“How do tornadoes form?” is one of the most commonly asked questions I get from friends, the media, and even fellow scientists outside of mesoscale meteorology.

Great question!  There’s much more to it than the “clashing of air masses,” which is a popular explanation that seems to be recycled every time a killer tornado makes the headlines. In fact, strongly clashing air masses often makes tornado formation less likely.  Much of what appears below is paraphrased or excerpted from the following publications in Weatherwise, Physics Today, and the Bulletin of the American Meteorological Society:

Markowski, P., and Y. Richardson, 2013: How to make a tornado. Weatherwise, July/August 2013, 12–19.

Markowski, P. M., and Y. P. Richardson, 2014: What we know and don’t know about tornado formation.  Physics Today, 67, 26–31.

Schultz, D. M., Y. P. Richardson, P. M. Markowski, and C. A. Doswell III, 2014: Tornadoes in the central United States and the “clash of air masses.” Bulletin of the American Meteorological Society, 95, 1704–1712.



You also might want to read this short piece by Prof. Richardson and me, published in The Conversation:







The vast majority of headline-making tornadoes are spawned by supercell thunderstorms, which are storms that are characterized by an updraft that rotates (counterclockwise rotation, if viewed from above, is most often observed, though some supercell updrafts rotate clockwise). Changes in wind speed and direction with altitude in the storm’s environment imply the presence of horizontal spin, similar to that of a football. As this spinning air is drawn into the storm’s updraft, the updraft rotates. Tornado formation is associated with a separate air stream, one that descends through a precipitation-driven downdraft and acquires horizontal spin by way of a horizontal variation of temperature along this air stream. If the air within this downdraft air stream is not too cold, this spinning air subsequently can be sucked upward by the supercell’s updraft, leading to a tornado.

What I tell kids

Scientists are still studying exactly how tornadoes form.  Although some of the details of tornado formation are still being investigated by scientists, we know the ingredients for tornadoes in a broad sense.  In general, thunderstorms occur when warm, humid air near the ground is overlaid by relatively cold air aloft (in other words, temperatures cool rapidly with height).  This variation of temperature with altitude results in what meteorologists refer to as “instability,” which gives energy to thunderstorms.  Tornadoes can form within thunderstorms when instability is accompanied by what meteorologists refer to as “wind shear”—large changes in wind speed or direction with altitude. On tornado outbreak days, the temperature can decrease with height by nearly 30°F per mile (10°C per kilometer) over a depth of several miles.  Moreover, wind speeds can change by over 50 mph over a similar depth.  Both rising, spinning, warm air and sinking cool air are important for tornado formation.

Thirty-second explanation of how tornadoes form in supercell thunderstorms (applicable to virtually all tornado outbreaks)

Thunderstorms occur when warm, humid air near the surface lies beneath a deep layer of air in which the temperature decreases rapidly with height. Such an atmosphere is said to be “unstable,” which simply means that it contains ample energy to fuel thunderstorm updrafts. When an unstable atmosphere also is characterized by unusually high humidity near the surface and large changes in wind with height (“wind shear”), especially when considerable variation in wind is found over just the lowest few thousand feet of the atmosphere, tornado-producing supercell storms are possible.


Here’s a lengthier explanation, for those who want more of the gory details (much of it excerpted from the Markowski–Richardson Physics Today article)

Step 1: The development of rotation aloft

All thunderstorms, regardless of their tornado potential, require an unstable atmosphere. In an unstable atmosphere, rising air becomes warmer than its surroundings, i.e., it becomes buoyant. Buoyancy causes the air to accelerate upward. The buoyancy is acquired primarily by the effects of condensation, which occurs as air becomes saturated during its ascent; if air ascends to sufficiently high altitudes within the storm’s updraft, freezing can contribute additional warmth. A parcel of air ascending through an updraft—we can envision a parcel as being a bubble of air that’s small enough to be considered to be fairly homogeneous, but much larger than microscopic—can be anywhere from a few degrees warmer than its surroundings to as much as 10–15°C (18–27°F) warmer in an atmosphere possessing extreme instability.

