess of favorable days that might never yield a thunderstorm. This raises the specter of the most challenging aspect of our understanding of SCS, initiation. Even on a day-to-day basis, severe-weather forecasters struggle to identify where and when thunderstorms may initiate, and if they do, whether they will be long-lived and severe. This challenge is magnified with climate scale data, as factors leading to initiation have been shown to be on the scale of less than 1 to 10s of kilometers (e.g., Weckwerth & Parsons, 2006), beyond the capabilities of climate models, and at the edge of skill for downscaled models with non-explicit microphysics. This presents a considerable hurdle to the field and an ongoing question as to whether the rate at which storms initiate will change. While recent work shows that initiation becomes less frequent for significant tornadic storms, it remains unclear how this may manifest for more moderate or marginal days (Hoogewind et al., 2017). It is also questionable whether initiation occurring identically matters in dynamically downscaled simulations. Provided that storms initiate realistically on a larger scale (as is the case when such models are used in the operational forecast context), and overly fine-scale spatial changes are not necessary, aggregate statistics and their secular changes will yield a useful understanding of the potential for SCS using this methodology.

Vertical resolution presents another challenge in understanding the response of SCS to changes in initiation. While reanalyses often have higher resolution as they reflect fixed versions of forecast models, many climate models have only limited vertical resolution between the boundary layer and 500 hPa. This can lead to insufficient sampling of the capping inversion that occurs within the profile to allow the buildup of instability. In regions of the United States such as southern Texas, or over much of the Great Plains during the summer, such capping is prevalent and important in its regulation of SCS severity, and similar patterns occur worldwide. If the atmosphere is uncapped, widespread storms can occur, which can quickly mix out thermal instability and lead to rain cooling of the surface, precluding further SCS development. However, if such an air mass is given time to recover, for example, the backing of winds from east to south in the United States and pre-moistening from earlier rain can lead to even more explosive convective development, as well as favorable conditions for the production of tornadoes.

Another driving evolution in recent years in the field has been the change in focus from exploring environments favorable to all SCS to leverage sample size from small observational datasets to refining these metrics to individual hazards such as tornadoes, convective wind gusts, and hail (e.g., Tippett et al., 2012, 2014; Allen et al., 2015a; Mohr et al., 2015; Muramatsu et al., 2016; Westermayer et al., 2017). While large-scale application of these refinements is still in progress, leveraging the differences that distinguish between SCS phenomena has yielded considerable gains and illustrated that not all of these phenomena are likely to respond to a warming or variable climate in the same way (Brooks, 2013). Building upon these differences, another pressing question is that of storm mode (Diffenbaugh et al., 2013; Allen et al., 2014b; Trapp & Hoogewind, 2016). Different structures of severe thunderstorms pose a likelihood of producing different hazards. For example, hail is produced in many thunderstorms, but the largest hail (>5 cm), typically associated with the greatest impact for the built and anthropogenic environment, is produced almost solely by supercells (Blair et al., 2017). In contrast, damaging winds can be produced by a variety of favorable environments, ranging from high-based thunderstorms in dry environments driven solely by buoyancy to those favorable to supercells, as well as in squall lines or mesoscale convective systems. The question of storm mode is not easily solved by querying the environment, and to some extent is dependent on the microphysics of the storm, below the grid-scaling of even the finest dynamic downscaling conducted.

# Assessing a Changing Frequency

The techniques that have been developed over the past decades provide a number of ways to identify the potential changes in the recent past and extrapolate to the future. To explore the state of the science on these methods, consideration is made of: past observational and environmental trends, the performance of climate models in resolving favorable SCS environments, and projected future changes in the context of both environments, and dynamically downscaled SCS proxies.

## Observed and Environmental Trends

A large number of studies globally have considered the observational records available to explore whether trends in the frequency of SCS are identifiable from historical records. The noted limitations in observational records, however, mean that ensuring that these trends are robust or meaningful is challenging, and regionally there is considerable variation. Thus both environmental and observational sources of evidence for trends are discussed to provide the necessary context. In terms of non-severe thunderstorms, there is little trend evident in thunder days in the United States, Australia, or Canada for the last half century (Kuleshov et al., 2002; Changnon, 2003; Dowdy & Kuleshov, 2014; Huryn et al., 2016).

Click to view larger

Figure 2. Number of days per year with at least one (black squares) and more than 30 (E)F1+ tornadoes (red triangles). Dots and lines are decadal means.

Adopted from Brooks et al. (2014), their Figure 4. Used with permission from the authors.

Hail has received greater attention globally owing to its frequency over all continents, rather than just the United States. Hail pad networks suffer from issues with trend analysis due to the spatial variability of the phenomenon (Xie et al., 2008; Webb et al., 2009; Punge & Kunz, 2016; Sanchez et al., 2017). This has led to contrasting viewpoints in different regions, where the observations appear to show local trends (e.g., Mezher et al., 2012; Hermida et al., 2013; Punge & Kunz, 2016, and illustrated by the summarized country-by-country trends in Table 3). Nationally, over the United States there is no trend in hail reports or associated environments that is not driven by reporting inhomogeneities and the large degree of variability in both severe and larger hail sizes (Brooks & Dotzek, 2008; Allen et al., 2015a; Allen & Tippett, 2015). However, on a regional basis there is a greater degree of variability of trends in station-derived hail occurrence (Changnon & Changnon, 2000), with potentially significant increases along the east coast and high plains. In Canada, Cao (2008) identified a positive trend in severe hail frequency; however, this trend is marginal, interspersed with large natural interannual variability and likely associated with reporting variability over the length of the dataset (Etkin & Brun, 1999). Despite this null result, the ten most active hail years over Canada were associated with warm temperature anomalies, which suggests that at least in part Canadian hail would be sensitive to warming temperatures.

A number of studies indicate robust downward trends in hail days and hailstorm frequency in China (Xie et al., 2008, 2010; Li et al., 2016), along with South Korea (Jin et al., 2017). Analysis of associated environments has suggested that this trend cannot be attributed to CAPE, but is sensitive to increasing melting level and changes to the column cloud liquid water. In Australia, a single-state study over New South Wales has also shown negative trends using hailstorm observations (Schuster et al., 2005), which seems to contrast the increasing frequency of reports. Over Europe, trends are extremely mixed and generally explored on a country-by-country basis rather than as a continent (Punge & Kunz, 2016). Nonetheless, negative trends in hail day frequency are found for the majority of regions, while only a few studies note increases (e.g., for France, Hermida et al., 2013). Similar results have been identified for South America (Mezher et al., 2012; Martins et al., 2017). Increasing intensity of hail has been identified in SW France with a decreasing frequency of hail (12%), which was hypothesized to be the result of a rising freezing altitude (320 meters) between 1988 and 2012 (Dessens et al., 2015), though this trend is found in hail sizes of 5–7mm contrasting increases for larger hail (11–21 mm), and insufficient data for severe hail. While there is uncertainty as to the trends that are occurring in hail globally, the high spatial variability of hail suggests it is reasonable that there would be local signals of both positive and negative trends (Hermida et al., 2013; Allen et al., 2015a), driven by long-term natural variability, non-meteorological inhomogeneous flaws of the data, or even differences in the potential response to a changing climate. Thus trends on regional scales are difficult to validate or dismiss without elongated datasets of observations, and varying spatial signals would be consistent with regional variations in the availability of both moisture and instability projected in climate models (e.g., Diffenbaugh et al., 2013).

Table 3. Summary of Results Describing Trends in Hail Occurrence and Size

Region/Country

Characteristic

Trend

Authors

United States

Occurrence, Size, Environments

No significant trends.

Changnon and Changnon (2003), Brooks and Dotzek (2008), Allen et al. (2015a), Allen and Tippett (2015), Allen et al. (2017)

Occurrence

No significant trends.

Etkin and Brun (1999), Cao (2008)

United Kingdom

Occurrence, Size

No significant trends.

Webb et al. (2009).

Netherlands

Occurrence

Significant increase projected.

Botzen et al. (2010), Botzen and Bouwer (2016)

France

Occurrence, Size, Environments

Mixed trends, decrease in small hail frequency, significance increase in larger hail frequency.

Vines (2001), Berthet et al. (2011), Hermida et al. (2013), Dessens et al. (2015), Sanchez et al. (2017)

Italy

Occurrence, Size,

Environments

Significant positive (central), and significant negative (north) in occurrence.

Piani et al. (2005), Eccel et al. (2012), Manzato (2012)

Germany

Occurrence, Environments

Significant increase in hail days, favorable environments.

Kunz et al. (2009), Kapsch et al. (2012), Mohr and Kunz (2013)

Czech Republic

Occurrence

Significant decrease in hail days.

Chroma et al. (2005)

Croatia

Occurrence

Increase in hail intensity, no change to frequency.

Pocakal (2011)

Romania

Occurrence

Mixed to negative trends.

Brucea et al. (2016)

Serbia

Occurrence

Significant negative trends in station hail.

Ćurić and Janc (2016)

Bulgaria

Occurrence

Significant negative trends in station hail.

Simeonov et al. (2009)

Southern Russia

Occurrence

Significant negative trend in station hail.

Malkarova (2011)

Australia

Occurrence, Environment

No significant trends to slight decreases.

Niall and Walsh (2005), Schuster et al. (2005), Allen and Karoly (2014)

China

Occurrence, Environment, Size

Significance decrease in frequency, no trend in size.

Zhang et al. (2008), Xie et al. (2010), Li et al. (2016), Ni et al. (2017)

South Korea

Occurrence, Environment

Significance decrease in frequency.

Jin et al. (2017)

Argentina

Occurrence, Environment

Mixed signals.

Mezher et al. (2012)

Brazil

Occurrence

No significant trends.

Martins et al. (2017)

Insurance-based trends appear to show contrasting patterns to the mixed signals in hail observations and environments, with consistent increases to damage of buildings, vehicles, and agriculture as a result of hail over Europe, and to a lesser extent the United States (e.g., Kunz et al., 2009; Barthel & Neumayer, 2012; Eccel et al., 2012; Sander et al., 2013; Hoeppe, 2016; Botzen & Bouwer, 2016). These increases in insured losses have generally been consistent with increasing numbers of SCS favorable weather patterns (Kapsch et al., 2012), or favorable SCS environments driven predominantly by increases to favorable thermodynamic conditions (Kunz et al., 2009; Eccel et al., 2012; Sander et al., 2013). However, despite the increases in favorable SCS conditions, these results have not been consistent with increases in the observed number of storms (Kunz et al., 2009; Eccel et al., 2012), and are sensitive to discontinuities in the data (e.g., Kunz et al., 2009). This result appears to be attributable in large part to increasingly favorable SEV, which has been identified to be driving the increase in large thunderstorm losses rather than increasing exposure (Sander et al., 2013).

Generalizing trend analysis to applications for SCS frequency, a number of studies have considered the trends in SCS environments for either SEV or its components. Globally, CAPE and CIN trends have been explored for the ERA-40 reanalysis, identifying CAPE trends in all seasons, and a decrease in CAPE during the autumnal months (Riemann-Campe et al., 2009). In terms of seasonality, this study was one of the earliest to show a consistent shift towards elongation of the season of SCS environments. Decreases to SEV over South America have also been suggested (Brooks & Dotzek, 2008). Over the United States, SEV environments from the NARR reanalysis have showed large interannual variability and negligible trends, contrasting the positive trend in SigSCS observations (Gensini & Ashley, 2011). Subsequently, Robinson et al. (2013) found a dearth of significant trends regionally in either SEV from the NCEP/NCAR reanalysis (1950–2009), or in dynamically downscaled data (Trapp et al., 2011; Robinson et al., 2013). In comparison, non-stationary extreme value models over the United States using SEV revealed appreciable increases to frequency (as assessed by decreasing 20-year return periods) over the center of the continent and gulf coast while frequency decreased and consequently increased return periods over the Midwest (Gilleland et al., 2013; Mannshardt & Gilleland, 2013). Retrospective analysis indicated these results were consistent with an increasing likelihood of rare environment high-impact events such as the 1999 and 2013 Moore, Oklahoma, tornadoes.

Outside of the United States, Allen and Karoly (2014) identified no trends outside of natural variability over Australia, while Mohr and Kunz (2013) and later Garcia-Ortega et al. (2014) identified increasing thermodynamic favorability for hailstone production over Germany, Europe and Spain. A limitation of the approach of Garcia-Ortega et al. (2014) was that the overall pattern shift was not particularly pronounced, and was extremely sensitive to an individual year of the record. Over France, analysis of atmospheric precursors to hail formation (1948–2015) illustrated trends in both the frequency of 500 mb troughs over southwestern Europe and thermal instability, particularly in terms of increased lapse rates (Sanchez et al., 2017). However, while such pattern change explains the increased frequency over the Pyrenees, there is little support for decreasing frequency due to increasing freezing level altitude (Dessens et al., 2015). A limiting factor for the use of environmental trends is that signals in SCS environments appear to contrast those of observations (Eccel et al., 2012), perhaps due to changes in the likelihood of initiation.

A key concept of these reanalysis-driven approaches to ascertaining trends in frequency is the use of favorable environments to approximate SCS occurrence. However, it is not clear that reanalyses are up to the task of estimating trends in convective environments on their own (Thorn & Vose, 2010; Allen et al., 2010). Reanalyses aim to provide an accurate rendition of the instantaneous climate state rather than the continuity of observations that would be necessary to produce a climate analysis quality record needed to generate an appropriate climate trend. However, the observations that are assimilated through time have changed in each of the reanalysis datasets, and thus analysis of trends through time can be problematic. For example, the assimilation of satellite data leads to significant discontinuities prior to its introduction in 1979, and with each subsequent new dataset. This is not to suggest that reanalyses cannot be used for this type of analysis, but authors should be aware of potential discontinuities in the record and implications for their results. Other approaches to overcome these issues involve providing a greater bounding of this uncertainty: the use of multiple reanalyses in an ensemble approach, or alternatively, ensemble realizations of the atmospheric state from an individual reanalysis (e.g., the NCEP/NCAR 20th-century reanalysis; Compo et al., 2011).

