SEARCH

SEARCH BY CITATION

Keywords:

  • hailstones;
  • climate models;
  • climate change;
  • UK

ABSTRACT

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Description of models
  5. 3 Hail climatology
  6. 4 Evaluation of hail model
  7. 5 Climate change impacts on hailstorms
  8. 6 Summary
  9. Acknowledgements
  10. References

Hailstorms can pose a significant threat to society, by damaging property and disrupting livelihoods. An understanding of how hailstorm characteristics may change under a warming climate is therefore important for assessing the risk of hail damage for the insurance industry. A simple model of hailstone formation has been driven using meteorological data produced by a regional climate model (RCM) to project how hailstorm numbers and hailstone sizes could change during the 21st century in the UK. Evaluation of the modelled hailstone sizes, numbers and spatial distributions showed that they agreed reasonably well with observations. The effect of climate change on the numbers of damaging hailstorms in the UK (hailstones with diameters greater than 15 mm) was then investigated. A downward trend in the total number of damaging hailstorms during the 21st century was projected, with statistically significant trends for hailstones with diameters between 21 and 50 mm. Melting of hailstones made little contribution to the projected reductions. The results are subject to large uncertainties, some of which originate with the convective parameterization scheme used by the climate model.


1 Introduction

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Description of models
  5. 3 Hail climatology
  6. 4 Evaluation of hail model
  7. 5 Climate change impacts on hailstorms
  8. 6 Summary
  9. Acknowledgements
  10. References

Hail is a solid form of precipitation associated with convective storms. Hailstones are formed in convective clouds by the freezing of supercooled water droplets onto ice particles by riming. Ice particles formed by riming are called hailstones if they have diameters larger than 5 mm (Pruppacher and Klett, 1997). Hail accounts for only a small amount of the total precipitation from the cloud, and so is almost always accompanied by rain (Morgan and Summers, 1986; Brimelow and Reuter, 2009). In terms of mass, hail production rarely exceeds 10% of the rainfall and is typically just a few percent (Morgan and Summers, 1986).

Hail is a highly localized phenomenon which is not normally recorded by automatic weather stations. Even if it were, the spatial distribution of weather stations in the UK is too coarse for accurate descriptions of hail events. In a few areas of Europe (e.g. parts of France, northern Italy and Spain), where the risk of damaging hail is high, networks of hailpads are maintained which can give a good indication of the distribution of hailstone sizes and the kinetic energy of the hailstones striking the ground (Berthet et al., 2011). Currently, numerical weather prediction (NWP) models cannot simulate hailstone sizes and numbers reliably.

Hailstorms pose a significant threat to modern society. In many parts of the world (e.g. Bangladesh, India, southern Africa, United States, Australia, southern Germany and Switzerland), large hailstones (diameters greater than 25 mm) are created in severe convective storms and are a cause of significant losses of crops and damage to cars and buildings (Cecil and Blankenship, 2012). Storms producing large damaging hailstones are rare in the UK, and almost always occur during the summer months (Webb et al., 2009). One of the most severe hailstorms recorded in the British Isles occurred in the southwest of England in July 1808 during an intense heat wave (Clark, 2004). Hailstones with diameters up to 70 mm were recorded, together with a few even larger stones of 100 mm diameter (Webb et al., 2009). Many buildings were badly damaged by the hail, and large areas of crops were destroyed. More recently, hailstones with diameters greater than 40 mm fell over parts of Leicestershire on 28 June 2012 causing significant damage to cars and property (Clark and Webb, 2013).

However, hailstones do not need to be large to cause damage and disruption. During June 1982, two localized but intense hailstorms produced pea-sized hail (5–10 mm diameter) which resulted in parts of Bristol and Ludlow, Shropshire, being buried in ice up to 10 cm deep (Meaden, 1982). In October 2008, a similar type of hailstorm struck the town of Ottery St Mary in east Devon (Grahame et al., 2009; Clark, 2011). A combination of mesoscale dynamics and local topographical effects produced a severe storm which was accompanied by unusually intense hail and rainfall. After 2 h, the hail on the ground was 10–20 cm deep, burying many parts of Ottery St Mary in thick ice and exacerbating the flooding caused by the intense rainfall which accompanied this storm (Clark, 2011).

Several studies have used long time series of surface-based hail observations to assess any trends in hail frequency, although the sparseness of the observations and data inhomogeneities mean changes in hail occurrence are difficult to quantify (Seneviratne et al., 2012). Webb et al. (2009) did not identify a trend in storms producing hailstones with diameters of 40 mm or greater between 1800 and 1999 over the UK, or in the number of damaging hailstorms (hailstones greater than 15 mm diameter) between 1930 and 2000. Berthet et al. (2011) examined data from a network of hail pads in three parts of France over a 22-year period, and found no change in the frequency of hailstorms, although the numbers of large hailstones had increased in April and May. Xie et al. (2008) did not detect a trend in annual hailstorm numbers in China between 1960 and the early 1980s, but a significant downward trend was found after this period. This trend was attributed to increases in the freezing level height, such that water droplets in clouds were less likely to freeze and form hail. Cao (2008) detected an upward trend in large hail over Ontario, Canada, between 1979 and 2002 although the numbers of events in individual years were highly variable. Mezher et al. (2012) found an increase in annual numbers of hail events in the north and south of Argentina, but a mixture of no change and downward trends in the centre. They also suggested that the downward trends in hail frequency were caused by increases in the freezing level height.