The vast majority of significant tornadoes (EF2 and stronger), and virtually all violent tornadoes (EF4–EF5), are spawned by supercell thunderstorms. Supercells are storms that have a rotating updraft. Horizontally oriented rotation is present in supercell environments owing to the variation of horizontal winds with height—vertical wind shear. For example, winds at the surface might be from the southeast, with winds aloft from the southwest, and they may have very different speeds. A difference of roughly 50 mph between the surface wind and winds at 18,000 feet is typically sufficient to create supercells. Parcels of air in such a wind field—warm, moist parcels that sustain the supercell’s updraft—possess what is known as streamwise vorticity. The term refers to spin (“vorticity”) that is aligned with the direction the air parcels are traveling (“streaming”). Air parcels that enter the supercell’s updraft spin like spiraling footballs. The horizontal spin becomes vertical as parcels are ingested into the updraft, and the collective influence of the spinning parcels results in a mesocyclone, or updraft-scale rotation about a vertical axis (Figure 1). Updrafts with mesocyclones can be visually stunning, with the updraft being striated like the threads of a screw and the rotation being plainly visible to the naked eye.

Paul Markowski photo
Our present understanding of how a tornado develops in a supercell thunderstorm. (a) A tornadic supercell near Deer Trail, Colorado, intercepted by the second Verification of the Origins of Rotation in Tornadoes Experiment (VORTEX2) on 10 June 2010 (photo by Bill Reid). The white arrows show the orientation of the vorticity vector, the yellow curved arrows indicate the sense of spin, and the red and blue lines indicate the paths taken by updraft and downdraft air parcels, respectively. In step 1 of tornadogenesis, the storm acquires large-scale rotation—a midlevel mesocyclone—by tilting the horizontal vorticity in winds entering the storm’s updraft. In step 2, buoyancy gradients due to relatively warm and cool air straddling either side of downdraft air parcels generate horizontal vorticity. That horizontal vorticity is then tilted upward by surrounding wind fields as the parcels descend. (b) A closeup of the region inside the dashed box in panel a. In step 3, conservation of angular momentum amplifies the now vertical vorticity as air converges toward the axis of rotation while being sucked upward by the strong mesocyclone above. (c) A nontornadic supercell with tornadogenesis failure in progress, intercepted by VORTEX2 near Panhandle, Texas, on 13 June 2009 (photo by Paul Markowski).

The mesocyclone that results from the tilting of streamwise vorticity into the vertical direction is typically strongest at 10,000–20,000 feet above the ground. Such midlevel mesocyclones are readily detectable by the National Weather Service WSR-88D radars, given their size (several miles in diameter) and altitude. However, the way by which a midlevel mesocyclone develops differs from how rotation develops next to the ground. When only the updraft is responsible for the tilting of horizontal spin toward the vertical, the horizontally spinning air parcels only develop vertically oriented spin as they rise away from the ground. So the updraft’s tilting of the streamwise vorticity that originates in the storm’s environment cannot produce a tornado, which is a violently spinning vortex in contact with the ground. The differences in the physical mechanisms for the development of rotation aloft (relatively easy to detect with WSR-88D) versus rotation next to the ground (difficult to detect with WSR-88D except at close range) are one reason for imperfect tornado warnings.

Step 2: The development of rotation next to the ground

The development of rotation next to the ground requires a downdraft. All thunderstorms have downdrafts in addition to their updrafts, and supercells are no different. The air that sinks within a downdraft is usually cooler than its surroundings. This is due to the evaporation of rain and, to a lesser extent, the melting of hail and snow. Once downdraft air reaches the ground, it spreads away from the storm as outflow. The leading edge of the outflow is called the gust front. If you’ve ever experienced the cool breeze that precedes the arrival of a thunderstorm (typically blowing toward you from the area of heavy rain), then you’ve experienced the outflow from a downdraft.

Supercell storms typically have an expansive region of downdraft and outflow that extends from northeast (ahead) of the updraft, around the updraft’s northern flank, and wraps around the western (rear) flank. Though non-supercell thunderstorms usually only feed off warm air from the environment, the supercell updrafts are strong enough to forcibly lift some of the air parcels from the cool outflow (such parcels are heavy and would not rise on their own if not for the strong sucking action of the supercell’s updraft), in addition to the warm parcels from the environment.