Summarizing, while reanalyses are not perfectly ideal for this application, there appear to be limited trends in the overall frequency of favorable SEV or SCS environments in the United States, which is not inconsistent with expectations that changes will take decades to manifest outside of variability (Trapp et al., 2009). However, there may be an increasing signal in the most extreme events (Sander et al., 2013). Over Europe the signal is more mixed, but at least regionally shows indications of thermodynamically driven increases. This is accompanied with observed evidence for decreasing frequency for smaller hail sizes across much of the globe, where hail observations less than severe (2 cm) are available, apparently driven by increasing freezing altitude. How this increasing freezing altitude impacts severe hail is still uncertain, though from a physical perspective these small increases would be unlikely to lead to noticeable melting. Trends in observed tornadoes and their associated environments already appear to be leading toward fewer tornado days but larger tornado outbreaks (e.g., Brooks et al., 2014; Tippett et al., 2016), which would be consistent with the shift toward the higher frequency of favorable SEV environments noted by Mannshardt and Gilleland (2013). Insured losses show increasing frequency (Hoeppe, 2016), a response to exposure and increasingly favorable weather patterns (e.g., Kapsch et al., 2012), and increasingly favorable thermodynamic conditions (e.g., Sander et al., 2013).

## Analyzing the Performance of Climate Models for Convection in the Present

Prior to considering future projections using regional or global climate models, first the degree to which these models are capable of representing the present climatology of severe thunderstorms must be considered. This is a key requirement if any projection into the 21st century can be considered realistic, or at the very least to enable understanding of the initial biases associated with such a projection. The fundamental issue which drives such challenges is that GCM grid resolution is incapable of directly resolving convective processes, and thus unable to provide a true depiction of SCS frequency. A further challenge to this validation is that the same day in a climate model does not correspond to the same day in a reanalysis, and thus comparison is limited to climatological statistical characteristics and their qualitative similarity (Marsh et al., 2007, 2009).

In comparison to convective processes, synoptic systems are relatively well simulated by GCMs (Griffiths et al., 1993; Niall & Walsh, 2005; Marsh et al., 2007; Van Klooster & Roebber, 2009; Trapp et al., 2009; Marsh et al., 2009), and the validations that have been performed for 20th-century performance of climate models in rendering SCS-favorable environments have generally been acceptable, particularly for S06 (Niall & Walsh, 2005; Marsh et al., 2009; Diffenbaugh et al., 2013; Allen et al., 2014a; Seeley & Romps, 2015; Púčik et al., 2017). At the coarsest resolutions, rendition of convective instability tends to be considerably underestimated (Ye et al., 1998). Individual model characteristics also influence the instability, with overestimates in models with anomalously high 20th-century boundary layer moisture (e.g., Niall & Walsh, 2005), and underestimation where convective parameterization schemes aim to eliminate CAPE (Marsh et al., 2007). Biases in terms of instability have been subsequently identified for regions over warm moist water or those in nearby proximity (Marsh et al., 2007; Diffenbaugh et al., 2013; Seeley & Romps, 2015) and over continents where excess drying occurs (Klein et al., 2006; Diffenbaugh et al., 2006; Trapp et al., 2007a; Allen et al., 2014a), though in some cases there are positive biases over land, particularly as horizontal resolution increases (e.g., Púčik et al., 2017). These differences can be exacerbated by misrepresentation of the land surface–atmospheric interactions, topographic smoothing, or behavior of convective parameterization schemes (Marsh et al., 2007; Trapp et al., 2007b; Allen et al., 2014a). Assessing ensembles of models also reveals large differences in ensemble members which are primarily driven by how instability is rendered (Figure 3) or potentially removed by convective schemes (Marsh et al., 2007; Diffenbaugh et al., 2013; Allen et al., 2014a; Seeley & Romps, 2015; Púčik et al., 2017).

Click to view larger

Figure 3. (Top) Mean SEV per MAM for (a)–(k) the years 1996–2005 in 11 CMIP5 GCMs and (l) the years 1999–2008 in SPARC radiosonde observations. (Bottom) As for top except JJA.

Adapted from Seeley and Romps (2015), their Figures 2 and 3. Used with permission from the authors.

Diurnal cycles also pose considerable challenges for climate models. Lags in maximum daytime temperature can produce cool temperature biases, potentially reducing CAPE in excess of 1000 Jkg-1 in some circumstances (Dai & Trenberth, 2004; Marsh et al., 2007). These biases suggest that instability is never allowed to build to the point at which severe convection would occur in the atmosphere, an issue driven by convective parameterization. This pattern has also been identified for both Europe (Marsh et al., 2009; Sanderson et al., 2015) and Australia (Allen et al., 2014a), where convection in climate models tends to be realized prematurely as early as the late morning, prior to when this mixing would occur in the real atmosphere (e.g., Figure 4), resulting in decreases in moisture in nearby grid-boxes, as well further depreciating CAPE.

Click to view larger

Figure 4. Temporal distribution of mean SEV environments per season for the respective models and reanalysis for all grid points over (a) Australia and (b) Eastern Australia as compared to severe thunderstorm reports (black lines and dots) over the comparable region for the period March 2003–April 2010. Climatology period corresponds to 20 warm seasons (September–April 1980–2000) for the ERA-Interim (black circles), CSIRO Mk3.6 (crossed circles), and CCAM (dotted circles). Times correspond to Australian Eastern Daylight Time or equivalently 1800, 0000, 0600, and 1200 UTC, respectively.

Adopted from Allen et al. (2014a), their Figure 2. Used with permission from the author.

Higher spatial resolution does seem to offset this problem, perhaps driven by improved rendition of CIN (Allen et al., 2014a). This suggests that for the purposes of estimating past convective climatology, convective parameterizations play an important role in climate model performance.

A general attribute of GCMs for convective parameters is a tendency for the base state of the climate models to be displaced on the climatological scale, through large differences in the boundary layer moisture fields, spatial displacement toward the equator, and spatial resolution (Trapp et al., 2007a; Diffenbaugh et al., 2013; Allen et al., 2014a; Seeley & Romps, 2015). This is particularly the case when lower spatial resolution leads to topographic features no longer realistically modifying flow patterns or providing sources of initiation (Allen et al., 2014a; Púčik et al., 2017). Validating the larger ensemble of CMIP5 models appears to provide a more consistent picture of overall change, even though individual models can have large differences in magnitude (Figure 3, Diffenbaugh et al., 2013; Seeley & Romps, 2015). Comparisons of interannual variability have only been briefly considered (Marsh et al., 2007), but generally this feature is well replicated, a result consistent with earlier models (Trapp et al., 2009) and the CMIP5 suite (Diffenbaugh et al., 2013). Seasonal biases have also received comparatively less attention, though the early studies noted a slight early summer bias (Marsh et al., 2007) and reasonable performance at single grid points (McMaster, 1999). Later analysis over Australia, Europe, and the United Kingdom (Marsh et al., 2009; Allen et al., 2014a; Sanderson et al., 2015) has shown generally good performance in rendering the annual cycle. One potential way to overcome these limitations is focusing on identifying climate models with optimized climatological performance during the climatologically active months (e.g., Allen et al., 2014a; Seeley & Romps, 2015), but this reduces ensemble size and thus the advantages of using a set of models with varying renditions of climatology. Nonetheless, the preponderance of evidence suggests that, provided biases are recognized, GCMs and regional simulations can provide suitable mean climatological performance in rendering vertical profiles, particularly for CAPE and S06, and to a lesser degree for CIN.

## Future Projections Using Proxy Environments

To explore future changes to SEV environments, there has been an increasing reliance on the output of consistently higher-resolution GCMs and more recently higher-resolution regional climate models (RCMs). Changes were initially assessed using individual models under single emissions scenarios (mostly highly warming), and later by applying ensembles with varying emissions projections for the late 21st century. The earliest changes for midlatitude convection were projected by coarse resolution models (3.3º × 5.6º) under doubling of CO2 over Australia (e.g., McMaster, 1999), which suggested decreases in the frequency of hail over the continent. This result is tempered as it did not include a coupled ocean model capable of simulating large-scale climate modulators such as El Nino–Southern Oscillation (ENSO). Later analysis using higher-resolution (1.875º × 0.635º) data suggested a decrease in hail frequency for isolated locations, driven by significant decreases in CAPE (Niall & Walsh, 2005). However, subsequent analysis identified unrealistically high representations of 20th-century instability from the same model family (CSIRO Mk3.6), and showed sharp changes into the 21st century (Allen et al., 2014a). This suggests that to some degree these changes may not reflect a realistic future projection, but rather are a result of model biases.

Defining this approach over the United States, Trapp et al. (2007a) assessed changes in SEV for four models between 1962–1989 and 2072–2099. Changes were driven by increases to the boundary layer atmospheric moisture, with a resulting 50%–60% increase in SCS frequency in May and June. However, these results were accompanied by large differences between the model simulations and sensitivity to 20th-century biases in SEV conditions. Further analysis exploring transient responses of SCS to climate change using five climate models for the period 1950–2099 further supported these results, with increases to SEV environments co-occurring with convective precipitation driven not by lapse rates or S06 but rather boundary layer moisture (Trapp et al., 2009). These results showed that the anticipated poleward shift of the midlatitude storm track and decreased meridional baroclinicity, thereby reducing the thermal wind gradient between the midlatitudes and poles, would suggest fewer favorable S06 environments. While increases in SEV environments were projected, there was little change apparent in the occurrences of initiating features such as fronts, troughs, and drylines (Trapp et al., 2009). Furthermore, the authors noted that the decreased frequency of continental extratropical cyclones were outside of the convective season. Using a later generation of climate models, changes to SEV environments explored whether projected increases were robust over a range of climate models (Diffenbaugh et al., 2013). Results from this large ensemble of CMIP5 models highlighted that a higher fraction of SEV days is driven by days with high CAPE and strong S06, and the increases of high CAPE days with low CIN (Figures 5, 6).

Click to view larger

Figure 5. Response of severe thunderstorm environments in the late-21st-century period of RCP8.5 during the winter (DJF), spring (MAM), summer (JJA), and autumn (SON) seasons. (A–D) Color contours show the difference in the number of days on which severe thunderstorm environments occur (NDSEV) between the 2070–2099 period of RCP8.5 and the 1970–1999 baseline, calculated as 2070–2099 minus 1970–1999. (E–H) Each gray line shows an individual model realization. For each realization, the anomaly in the regional average NDSEV value over the eastern United States (105°W–67.5°W, 25°N–50°N; land points only) is calculated for each year in the 21st century, with the anomaly expressed as a percentage of the 1970–1999 baseline mean value. The black line shows the mean of the individual realizations.

Adopted from Diffenbaugh et al. (2013), their Figure 1. Used with permission from the authors. Copyright 2013 National Academy of Sciences.

Click to view larger

Figure 6. Change in the frequency of occurrence of daily CAPE and CIN in the winter (DJF; A), spring (MAM; B), summer (JJA; C), and autumn (SON; D) seasons in the late-21st-century period of RCP8.5. Occurrences are counted for land grid points in the eastern United States (105°W–67.5°W, 25°N–50°N; land points only) for days in which the SEV threshold is met. The absolute difference in the ensemble-mean number of occurrences between the 1970–1999 and 2070–2099 periods, calculated as 2070–2099 minus 1970–1999, is shown for each season.

Adopted from Diffenbaugh et al. (2013), their Figure 5. Used with permission from the authors. Copyright 2013 National Academy of Sciences.

This was interpreted to be indicative of a robust increase in the number of SCS days, under both highly warming and lower warming scenarios, and likely to be evident by the mid-21st century.

Generalizing these seminal studies to reflect the broader set of assessments on continental scales shows that increases in SEV environments in climate model projections for the late 21st century are primarily driven by increases in instability for the United States and North America (Del Genio et al., 2007; Trapp et al., 2007b, 2009; Diffenbaugh et al., 2013; Gensini et al., 2014b; Paquin et al., 2014; Seeley & Romps, 2015; Li & Colle, 2016), Australia (Allen et al., 2014b), Europe (Marsh et al., 2009; Viceto et al., 2017; Púčik et al., 2017) and Japan (Muramatsu et al., 2016). These increases are primarily a response to increasing moisture content in the lower troposphere (Trapp et al., 2007b, 2009; Diffenbaugh et al., 2013; Allen et al., 2014b; Paquin et al., 2014; Seeley & Romps, 2015; Púčik et al., 2017) associated with the increasing saturation capacitance of air in line with arguments such as Clausius-Clapeyron scaling over the midlatitudes with temperature and observed trends of moistening (Peterson et al., 2011; Romps, 2016; Agard & Emanuel, 2017). In contrast to the consensus of increasing instability, changes to shear show greater variation between the continents. Over the United States, earlier studies highlighted expected decreases to shear as a response to the poleward displacement of the jet streams (Trapp et al., 2007a; Del Genio et al., 2007; Trapp et al., 2009; Van Klooster & Roebber, 2009), a result echoed for both Europe and Australia (Marsh et al., 2009; Allen et al., 2014b). Subsequent exploration by Diffenbaugh et al. (2013) and Gensini et al. (2014b) identified that decreases in DLS were mostly confined to lower CAPE days, suggesting these changes may have limited implications for SCS frequency. A contrasting result for Europe by Púčik et al. (2017) demonstrated no robust changes to S06 for either low- or high-instability environments. This suggests that careful attention to the details of the interplay of covariate SEV environments is needed to understand how these changes manifest. The juxtaposition between S06 and CAPE when combined with the annual cycle of these quantities also poses challenges (Brooks et al., 2007; Brooks & Dotzek, 2008; Brooks, 2013). This characteristic combined with the increase of instability in spring and fall has been suggested to lead to an extension of the severe season into these months when shear is generally abundant (Diffenbaugh et al., 2013; Gensini et al., 2014b). Recent results by Hoogewind et al. (2017) point to this lengthening of the favorable SCS environment season being as much as a month earlier in the spring, and half a month in the fall. Thus, from the perspective of favorable SCS environments there is a growing consensus that the frequency of severe thunderstorm favorable environments over each of the three continents described here is increasing, and the length of the SCS season is increasing over the United States.