Other studies have used insurance loss data as a proxy for days with damaging hail. Kunz et al. (2009) found that the proportion of severe storms accompanied by hail appeared to have increased between 1974 and 2003 over a region of southwest Germany, although there was no indication of a general increase in the number of convective storms. A later study by Kapsch et al. (2012) for the same region identified weather types associated with severe hail using insurance loss data, and used model simulations to examine how these weather types could change during the first half of the 21st century. The model simulations projected a small increase in the numbers of weather types associated with hail, leading Kapsch et al. (2012) to suggest that hailstorms would increase between 7 and 15% for the period 2031–2045 compared to 1971–2000.

There have been very few climate model-based studies explicitly simulating hail and how its characteristics could change in the future, and there is often a lack of agreement between these studies (Seneviratne et al., 2012). Leslie et al. (2008) used a series of nested climate models and a cloud-ice model to simulate hailstone formation over part of southeast Australia between 2000 and 2050. Their simulations suggested a gradual increase in the frequency of hailstorms and the largest hailstone sizes. A different conclusion was reached by Niall and Walsh (2005). They established an empirical relationship between convectively available potential energy (CAPE) and hail occurrence in southeast Australia. Simulated CAPE values from a climate model with double the pre-industrial levels of CO2 were generally smaller than present-day values leading Niall and Walsh (2005) to propose that fewer hailstorms might occur over southeast Australia in the future.

An important question for the UK is whether the frequency of hailstorms and the size of the hailstones will increase or decrease under a warming climate. Currently, it is not clear what the impact of climate change will be on hailstorms in the UK. The analysis of Webb et al. (2009) indicates no trend in the numbers of storms producing damaging hailstones during the 19th and 20th centuries. However, the variability in observed numbers of hail events and hail stone sizes is very high.

The objective of this study is to evaluate a simple model of hail production and then project how hailstorm numbers over the UK could change during the 21st century. Only hailstorms over land are considered. In Section 'Description of models', the RCM and hail model are described and the observations used to evaluate the hail model are presented in Section 'Hail climatology'. The hail model is evaluated in Section 'Evaluation of hail model', and the projected changes in hailstorm numbers are discussed in Section 'Climate change impacts on hailstorms'. A summary of the results and a discussion of the uncertainties are given in Section 'Summary'.

2 Description of models

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Description of models
  5. 3 Hail climatology
  6. 4 Evaluation of hail model
  7. 5 Climate change impacts on hailstorms
  8. 6 Summary
  9. Acknowledgements
  10. References

2.1 Regional climate model simulations

RCMs are widely used to downscale meteorological fields from coarse global models and to provide projections of local climate change. They have higher resolutions than general circulation models (GCMs) and so give a better representation of physical and dynamical processes (Kendon et al., 2012). Ideally, an ensemble of RCMs would have been used to drive the hail model to enable uncertainties in the projections of hail to be assessed. However, the meteorological data used by the hail model (vertical profiles of temperature and humidity, wet bulb freezing level, CAPE and convective rainfall) are generally not available on the subdaily time scales required.

In this study, meteorological data from two simulations using the Hadley Centre regional climate model HadRM3 have been used (Burke and Brown, 2010). HadRM3 has a horizontal resolution of 25 km and 19 vertical levels which represent the atmosphere up to 10 hPa (approximately 40 km). The regional model was used to simulate the climate of Europe in both cases. First, the HadRM3 model was forced at the boundaries of the domain using the ERA-40 reanalysis (Uppala et al., 2005). Secondly, the regional model was forced using data from the global climate model HadCM3 for the period 1949–2099; this RCM simulation was one of those used to generate the UK Climate Projections 2009 (UKCP09; Murphy et al., 2009). Up to the year 2000, the RCM was driven using a historical series of greenhouse gas and ozone concentrations, sulphur emissions and reconstructed variations in solar output and volcanic aerosol levels. After 2000, the regional model was driven by greenhouse gas concentrations and sulphur emissions specified in the A1B scenario from the Special Report on Emissions Scenarios (SRES; Nakićenović and Swart, 2000). The A1B scenario assumes that future world energy production will be generated from an equal mix of fossil fuels and renewable resources, and is often referred to as a ‘medium’ emissions scenario. These two simulations will be referred to as RCM-ERA and RCM-GCM. Meteorological data required to drive the hail model were archived as 6-h mean values for an area enclosing the British Isles from both simulations.

Before running the hail model, the RCM data were subselected in several ways. First, the potential for hail production was only considered when convective rainfall was greater than 2 mm day−1, as hailstorms are always accompanied by rain (Morgan and Summers, 1986). This check excludes any small convective clouds that produce low amounts of precipitation and would not generate hail. Secondly, the hail model is only called if the convective cloud base is less than 5000 m above ground level and the cloud depth is greater than 3500 m (Hand and Cappelluti, 2011). These conditions mean that a good proportion of the cloud layer has temperatures below freezing (0 °C), and provides an increased likelihood of strong updrafts and high liquid water content. Finally, the convective available potential energy (CAPE), as calculated within the RCM, had to exceed a threshold of 500 J kg−1 before the hail model was called. This threshold is arbitrary but forms part of current best forecasting practice at the Met Office to identify severe convective storms that could produce hail in the weather forecast models. The calculation of the CAPE diagnostic within the RCM is based on a ‘dilute’ formulation which includes a small amount of mixing between a rising air parcel and the surrounding air (Zhang, 2009).

A comparison of the spatial distribution of modelled hailstorms, the numbers of storms and distribution of hailstone sizes from these two simulations will indicate whether any errors in the climatology of the global climate model have a significant effect on the hail simulations. This comparison is important, since any errors in the large-scale atmospheric flow in the GCM will also be present in the RCM simulation, and the GCM-driven regional model data would be used to examine any future changes in hailstorms.

