Presentation given at: III Jornades de Meteorologia Eduard Fontsere, sponsored by the Catalan Meteorological Society, 15-16 November 1997, Barcelona, Spain. The viewgraphs from this presentation can be viewed here.

Severe Storms


Charles A. Doswell III

NOAA, National Severe Storms Laboratory

Norman, Oklahoma U.S.A.

1. Introduction

This presentation is focused on severe convective storms, where the notion of a severe storm is herein limited to storms producing one or more of: potentially damaging non-tornadic winds, large hail, or tornadoes. Although heavy precipitation is a serious problem often associated with convection, it has been covered elsewhere (Doswell 1997). Severe convective events are associated with the strongest vertical motions observed in any atmospheric phenomenon. Simple parcel theory produces the concept of convective available potential energy (CAPE ):


where g is the acceleration due to gravity, LFC denotes the level of free convection for an ascending parcel, EL is the equilibrium level for that parcel, Tv denotes the virtual temperature, and the overbar stands for the environment while the prime indicates the value for the ascending parcel. CAPE represents the positive area on a thermodynamic diagram, wherein area is equivalent to energy, so this energy is available to accelerate an ascending parcel. It is easy to show from pure parcel theory that the peak vertical motion, wmax, associated with a given amount of CAPE is simply

For a CAPE value of 1000 J kg-1 (or, equivalently, 1000 m2 s-2), then, pure parcel theory predicts a value for wmax of about 45 m s-1. Alternatively, when wmax is 10 m s-1, the required CAPE is only 50 J kg-1. Although pure parcel theory is only an abstraction, convective vertical motions are driven primarily by CAPE, with only a few exceptions to be noted below.

Another important factor in severe convection is downdrafts. It is possible to define a downdraft equivalent to CAPE, sometimes referred to as downdraft CAPE (or DCAPE):


where LFS is the level of free sink (the downdraft equivalent of the LFC), and "sfc " is the surface, which presumably is where the potential damage from downdraft-induced winds is to occur. Since the buoyancy is negative, positive DCAPE is associated with descending parcels. The same relationship for the peak vertical motions holds for downdrafts, with the appropriate sign change, of course.

For strong updrafts in a conditionally unstable atmosphere (i.e., where the lapse rate exceeds moist adiabatic but is less than dry adiabatic), the presence of moisture is needed for positive buoyancy. For strong downdrafts, it is negative buoyancy that must be maintained. Severe convective weather is strongly connected to processes that result in rapid ascent or rapid descent, or both, as I shall discuss.

Another important issue in severe convection is vertical wind shear. Depending on the vertical coordinate, this is depicted by either or by , where p is pressure and z is geometric height, and the Vh is the horizontal wind vector (on either a p-surface or a z-surface). Vertical wind shear is critical in organizing convection into large, long-lived structures, and is almost certainly a factor in many tornado events.

2. Severe windstorms

I have written about this elsewhere (Doswell 1994), so I will attempt to be brief. The basic process for developing potentially damaging winds in a non-tornadic convective storm is through development of intense downdrafts. The primary factors for this are negative buoyancy and water loading of the downdraft column. In some supercell situations, downward-directed perturbation pressure gradient forces may be at work (see Klemp and Rotunno 1983), but this is not likely to be important outside of supercell events. Only in cases where an intense low-level mesocyclone develops is there much potential for downward-directed pressure gradient forces to be significant. Further, the inflow in some supercells can attain damaging proportions; again, this is confined to supercell storms. Not all supercells have such strong inflows, of course.

The ambient stratification can be a large factor in developing strong downdrafts, since it is easier to develop negative buoyancy if the environment shows high lapse rates. It is not uncommon over some regions for the development of deep surface-based layers with near-adiabatic lapse rates (Brown et al. 1982). In such cases, the updrafts associated with the convection are often high-based and relatively weak, but the downdrafts can become quite substantial. When lapse rates approach dry adiabatic, the only factor limiting the descent of a negatively buoyant parcel is entrainment.