For tornadoes, the response is somewhat uncertain owing to the small grid scale of these events. Increasing instability in the early spring and late fall appear likely to increase the frequency of tornadic events in periods which are normally associated with favorable S06 and 0–1 km shear (S01; Diffenbaugh et al., 2013). An alternative interpretation here is that this may influence the climatology of tornadoes outside of the Great Plains, where S01 dominates the formation and buoyancy is comparatively low or limited (e.g., the southeast United States; Sherburn et al., 2016). Unlike over the United States, changes to cool season tornadoes have received attention in Australia (Timbal et al., 2010). These events occur in low instability but high S01 during the winter months (Hanstrum et al., 2002; Kounkou et al., 2009) and are likely similar to those described in the southeast United States (Sherburn et al., 2016), and thus any small change to instability may reflect a considerably magnified hazard. However, model projections for Australia suggest that the thermal characteristics of the lower troposphere are likely to stabilize, leading to a diminished threat despite favorable shear (Timbal et al., 2010). Another possibility is that this reflects a potential shift in environments toward a mechanism similar to the warm season, rather than extreme low instability, and thus the overall effect may balance. Elsewhere globally, Muramatsu et al. (2016) explored changes to favorable tornadic environments over Japan and the United States using a 20km RCM. Over the United States, increases to favorable environments over the great plains were found, while for Japan increases to tornadic occurrence, particularly over the southwest, were projected.

Assessments of the response of hail to warming are also relatively limited. Increasing mid-tropospheric temperature by 1ºC, for example, on hail days would decrease hail probability by 7%, which would indicate for a strongly warming scenario considerable decreases in hail frequency were likely (Billet et al., 1997; Van Klooster & Roebber, 2009). A simple empirical hail nomogram, with hail estimation driven by the projected environment changes from an RCM over the United Kingdom identified decreasing trend in hail events between 21 and 50 mm, but little spatial change in occurrence (Sanderson et al., 2015). These changes appeared to be driven by weaker overall updrafts for hail production, but associated with a large degree of variability in these trends and a tendency for the model historical climatology to be overestimated and considerable sensitivity to the convective parameterization scheme used. More recently, Brimelow et al. (2017) explored the changes to hail frequency and size over North America using a single-column hail growth model coupled to a steady-state cloud model (HAILCAST) to ingest environmental profiles from 50km RCM output from NARCCAP. Testing for the 20th century showed good consistency with earlier environmentally derived hailstone climatologies (e.g., Cintineo et al., 2012; Allen et al., 2015a) in simulating the present climate. While this method is novel, the growth of hailstones within this model does have some limitations, as the evolution of hail growth is discontinuous and relies on the steady-state convective clouds simulated from the ingested RCM data. HAILCAST in this application does not continually simulate storms, but rather assesses potential for hail growth based on each environmental six-hourly analysis provided a convective cloud initiates. Consequently, this can lead to underestimation of hail sizes, as the largest hailstones do not appear to grow to sizes consistent with those observed climatologically (e.g., Allen et al., 2017), or there appears to be limited signal in this larger-size population. The projections suggest shifts toward fewer days with smaller hail over the southern United States, with increases particularly for hail between 20 and 40 mm during the spring and summer over the northern plains of the United States and the southern parts of the Canadian plains through Alberta.

From an international perspective, changes over Europe by the end of the 21st century suggest decreasing CAPE in the warm season and increased CAPE in the cool season (Marsh et al., 2009). This subsequently leads to fewer low-CAPE and high-S06 environments and a greater number of high-CAPE, lower-S06 environments, though the overall tendency for SEV increased. Projecting favorable hail circulation patterns over Germany using an ensemble of RCM simulations to the mid-21st century and a logistic hail regression using both environments and circulation patterns suggests increases of 7%–15% in hail days, but with wide ensemble spread (Kapsch et al., 2012; Mohr et al., 2015). Weaknesses to both of these approaches are that weather types are used as the sole predictor for favorable hail days, which potentially neglects environmental parameters that are favorable to the development of hailstorms (Púčik et al., 2017). In a state-of-the-art study, Púčik et al. (2017) used an ensemble of 14 RCMs over Europe with 0.44° resolution, applying a lifted index and S06 coinciding with model precipitation to project SCS threat by the mid and late 21st century. Exploring a middle and high-end warming scenario identified robust increases in unstable environments across central and south-central Europe coinciding with an increase in moisture, particularly in the late-21st-century highly warming scenario (Figure 7).

Click to view larger

Figure 7. Ensemble mean value of annual number of unstable environments for (a) the historical 1971–2000 period, and the mean change in the number of environments between the future and historical period for the periods (b) 2021–2050 and (c) 2071–2100 in the RCP4.5 scenario, and (d) 2021–2050 and (e) 2071–2100 in the RCP8.5 scenario.

Adopted from Púčik et al. (2017), their Figure 5. Used with permission from the authors.

Little to no change in S06 was identified over the majority of the domain. Increases of covariates of 100% in the highly warmed scenario (RCP8.5) compared to 30%–50% in the moderate warming (RCP4.5), with increases limited to the 10%–25% range prior to midcentury, suggesting that changes to SCS will likely be slow to occur relative to natural variability (Trapp et al., 2009). As with US studies, large ensemble spread in projections was identified to be associated with instability differences rather than S06 and precipitation between models. Elsewhere, over Australia, there are projected robust increases to favorable SCS environments (Figure 8), particularly over the eastern third of the continent, driven by increasing instability, though the degree of increase is sensitive to model resolution (Allen et al., 2014b).

Click to view larger

Figure 8. Differences between the mean seasonal frequency of SEV environments for the 21st-century period and the 20th-century period over the Australian continent for (left) CSIRO Mk3.6 and (right) CCAM. Stippling is indicative of significant increases to the 21st-century mean above the 97.5th percentile, while hatching indicates significant decreases below the 2.5th percentile as determined using a bootstrapping procedure.

Adopted from Allen et al. (2014b), their Figure 12. Used with permission from the author.

Limitations to many of these environment-based approaches are that the results are nonconditional, or rely on the use of convective precipitation derived from the parameterization scheme as a way to infer convective initiation (Trapp et al., 2009; Allen et al., 2014b; Púčik et al., 2017). However, the frequency of initiating factors such as synoptic troughs or frontal boundaries appears to show little change in future projections (Trapp et al., 2009; Van Klooster & Roebber, 2009). These limitations of the environmental approach argue for the use of dynamic downscaling to simulate the frequency of initiation (Hoogewind et al., 2017). While validation can provide an idea of potential of biases, a high degree of model-to-model variability has also been identified as a potential weakness in the analysis of future scenarios for severe convection (Gettleman et al., 2002; Niall & Walsh, 2005; Marsh et al., 2007; Diffenbaugh et al., 2013; Paquin et al., 2014; Seeley & Romps, 2015; Sanderson et al., 2015). Whether such changes are distinguishable from natural variability is also a question that must be asked, especially when interannual variability is so large (e.g., tornadoes), and the future periods considered are relatively brief (Van Klooster & Roebber, 2009; Allen et al., 2014b).

## Decreasing the Scale: Dynamic Downscaling in the Present

The conditional likelihood relationship between an environment and SCS has led to alternate approaches to address the occurrence of storms both in the present and future climate explicitly (Griffiths et al., 1993; Trapp et al., 2007b). As noted in a large number of studies, a major limitation of GCMs is that convective processes are parameterized, rather than storms being simulated directly (Gensini & Mote, 2014). Thus, obtaining a picture of SCS equivalent to the resolution used in operational weather forecasting (Corfidi, 2017) requires some form of dynamical downscaling, akin to extremely high resolution RCMs. This concept is not unique to applications for SCS, but rather has its roots in downscaling for assessing projected changes in precipitation from climate models (e.g., Giorgi, 1990; Mahoney et al., 2013). The concept of applying these techniques to severe thunderstorms arose in the early 1990s (Griffiths et al., 1993), which suggested the use of regional high resolution nesting over an area of interest from GCM lateral boundary conditions. However, the scale of SCS meant that computationally, this problem was prohibitive.

Trapp et al. (2007b) were the first to apply this concept to SCS using the NCEP/NCAR reanalysis to provide boundary and initial conditions to a two-way nested Weather Research and Forecasting model (WRF) simulation downscaled to 3km from the original 2.5˚ data. Rather than using continuous integration, 30-hour integrations with continual restarts for multiple events were applied. Limited by computational power, the study only explored two case studies, but the results revealed a similar number of proxy tornado events at a range of WRF downscaled resolutions (using updraft strength and rotation as defined by vorticity) to the events, albeit with a temporal lag, and a sensitivity of morphology and intensity to the spatial scale of modeling. Although this approach yielded little insight into a changing climate, it provided an essential framework to address the question of whether the changes seen to SEV environments reflect a realistic scenario on the storm scale. Testing of approaches using an intermediary RCM or stepwise domains in the downscaling process had negative implications due to feedbacks from the model convective parameterization schemes at coarser resolutions, but suggested that direct simulation from the higher resolution data is more representative than gradually scaling. Building on this result, Trapp et al. (2011) used 10-year periods of reanalysis data dynamically downscaled to 4.25 km without domain stepping for 1991–2000 using a proxy of 2–5 km updraft helicity juxtaposed with strong reflectivity (effectively conditioning for supercell storms). This proxy was identified to produce a realistic diurnal cycle and a good match to observed SCS. Trend analysis of these data showed a flat frequency from the proxy and considerably higher values for the SEV environmental proxies with little trend, both of which contrasted the issue prone observed record particularly for this period (Verbout et al., 2006; Allen & Tippett, 2015).

As an alternative to using arbitrary thresholds for proxies to SCS observations, artificial neural networks have also been used to identify event occurrence from dynamically downscaled NCEP-NCAR reanalysis (Robinson et al., 2013). Analyzing a 20-year period revealed no significant regional trends in proxy SCS frequency or in SEV environments, consistent with suggestions that several decades may be needed before trends in SCS can be distinguished from large natural variability (Trapp et al., 2009). Generally, the spatial distribution of SCS occurrence was well replicated, but the climatology is biased toward supercells by including the updraft helicity condition. While storms that rotate are responsible for a disproportionate share of the largest hail and tornadoes, this condition neglects other mechanisms within Quasi-Linear Convective Systems (Trapp et al., 2005) that produce weaker tornadoes with reasonable regularity, and gust-front driven systems are also the most frequent producer of damaging winds (Johns & Hirt, 1987; Smith et al., 2013). The depiction of storm mode is also sensitive to the scale to which downscaling occurs. Models with broader horizontal resolution operationally have been shown to be biased toward linear convective modes rather than rotating storms. Dynamically downscaled cases also display this problem, whereby a significant outbreak of tornadoes on May 3, 1999, was depicted as a squall line rather than the discrete storms observed (Robinson et al., 2013). Another limitation of this approach is that different model or reanalysis averaged conditions ingested by the downscaling procedure may lead to a varied climatology (Thorne & Vose, 2010; Diffenbaugh et al., 2013; Allen et al., 2014a).

Verifying the skill of dynamic downscaling using GCM data was an important step before any future projection could be considered. Gensini and Mote (2014) applied data from the Community Climate Modeling System version 3 (CCSM3) to drive hourly output from a 4 km dynamically downscaled WRF simulation. Unlike the approach of Trapp et al. (2007b, 2011) and Robinson et al. (2013), continuous integration of the simulation provided atmospheric memory of the convective evolution from day to day, and consequentially, memory of land surface features such as precipitation-driven evaporative fluxes. Applying proxies equivalent to earlier studies, GCM-driven downscaled climatologies were found to represent the late 20th century well. One deficiency of this approach was that the derived report frequencies lagged the observed diurnal cycle of the climatology somewhat, suggesting that the model initiation was later. Furthermore, the continuous integration can also incur integration errors and is computationally inefficient, leading to a shorter climatological period analyzed, and thus making the results sensitive to natural variability. Similar experiments for the Iberian Peninsula (Viceto et al., 2017) downscaling reanalysis data and climate model data to evaluate environment climatology have also illustrated the usefulness of this approach to better understand changes to SCS favorable environments while appropriately including terrain features. Despite some limitations, the performance of these techniques suggests that dynamic downscaling is a viable and important avenue for exploring future projections.