2.2 Hail model

The potential formation of hail in convective clouds is assessed using an empirical model developed by Fawbush and Miller (1953) which estimates the maximum hailstone diameter that could form under a given set of conditions. Data from 274 vertical soundings of temperature and humidity and surface-based observations of hail made in the United States between 1950 and 1953 were used to create the model. The hailstone diameters ranged from 3 to 100 mm, with 82% less than or equal to 25 mm, so that the hailstone data used were biased toward smaller sizes. Nevertheless, testing of the hail model by Fawbush and Miller (1953) using an independent sample of 219 observations showed that hailstone sizes were correctly forecast 77% of the time, with a tendency to over predict hailstone sizes.

From the vertical soundings, the convective condensational level (CCL), the level of the base of the cloud, is estimated and then two temperature differences are calculated. An air parcel is raised from the CCL along a saturated adiabat until it reaches the pressure of the −5 °C isotherm. The −5 °C level is within the region of a cloud where hail is formed (Fawbush and Miller, 1953). The temperature difference between the air parcel and its environment is a measure of the instability, CAPE and vertical velocity of the parcel. The greater the temperature difference, the larger the hailstones that could be produced.

A measure of the depth of cloud below the −5 °C level is calculated by descending along a dry adiabat from the −5 °C isotherm to the same pressure as the CCL and taking the temperature difference between these two points. A relatively deep cloud layer between the CCL and the −5 °C isotherm will provide ideal conditions for hail growth, provided the cloud top is sufficiently cold for ice to form, and the deeper the layer, the more chance the hailstones within the cloud have to grow to bigger sizes (Hand and Cappelluti, 2011). Fawbush and Miller (1953) produced a nomogram so that the maximum hailstone sizes could be estimated as a continuous function of the two temperature proxies. This nomogram has been discretized by Hand and Cappelluti (2011) so that the maximum hailstone diameter is found from a lookup table which is displayed graphically in Figure 1. Large values of both proxy temperatures are needed for very large hailstones to form. The smallest hailstone diameter calculated is 10 mm, and the diameters of larger hailstones are only calculated to the nearest 5 mm. Graupel and soft hail (diameters of 2 and 5 mm) are also simulated.

image

Figure 1. Maximum hailstone diameters (in mm) as a function of the cloud depth and CAPE (instability) proxy temperatures (in °C) from the Fawbush–Miller hail model. The maximum hailstone diameter that could be produced is 120 mm. Only hailstone diameters of 10 mm or larger are shown. This figure was derived from ‘Table 1’ of Hand and Cappelluti (2011).

Download figure to PowerPoint

Melting of hailstones is parameterized using the height of the wet bulb freezing level and the initial hailstone diameter using results from Miller (1972) as implemented by Hand and Cappelluti (2011). If the height of the wet bulb freezing level is less than 3350 m it is assumed that no melting occurs. The hailstones are assumed to melt completely (regardless of their size) before they reach the surface if the wet bulb freezing level exceeds 4400 m (Fawbush and Miller, 1953). In reality, the rate of melting will vary as the hailstone diameter decreases and would also depend on the environment through which the hailstone is falling (Fraile et al., 2003).

The Fawbush–Miller hail model has been adapted to use meteorological data produced by weather forecast models (Hand and Cappelluti, 2011) and has been used to provide guidance for forecasters at the Met Office for the past 10 years. The hail model was driven using 5 years of archived weather forecast model data to create a global hail climatology (Hand and Cappelluti, 2011). The modelled climatology compared well with regional climatologies derived from observations. Despite the simplicity of the Fawbush–Miller hail model, the results of Hand and Cappelluti (2011) show that it is applicable outside of the United States.

The Fawbush–Miller hail model performs best when used in conjunction with other calculations that determine whether convective instability can be released and whether the convective environment would support the generation of hail. The hail model is used to estimate maximum hailstone sizes, not whether hail could form or not. This point is discussed further in Section 'Summary'. Additionally, the Fawbush–Miller hail model does not give any indication of the intensity of the hail fall and so cannot distinguish a light shower of hail from an intense event like the one which occurred in Ottery St Mary in October 2008.

3 Hail climatology

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Description of models
  5. 3 Hail climatology
  6. 4 Evaluation of hail model
  7. 5 Climate change impacts on hailstorms
  8. 6 Summary
  9. Acknowledgements
  10. References

Most hail climatologies have been constructed using reports by staff at weather observation sites, meteorological journals, newspapers, private diaries and members of the public. The earliest known hail climatology which included the UK appeared in an unpublished report by Williams (1973). A map showing the average number of days with hail over land from this report has been reproduced by Hand and Cappelluti (2011). This climatology was constructed using all reports of hail regardless of size and suggested hail would occur, on average, just 1 day per year at any given location in Britain, but was based on limited data.

In this study, an observation-based climatology of hailstones for the UK (Webb et al., 2009) has been used to evaluate the hail model. This climatology was constructed using a wide range of sources published between 1800 and 2001, and observations of damaging hail (hailstones with diameters greater than 15 mm) have been mapped to the 25 km grid used by the RCM (Figure 2(a)). The largest numbers of storms are found in the southeast of the UK, and the fewest in Scotland and Ireland. There is a bias towards areas of high population, with a relatively high frequency of hailstorms around Manchester, Birmingham and London, and a low incidence in Wales (Figure 2(a)). Some severe hailstorms may not have been recorded; e.g. those in sparsely populated areas. Nevertheless, the hail climatology of Webb et al. (2009) provides the best description of the characteristics of hailstorms in the UK. Webb et al. (2009) also analysed the monthly and diurnal cycles of damaging hailstones and their size distributions, all of which will be used to evaluate modelled hail.

image

Figure 2. Spatial distributions of damaging hailstorms on the 25 km grid used by the RCM. The data have been scaled so that they have units of storms per 1000 km2 per 100 years. (a) Number of storms from the hail climatology created by Webb et al. (2009). (b) Simulated number of storms when the hail model is driven by RCM data generated using boundary conditions from ERA-40. (c) as (b) but using boundary conditions from a global climate model simulation.