Water loading is not a very efficient way to generate downdrafts. It takes substantial condensed water content to be as important as negative buoyancy. It was shown in Doswell (1994) that the vertical momentum equation can be written as


where is the environmental density, l is the condensed water mixing ratio (dimensionless), and the rest of the symbols are as defined above. Thus, the contribution from buoyancy is equal to that from condensed water when


For a typical value ( ), it can be seen that one degree of buoyancy difference is roughly equivalent to a mixing ratio of about 3.5 g kg-1 of condensed water, a large value.

Nevertheless, condensed water can be very important in some situations, because the availability of condensed water allows a downdraft to descend more or less reversibly along a moist adiabat, instead of following a dry adiabatic descent path. In situations with conditionally unstable environmental lapse rates, this allows a condensed water-filled downdraft to maintain negative buoyancy when a dry downdraft rapidly would lose its negative buoyancy by adiabatic compression.

3. Severe hailstorms

Hail formation is not so easily understood as is the production of strong downdrafts, but the basic idea is relatively simple. Whereas intense downdrafts are needed to produce potentially damaging winds at the surface, it is quite obvious that an intense updraft is needed to form large hailstones. Although the physics of real hailstones can be complicated by many factors, in order that a hailstone grow to large (> 2 cm diameter) size, a strong updraft is needed. It is worth noting that in supercells, a significant part of the total updraft speed may come from the vertical pressure gradient forces arising from the interaction between the updraft and the shear (see Rotunno and Klemp 1982). Not all supercells occur in regions with high CAPE. The larger the hailstone, the stronger the implied updraft. The terminal velocity of real hailstones is complicated by the morphology of the actual stones, but a falling ice sphere 8 cm in diameter has a terminal velocity in the range of 40-60 m s-1. Some growth may occur during descent, but some melting also occurs near the surface, so the terminal velocity of the stone is probably not far from the peak updraft speed in the convective storm.

The old idea of "recycling" of hailstones (e.g., Ludlam 1963) to account for the layered structure of some hailstones is no longer considered to be necessary to account for this layering (Nelson 1983). It is now felt that some minor oscillations in height during the hailstone trajectory, plus inhomogeneities in the updraft thermodynamics likely could account for the multiple transitions from dry to wet growth phases and it is thereby not necessary for a real hailstone trajectory to include large height oscillations. However, it is quite possible that the embryos for hailstones may be "recycled" from other nearby cells or from other parts of the same convective cell.

Clearly, the issue of hailstone growth depends on microphysical factors that are not observed very often. In consequence, it is not obvious why some storms produce copious amounts of large hail and other, seemingly similar storms produce little or no large hail. Although having a strong updraft certainly is a necessary condition for the production of large hail, it is not a sufficient condition. Much remains to be learned about hail production in convective storms.

4. Tornadoes and tornadic storms

Tornadoes are relatively rare throughout most of the world. A few places, notably the plains of North America, have many more tornadoes than the rest of the planet combined. Not only are North American tornadoes more common than elsewhere, but they attain intensities seldom seen outside of North America. Some of this might well be related to their relatively high frequency. Even in the United States, where tornado frequency is the highest in the world, tornado frequency decreases as the intensity scale increases. Thus, out of approximately 1000 tornadoes observed annually in the United States, only a handful are truly violent. Even a "weak" tornado is not a trifling event, of course, with windspeeds on the order of, say, 40 m s-1. Generally speaking, the greater the intensity, the more likely a tornado is to be both large and long-lasting. This relationship is far from perfect, but there is undeniably this tendency in the observations (see Kelly et al. 1978; Grazulis 1993). Thus, a substantial fraction of the damage and casualties is due to a small number of especially dangerous events. These tornadoes tend to occur during days that have been characterized as "synoptically evident" (Doswell et al. 1993).

Such tornadoes are almost uniquely associated with supercell storms. By far the majority of tornado days do not fit this classic mold, however. Many tornadoes occur in conditions that do not seem so evident. Moreover, it has become clear that many tornadoes do not occur in supercell storms (Doswell and Burgess 1993). Thus, it seems that the majority of tornadoes occur in situations that do not exhibit strong signatures of their likelihood in the synoptic-scale data.