Outflow air parcels that are drawn toward the updraft gradually descend as they travel toward the updraft because they are cooler than the environment. Those parcels that travel along the immediate cool side of the gust front experience a horizontal temperature gradient en route, with warm air to the parcel’s left and cool air to the parcel’s right, with respect to the direction that the parcels are traveling (from right to left in Figure 1). The horizontal temperature gradient generates so-called baroclinic vorticity about a horizontal axis—essentially a torque is being applied to the parcels by virtue of the fact that relatively warm air rises and relatively cool air sinks. For example, the horizontal spin that is produced on the flanks of a small pour of milk into a glass of water is the result of the same dynamics: the horizontal density difference between the (heavier) milk and (lighter) water generates horizontal spin just like the horizontal temperature difference between (heavier) cool air and (lighter) warm air.

The horizontal spin of the parcels that gradually descend within the downdraft and outflow can become greatly intensified within the sinking air stream, and even can be tilted upward very near to the surface. Thus, vertical rotation can be acquired next to the ground within the outflow, i.e., within air parcels that have a prior history of descent and baroclinic vorticity generation. In contrast, recall that air parcels possessing only environmental horizontal vorticity and tilted only by an updraft acquire significant vertical vorticity only after the parcels have risen a considerable distance away from the ground.

Step 3: The intensification of near-ground rotation

Though the development of near-ground rotation in supercells is a prerequisite for tornadogenesis, in recent years we’ve learned that most supercells develop near-ground rotation yet are nontornadic, i.e., the rotation fails to reach tornado strength. The vertical vorticity that arises next to the ground in Step 2 is roughly one-hundredth that of a tornado. Tornadogenesis requires a dramatic intensification of the vertical vorticity acquired in Step 2.

The intensification occurs by way of the “figure-skater effect”—referred to by scientists as the conservation of angular momentum or vorticity stretching. A figure skater spins faster as she draws her arms closer to her axis of rotation. The same principle applies to the spin about a vertical axis produced in Step 2. If the spinning air can be converged—drawn inward toward its axis of rotation—it will spin faster.

The convergence of the spinning air depends on the extent to which the spinning air can rise: Air that is accelerated upward is unavoidably required to be associated with the convergence of air below (if not, a vacuum would develop!). Recall that the parcels that have vertical vorticity next to the ground are parcels that previously descended through a downdraft. In other words, these parcels are cooler than the environment. In order for them to be accelerated upward, and in doing so, promote convergence and the rapid intensification of rotation to tornado strength, either the parcels must not be too cold or the supercell’s updraft must have unusually strong “suction” just above the ground (to the scientists, it’s a strong upward-directed pressure-gradient force). The suction is associated with the rotation in the overlying updraft. More will be said about this in the next section.

One of the principal findings of the VORTEX project was that the outflow of tornadic supercells tends to not be as cold as the outflow of nontornadic supercells. In tornadic supercells, the outflow/downdraft air that bears the rotation that is intensified to tornado strength is sometimes just a few degrees colder than the environment. The “suction” tends to be strong as well. The combination of only slightly cold air and strong suction from above makes it likely that upward accelerations and convergence near the ground will be sufficiently strong to intensify vertical vorticity to tornado strength. The parcels of air spin faster as they near the axis of rotation, and they ascend rapidly as well. Conversely, in nontornadic supercells, the outflow/downdraft air can be up to 5–10°C (9–18°F) colder than the environment, which implies that the air is heavy and resists upward acceleration. The suction acting on the near-ground rotation is often weak as well, either because the rotation is shunted away from the strongest suction by the cold outflow, or because the suction is just weak overall (more on this below). The bottom line is that upward accelerations of the cold, heavy air parcels near the ground are inhibited, and air simply spreads away from the storm along the ground. The lack of a strong figure-skater effect in this case results in the near-ground rotation remaining well below tornado strength.

Tornado forecasting and nowcasting

In the past decade, forecasters have become skillful at discriminating between the environments capable of supporting strong-to-violent (EF2+) tornadoes and environments incapable of supporting such tornadoes. For example, large outbreaks are now routinely predicted by the Storm Prediction Center days in advance, “high-risk” outlooks capture most major tornado events, and strong-to-violent tornadoes rarely occur outside of tornado watches. The fact that schools sometimes are dismissed early on tornado outbreak days testifies to the skill and public confidence in today’s forecasts.