Download figure to PowerPoint

4 Evaluation of hail model

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Description of models
  5. 3 Hail climatology
  6. 4 Evaluation of hail model
  7. 5 Climate change impacts on hailstorms
  8. 6 Summary
  9. Acknowledgements
  10. References

Evaluation of hail from models is difficult. The numbers and sizes of modelled hailstones are dependent on the ability of a climate model to simulate convective storms of sufficient intensity to produce hail. One source of error originates with the convective parameterization scheme used by climate models. The resolution of most climate models means that they cannot resolve convection directly, but requires a parameterization scheme to calculate the average effects of convection on temperature, moisture and momentum within a model grid cell (Kendon et al., 2012). Many convective parameterizations have been developed to represent tropical convection in coarse-resolution GCMs where many of the underlying assumptions are most valid; they are less applicable in higher resolution RCMs and at higher latitudes. Even if the parameterization worked perfectly, it is not designed to represent individual storms (Kendon et al., 2012). Hence, the number of convective storms and their intensities could be over- or underestimated, which will also affect the number of hail events and hailstone sizes simulated. Additionally, when comparing modelled and observed hail climatologies, the hail events may not have occurred at the same time or location. The reliability of the modelled hail is assessed in Section 'Reliability of modelled hailstorms'

There is a lack of observations of hailstones in some areas and inhomogeneities in reporting. The biggest hailstones reported may not be the largest produced by a storm. Consequently, the only aspect of the hail model (at least for the UK) which can be evaluated is the number of days per year when hailstones within a given size range or above a specified diameter have fallen to the ground and the climatological distribution of hailstone sizes.

The analysis of UK hailstorms by Webb et al. (2009) mainly focused on hailstones whose diameters were greater than 15 mm, and a similar approach will be used here. The hailstone size ranges which will be considered are shown in Table 1. In Section 'Spatial distribution of hailstorms' the spatial distribution of hailstones simulated over the period 1971–2000 using data from the RCM-ERA and RCM-GCM simulations will be compared with the observations. A comparison of the two modelled hail distributions will show whether any errors in the climatology of the global climate model have a significant effect on the hail simulations. Modelled hail is verified in Section 'Reliability of modelled hailstorms' and the modelled and observed climatologies are compared in Section 'Comparison of modelled and observed hail climatologies'

Table 1. Hailstone diameter ranges used in this study.
Hailstone diameter range (mm)
  1. The range 15+ refers to all hailstones with sizes greater than 15 mm, which are classed as damaging.

15+10–1516–2021–30
31–4041–5051–6061–75

4.1 Spatial distribution of hailstorms

The simulated spatial distribution of hailstorms where the hailstone diameters are greater than 15 mm using the RCM-ERA data is shown in Figure 2(b). It is in reasonable agreement with the hail climatology (Figure 2(a)). There are two areas with maximum hailstorm numbers located over south central Britain and the Midlands in the simulated distribution, but their positions are offset compared to the maxima in the observations. The use of a convective parameterization scheme means climate models cannot simulate the large scale organization of convection, and the propagation of convective clouds and mesoscale convective systems through time, constituting a possible cause for the offset. The RCM cannot resolve small orographic features which can initiate convection and hailstorms (Kunz et al., 2009). The offset may be partly caused by biases in the reporting of large hailstones towards highly populated areas.

The simulated numbers of hailstorms are a factor of 5–6 larger than those in the climatology of Webb et al. (2009). The relatively low vertical resolution of the RCM means that too many deep convective clouds (and hence storms producing hail) may be simulated. Innes et al. (2001) used two versions of the atmospheric component of the GCM HadCM3 (on which the HadRM3 model is based) to study tropical convection when the number of vertical levels was increased from 19 to 30. When the vertical resolution was higher, the distribution of cloud types changed from a bimodal distribution (with shallow and deep convective clouds) to a trimodal one, with an extra cloud type which extended up to the mid-troposphere and fewer numbers of deep convective clouds.

The simulated hailstorm distribution using the RCM-GCM data is shown in Figure 2(c). There are once again two maxima: one is located over southeast England, but is located slightly too far to the southeast. A second maximum in large hailstone numbers is seen on the south coast of Wales. This simulation does not reproduce the second maximum in the Midlands seen in the observations, and the small maximum over Wales is not seen in the observations. However, only a small number of hailstorms are simulated over northern England and Wales, and very few over Scotland and Ireland, in agreement with the observations. Overall, the spatial distribution of large hail in the RCM-GCM simulation does not agree very well with the observations, although it does simulate a maximum in hailstorm numbers over southeast England, but the RCM-ERA results are in closer agreement.

4.2 Reliability of modelled hailstorms

Individual modelled hail events using RCM-ERA data were verified with a 2 × 2 contingency table (Wilks, 2011). Observations of hailstones made between 1960 and 2000 in Britain were used (Webb et al., 2009). Britain was treated as a single area, so that if large hail in the model and observations occurred on the same day that result would be regarded as a success. From this verification, the bias is 3.9, indicating about four times as many hail events are simulated as are observed. The false alarm ratio is 0.84, showing that a fairly large fraction of the simulated hail events did not occur. However, the hit rate is 0.63 and the probability of false detection is 0.08, indicating that more than half of the observed hail events were simulated and large hail was simulated on only a small fraction of days with no large hail. The Heidke skill score and true skill statistic have values of 0.23 and 0.55, respectively, indicating the hail model does have some skill.