It is important to understand this, as the tornado threat is considered outside of the United States. Tornadoes are not common anywhere, but neither can their occurrence be excluded categorically over a large part of the world. In a place where tornadoes are rare, the most likely tornado occurrences are "weak" tornadoes; perhaps only one tornado in 100 (or more) are intense. This should not be taken to imply that all tornadoes outside of the United States are going to be small and weak. The factors associated with producing tornadoes are poorly understood.

For supercell-associated events, the chances of a significant tornado are substantial, so in large measure, the real threat from a rare tornado is at its greatest when supercells are possible. Recent research has shown that supercells arise when convection develops in an environment with substantial vertical wind shear. One useful measure of this shear is the storm-relative environmental helicity (H) defined by


where k is the vertical unit vector and C is the storm motion vector. This quantity is shown by Droegemeier et al. (1993) to be a good predictor of convective storm rotation, which is characteristic of supercells (Doswell and Burgess 1993). A word of caution here: no single variable is going to be a cure-all for forecasting. A convective storm developing in a high-helicity environment is quite likely to become supercellular, but it should be noted that not all supercells become tornadic; in fact, perhaps only 20% of supercells produce tornadoes. However, more than 90% of supercells do produce some form of severe weather (including heavy precipitation). It is not clearly understood why so few mesocyclonic storms fail to become tornadic (but see Brooks et al. 1994), but most of the strong tornadoes and virtually all of the violent tornadoes are from supercells.

On the other hand, non-supercell tornadoes (NSTs) can be significant events, as no tornado can be taken lightly. Lee and Wilhelmson (1997) have presented a detailed numerical simulation-based study of one mechanism for forming NSTs. Others may well exist, but have not received a great deal of attention. Many tornadoes around the world are, of course, NSTs; this is likely consistent with tornadoes in the United States as well. Such storms are correspondingly difficult to predict, since the mechanisms for producing them are less well-known than in the case of supercells.

5. Discussion

This presentation has so far considered some basic ideas of severe storms. It remains to be shown how this might be of relevance to forecasting. My view of forecasting is spelled out in Doswell et al. (1996), albeit in terms of heavy precipitation foreasts. To the extent that the meteorological science understands an event, that knowledge can be used to define the essential ingredients for that event, without which that event is not likely to occur.

Since this presentation concerns severe convection, it is obvious that deep convection must occur before any consideration of its severity can be done. Generally, deep, moist convection requires some buoyancy (generally for lower tropospheric parcels) and a process to lift parcels to their LFCs (see Johns and Doswell 1991). If it can be established that deep, moist convection is possible, then it becomes necessary to consider the possibility of severe weather associated with that convection.

In this context, it is possible to describe in these simple terms what is necessary (but not always sufficient) for the events I have been discussing. For windstorms, most events are associated with strong downdrafts and their associated outflows. Thus, any process promoting strong downdrafts makes damaging nontornadic winds more likely; for example, high lapse rates promote strong downdrafts even when the updrafts are weak. In a few cases, of course, supercell inflows are going to become damaging. However, the general threat for severe weather is high in supercell situations, so this sort of relatively rare event (damaging inflow winds) are unlikely to be a complete surprise. It is not known why a few inflows become strong enough to cause damage, so it is not possible to specify an ingredient. A similar statement can be made for those occasions with supercells where the so-called "occlusion downdraft" becomes strong owing to downward-directed perturbation pressure gradient forces. It is not obvious how to predict this, but given the existence of a supercell, it is unlikely that forecasters would not already have a high expectation of severe weather.

For hailstorms, it is clear that science leaves considerable room for doubt. The occurrence of large hail requires a strong updraft, but it is not at all obvious how to recognize those strong updrafts that will not produce large hail. This suggests that empiricisms might be useful in forecasting even when it is not yet understood how they work. The challenge is to limit the occasions when false alarms will occur: intense updrafts that fail to produce large hail. Perhaps successful empirical studies will point the way to developing new understanding for why they work. This is desireable, since a major flaw in purely empirical studies is that it is difficult to know when they will fail.