In discriminating between tornadic and nontornadic supercell environments, considerable attention is paid to the relative humidity and vertical wind shear in roughly the lowest half-mile, both of which are relatively easy to diagnose in real time and are fairly well-predicted by operational numerical weather prediction models (this is what enables skillful outlooks to be made days in advance in some situations). Assuming that conditions will be present to support supercell thunderstorms in general, i.e., that the environment has sufficient wind shear and instability to favor rotating updrafts, tornadogenesis becomes increasingly likely as the low-level wind shear and relative humidity increase. On tornado outbreak days, the lower atmosphere can be so humid that cloud bases are just a couple thousand feet above the ground (the cloud base lowers as the relative humidity increases). The wind shear can be so extreme that winds can vary by 50 mph between the ground and cloud base.

Enhanced low-level wind shear usually implies enhanced low-level streamwise vorticity, and the tilting of the enhanced streamwise vorticity promotes a stronger mesocyclone in the updraft that overlies the near-ground rotation that develops in Step 2. The stronger mesocyclone is associated with lower pressure. (Think of stirring a cup of coffee—the faster you stir, the greater the fluid drop/pressure drop at the center of rotation.) The lower the pressure that overlies the near-ground rotation, the stronger the suction will be. As for the effect of enhanced relative humidity, as the low-level relative humidity increases, the outflow/downdraft air parcels tend to be less cold because evaporation is suppressed. The combination of the strong suction and air parcels that are only slightly colder than the environment strongly favors the intense near-ground upward accelerations and associated convergence required in Step 3.

Though analyzing low-level wind shear and relative humidity has worked well for identifying environments capable of supporting strong-to-violent tornadoes, we as of yet have little ability to discriminate between weak-tornado (EF0–EF1) supercell environments and nontornadic supercell environments. Many weak tornadoes also occur in non-supercell thunderstorms such as squall lines. And waterspouts and landspouts (which also tend to be weak) can develop from otherwise benign cumulus congestus clouds. Though there is some evidence suggesting an enhanced tornado threat in squall lines when the instability and low-level shear are exceptionally large, identifying environments favorable for waterspouts and landspouts has proven even more difficult.

Active research areas—what are scientists focusing on now?

Unfortunately, even if the environment is known to be extremely favorable for supercell tornadoes, forecasters have a limited ability to say when or if a specific storm will produce a tornado. Even on tornado outbreak days, not all the supercells are tornadic. Moreover, tornadic supercells are not tornadic all the time. So researchers have been investigating triggers for tornadogenesis, such as small-scale downdraft surges and descending precipitation shafts on the supercell’s rear flank, as well as the processes that sustain tornadoes once they form. But as of now, if a tornado is occurring, forecasters have practically no ability to provide guidance to the public on the tornado’s current intensity (spotter reports are about the only source of information), future intensity, or expected duration.

Another active area of research looks at the precipitation characteristics of supercells, like the size distributions of raindrops and hailstones, and how those characteristics affect supercell downdraft regions, vorticity generation, and, ultimately, the formation and maintenance of tornadoes. Significant shortcomings in both simulations and observations of the precipitation characteristics and buoyancy fields of thunderstorms make the topic especially challenging. Our understanding of the role of surface friction on tornadogenesis also is incomplete and is similarly hampered by both our simulation and observing capabilities. And our knowledge of the effects of terrain on tornadoes and their parent thunderstorms is mediocre at best.

A few technological advances in tornado short-term forecasting are worth mentioning. The recent dual-polarization upgrade of National Weather Service radars improves the characterization of what is sampled by the radar beam. Instead of merely identifying where precipitation exists, the radars now can identify whether the precipitation comprises large or small hail, large or small raindrops, or even debris. Such observations benefit research on how a storm’s precipitation characteristics might influence tornado development. Identifying debris obviously does not improve tornado warning lead times, but a late warning is better than no warning.

Given the average spacing of about 250 km between National Weather Service radars, most tornadoes occur too far from a radar to be resolved or are simply overshot by the radar beam. Mesocyclones are relatively easy to detect aloft, but detecting a mesocyclone is not the same as detecting a tornado. One solution, though likely many years away on a national scale, might be gap-filling radars—low-cost, low-power radars used to augment the existing radar network. A demonstration network already exists in the Dallas–Fort Worth area.

Lastly, NOAA scientists are exploring the feasibility of a concept called “Warn-On-Forecast” to make thunderstorm-specific predictions as opposed to diagnoses. The idea is to use multiple extremely short-range, high-resolution computer simulations that are updated in real time with observations. The intrinsic limits on the predictability of thunderstorms are a daunting challenge to face. But if successful, Warn-On-Forecast could dramatically increase lead times for severe weather warnings.