## Future Projections on the Convective Scale

While Trapp et al. (2007b) were the first to apply the methodology to downscale future climate projections, the first study looking at future severe thunderstorm environments focused on summer hailstorms around Sydney, Australia (Leslie et al., 2008). Coupled GCM ensemble output from a 1.5˚ horizontal grid was nest downscaled to 1 km, which was more appropriate for hailstorms, but perhaps not fully capable of representative of hail swaths (10 m to thousands of meters wide: Changnon, 1977; Dessens, 1986). Key to this approach was the use of an explicit 10-ice phase microphysical model including graupel (<0.5 cm), small hail (0.5–2 cm), and large hail (>2 cm) to simulate hail growth and decay. Giant or extreme hail was assumed when concentrations per diameter size bin were small. Testing this approach using three historic hailstorms suggested that performance was comparable to observations, and a highly warming scenario (A1B) was used to explore the change 1970–2000 as compared to 2001–2050, identifying a 20% increase in the number of large hail events exceeding 2 cm, despite a similar number of days occurring per year. The approach however was computationally expensive and only evaluated hail events when ensemble 5 km runs suggested highly restrictive environmental conditions were favorable to hailstorm development.

Future projections using the technique over the United States first considered changes in explicitly simulated Colorado hailstorms between 1971–2000 and 2041–2070 using 1.3 km downscaled WRF simulations derived from 50 km RCM data (Mahoney et al., 2012). This analysis identified two potential impacts of a warming climate on hail: stronger thunderstorms in a warming climate leading to increased production of hail aloft, but a reduced frequency in smaller hailstones reaching the ground as a response to increasing freezing-level altitude (300–500 m) consistent with noted observational trends seen in Europe and China (e.g., Dessens et al., 2015). However, this reduction in smaller hailstones may lead to increased precipitation as a result, but larger stones melt less when falling through warm layers. Caveats to the findings include a marked sensitivity to the choice of microphysics scheme and overall parameterizations used in the model as to whether large hail grows at all, similar to the sensitivity noted by Adams-Selin and Ziegler (2016) in an operational context using 1-D simulated hail growth from HAILCAST.

Returning to approaches similar to those described by Trapp et al. (2007b), Gensini and Mote (2015) explored the change between two 10-year periods (1980–1990 and 2080–2090), using CCSM3 for the boreal spring using a highly warming projection (A2). Continuously integrating over each spring, they identified increases in proxy SCS reports driven predominantly by the early parts of the spring, with a 70% increase in March, a 15% increase in April, and no substantial change in May. Individual years were shown to be associated with a considerable increase in variability of frequency, with a doubling of the standard deviation in reports and a substantial increase in the coefficient of variation from 0.29 to 0.43 (Figure 9).

Click to view larger

Figure 9. Cumulative frequency of historical (black) and future (red) period synthetic hazardous convective weather reports. Thick black and red lines indicate averages for their respective period.

Adopted from Gensini and Mote (2015), Climatic Change, their Figure 3, used with permission of Springer.

This change in variability is consistent with patterns already observed for tornadoes (Brooks et al., 2014; Tippett, 2014). Increases are also diurnally sensitive, with the 21-05UTC period indicating an increasingly persistent evening threat for severe occurrence (Figure 10).

Click to view larger

Figure 10. Diurnal frequency comparison of historical (blue) and future (red) severe weather reports. Error bars indicate the standard error.

Adopted from Gensini and Mote (2015), Climatic Change, their Figure 5, used with permission of Springer.

Environmental changes appear to be the primary driver of 80% of the future variance, with the strongest signal associated with a 236% increase in the frequency of CAPE > 2000 Jkg-1 and low-level moisture in already S06 favorable environments (Figure 11).

Click to view larger

Figure 11. Average frequency of March–May historical (a) CAPE≥2000 Jkg−1 and (b) the product of CAPE and 0–6-km BWD exceeding 20,000. Panels (c) and (d) represent the future period respectively. In the case of CAPE×0–6-km BWD, CAPE is constrained to only evaluate events with ≥100 J kg−1.

Adopted from Gensini and Mote (2015), Climatic Change, their Figure 10, used with permission of Springer.

A limitation of the approach of this study is the extreme shortness of the record, as it fails to capture potential variability driven by natural sources, while the 4 km grid scale may also tend to weaken potential supercell updraft intensity and helicity, favoring an MCS mode that may have different probabilities of producing severe phenomena. Hoogewind et al. (2017) followed this study, exploring how GCM rendered favorable SCS environments compared to the changes seen in the number of dynamically downscaled storms over the complete calendar year for a 30-year period (Figure 12).

Click to view larger

Figure 12. Mean seasonal response in NDSEV from GFDL CM3 and synthetic HCW days from WRF in the future (2071–2100) relative to the historical (1971–2000) period. Stippling indicates where the distribution of seasonal means between the two periods is statistically significant from one another at the 95% confidence level using the Mann-Whitney U test.

Adopted from Hoogewind et al. (2017), their Figure 5. Used with permission from the authors.

Applying the continual restart approach of Trapp et al. (2007b), Hoogewind et al. (2017) used land-surface memory at initialization derived from the GCM driver to overcome the limitations of noncontinuous integration. Comparing the annual cycle, for example, showed that changes in favorable SCS environments from GCMs are two to four times larger than those identified using the downscaled WRF proxies, while frequency in the summer months declined due to increasing CIN (Figure 13).

Click to view larger

Figure 13. (a) Mean CONUS grid point frequency (east of 105°W, land points only) by Julian date (smoothed with Gaussian filter, sigma=15 days) for NDSEV (solid line) and WRF synthetic HCW (dashed). Smoothed 95% confidence intervals are shaded. The historical period (1971–2000) is represented in blue and the future RCP8.5 scenarios (2071–2100) by the red lines, respectively. (b) Cumulative grid point frequency, with bootstrapped 95% confidence intervals shaded.

Adopted from Hoogewind et al. (2017), their Figure 8. Used with permission from the authors.

In each case, however, there is a projected increase in the length of the season of as much as a month in the spring and half a month in the fall. The results of these two studies represent important contributions to the state of the science. However, both approaches are limited by failing to consider the overall influence of changes to the convective mode, where the type of severe thunderstorm report might change as storms change from rotating supercells to forward propagating MCS systems. Furthermore, neither study considered solution variation due to the driving GCM to provide an assessment of uncertainty or robustness, as a result of computational constraints.

Exploring the problem through sensitivity to convective mode, Trapp and Hoogewind (2016) used downscaling to assess whether supercells in the present become squall lines in the future that do not produce tornadoes. Known intense tornado events for three cases were placed into the realm of a warmed climate using the concept of climate deltas, also known as pseudo-global warming (PGW; Lackmann, 2013), where changes by the late 21st century as determined from climate models are applied to the initial and boundary conditions of a downscaled WRF simulation using NAM analysis of known tornadic events. The downscaling involved multiple nests to 1 km and avoidance of convective parameterization to ensure storm fidelity. In all scenarios, CAPE was found to increase, but the peak updraft speed estimated in each case yielded a lower fractional proportion of CAPE relative to the control simulation. This finding is particularly important because it implies that larger CAPE does not necessarily lead to stronger overall updrafts, and thus there is reduced realization of potential buoyancy, which contrasts with many of the earlier SEV/environment-based studies. One suggested cause for this reduced realization was the entrainment of drier midlevel air diluting the buoyancy of the updraft (which was not evaluated in this case); it is also possible that the increased condensate may lead to a greater precipitation loading term within the updraft, reducing the updraft velocity. CIN was another important factor as a response to the PGW approach, with CIN leading to either delay or nonoccurrence of supercells under PGW simulations, with failure to initiate in many ensemble members. In contrast to the thermodynamic response, rotational peak within the supercells appeared to show little sensitivity to the PGW approach. Based on this set of three cases, PGW did not appear to lead to a convective mode change, but it should be noted that this approach is limited by the number of realizations and an extremely limited set of storms, which may not be reflective of this response through all SCS.

Clearly, dynamic downscaling has led to important advancements in our ability to project SCS under future climates, but further work is needed to explore a wider range of parent climate models and longer integrations, similar to the approaches of Gensini and Mote (2015) and Hoogewind et al. (2017), to better contextualize the results of these forward projections.

## The Concerns of Climate Variability

A complicating factor in attributing changes in SCS to either near or future climate change is the influence of climate variability on the large-scale systems that produce these environments (e.g., Diffenbaugh et al., 2008; Tippett et al., 2015). These contributions can range from the seasonal (ENSO, Artic Oscillation) to subseasonal (Madden-Julian Oscillation, Global Wind Oscillation) scales. A number of potential influences exist, though to date only limited consideration of these mechanisms has been undertaken, and no such consideration has been made of their influence on future climate projections and their influence on SCS. Thus, the limited periods which have been considered in terms of future projections represent an apparent limitation to analysis of the influence of a warmed climate on SCS. For example, no study has explored more than 30 years as a future projection time frame, or considered these influences in terms of dynamic downscaling.

Globally, both ENSO and the North Atlantic Oscillation have shown correlations to instability and shear (e.g., Riemann-Campe et al., 2011; Cook & Schaefer, 2008; Lee et al., 2013; Allen & Karoly, 2014; Allen et al., 2015b; Lee et al., 2016; Cook et al., 2017; Lepore et al., 2017), with ENSO playing a large role in North America and Australia. On more limited scales, the convective influences on Rossby Waves and moisture by the Madden Julian Oscillation (Thompson & Roundy, 2013; Barrett & Gensini, 2013; Barrett & Henley, 2015) and the source of available moisture can also play an important role (e.g., the Gulf of Mexico: Jung & Kirtman, 2016; Molina et al., 2016). Other remote sea-surface temperature influences have also been identified to contribute to interannual variability, though the specific mechanism by which these systems modulate the conditions favorable to SCS is as yet unclear (e.g., Elsner & Widen, 2014; Lee et al., 2016; Elsner et al., 2016). The Global Wind Oscillation has also shown promise as an indicator of variability, particularly on the subseasonal scale, encapsulating the overall net influence of jet stream extensions and wave breaking events, the latter of which tend to favor substantial tornado and hail activity over North America (Gensini & Marinaro, 2016; Gensini & Allen, 2018). Climate models in general simulate sources of natural variability such as ENSO relatively poorly. Even so, recent studies exploring the potential responses of signals like ENSO have shown greater likelihood of both El Niño and La Niña conditions (e.g., Cai et al., 2015). It is also unknown whether responses to variability will remain the same. For example, it is unclear if the effects of El Niño will be enhanced or mediated when the warm sea-surface temperature state characteristic of this signal is projected onto a globally warmed ocean. Another potential contribution may come from expected “super droughts” associated with continental drying over the Great Plains, for example (Cook et al., 2015), where little to no rainfall is recorded despite the source of rainfall in these regions being mostly convectively driven, and thus this may influence summer SCS. Based on these drivers of variability, it remains unknown whether future projections reflect conditions that are favorable to one source of natural variability over another, or whether they are not representative of the overall changes to a future climate.

# Indirect Changes to Severe Thunderstorm Impacts

An overlooked factor in considering the impact of a changing climate on SCS is signals driven by anthropogenic modification of the built environment, and thus indirect impacts. For example, a number of authors have identified that expanding urban centers, along with associated infrastructure such as reservoirs and resulting land usage changes, can have a marked influence on the frequency of rainfall and thunderstorm occurrence (Huff & Changnon, 1972; Carleton et al., 2001; Shepherd, 2005; Gero et al., 2006; Frye & Mote, 2010; Nigoyi et al., 2011; Ashley et al., 2012; Degu & Hossain, 2012; Haberlie et al., 2015, 2016). Beyond the urban influence, how lakes or other coastal water bodies respond to a warming climate may produce a significant contribution to SCS frequency beyond or in addition to the changes seen directly, for example adding to local moistening of the boundary layer (Laird et al., 2001; Durkee et al., 2014) or enhancing initiating circulations such as sea breezes (Hill et al., 2010). Changing patterns of irrigation or responses to large-scale droughts may also have the potential to influence severe thunderstorm frequency and modulate the overall likelihood (DeAngelis et al., 2010). It thus appears likely that if such changes were to continue to occur to urban, agricultural, or coastal regions or to vary through time, this could also influence the frequency of SCS and lead to localized enhancement or decreases in risk in addition to the influence of a changing climate. These co-mitigant factors, given that they intersect with populated areas, are also important to the insurance sector (Sander et al., 2013) and can lead to increasing exposure of at-risk populations, as illustrated in Figure 14 (Strader et al., 2017).

Click to view larger

Figure 14. Illustration of EF2+ tornado climatology and the projected growth of US housing unit exposure surface by 2100 as compared to 2010. (a) 2010 housing unit (HU) exposure surface, (b) 2100 (A2) HU exposure surface, (c) all EF2+ tornado paths from 1950 to 2015, and (d) low-pass filtered EF2+ tornado path frequency from 1950 to 2015.

Adapted from Figure 1, Strader et al. (2017), Climatic Change, used with permission of Springer.

Thus, consequences of these factors may further enhance frequency, increase insured and uninsured losses, or lead to a greater number of fatalities.

# Discussion

SCS show considerable sensitivity to a warming, changing, or more variable climate. The stifling factors for evolution in this field have been computational limitations as they relate to grid size and our ability to resolve the appropriate phenomena (Gates et al., 1990; Griffiths et al., 1993; Cubasch et al., 2001). This has led to a spectrum of approaches to the problem, utilizing the available but limited observations along with proxies based on favorable formative conditions or direct simulations of SCS. While projections generally call for an increasing likelihood of SCS as a response to increasing thermodynamic instability (e.g., Diffenbaugh et al., 2013; Hoogewind et al., 2017), the degree to which the range of climate modeling outcomes (those with greater or less warming) will influence the occurrence of SCS (e.g., Seeley & Romps, 2015), and for which seasons this result is applicable, are as yet unclear. There are large differences within the CMIP5 ensemble for environments, which contributes to some degree of uncertainty. Despite this limitation, an increase in the frequency of SCS favorable environments (e.g. SEV) in the spring appears likely, accompanied by perhaps a less significant increase, or even decrease, into the summer months. These changes manifest most strongly as the season elongates, juxtaposing a greater degree of instability on the deep-layer vertical wind shear favorable conditions of the late winter and early spring (Diffenbaugh et al., 2013; Gensini & Mote, 2015; Hoogewind et al., 2017).