4.3 Comparison of modelled and observed hail climatologies

The modelled hailstone sizes and distributions were further evaluated by comparison with other data presented by Webb et al. (2009) using the hailstone diameter ranges listed in Table 1, as shown in Figure 3. The simulated percentages of hailstorms occurring in each month are shown in Figure 3(a) and (b) for hailstones with diameters greater than 15 and 40 mm, respectively, together with the observed percentages from Webb et al. (2009). The modelled seasonal cycle of large hail (> 15 mm diameter) is slightly over amplified compared with the observations, with too many storms during June, July and August and none at all in the winter months.

image

Figure 3. Comparison of modelled and observed hail distributions. In all four panels, the observed hail distributions from Webb et al. (2009) are compared with modelled hail using RCM data which was forced with ERA-40 and GCM boundary conditions. Panels (a) and (b) show the percentages of hailstorms in each month when the hailstone diameters are greater than 15 and 40 mm, respectively. Panel (c) illustrates the diurnal cycle of hailstorm numbers and shows the percentages of storms in each 6-h period (times are UTC). Panel (d) shows the percentages of hailstorms with hailstone diameters within the given size ranges (in mm). The hail model did not simulate any storms with hailstone diameters larger than 75 mm.

Download figure to PowerPoint

Interestingly, hailstones modelled using RCM-GCM data appear to be in closer agreement with the observations than when the RCM-ERA data are used, indicating some robustness in the RCM-GCM results and confidence in the future projections. To investigate this result, the number of times CAPE exceeded the threshold of 500 J kg−1 in each month in the RCM-ERA and RCM-GCM simulations was compared with the number of exceedances in the ERA-Interim reanalysis (Dee et al., 2011) over the part of central England. Both RCM simulations underestimate the number of exceedances by a factor of 3. The maximum number of exceedances occurs in June in the RCM-ERA simulation, but during July in the RCM-GCM and ERA-Interim data. CAPE in the RCM-ERA simulation almost never exceeds the threshold during September and not at all during winter, whereas a small number of exceedances occur in the RCM-GCM and ERA-Interim data. This result partly explains why hail simulated using the RCM-ERA data has a large maximum in June and little hail in September (Figure 3(a) and (b)).

The diurnal cycle of hailstorm numbers which produce hailstones with diameters greater than 30 mm is shown in Figure 3(c). In this case, the diurnal cycles of modelled and observed hail agree reasonably well, and the diurnal cycle from the RCM-ERA data is in closer agreement with the observations than the cycle from the RCM-GCM data. Modelled numbers of large hailstones are overestimates during 6–12 h and underestimates in the final part of the day (18:00–00:00 h). Convection in many climate models reaches maximum strength too early in the day (Stratton and Stirling, 2012), which is reflected in the diurnal distribution of large hailstone numbers, particularly in the RCM-GCM data.

The percentage of hailstones within specific diameter ranges from the hail model and the observations is compared in Figure 3(d). Neither version of the hail model reproduces the observed distribution exactly. The maximum number of large hailstones appears in the 21–30 mm size range in the observations, which is likely due to under-reporting of the occurrence of smaller hailstones. In contrast, the simulated hail size distributions have a large maximum in the 10–15 mm size, and a secondary maximum in the 31–40 mm size range (RCM-ERA) and the 21–30 mm size range (RCM-GCM). Hail stones with diameters larger than 75 mm were not simulated at any time.

5 Climate change impacts on hailstorms

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Description of models
  5. 3 Hail climatology
  6. 4 Evaluation of hail model
  7. 5 Climate change impacts on hailstorms
  8. 6 Summary
  9. Acknowledgements
  10. References

The effect of a warming climate on the numbers of storms producing hail was then investigated using the RCM-GCM simulation. The spatial distribution of damaging hailstorms was similar to that shown in Figure 2(c) for three time periods 2010–2039, 2040–2069 and 2070–2099, although the number of storms did vary. The maximum in hailstorm numbers remained over southeast England, and few storms were simulated over Scotland and Ireland. For the period 2070–2099, the two maxima in storm numbers were largely absent.

The numbers of storms per year producing hailstones in each size range listed in Table 1 are shown in Figure 4. Soft hail and graupel (sizes 2–5 mm) are also included. The numbers of storms are highly variable, and the largest proportion of hailstones has diameters of 10–15 mm. Overall, there appears to be a downward trend in the numbers of hailstones in most categories. The data for the size range 61–75 mm are not shown as very few of these events were simulated.

image

Figure 4. Simulated numbers of storms per year producing hailstones with diameters in the given size range (mm) between 1971 and 2099. The bottom right panel shows the numbers of damaging hailstones, which have diameters greater than 15 mm. The straight lines represent linear least squares fit to the data; trends which are significant at the 5% level are shown by solid lines, and dashed lines indicate trends which are not significant.

Download figure to PowerPoint

A straight line was fitted using a least squares method to the numbers of hailstorms and these straight lines are shown in Figure 4 for all size categories. Overall, there is a downward trend in all hail sizes, except for the 2–5 mm category, where an upward trend was found. These trends were tested for significance using the nonparametric Kendall τ test (Kendall and Stuart, 1983) for a range of time periods. The trends were significant at the 5% level for size ranges 21–30, 31–40 and 41–50 mm between 1971 and 2099 and 2010 and 2099. The downward trends in the size ranges 10–15, 16–20 and 51–60 mm were very small and not significant at the 5% level. Very few storms producing the hailstones in the largest size category (61–75 mm diameter) were simulated, and an assessment of any trend in their numbers could not be made. The model projects an increase in the number of storms producing soft hail and graupel (with diameters of 2 or 5 mm), but this trend is not significant at the 5% level.

Overall, there is a downward trend in the numbers of storms producing damaging hailstones during the 21st century. An analysis of CAPE diagnosed directly from the RCM-GCM simulation for central England showed that the number of days when the threshold of 500 J kg−1 was exceeded increased slightly between 1980 and 2099. Despite this fact, the number of damaging hailstorms is projected to decrease. One reason could be as the tropopause rises under a warming climate, the depth of convective clouds increases, resulting in higher CAPE values but weaker updrafts. This change would be reflected in a lower value of the CAPE proxy used by the hail model, as this proxy is calculated using conditions between the convective cloud base and the −5 °C isotherm. The freezing level height was also found to increase during the simulation, meaning that temperatures within the lower parts of convective clouds could be warmer. More vigorous convection would be needed to raise air parcels to altitudes where water droplets can freeze and form hailstones.

The distributions of values of both proxies in four 30-year periods (1971–2000, 2010–2039, 2040–2069 and 2070–2099) were investigated, and are shown in Figure 5 where the percentages of hailstorms within each proxy temperature range are displayed.

image

Figure 5. Distribution of CAPE proxy temperatures (upper panel) and cloud depth proxy temperatures (lower panel) for the four periods 1971–2000, 2010–2039, 2040–2069 and 2070–2099. The CAPE proxy temperatures are calculated to the nearest 0.5 °C, but only the whole numbers are shown on the x-axis.

Download figure to PowerPoint

For the periods 1971–2000 and 2010–2039, the distributions of the two proxies are very similar, although the number of storms has increased during the latter period. During the period 2040–2069, there is an increase in the percentages of storms with smaller CAPE proxy temperatures (between 3.5 and 5 °C), and a decrease in storms with larger proxy temperatures (between 8 and 11 °C). This reduction in the CAPE proxy suggests updrafts within the part of the cloud where hail is formed are weaker, hence the shift in hailstone diameters towards smaller values. However, this result does not mean that the overall convective activity is weakening or reducing in frequency

During 2070–2099, there is a larger shift in both proxy temperatures towards smaller values. For the CAPE proxy temperatures, the maximum number of storms still occurs at temperatures of 4 and 7.5 °C, but there are more storms with proxy temperatures between 3 and 5 °C, and fewer storms with proxy temperatures of 8 °C or more. A larger proportion of storms have a proxy cloud depth temperature of 20 °C, and fewer storms occur at 35 °C. The simulated reduction in hailstone sizes during the 21st century is therefore mostly caused by smaller values of the CAPE proxy temperature, with the reduction in the cloud depth proxy only becoming important during 2070–2099. The shift in CAPE proxy temperatures to smaller values by the end of the 21st century suggests less vigorous convection which causes a reduction in hailstone sizes. This result does not mean that convective activity overall is declining, rather the conditions suitable for large hail formation do not occur as frequently. The depth of cloud beneath the −5 °C isotherm is also becoming smaller, presumably due to higher and colder convective cloud bases as a consequence of a drier boundary layer. This result does not necessarily mean that the overall cloud depth is decreasing, but the portion of the cloud where temperatures are optimum for water droplets to freeze and form hail is decreasing in depth.

It is possible that increased melting of the hailstones could also contribute to the simulated reduction in hailstone sizes. The impact of melting was investigated by repeating the hail model simulations and removing the melting calculation. There was a small increase in the total number of hailstorms when the melting term was removed. A slight reduction in the number of storms with hailstones between 10 and 15 mm, and a small increase in storms with hailstone diameters greater than 30 mm were simulated. However, these changes were small and were not significant. Overall, the parameterized melting had only a small effect on the projected future reduction in hailstone sizes.

6 Summary

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Description of models
  5. 3 Hail climatology
  6. 4 Evaluation of hail model
  7. 5 Climate change impacts on hailstorms
  8. 6 Summary
  9. Acknowledgements
  10. References

This study has used a hail model driven by meteorological data from a RCM to project how hailstone numbers and sizes over the UK could change during the 21st century. The simulated spatial pattern of damaging hailstorms over the UK (where the hailstones have diameters greater than 15 mm) agreed reasonably well with the observations, although the number of hailstorms simulated is larger by a factor of 5–6. Verification of modelled hailstorms using a 2 × 2 contingency table showed that the model does have some skill. The distributions of hailstone sizes were found to agree satisfactorily with those derived from surface observations. Overall, results using the RCM-GCM simulation were in closer agreement with the observations than the RCM-ERA simulation, indicating some robustness in the results and giving some confidence in the projections of future large hailstone numbers.

The effect of climate change on hailstorms in the UK during the 21st century was then investigated. The spatial distribution of hailstorms did not change – the maximum remained over southeast England, and few storms were simulated over Scotland and Ireland. The numbers of damaging hailstorms (where the hailstone diameters were greater than 15 mm) were projected to decrease by a factor of 2 during the 21st century, and this trend was significant at the 5% level. Downward trends in storms producing hailstones with diameters between 21 and 50 mm were significant at the 5% level for the periods 1971–2099 and 2010–2099. Very few trends in other size ranges and time periods were significant at the 5% level. A small increase in the numbers of small hail and graupel (2 and 5 mm diameters) was projected, but the trend was also not significant at the 5% level. This reduction in numbers of damaging hailstones was caused mostly by smaller values of the CAPE proxy temperatures in the future climate. Increases in the wet bulb freezing level may also have contributed to a reduction in the number of modelled hailstones. An increase in the number of hailstorms with small values of the cloud depth proxy was simulated at the end of the 21st century, which would also result in smaller numbers of damaging hailstones. The parameterized melting of hailstones was found to have a small effect on the simulated reduction in hailstone numbers.

The production of hailstorms has been simulated using a simple model which estimates the largest hailstone size for the given conditions. Only the simulated number of days when hailstorms producing maximum hailstone sizes within given size ranges has been evaluated. In reality, vertical velocities within convective clouds will have a range of values so that a distribution of hailstone sizes would be supported. Ideally, a very high resolution climate model which can treat convection explicitly would be used with a sophisticated microphysical parameterization to simulate hailstone development and evolution and verified against observations.

There is considerable uncertainty in the results presented here. The Fawbush–Miller hail model is a simple one and was developed based on 274 upper air soundings made in the United States for which hailstones of known sizes were formed. In particular, the role of the cloud depth proxy in controlling hailstone sizes has not been confirmed by other studies. However, a simple linear regression analysis showed that large hail formation was more closely correlated with the CAPE proxy than the cloud depth proxy. The distribution of hailstone sizes in individual years differs considerably in both the observations and simulations, so that only the long-term climatological properties of hailstone sizes and distributions have been evaluated.

The low vertical resolution of the HadRM3 model means that it simulates large numbers of deep convective storms which can produce large hail. Experiments with a similar model with higher vertical resolution simulated fewer deep convective clouds (Innes et al., 2001). The convective parameterization scheme used in the model means only the average effects of convection can be simulated at a given location, and the propagation and evolution of convective systems in time cannot be simulated (Kendon et al., 2012).

Only a single model simulation of future climate has been used in this study. Ideally, an ensemble of RCM simulations would have been used. Other models could project very different changes in convective activity under a warming climate and hence hailstone sizes and numbers. Simulations using different emissions scenarios, with higher or lower levels of carbon dioxide and other greenhouse gases, may also yield different trends in projected hailstorm numbers during the 21st century.

Cloud feedbacks in response to a warming climate are still highly uncertain and differ considerably between models (Dufresne and Bony, 2008). A study comparing the soil moisture–precipitation feedback via moist convection over the Alpine region using a model with parameterized convection (25 km resolution) and at cloud-resolving resolution (2.2 km) showed that the feedback changed sign between these resolutions (Hohenegger et al., 2009). These authors also showed that the use of an alternative convective parameterization scheme in the 25 km model changed the sign of the feedback. These results suggest that the sign of projected changes in numbers of large hailstones could change if either a different convective parameterization scheme was employed or a much higher resolution climate model was used.

Acknowledgements

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Description of models
  5. 3 Hail climatology
  6. 4 Evaluation of hail model
  7. 5 Climate change impacts on hailstorms
  8. 6 Summary
  9. Acknowledgements
  10. References

Financial support for this work was provided by the AXA Research Fund. Additional support was provided by the Joint DECC/Defra Met Office Hadley Centre Climate Programme (GA01101).

References

  1. Top of page
  2. ABSTRACT
  3. 1 Introduction
  4. 2 Description of models
  5. 3 Hail climatology
  6. 4 Evaluation of hail model
  7. 5 Climate change impacts on hailstorms
  8. 6 Summary
  9. Acknowledgements
  10. References
  • Berthet C, Dessens J, Sanchez JL. 2011. Regional and yearly variations of hail frequency and intensity in France. Atmos. Res. 100: 391400.
  • Brimelow JC, Reuter GW. 2009. Explicit forecasts of hail occurrence and expected hail size using the GEM–HAILCAST system with a rainfall filter. Weather Forecast. 24: 935945, DOI: 10.1175/2009WAF2222138.1.
  • Burke EJ, Brown SJ. 2010. Regional drought over the UK and changes for the future. J. Hydrol. 394: 471485.
  • Cao Z. 2008. Severe hail frequency over Ontario, Canada: recent trend and variability. Geophys. Res. Lett. 35: L14803, DOI: 10.1029/2008GL034888.
  • Cecil DJ, Blankenship CB. 2012. Toward a global climatology of severe hailstorms as estimated by satellite passive microwave imagers. J. Clim. 25: 687703.
  • Clark C. 2004. The heatwave over England and the great hailstorm in Somerset, July 1808. Weather 59: 172176.
  • Clark MR. 2011. An observational study of the exceptional ‘Ottery St Mary’ thunderstorm of 30 October 2008. Meteorol. Appl. 18: 137154.
  • Clark MR, Webb JDC. 2013. A severe hailstorm across the English Midlands on 28 June 2012. Weather 68: 284291.
  • Dee DP, Uppala SM, Simmons AJ, Berrisford P, Poli P, Kobayashi S, Andrae U, Balmaseda MA, Balsamo G, Bauer P, Bechtold P, Beljaars ACM, van de Berg L, Bidlot J, Bormann N, Delsol C, Dragani R, Fuentes M, Geer AJ, Haimberger L, Healy SB, Hersbach H, Hólm EV, Isaksen L, Kållberg P, Köhler M, Matricardi M, McNally AP, Monge-Sanz BM, Morcrette J-J, Park B-K, Peubey C, de Rosnay P, Tavolato C, Thépaut J-N, Vitart F. 2011. The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q. J. R. Meteorol. Soc. 137: 553597.
  • Dufresne J-L, Bony S. 2008. An assessment of the primary sources of spread of global warming estimates from coupled atmosphere–ocean models. J. Clim. 21: 51355144.
  • Fawbush EJ, Miller RC. 1953. A method for forecasting hailstone size at the Earth's surface. Bull. Am. Meteorol. Soc. 34: 235244.
  • Fraile R, Castro A, López L, Sánchez JL, Palencia C. 2003. The influence of melting on hailstone size distribution. Atmos. Res. 67–68: 203213.
  • Grahame N, Riddaway B, Eadie A, Hall B, McCallum E. 2009. Exceptional hailstorm hits Ottery St Mary on 30 October 2008. Weather 64: 255263.
  • Hand WH, Cappelluti G. 2011. A global hail climatology using the UK Met Office convection diagnosis procedure (CDP) and model analyses. Meteorol. Appl. 18: 446458.
  • Hohenegger C, Brockhaus P, Bretherton CS, Schär C. 2009. The soil moisture–precipitation feedback in simulations with explicit and parameterized convection. J. Clim. 22: 50035020.
  • Innes PM, Slingo JM, Woolnough SJ, Neale RB, Pope VD. 2001. Organization of tropical convection in a GCM with varying vertical resolution; implications for the simulation of the Madden-Julian Oscillation. Clim. Dyn. 17: 777793.
  • 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. J. Geophys. Res. 117: D15107, DOI: 10.1029/2011JD017185.
  • Kendall MG, Stuart A. 1983. The Advanced Theory of Statistics, Volume 2: Inference and Relationship, 4th edn. Charles Griffin: London, UK.
  • Kendon E, Roberts NM, Senior CA, Roberts MJ. 2012. Realism of rainfall in a very high resolution climate model. J. Clim. 25: 57915806.
  • Kunz M, Sander J, Kottmeier C. 2009. Recent trends of thunderstorm and hailstorm frequency and their relation to atmospheric characteristics in southwest Germany. Int. J. Climatol. 29: 22832297.
  • Leslie LM, Leplastrier M, Buckley BW. 2008. Estimating future trends in severe hailstorms over the Sydney Basin: a climate modelling study. Atmos. Res. 87: 3751.
  • Meaden GT. 1982. Hailstorms in north-east Bristol, 18 June1982, and Ludlow, 26 June 1982. J. Meteorol. 7: 224225.
  • Mezher RN, Doyle M, Barros V. 2012. Climatology of hail in Argentina. Atmos. Res. 114–115: 7082.
  • Miller RC. 1972. Notes on analysis and severe storm forecasting procedures of the Air Force Global Weather Center. AWS Technical Report 200 (Rev), Headquarters, Air Weather Service. Scott AFB, IL, 106.
  • Morgan GM, Summers PW. 1986. Hailfall and hailstorm characteristics. In Thunderstorm Morphology and Dynamics, 2nd edn, Kessler E (ed). University of Oklahoma Press: Norman, OK; 237257.
  • Murphy JM, Sexton DMH, Jenkins GJ, Boorman PM, Booth BBB, Brown CC, Clark RT, Collins M, Harris GR, Kendon EJ, Betts RA, Brown SJ, Howard TP, Humphrey KA, McCarthy MP, McDonald RE, Stephens A, Wallace C, Warren R, Wilby R, Wood RA. 2009. UK Climate Projections Science Report: Climate Change Projections. Met Office Hadley Centre: Exeter, UK.
  • Nakićenović N, Swart R. 2000. Special Report on Emissions Scenarios: A special report of Working Group III of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, UK.
  • Niall S, Walsh K. 2005. The impact of climate change on hail storms in southeast Australia. Int. J. Climatol. 25: 19331952.
  • Pruppacher HR, Klett JD. 1997. Microphysics of Clouds and Precipitation, 2nd edn. Kluwer Academic Publishers: Dordrecht, The Netherlands.
  • Seneviratne SI, Nicholls N, Easterling D, Goodess CM, Kanae S, Kossin J, Luo Y, Marengo J, McInnes K, Rahimi M, Reichstein M, Sorteberg A, Vera C, Zhang X. 2012. Changes in climate extremes and their impacts on the natural physical environment. In: Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, Field CB, Barros V, Stocker TF, Qin D, Dokken DJ, Ebi KL, Mastrandrea MD, Mach KJ, Plattner G-K, Allen SK, Tignor M, Midgley PM (eds). A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change (IPCC). Cambridge University Press, Cambridge, UK; New York, NY, Ch. 3, 109–230.
  • Stratton RA, Stirling AJ. 2012. Improving the diurnal cycle of convection in GCMs. Q. J. R. Meteorol. Soc. 138: 11211134.
  • Uppala SM, KÅllberg PW, Simmons AJ, Andrae U, Da Costa Bechtold V, Fiorino M, Gibson JK, Haseler J, Hernandez A, Kelly GA, Li X, Onogi K, Saarinen S, Sokka N, Allan RP, Andersson E, Arpe K, Balmaseda MA, Beljaars ACM, Van De Berg L, Bidlot J, Bormann N, Caires S, Chevallier F, Dethof A, Dragosavac M, Fisher M, Fuentes M, Hagemann S, Hólm E, Hoskins BJ, Isaksen L, Janssen PAEM, Jenne R, Mcnally AP, Mahfouf J-F, Morcrette J-J, Rayner NA, Saunders RW, Simon P, Sterl A, Trenberth KE, Untch A, Vasiljevic D, Viterbo P, Woollen J. 2005. The ERA-40 re-analysis. Q. J. R. Meteorol. Soc. 131: 29613012.
  • Webb JDC, Elsom DM, Meaden GT. 2009. Severe hailstorms in Britain and Ireland, a climatological survey and hazard assessment. Atmos. Res. 93: 587606.
  • Wilks DS. 2011. Statistical Methods in the Atmospheric Sciences, 3rd edn. Elsevier: Amsterdam, The Netherlands.
  • Williams L. 1973. Hail and Its Distribution. Studies of the Army Aviation (V/STOL Environment), Army Engineer Topographic Laboratories Rep. 8, ETL-SR-73-3, 27.
  • Xie B, Zhang Q, Wang Y. 2008. Trends in hail in China during 1960–2005. Geophys. Res. Lett. 35: L13801, DOI: 10.1029/2008GL034067.
  • Zhang JG. 2009. Effects of entrainment on convective available potential energy and closure assumptions in convection parameterization. J. Geophys. Res. 114: D07109, DOI: 10.1029/2008JD010976.