Tornadoes will continue to be a challenge. Even when a storm has been recognized as supercellular, it is still difficult to know whether or not it will become tornadic. NSTs represent a further challenge to forecasting. Until the processes are understood better than they are at present, it is likely that the task of tornado forecasting will consist of predicting the likelihood of supercells and using empirical techniques for the rest. As an example of the latter, it has been shown that some NSTs are associated with terrain-induced mesoscale structures (see Brady and Szoke 1989). These terrain-associated occurrences almost certainly can be reasonably well-forecast even if the details of how the event occurs are not perfectly well-understood.

Severe convection represents a difficult forecasting problem for a number of reasons. Severe events are rare even in places where they are relatively common. They often are highly localized in time and space, and poorly represented in synoptic-scale data. Using numerical models for explicit prediction of the severe weather associated with convective storms is likely to be difficult, if not impossible, and certainly is a long ways from being useful in operations (Brooks et al. 1992). Human forecasters will need to employ ingredients-based methods, using numerical model guidance to address the relevant physical variables, perhaps modified by empirical methods to reduce the "false alarms."


Brady, R.H., and E.J. Szoke, 1989: A case study of non-mesocyclone tornado development in northeast Colorado; Similarities to waterspout formation. Mon. Wea. Rev., 117, 843-856.

Brooks, H.E., C.A. Doswell III and R.A. Maddox, 1992: On the use of mesoscale and cloud-scale models in operational forecasting. Wea. Forecasting, 7, 120-132.

______, ______, and J. Cooper, 1994: On the environments of tornadic and nontornadic mesocyclones. Wea. Forecasting, 9, 606-618.

Doswell, C.A. III, and D.W. Burgess, 1993: Tornadoes and tornadic storms. A review of conceptual models. The Tornado: Its Structure, Dynamics, Hazards, and Prediction (Geophys. Monogr. 79), Amer. Geophys. Union, 161-172.

______, R.H. Johns and S.J. Weiss, 1993: Tornado forecasting: A review. The Tornado: Its Structure, Dynamics, Hazards, and Prediction (Geophys. Monogr. 79), Amer. Geophys. Union, 557-571.

______, 1994: Extreme convective windstorms: Current understanding and research. Report of the proceedings, U.S.-Spain Workshop on Natural Hazards (Barcelona, Spain), National Science Foundation, 44-55.

______, 1997: Flash flood forecasting - Techniques and limitations. Elsewhere in this volume.

Droegemeier, K.K., S.M. Lazarus, and R. Davies-Jones, 1993: The influence of helicity on numerically-simulated convective storms. Mon. Wea. Rev., 121, 2005-2029.

Johns, R.H., and C.A. Doswell III, 1992: Severe local storms forecasting. Wea. Forecasting, 7, 588-612.

Kelly, D.L., J.T. Schaefer, R.P. McNulty, C.A. Doswell III and R.F. Abbey, Jr., 1978: An augmented tornado climatology. Mon. Wea. Rev., 106, 1172-1183.

Klemp, J.B., and R. Rotunno, 1983: A study of the tornadic region within a supercell thunderstorm. J. Atmos. Sci., 40, 359-377.

Lee, B.D., and R. B. Wilhelmson, 1993: The numerical simulation of non-supercell tornadogenesis. Part I: Initiation and evolution of pretornadic misocyclone circulations along a dry outflow boundary. J. Atmos. Sci., 54, 32-60.

Ludlam, F.H., 1963: Severe local storms: A review. Meteor. Monogr., 5, No. 27, Amer. Meteor. Soc., 1-30.

Nelson, S.P., 1983: The influence of storm flow structure on hail growth. J. Atmos. Sci., 40, 1965-1983.

Rotunno, R., and J.B. Klemp, 1982: The influence of the shear-induced vertical pressure gradient on thunderstorm motion. Mon. Wea. Rev., 110, 136-151.