Looking to future developments, the vast majority of the literature exploring future climate projections does not explore regions outside of North America, Europe, or Australia, and even over Europe and Australia this work is confined to a few studies. SCS are a global phenomenon, occurring on each of the inhabited continents. Thus, a pressing need for research is application of techniques already demonstrated to be successful in North America, and avoidance of some of the associated pitfalls. Part of the challenge to producing such analyses is the weakness in climatological observational records of SCS or environmental proxy studies of frequency outside of North America, but these problems are being addressed by a growing number of researchers, which bodes well for future investigations (e.g., Allen & Allen, 2016; Ni et al., 2017; Groenemeijer et al., 2017). As a future and idealized direction, continued use of the approach of dynamic downscaling and expanding it to multi-model ensembles of convective resolving models with high vertical resolution to explore the influence of a changing climate over each of the continents would be a lofty aspiration. However, at the current juncture this remains a computationally intensive option unless novel approaches are developed. Furthermore, there is likely a need to consider the impact of upscale feedbacks of changes to the initiated convection (e.g., diabatic enhancement of Rossby waves), as current approaches have been one-way, using a global mesh with locally higher resolution or similar. At the same time, understanding the greater role that humans play in increasing our exposure to risk along with indirect modifications of the environment is also important. As SCS generate average annual losses of greater than \$10 billion USD each year over the United States alone, this points to a need to better depict and constrain our global understanding of how severe convective storms respond to a changing or more variable climate system.

For areas that reflect various elements that contribute to this article, see Tables 1, 2, and 3.

Allen, J. T., Karoly, D. J., & Walsh, K. J. (2014a). Future Australian severe thunderstorm environments. Part I: A novel evaluation and climatology of convective parameters from two climate models for the late twentieth century. Journal of Climate, 27, 3827–3847.Find this resource:

Allen, J. T., Karoly, D. J., & Walsh, K. J. (2014b). Future Australian severe thunderstorm environments. Part II: The influence of a strongly warming climate on convective environments. Journal of Climate, 27, 3848–3868.Find this resource:

Brooks, H. E. (2013). Severe thunderstorms and climate change. Atmospheric Research, 123, 129–138.Find this resource:

Brooks, H. E., Carbin, G. W., & Marsh, P. T. (2014). Increased variability of tornado occurrence in the United States. Science, 346, 349–352.Find this resource:

Brooks, H. E., Lee, J. W., & Craven, J. P. (2003). The spatial distribution of severe thunderstorm and tornado environments from global reanalysis data. Atmospheric Research, 67–68, 73–94.Find this resource:

Diffenbaugh, N. S., Scherer, M., & Trapp, R. J. (2013). Robust increases in severe thunderstorm environments in response to greenhouse forcing. Proceedings of the National Academy of Sciences USA, 110, 16,361–16,366.Find this resource:

Gensini, V. A., & Mote, T. L. (2015). Downscaled estimates of late 21st century severe weather from CCSM3. Climatic Change, 129, 307–321.Find this resource:

Griffiths, D., Colquhoun, J., Batt, K., & Casinader, T. (1993). Severe thunderstorms in New South Wales: Climatology and means of assessing the impact of climate change. Climatic Change, 25, 369–388.Find this resource:

Hoogewind, K. A., Baldwin, M. E., & Trapp, R. J. (2017). The impact of climate change on hazardous convective weather in the United States: Insight from high-resolution dynamical downscaling. Journal of Climate. Advance online publication.Find this resource:

Marsh, P. T., Brooks, H. E., & Karoly, D. J. (2007). Assessment of the severe weather environment in North America simulated by a global climate model. Atmospheric Science Letters, 8.Find this resource:

Púčik, T., Groenemeijer, P., Rädler, A. T., Tijssen, L., Nikulin, G., Prein, A. F., . . .Teichmann, C. (2017). Future changes in European severe convection environments in a regional climate model ensemble. Journal of Climate, 30, 6771–6794.Find this resource:

Tippett, M. K., Allen, J. T., Gensini, V. A., & Brooks, H. E. (2015). Climate and hazardous convective weather. Current Climate Change Reports, 1, 60–73.Find this resource:

Trapp, R. J., Diffenbaugh, N. S., Brooks, H. E., Baldwin, M. E., Robinson, E. D., & Pal, J. S. (2007a). Changes in severe thunderstorm frequency during the 21st century due to anthropogenically enhanced global radiative forcing. Proceedings of the National Academy of Sciences, USA, 104.Find this resource:

Trapp, R. J., Diffenbaugh, N. S., & Gluhovsky, A. (2009). Transient response of severe thunderstorm forcing to elevated greenhouse gas concentrations. Geophysical Research Letters, 36, L01703.Find this resource:

Trapp, R. J., Halvorson, B. A., & Diffenbaugh, N. S. (2007b). Telescoping, multimodal approaches to evaluate extreme convective weather under future climates. Journal of Geophysical Research, 112, D20109.Find this resource:

Trapp, R. J., & Hoogewind, K. A. (2016). The realization of extreme tornadic storm events under future anthropogenic climate change. Journal of Climate, 29, 5251–5265.Find this resource:

Trapp, R. J., Robinson, E. D., Baldwin, M. E., Diffenbaugh, N. S., & Schwedler, B. R. J. (2011). Regional climate of hazardous convective weather through high-resolution dynamical downscaling. Climate Dynamics, 37, 677–688.Find this resource:

## References

Adams-Selin, R. D., & Ziegler, C. L. (2016). Forecasting hail using a one-dimensional hail growth model within WRF. Monthly Weather Review, 144, 4919–4939.Find this resource:

Admirat, P., Goyer, G. G., Wojtiw, L., Carte, E. A., Roos, D., & Lozowki, E. P. (1985). A comparative study of hailstorms in Switzerland, Canada and South Africa. International Journal of Climatology, 5, 35–51.Find this resource:

Agard, V., & Emanuel, K. (2017). Clausius–Clapeyron scaling of peak CAPE in continental convective storm environments. Journal of the Atmospheric Sciences, 74, 3043–3054.Find this resource:

Allen, J., & Karoly, D. (2014). A climatology of Australian severe thunderstorm environments 1979–2011: Inter-annual variability and the ENSO influence. International Journal of Climatology, 34, 81–97.Find this resource:

Allen, J., Karoly, D., & Mills, G. (2011). A severe thunderstorm climatology for Australia and associated thunderstorm environments. Australian Meteorological Oceanographic Journal, 61, 143–158.Find this resource:

Allen, J. T. (2017). Atmospheric hazards: Hail potential heating up. Nature Climate Change, 7, 474–475.Find this resource:

Allen, J. T., & Allen, E. R. (2016). A review of severe thunderstorms in Australia. Atmospheric Research, 178–179, 347–366.Find this resource:

Allen, J. T., Karoly, D. J., & Walsh, K. J. (2014a). Future Australian severe thunderstorm environments. Part I: A novel evaluation and climatology of convective parameters from two climate models for the late twentieth century. Journal of Climate, 27, 3827–3847.Find this resource:

Allen, J. T., Karoly, D. J., & Walsh, K. J. (2014b). Future Australian severe thunderstorm environments. Part II: The influence of a strongly warming climate on convective environments. Journal of Climate, 27, 3848–3868.Find this resource:

Allen, J. T., Pezza, A. B., & Black, M. T. (2010). Explosive cyclogenesis: A global climatology comparing multiple reanalyses. Journal of Climate, 23, 6468–6484.Find this resource:

Allen, J. T., & Tippett, M. K. (2015). Characteristics of the United States Hail Observations Dataset 1955–2014. Electronic Journal of Severe Storms Meteorology, 10, 1–31.Find this resource:

Allen, J. T., Tippett, M. K., & Sobel, A. H. (2015a). An empirical model relating U.S. monthly hail occurrence to large-scale meteorological environment. Journal of Advances in Modeling Earth Systems, 7, 226.Find this resource:

Allen, J. T., Tippett, M. K., Kaheil, Y., Sobel, A. H., Lepore, C., Nong, S., & Muehlbauer, A. (2017). An extreme value model for U.S. hail size. Monthly Weather Review, 145, 4501–4519.Find this resource:

Allen, J. T., Tippett, M. K., & Sobel, A. H. (2015b). Influence of the El Niño/Southern Oscillation on tornado and hail frequency in the United States. Nature Geoscience, 8, 278–283.Find this resource:

Antonescu, B., Schultz, D. M., Holzer, A., & Groenemeijer, P. (2017). Tornadoes in Europe: An underestimated threat. Bulletin of the American Meteorological Society, 98, 713–728.Find this resource:

Antonescu, B., Schultz, D. M., Lomas, F., & Kühne, T. (2016). Tornadoes in Europe: Synthesis of the observational datasets. Monthly Weather Review, 144, 2445–2480.Find this resource:

Ashley, W. S., Bentley, M. L., & Stallins, J. A. (2012). Urban-induced thunderstorm modification in the Southeast United States. Climatic Change, 113, 481–498.Find this resource:

Balling, R. C., & Cerveny, R. S. (2003). Compilation and discussion of trends in severe storms in the United States: Popular perception vs. climate reality. Natural Hazards, 29, 103–112.Find this resource:

Banafsheh, R., Asl, S. J. B., & Ghaderi, F. (2016). Synoptic analysis and humidity flux identification of days with hail fall in the West Iran (Case study: Zab Catchment, west Azerbaijan). Journal of Current Research in Science, 2, 211–224.Find this resource:

Barrett, B. S., & Gensini, V. A. (2013). Variability of central United States April–May tornado day likelihood by phase of the Madden-Julian Oscillation. Geophysical Research Letters, 40, 2790–2795.Find this resource:

Barrett, B. S., & Henley, B. N. (2015). Intraseasonal variability of hail in the contiguous United States: relationship to the Madden-Julian oscillation. Monthly Weather Review, 143, 1086–1103.Find this resource:

Barthel, F., & Neumayer, E. (2012). A trend analysis of normalized insured damage from natural disasters. Climatic Change, 113, 215–237.Find this resource:

Berthet, C., Dessens, J., & Sanchez, J. L. (2011). Regional and yearly variations of hail frequency and intensity in France. Atmospheric Research, 100, 391–400.Find this resource:

Bhan, S. C., Paul, S., Chakravarthy, K., Saxena, R., Ray, K., & Gopal, N. K. (2016). Climatology of tornadoes over northwest India and Pakistan; and meteorological analysis of recent tornadoes over the region. Journal of the Indian Geophysical Union, 20, 75–88.Find this resource:

Billet, J., DeLisi, M., & Smith, B. G. (1997). Use of regression techniques to predict hail size and the probability of large hail. Weather and Forecasting, 12, 154–164.Find this resource:

Blair, S. F., Laflin, J. M., Cavanaugh, D. E., Sanders, K. J., Currens, S. R., Pullin, J. I., . . . Mallinson, H. M. (2017). High-resolution hail observations: Implications for NWS warning operations. Weather and Forecasting, 32, 1101–1119.Find this resource:

Blamey, R. C., Middleton, C., Lennard, C., & Reason, C. J. C. (2017). A climatology of potential severe convective environments across South Africa. Climate Dynamics, 49, 2161–2178.Find this resource:

Botzen, W. J. W., & Bouwer, L. M. (2016). Weather indicators for insured hailstorm damage to motor vehicles and potential climate change impacts. The Geneva Papers on Risk and Insurance-Issues and Practice, 41, 512–527.Find this resource:

Botzen, W. J. W., Bouwer, L. M., & van den Bergh, J. C. J. M., (2010). Climate change and hailstorm damage: Empirical evidence and implications for agriculture and insurance. Resource and Energy Economics, 32, 341–362.Find this resource:

Brimelow, J. C., Burrows, W. R., & Hanesiak, J. M. (2017). The changing hail threat over North America in response to anthropogenic climate change. Nature Climate Change, 7, 516–522.Find this resource:

Brooks, H. E. (2009). Proximity soundings for Europe and the United States from reanalysis data. Atmospheric Research, 93, 546–553.Find this resource:

Brooks, H. E. (2013). Severe thunderstorms and climate change. Atmospheric Research, 123, 129–138.Find this resource:

Brooks, H. E., Anderson, A. R., Riemann, K., Ebbers, I., & Flachs, H. (2007). Climatological aspects of convective parameters from the NCAR/NCEP reanalysis. Atmospheric Research, 83, 294–305.Find this resource:

Brooks, H. E., & Doswell, C. A. (2001). Some aspects of the international climatology of tornadoes by damage classification. Atmospheric Research, 56, 191–201.Find this resource:

Brooks, H. E., & Dotzek, N. (2008). The spatial distribution of severe convective storms and an analysis of their secular changes. In H. F. Diaz & R. Murnane (Eds.), Climate extremes and society (pp. 35–54). New York, NY: Cambridge University Press.Find this resource:

Brooks, H. E., Lee, J. W. & Craven, J. P. (2003). The spatial distribution of severe thunderstorm and tornado environments from global reanalysis data. Atmospheric Research, 67–68, 73–94.Find this resource:

Brooks, H. E., Carbin, G. W., & Marsh, P. T. (2014). Increased variability of tornado occurrence in the United States. Science, 346, 349–352.Find this resource:

Brown, B. G., & Murphy, A. H. (1996). Verification of aircraft icing forecasts: The use of standard measures and meteorological covariates. Preprints, 13th Conf. Probability and Statistics in the Atmospheric Sciences, San Francisco, California, USA, American Meteorological Society, pp. 251–252.Find this resource:

Burcea, S., Cică, R., & Bojariu, R. (2016). Hail climatology and trends in Romania: 1961–2014. Monthly Weather Review, 144, 4289–4299.Find this resource:

Cai, W., Santoso, A., Wang, G., Yeh, S. W., An, S. I., Cobb, K. M., . . . Lengaigne, M. (2015). ENSO and greenhouse warming. Nature Climate Change, 5, 849–859.Find this resource:

Cao, Z. (2008). Severe hail frequency over Ontario, Canada: Recent trend and variability. Geophysical Research Letters, 35, L14803.Find this resource:

Carleton, A. M., Adegoke, J., Allard, J., Arnold, D. L., & Travis, D. J. (2001). Summer season land cover—convective cloud associations for the Midwest US “Corn Belt.” Geophysical Research Letters, 28, 1679–1682.Find this resource:

Cecil, D. J., & Blankenship, C. B. (2012). Toward a global climatology of severe hailstorms as estimated by satellite passive microwave imagers. Journal of Climate, 25, 687–703.Find this resource:

Changnon, S. A., Jr. (1977). The scales of hail. Journal of Applied Meteorology, 16, 626–648.Find this resource:

Changnon, S. A. (2003). Shifting economic impacts from weather extremes in the United States: A result of societal changes, not global warming. Natural Hazards, 29, 273–290.Find this resource:

Changnon, S. A., & Changnon, D. (2000). Long-term fluctuations in hail incidences in the United States. Journal of Climate, 13, 658–664.Find this resource:

Cheng, V. Y., Arhonditsis, G. B., Sills, D. M., Auld, H., Shephard, M. W., Gough, W. A., & Klaassen, J. (2013). Probability of tornado occurrence across Canada. Journal of Climate, 26, 9415–9428.Find this resource:

Cheng, V. Y. S., Arhonditsis, G. B., Sills, D. M. L., Gough, W. A., & Auld, H. (2015). A Bayesian modelling framework for tornado occurrences in North America. Nature Communications, 6, 6599.Find this resource:

Cheng, V. Y. S., Arhonditsis, G. B., Sills, D. M. L., Gough, W. A., & Auld, H. (2016). Predicting the climatology of tornado occurrences in North America with a Bayesian hierarchical modeling framework. Journal of Climate, 29, 1899–1917.Find this resource:

Chromá, K., Brázdil, R., & Tolasz, R. (2005). Spatio-temporal variability of hailstorms for Moravia and Silesia in the summer half-year of the period 1961–2000. Meteorologický časopis, 8, 65–74.Find this resource:

Cintineo, J. L., Smith, T. M., Lakshmanan, V., Brooks, H. E., & Ortega, K. L. (2012). An objective high-resolution hail climatology of the contiguous United States. Weather and Forecasting, 27, 1235–1248.Find this resource:

Compo, G. P., Whitaker, J. S., Sardeshmukh, P. D., Matsui, N., Allan, R. J., Yin, X., . . . Brönnimann, S. (2011). The twentieth century reanalysis project. Quarterly Journal of the Royal Meteorological Society, 137(654), 1–28.Find this resource:

Cook, A. R., Leslie, L. M., Parsons, D. B., & Schaefer, J. T. (2017). The impact of the El Niño Southern Oscillation (ENSO) on winter and early spring U.S. tornado outbreaks. Journal of Applied Meteorology and Climatology, 56, 2455–2478.Find this resource:

Cook, A. R., & Schaefer, J. T. (2008). The relation of El Niño–Southern Oscillation (ENSO) to winter tornado outbreaks. Monthly Weather Review, 136, 3121–3137.Find this resource:

Cook, B. I., Ault, T. R., & Smerdon, J. E. (2015). Unprecedented 21st-century drought risk in the American Southwest and Central Plains. Scientific Advances, 1, e1400082.Find this resource:

Corfidi, S. (2017). Forecasting severe convective storms. Oxford research encyclopedia of climate science.Find this resource:

Craven, J. P., Jewell, R. E., & Brooks, H. E. (2002). Comparison between observed convective cloud-base heights and lifting condensation level for two different lifted parcels. Weather and Forecasting, 17, 885–890.Find this resource:

Cubasch, U., & Meehl, G. A. (2001). Projections of future climate change. In J. T. Houghton et al. (Eds.), Climate change 2001: The scientific basis (pp. 525–582). New York, NY: Cambridge University Press.Find this resource:

Ćurić, M., & Janc, D. (2016). Hail climatology in Serbia. International Journal of Climatology, 36, 3270–3279.Find this resource:

Dai, A. (2006). Recent climatology, variability, and trends in global surface humidity. Journal of Climate, 19, 3589–3606.Find this resource:

Dai, A. G., & Trenberth, K. E. (2004). The diurnal cycle and its depiction in the Community Climate System Model. Journal of Climate, 17, 930–951.Find this resource:

DeAngelis, A., Dominguez, F., Fan, Y., Robock, A., Kustu, M. D., & Robinson, D. (2010). Evidence of enhanced precipitation due to irrigation over the Great Plains of the United States. Journal of Geophysical Research, 115, D15115.Find this resource:

Degu, A. M., & Hossain, F. (2012). Investigating the mesoscale impact of artificial reservoirs on frequency of rain during growing season. Water Resource Research, 48, W05510.Find this resource:

Del Genio, A. D., Yao, M.-S., & Jonas, J. (2007). Will moist convection be stronger in a warmer climate? Geophysical Research Letters, 34, L16703.Find this resource:

Dennis, E. J., & Kumjian, M. R. (2017). The impact of vertical wind shear on hail growth in simulated supercells. Journal of the Atmospheric Sciences, 74, 641–663.Find this resource:

Dessens, J. (1986). Hail in south-western France. I: Hailfall characteristics and hailstorm Environment. Journal of Climatology and Applied Meteorology, 25, 35–47.Find this resource:

Dessens, J. (1995). Severe convective weather in the context of a nighttime global warming. Geophysical Research Letters, 22, 1241–1244.Find this resource:

Dessens, J., Berthet, C., & Sanchez, J. L. (2015). Change in hailstone size distributions with an increase in the melting level height. Atmospheric Research, 158–159, 245–253.Find this resource:

Diffenbaugh, N. S., Ashfaq, M., Shuman, B., Williams, J. W., & Bartlein, P. J. (2006). Summer aridity in the United States: Response to mid-Holocene changes in insolation and sea surface temperature. Geophysical Research Letters, 33.Find this resource:

Diffenbaugh, N. S., Scherer, M. & Trapp, R. J. (2013). Robust increases in severe thunderstorm environments in response to greenhouse forcing. Proceedings of the National Academy of Sciences USA, 110, 16,361–16,366.Find this resource:

Diffenbaugh, N. S., Trapp, R. J., & Brooks, H. E. (2008). Challenges in identifying influences of global warming on tornado activity. Eos Transactions, 89, 553–554.Find this resource:

Doswell, C. (2001). Severe convective storms: An overview. In C. Doswell (Ed.), Severe convective storms (pp. 1–26). Boston, MA: American Meteorological Society.Find this resource:

Doswell, C. A. (2016). History of convective storm science. Oxford research encyclopedia of climate science.Find this resource:

Doswell, C. A., & Bosart, L. F. (2001). Extratropical synoptic-scale processes and severe convection. Meteorological Monographs, 50, 27–70.Find this resource:

Doswell, C. A., Brooks, H. E., & Kay, M. P. (2005). Climatological estimates of daily nontornadic severe thunderstorm probability for the United States. Weather and Forecasting, 20, 577–595.Find this resource:

Doswell, C. A., Brooks, H. E., & Maddox, R. A. (1996). Flash flood forecasting: An ingredients-based methodology. Weather and Forecasting, 11, 560–580.Find this resource:

Doswell, C. A., & Schultz, D. M. (2006). On the use of indices and parameters in forecasting severe storms. Electronic Journal of Severe Storms Meteorology, 1, 1–22.Find this resource:

Dotzek, N., Groenemeijer, P., Feuerstein, B., & Holzer, A. M. (2009). Overview of ESSL's severe convective storms research using the European Severe Weather Database ESWD. Atmospheric Research, 93, 575–586.Find this resource:

Dowdy, A. J., & Kuleshov, Y. (2014). Climatology of lightning activity in Australia: Spatial and seasonal variability. Australian Meteorological and Oceanographic Journal, 6, 9–14.Find this resource:

Durkee, J., Degu, A. M., Hossain, F., Mahmood, R., Winchester, J., & Chronis, T. (2014). Investigating the effect of land between the lakes on storm patterns. Journal of Applied Meteorology and Climatology, 53, 1506–1524.Find this resource:

Eccel, E., Cau, P., Riemann-Campe, K., & Biasioli, F. (2012). Quantitative hail monitoring in an alpine area: 35-year climatology and links with atmospheric variables. International Journal of Climatology, 32, 503–517.Find this resource:

Elsner, J. B., Elsner, S. C., & Jagger, T. H. (2015). The increasing efficiency of tornado days in the United States. Climate Dynamics, 45, 651–659.Find this resource:

Elsner, J. B., Jagger, T. H., & Fricker, T. (2016). Statistical models for tornado climatology: Long and short-term views. PloS one, 11, e0166895.Find this resource:

Elsner, J. B., & Widen, H. M. (2014). Predicting spring tornado activity in the central Great Plains by 1 March. Monthly Weather Review, 142, 259–267.Find this resource:

Emanuel, K. A. (1994). Atmospheric convection. New York, NY: Oxford University Press.Find this resource:

Etkin, D., & Brun, S. E. (1999). A note on Canada’s hail climatology: 1977–1993. International Journal of Climatology, 19, 1357–1373.Find this resource:

Etkin, D., Brun, S. E., Shabbar, A., & Joe, P. (2001). Tornado climatology of Canada revisited: Tornado activity during different phases of ENSO. International Journal of Climatology, 21, 915–938.Find this resource:

Fujita, T. T. (1963). Analytical mesometeorology: A review. Severe Local Storms, Meteorological Monograph, 5, 77–125.Find this resource:

Fujita, T. T. (1973). Tornadoes around the world. Weatherwise, 56, 78–83.Find this resource:

Frisby, E. M., & Sansom, H. W. (1967). Hail incidence in the tropics. Journal of Applied Meteorology, 6, 339–354.Find this resource:

Frye, J. D., & Mote, T. L. (2010). Convection initiation along soil moisture boundaries in the Southern Great Plains. Monthly Weather Review, 138, 1140–1151.Find this resource:

García-Ortega, E., Hermida, L., Hierro, R., Merino, A., Gascón, E., Fernández-González, S., . . . López, L. (2014). Anomalies, trends and variability in atmospheric fields related to hailstorms in north-eastern Spain. International Journal of Climatology, 34, 3251–3263.Find this resource:

Gates, W. L., Rowntree, P. R., & Zeng, Q. C. (1990). Validation of climate models. In J. T. Houghton, G. J. Jenkins, & J. J. Ephraums (Eds.), Climate change: The IPCC scientific assessment (pp. 93–130). Cambridge, UK: Cambridge University Press.Find this resource:

Gensini, V. A., & Ashley, W. S. (2011). Climatology of potentially severe convective environments from the North American regional reanalysis. Electronic Journal of Severe Storms Meteorology, 6, 1–40.Find this resource:

Gensini, V. A., & Marinaro, A. (2016). Tornado frequency in the United States related to global relative angular momentum. Monthly Weather Review, 144, 801–810.Find this resource:

Gensini, V. A., & Allen, J. T. (2018). US hail frequency and the Global Wind Oscillation. Geophysical Research Letters, 45(3), 1611–1620.Find this resource:

Gensini, V. A., & Mote, T. L. (2014). Estimations of hazardous convective weather in the United States using dynamical downscaling. Journal of Climate, 27, 6581–6598.Find this resource:

Gensini, V. A., Mote, T. L., & Brooks, H. E. (2014a). Severe thunderstorm reanalysis environments and collocated radiosonde observations. Journal of Applied Meteorology and Climatology, 53, 742–751.Find this resource:

Gensini, V. A., Ramseyer, C. A., & Mote, T. L. (2014b). Future convective environments using NARCCAP. International Journal of Climatology, 34, 1699–1705.Find this resource:

Gensini, V. A., & Mote, T. L. (2015). Downscaled estimates of late 21st century severe weather from CCSM3. Climatic Change, 129, 307–321.Find this resource:

Gero, A. F., Pitman, A. J., Narisma, G. T., Jacobson, C., & Pielke, R. A. (2006). The impact of land cover change on storms in the Sydney Basin, Australia. Global and Planetary Change, 54, 57–78.Find this resource:

Gettleman, A., Seidel, D., Wheeler, M., & Ross, R. (2002). Multi-decadal trends in tropical convective available potential energy. Journal of Geophysical Research, 107(D21), 4606–4614.Find this resource:

Gilleland, E., Brown, B. G., & Ammann, C. M. (2013). Spatial extreme value analysis to project extremes of large-scale indicators for severe weather. Environmetrics, 24, 418–432.Find this resource:

Giorgi, E. (1990). Simulation of regional climate using a limited area model nested in a general circulation model. Journal of Climate, 3, 941–963.Find this resource:

Goliger, A., & Milford, R. (1998). A review of worldwide occurrence of tornadoes. Journal of Wind Engineering and Industrial Aerodynamics, 74, 111–121.Find this resource:

Grazulis, T. P. (1993). A 110-year perspective of significant tornadoes. In C. Church, D. Burgess, C. Doswell, & R. Davies-Jone (Eds.), The tornado: Its structure, dynamics, prediction, and hazards (pp. 467–474). Washington, DC: American Geophysical Union.Find this resource:

Griffiths, D., Colquhoun, J., Batt, K., & Casinader, T. (1993). Severe thunderstorms in New South Wales: Climatology and means of assessing the impact of climate change. Climatic Change, 25, 369–388.Find this resource:

Groenemeijer, P., & Kühne, T. (2014). A climatology of tornadoes in Europe: Results from the European Severe Weather Database. Monthly Weather Review, 142, 4775–4790.Find this resource:

Groenemeijer, P., Púčik, T., Holzer, A. M., Antonescu, B., Riemann-Campe, K., Schultz, D. M., . . . Sausen, R. (2017). Severe convective storms in Europe: 10 years of research at the European Severe Storms Laboratory. Bulletin of the American Meteorological Society, 98, 2641–2651.Find this resource:

Haberlie, A. M., Ashley, W. S., Fultz, A. J., & Eagan, S. M. (2016). The effect of reservoirs on the climatology of warm‐season thunderstorms in Southeast Texas, USA. International Journal of Climatology, 36, 1808–1820.Find this resource:

Haberlie, A. M., Ashley, W. S., & Pingel, T. J. (2015). The effect of urbanisation on the climatology of thunderstorm initiation. Quarterly Journal of the Royal Meteorological Society, 141, 663–675.Find this resource:

Hales, J. E. (1993). Biases in the severe thunderstorm database ramifications and solutions. Preprints, 13th Conference on Weather Analysis and Forecasting, Vienna, Virginia, American Meteorological Society, pp. 504–507.Find this resource:

Hand, W. H., & Cappelluti, G. (2011). A global hail climatology using the UK Met Office convection diagnosis procedure (CDP) and model analyses. Meteorological Applications, 18, 446–458.Find this resource:

Hanstrum, B. N., Mills, G. A., Watson, A., Monteverdi, J. P., & Doswell, C. A. (2002). The cool-season tornadoes of California and southern Australia. Weather and Forecasting, 17, 705–722.Find this resource:

Heaton, M. J., Katzfuss, M., Ramachandar, S., Pedings, K., Gilleland, E., Mannshardt-Shamseldin, E., & Smith, R. L. (2011). Spatio-temporal models for large scale indicators of extreme weather. Environmetrics, 22, 294–303.Find this resource:

Hermida, L., Sanchez, J. L., Lopez, L., Berthet, C., Dessens, J., Garcia-Ortega, E., & Merino, A. (2013). Climatic trends in hail precipitation in France: spatial, altitudinal, and temporal variability. The Scientific World Journal, 2013. Online publication.Find this resource:

Hill, C. M., Fitzpatrick, P. J., Corbin, J. H., Lau, Y. H., & Bhate, S. K. (2010). Summertime precipitation regimes associated with the sea breeze and land breeze in Southern Mississippi and Eastern Louisiana. Weather and Forecasting, 25, 1755–1779.Find this resource:

Hoeppe, P. (2016). Trends in weather related disasters—Consequences for insurers and society. Weather and Climate Extremes, 11, 70–79.Find this resource:

Hoogewind, K. A., Baldwin, M. E., & Trapp, R. J. (2017). The impact of climate change on hazardous convective weather in the United States: Insight from high-resolution dynamical downscaling. Journal of Climate. Advance online publication.Find this resource:

Houze, R. A. (1997). Stratiform precipitation in regions of convection: A meteorological paradox? Bulletin of the American Meteorological Society, 78, 2179–2196.Find this resource:

Huff, F. A., & Changnon, S. A. (1972). Climatological assessment of urban effects on precipitation at St Louis. Journal of Applied Meteorology, 11, 823–842.Find this resource:

Huryn, S. M., Gough, W. A., & Butler, K. (2016). A review of thunderstorm trends across southern Ontario, Canada. Atmosphere-Ocean, 54, 519–528.Find this resource:

Jin, H.-G., Lee, H., Lkhamjav, J., & Baik, J.-J. (2017). A hail climatology in South Korea. Atmospheric Research, 188, 90–99.Find this resource:

Johns, R. H., & Hirt, W. D. (1987). Derechos: Widespread convectively induced windstorms. Weather and Forecasting, 2, 32–49.Find this resource:

Jung, E., & Kirtman, B. P. (2016). Can we predict seasonal changes in high impact weather in the United States? Environmental Research Letters, 11, 74018.Find this resource:

Kalnay, E., Kanamitsu, M., Kistler, R., Collins, W., Deaven, D., Gandin, L., . . . Joseph, D. (1996). The NCEP/NCAR 40-year reanalysis project. Bulletin of the American Meteorological Society, 77, 437–472.Find this resource:

Kapsch, M.-L., Kunz, M., Vitolo, R., & Economou, T. (2012). Long-term trends of hail-related weather types in an ensemble of regional climate models using a Bayesian approach. Journal of Geophysical Research, 117, D15107.Find this resource:

Kelly, D. L., Schaefer, J. T., & Doswell, C. A. (1985). Climatology of non-tornadic severe thunderstorm events in the United States. Monthly Weather Review, 113, 1997–2014.Find this resource:

King, P. S. W. (1997). On the absence of population bias in the tornado climatology of southwestern Ontario. Weather and Forecasting, 12, 939–946.Find this resource:

Klein, S. A., Jiang, X., Boyle, J., Malyshev, S., & Xie, S. (2006). Diagnosis of the summertime warm and dry bias over the U.S. Southern Great Plains in the GFDL climate model using a weather forecasting approach. Geophysical Research Letters, 33. Online publication.Find this resource:

Kounkou, R., Mills, G., & Timbal, B. (2009). A reanalysis climatology of cool-season tornado environments over southern Australia. International Journal of Climatology, 29, 2079–2090.Find this resource:

Kuleshov, Y., Hoedt, G. D., Wright, W., & Brewster, A. (2002). Thunderstorm distribution and frequency in Australia. Australian Meteorological Magazine, 51, 145–154.Find this resource:

Kunz, M., Sander, J., & Kottmeier, C. (2009). Recent trends of thunderstorm and hailstorm frequency and their relation to atmospheric characteristics in southwest Germany. International Journal of Climatology, 29, 2283–2297.Find this resource:

Lackmann, G. M. (2013). The south-central U.S. flood of May 2010: Present and future. Journal of Climate, 26, 4688–4709.Find this resource:

Laird, N. F., Kristovich, D. A. R., Liang, X., Arritt, R. W., & Labas, K. (2001). Lake Michigan lake breezes: Climatology, local forcing, and synoptic environment. Journal of Applied Meteorology, 40, 409–424.Find this resource:

Lanicci, J. M., & Warner, T. T. (1991). A synoptic climatology of the elevated mixed-layer inversion over the southern Great Plains in spring. Part I: Structure, dynamics, and seasonal evolution. Weather and Forecasting, 6, 198–213.Find this resource:

Lee, J. W. (2002). Tornado proximity soundings from the NCEP/NCAR reanalysis data (Unpublished master's thesis). University of Oklahoma, Norman.Find this resource:

Lee, S. K., Atlas, R., Enfield, D., Wang, C., & Liu, H. (2013). Is there an optimal ENSO pattern that enhances large-scale atmospheric processes conducive to tornado outbreaks in the United States? Journal of Climate, 26, 1626–1642.Find this resource:

Lee, S.-K., Wittenberg, A. T., Enfield, D. B., Weaver, S. J., Wang, C., & Atlas, R. (2016). US regional tornado outbreaks and their links to spring ENSO phases and North Atlantic SST variability. Environmental Research Letters, 11, 044008.Find this resource:

Lepore, C., Tippett, M. K., & Allen, J. T. (2017). ENSO-based probabilistic forecasts of March–May U.S. tornado and hail activity. Geophysical Research Letters, 44, 9093–9101.Find this resource:

Leslie, L. M., Leplastrier, M., & Buckley, B. W. (2008). Estimating future trends in severe hailstorms over the Sydney Basin: A climate modelling study. Atmospheric Research, 87, 37–51.Find this resource:

Li, H., & Colle, B. A. (2016). Future changes in convective storm days over the northeastern United States using linear discriminant analysis applied to CMIP5 predictions. Journal of Climate, 29, 4327–4345.Find this resource:

Li, M., Zhang, Q., & Zhang, F. (2016). Hail day frequency trends and associated atmospheric circulation patterns over China during 1960–2012. Journal of Climate, 29, 7027–7044.Find this resource:

Long, J. A., & Stoy, P. C. (2014). Peak tornado activity is occurring earlier in the heart of “Tornado Alley.” Geophysical Research Letters, 41, 6259–6264.Find this resource:

Lu, M., Tippett, M., & Lall, U. (2015). Changes in the seasonality of tornado and favorable genesis conditions in the central United States. Geophysical Research Letters, 42, 4224–4231.Find this resource:

Mahoney, K., Alexander, M., Scott, J. D., & Barsugli, J. (2013). High-resolution downscaled simulations of warm-season extreme precipitation events in the Colorado Front Range under past and future climates. Journal of Climate, 26, 8671–8689.Find this resource:

Mahoney, K., Alexander, M. A., Thompson, G., Barsugli, J. J., & Scott, J. D. (2012). Changes in hail and flood risk in high-resolution simulations over Colorado’s mountains. Nature Climate Change, 2, 125–131.Find this resource:

Malkarova, A. (2011). Estimation of physical efficiency of hail protection accounting for changes in hail climatology. Russian Meteorology and Hydrology, 36, 392–398.Find this resource:

Mannshardt, E., & Gilleland, E. (2013). Extremes of severe storm environments under a changing climate. American Journal of Climate Change, 2, 47–61.Find this resource:

Manzato, A. (2012). Hail in northeast Italy: Climatology and bivariate analysis with the sounding-derived indices. Journal of Applied Meteorology and Climatology, 51, 449–467.Find this resource:

Marsh, P. T., Brooks, H. E., & Karoly, D. J. (2007). Assessment of the severe weather environment in North America simulated by a global climate model. Atmospheric Science Letters, 8.Find this resource:

Marsh, P. T., Brooks, H. E., & Karoly, D. J. (2009). Preliminary investigation into the severe thunderstorm environment of Europe simulated by the Community Climate System Model 3. Atmospheric Research, 93, 607–618.Find this resource:

Martins, J. A., Brand, V. S., Capucim, M. N., Felix, R. R., Martins, L. D., Freitas, E. D., . . . Cecil, D. J. (2017). Climatology of destructive hailstorms in Brazil. Atmospheric Research, 184, 126–138.Find this resource:

McMaster, H. J. (1999). The potential impact of global warming on hail losses to winter cereal crops in New South Wales. Climatic Change, 43, 455–476.Find this resource:

McMaster, H. (2001). Hailstorm risk assessment in rural New South Wales. Natural Hazards, 24(2), 187–196.Find this resource:

Mesinger, F., DiMego, G., Kalnay, E., Mitchell, K., Shafran, P. C., Ebisuzaki, W., . . . Shi, W. (2006). North American regional reanalysis. Bulletin of the American Meteorological Society, 87, 343–360.Find this resource:

Mezher, R. N., Doyle, M., & Barros, V. (2012). Climatology of hail in Argentina. Atmospheric Research, 114–115, 70–82.Find this resource:

Mohr, S., & Kunz, M. (2013). Recent trends and variabilities of convective parameters relevant for hail events in Germany and Europe. Atmospheric Research, 123, 211–228.Find this resource:

Mohr, S., Kunz, M., & Keuler, K. (2015). Development and application of a logistic model to estimate the past and future hail potential in Germany. Journal of Geophysical Research Atmospheres, 120, 3939–3956.Find this resource:

Molina, M. J., Timmer, R. P., & Allen, J. T. (2016). Importance of the Gulf of Mexico as a climate driver for U.S. severe thunderstorm activity. Geophysical Research Letters, 43, 12,295–12,304.Find this resource:

Moore, T. W. (2017). On the temporal and spatial characteristics of tornado days in the United States. Atmospheric Research, 184, 56–65.Find this resource:

Muramatsu, T., Kato, T., Nakazato, M., Endo, H., & Kitoh, A. (2016). Future change of tornadogenesis: Favorable environmental conditions in Japan estimated by a 20-km-mesh atmospheric general circulation model. Journal of the Meteorological Society of Japan, 94A, 105–120.Find this resource:

Ni, X., Liu, C., Cecil, D. J., & Zhang, Q. (2017). On the hail detection using satellite passive microwave radiometers and precipitation radar. Journal of Applied Meteorology and Climatology. Advance online publication.Find this resource:

Ni, X., Zhang, Q., Liu, C., Li, X., Zou, T., Lin, J., . . . Ren, Z. (2017). Decreased hail size in China since 1980. Scientific Reports, 7, 10913.Find this resource:

Niall, S., & Walsh, K. (2005). The impact of climate change on hail storms in southeast Australia. International Journal of Climatology, 25, 1933–1952.Find this resource:

Niyogi, D., Pyle, P., Lei, M., Arya, S. P., Kishtawal, C. M., Shepherd, M., . . . Wolfe, B. (2011). Urban modification of thunderstorms: An observational storm climatology and model case study for the Indianapolis urban region. Journal of Applied Meteorology and Climatology, 50, 1129–1144.Find this resource:

Nizamuddin, S. (1993). Hail occurrences in India. Weather, 48, 90–92.Find this resource:

Paquin, D., de Elía, R., & Frigon, A. (2014). Change in North American atmospheric conditions associated with deep convection and severe weather using CRCM4 climate projections. Atmosphere-Ocean, 52, 175–190.Find this resource:

Peterson, R. E., & Mehta, K. C. (1981). Climatology of tornadoes of India and Bangladesh. Meteorology and Atmospheric Physics, 29, 345–356.Find this resource:

Peterson, T. C., Willett, K. M., & Thorne, P. W. (2011). Observed changes in surface atmospheric energy over land. Geophysical Research Letters, 38, L16707.Find this resource:

Piani, F., Crisci, A., De Chiara, G., Maracchi, G., & Meneguzzo, F. (2005). Recent trends and climatic perspectives of hailstorms frequency and intensity in Tuscany and Central Italy. Natural Hazards Earth System Science, 5, 217–224.Find this resource:

Počakal, D. (2011). Hailpad data analysis for the continental part of Croatia. Meteorologische Zeitschrift, 20, 441–447.Find this resource:

Púčik, T., Groenemeijer, P., Rädler, A. T., Tijssen, L., Nikulin, G., Prein, A. F., . . . Teichmann, C. (2017). Future changes in European severe convection environments in a regional climate model ensemble. Journal of Climate, 30, 6771–6794.Find this resource:

Púčik, T., Groenemeijer, P., Ryva, D., & Kolar, M. (2015). Proximity soundings of severe and nonsevere thunderstorms in central Europe. Monthly Weather Review, 143, 4805–4821.Find this resource:

Punge, H. J., & Kunz, M. (2016). Hail observations and hailstorm characteristics in Europe: A review. Atmospheric Research, 176–177, 159–184.Find this resource:

Rasmussen, E. N. (2003). Refined supercell and tornado forecast parameters. Weather and Forecasting, 18, 530–535.Find this resource:

Rasmussen, E. N., & Blanchard, D. O. (1998). A baseline climatology of sounding-derived supercell and tornado forecast parameters. Weather and Forecasting, 13, 1148–1164.Find this resource:

Rasmussen, K. L., Zuluaga, M. D., & Houze, R. A. (2014). Severe convection and lightning in subtropical South America. Geophysical Research Letters, 41, 7359–7366.Find this resource:

Riemann-Campe, K., Blender, R., & Fraedrich, K. (2011). Global memory analysis in observed and simulated CAPE and CIN. International Journal of Climatology, 31, 1099–1107.Find this resource:

Riemann-Campe, K., Fraedrich, K., & Lunkeit, F. (2009). Global climatology of Convective Available Potential Energy (CAPE) and Convective Inhibition (CIN) in ERA-40 reanalysis. Atmospheric Research, 93, 534–545.Find this resource:

Robinson, E. D., Trapp, R. J., & Baldwin, M. E. (2013). The geospatial and temporal distributions of severe thunderstorms from high-resolution dynamical downscaling. Journal of Applied Meteorology and Climatology, 52, 2147–2161.Find this resource:

Romero, R., Gayà, M., & Doswell, C. A. (2007). European climatology of severe convective storm environmental parameters: A test for significant tornado events. Atmospheric Research, 83, 389–404.Find this resource:

Romps, D. M. (2016). Clausius–Clapeyron scaling of CAPE from analytical solutions to RCE. Journal of the Atmospheric Sciences, 73, 3719–3737.Find this resource:

Sanchez, J. L., Merino, A., Melcón, P., García-Ortega, E., Fernández-González, S., Berthet, C., & Dessens, J. (2017). Are meteorological conditions favoring hail precipitation change in Southern Europe? Analysis of the period 1948–2015. Atmospheric Research, 198, 1–10.Find this resource:

Sander, J., Eichner, J. F., Faust, E., & Steuer, M. (2013). Rising variability in thunderstorm-related U.S. losses as a reflection of changes in large-scale thunderstorm forcing. Weather, Climate and Society, 5, 317–331.Find this resource:

Sanderson, M. G., Hand, W. H., Groenemeijer, P., Boorman, P. M., Webb, J. D. C., & McColl, L. J. (2015). Projected changes in hailstorms during the 21st century over the UK. International Journal of Climatology, 35, 15–24.Find this resource:

Schuster, S. S., Blong, R. J., & Speer, M. S. (2005). A hail climatology of the Greater Sydney Area and New South Wales. International Journal of Climatology, 25, 1633–1650.Find this resource:

Seeley, J. T., & Romps, D. M. (2015). The effect of global warming on severe thunderstorms in the United States. Journal of Climate, 28, 2443–2458.Find this resource:

Shepherd, J. M. (2005). A review of current investigations of urban-induced rainfall and recommendations for the future. Earth Interactions, 9, 1–27.Find this resource:

Sherburn, K. D., Parker, M. D., King, J. R., & Lackmann, G. M. (2016). Composite environments of severe and nonsevere high-shear, low-CAPE convective events. Weather and Forecasting, 31, 1899–1927.Find this resource:

Silva Dias, M. A. (2011). An increase in the number of tornado reports in Brazil. Weather, Climate, and Society, 3, 209–217.Find this resource:

Simeonov, P., Bocheva, L., & Marinova, T. (2009). Severe convective storms phenomena occurrence during the warm half of the year in Bulgaria (1961–2006). Atmospheric Research, 93, 498–505.Find this resource:

Smith, B. T., Castellanos, T. E., Winters, A. C., Mead, C. M., Dean, A. R., & Thompson, R. L. (2013). Measured severe convective wind climatology and associated convective modes of thunderstorms in the contiguous United States, 2003–2009. Weather and Forecasting, 228, 229–236.Find this resource:

Smith, S. B., Reuter, G. W., & Yau, M. K. (1998). The episodic occurrence of hail in central Alberta and the Highveld of South Africa: Research note. Atmosphere-Ocean, 36, 169–178.Find this resource:

Soderholm, J. S., McGowan, H., Richter, H., Walsh, K., Weckwerth, T. M., & Coleman, M. (2017). An 18-year climatology of hailstorms trends and related drivers across Southeast Queensland, Australia. Quarterly Journal of the Royal Meteorological Society, 143, 1123–1135.Find this resource:

Strader, S. M., Ashley, W., Pingel, T., & Krmenec, A. (2017). Projected 21st century changes in tornado exposure, risk, and disaster potential. Climatic Change, 141, 301–313.Find this resource:

Taylor, K. E., Stouffer, R. J., & Meehl, G. A. (2012). An overview of CMIP5 and the experiment design. Bulletin of the American Meteorological Society, 93, 485–498.Find this resource:

Thompson, D. B., & Roundy, P. E. (2013). The relationship between the Madden-Julian oscillation and U.S. violent tornado outbreaks in the spring. Monthly Weather Review, 141, 2087–2095.Find this resource:

Thorne, P. W., & Vose, R. S. (2010). Reanalyses suitable for characterizing long-term trends. Bulletin of the American Meteorological Society, 91, 353–361.Find this resource:

Timbal, B., Kounkou, R., & Mills, G. (2010). Changes in the risk of cool-season tornadoes over southern Australia due to model projections of anthropogenic warming. Journal of Climate, 23, 2440–2449.Find this resource:

Tippett, M., Sobel, A., Camargo, S., & Allen, J. T. (2014). An empirical relation between U.S. tornado activity and monthly environmental parameters. Journal of Climate, 27, 2983–2999.Find this resource:

Tippett, M. K. (2014). Changing volatility of U.S. annual tornado reports. Geophysical Research Letters, 41, 6956–6961.Find this resource:

Tippett, M. K., Allen, J. T., Gensini, V. A., & Brooks, H. E. (2015). Climate and hazardous convective weather. Current Climate Change Reports, 1, 60–73.Find this resource:

Tippett, M. K., & Cohen, J. E. (2016). Tornado outbreak variability follows Taylor’s power law of fluctuation scaling and increases dramatically with severity. Nature Communications, 7, 10668.Find this resource:

Tippett, M. K., Lepore, C., & Cohen, J. E. (2016). More tornadoes in the most extreme U.S. tornado outbreaks. Science, 354, 1419–1423.Find this resource:

Tippett, M. K., Sobel, A. H., & Camargo, S. J. (2012). Association of U.S. tornado occurrence with monthly environmental parameters. Geophysical Research Letters, 39, L02801.Find this resource:

Trapp, R. J., Tessendorf, S. A., Savageau, E. G., & Brooks, H. E. (2005). Tornadoes in squall lines and bow echoes. Part I: Climatological distribution. Weather and Forecasting, 40, 23–34.Find this resource:

Trapp, R. J., Wheatley, D. M., Atkins, N. T., Przybylinski, R. W., & Wolf, R. (2006). Buyer beware: Some words of caution on the use of severe wind reports in post-event assessment and research. Weather and Forecasting, 21, 408–415.Find this resource:

Trapp, R. J., Diffenbaugh, N. S., Brooks, H. E., Baldwin, M. E., Robinson, E. D., & Pal, J. S. (2007a). Changes in severe thunderstorm frequency during the 21st century due to anthropogenically enhanced global radiative forcing. Proceedings of the National Academy of Sciences, USA, 104.Find this resource:

Trapp, R. J., Halvorson, B. A., & Diffenbaugh, N. S. (2007b). Telescoping, multimodal approaches to evaluate extreme convective weather under future climates. Journal of Geophysical Research, 112, D20109.Find this resource:

Trapp, R. J., Diffenbaugh, N. S., & Gluhovsky, A. (2009). Transient response of severe thunderstorm forcing to elevated greenhouse gas concentrations. Geophysical Research Letters, 36, L01703.Find this resource:

Trapp, R. J., Robinson, E. D., Baldwin, M. E., Diffenbaugh, N. S., & Schwedler, B. R. J. (2011). Regional climate of hazardous convective weather through high-resolution dynamical downscaling. Climate Dynamics, 37, 677–688.Find this resource:

Trapp, R. J., & Brooks, H. E. (2013). Regional characterization of tornado activity. Journal of Applied Meteorology and Climatology, 52, 654–659.Find this resource:

Trapp, R. J., & Hoogewind, K. A. (2016). The realization of extreme tornadic storm events under future anthropogenic climate change. Journal of Climate, 29, 5251–5265.Find this resource:

Trapp, R. J. (2017). Formation and development of convective storms. Oxford research encyclopedia of climate science.Find this resource:

Trenberth, K. E., & Guillemot, C. J. (1998). Evaluation of the atmospheric moisture and hydrological cycle in the NCEP/NCAR reanalyses. Climate Dynamics, 14, 213–231.Find this resource:

Van Klooster, S. L., & Roebber, P. J. (2009). Surface-based convective potential in the contiguous United States in a business-as-usual future climate. Journal of Climate, 22, 3317–3330.Find this resource:

Verbout, S. M., Brooks, H. E., Leslie, L. M., & Schultz, D. M. (2006). Evolution of the US tornado database: 1954–2003. Weather and Forecasting, 21, 86–93.Find this resource:

Viceto, C., Marta-Almeida, M., & Rocha, A. (2017). Future climate change of stability indices for the Iberian Peninsula. International Journal of Climatology. Advance online publication.Find this resource:

Vines, F. (2001). Climatology of hail in France. Atmospheric Research, 56, 309–323.Find this resource:

Webb, J. D. C., Elsom, D. M., & Meaden, G. T. (2009). Severe hailstorms in Britain and Ireland, a climatological survey and hazard assessment. Atmospheric Research, 93, 587–606.Find this resource:

Weckwerth, T. M., & Parsons, D. B. (2006). A review of convection initiation and motivation for IHOP_2002. Monthly Weather Review, 134, 5–22.Find this resource:

Westermayer, A., Groenemeijer, P., Pistotnik, G., Sausen, R., & Faust, E. (2017). Identification of favorable environments for thunderstorms in reanalysis data. Meteorologische Zeitschrift, 26, 59–70.Find this resource:

Willett, K. M., Phillip, P. D., Thorne, P. W., & Gillett, N. P. (2010). A comparison of large scale changes in surface humidity over land in observations and CMIP3 general circulation models. Environmental Research Letters, 5/025210.Find this resource:

Xie, B., Zhang, Q., & Wang, Y. (2008). Trends in hail in China during 1960–2005. Geophysical Research Letters, 35, L13801.Find this resource:

Xie, B., Zhang, Q., & Wang, Y. (2010). Observed characteristics of hail size in four regions in China during 1980–2005. Journal of Climate, 23, 4973–4982.Find this resource:

Yao, Y., Yu, X., Zhang, Y., Zhou, Z., Xie, W., Lu, Y., . . . Wei, L. (2015). Climate analysis of tornadoes in China. Journal of Meteorological Research, 29, 359–369.Find this resource:

Ye, B., Del Genio, A. D., & Lo, K. K.-W. (1998). CAPE variations in the current climate and in a climate change. Journal of Climate, 11, 1997–2015.Find this resource:

Zhang, C., Zhang, Q., & Wang, Y. (2008). Climatology of hail in China: 1961–2005. Journal of Applied Meteorology and Climatology, 47, 795–804.Find this resource:

Zipser, E. J., Cecil, D. J., Liu, C., Nesbitt, S. W., & Yorty, D. P. (2006). Where are the most intense thunderstorms on Earth? Bulletin of the American Meteorological Society, 87, 1057–1071.Find this resource:

Zixiu, X., Pengyun, W., & Xuefang, L. (1993). Tornadoes of China. In C. Church, D. Burgess, C. Doswell, & R. Davies-Jone (Eds.), The tornado: Its structure, dynamics, prediction, and hazards (pp. 435–444). Washington, DC: American Geophysical Union.Find this resource: