Show Summary Details

Page of

Printed from Oxford Research Encyclopedias, Planetary Science. Under the terms of the licence agreement, an individual user may print out a single article for personal use (for details see Privacy Policy and Legal Notice).

date: 05 March 2021

Dust Devils on Earth and Marsfree

  • Matthew R. BalmeMatthew R. BalmeThe Open University, School of Physical Sciences


Dust devils are rotating columns or cones of air, loaded with dust and other fine particles, that are most often found in arid or desert areas. They are common on both Mars and Earth, despite Mars’ very thin atmosphere. The smallest and least intense dust devils might last only a few 10s of seconds and be just a meters or two across. The largest dust devils can persist for hours and are intensely swirling columns of dust with “skirts” of sand at their base, 10s or more meters in diameter and hundreds of meters high; even larger examples have been seen on Mars. Dust devils on Earth have been documented for thousands of years, but scientific observations really began in the early 20th century, culminating in a period of intense research in the 1960s. The discovery of dust devils on Mars was made using data from the NASA Viking lander and orbiter missions in the late 1970s and early 1980s and stimulated a renewed scientific interest in dust devils. Observations from subsequent lander, rover, and orbital missions show that Martian dust devils are common but heterogeneously distributed in space and time and have a significant effect on surface albedo (often leaving “tracks” on the surface) but do not appear to be triggers of global or major dust storms. An aspiration of future research is to synthesize observations and detailed models of dust devils to estimate more accurately their role in dust lifting at both local and global scales, both on Earth and on Mars.


Dust devils—also known as whirlwinds, willie-willies, or sand/dust whirls—are rotating columns or cones of air, loaded with dust and other fine particles (Fig. 1) and are most often found in arid or desert areas. They differ from tornadoes, their larger and more dangerous vortical cousins, in the immediate source of the energy that drives them—insolation upon the ground is the source for dust devils, whereas latent heat in moisture is the source for tornadoes. Furthermore, dust devils are associated with clear skies rather than clouds. They are common on Earth and also, notwithstanding that planet’s thin atmosphere, on Mars (Fig. 2).

Figure 1. Well-developed dust devils on Earth. Some dust devils do not possess such well-defined structure and simply appear as rotating clouds of dust. (a) Large (see vehicle for scale) dust devil in Nevada, USA. Image credit M. Balme. (b) ~5 m diameter dust devil in Arizona, USA.

Image credit L. Neakrase.

A commonly used definition for a dust devil is provided by Oke, Tapper, and Dunkerley(2007):

A vortex rotating in either direction faster than its motion downwind. It must have a coherent columnar or cone shape extending in a vertical direction above the surface. It must carry dust to a height of at least 1 m and last for at least 10 s. It is distinct from a dust blow in that it maintains a fast rotating coherent structure with vertical extension and distinct from a small dust whirl in its size and duration (p. 215).

Figure 2. Dust devils on Mars. (a) Orbital HiRISE (High Resolution Imaging Science Experiment; McEwen et al., 2007) camera observation of ~30 m diameter dust devil. The sinuous dust column leaves a shadow on the surface that allows the height of the dust devil to be estimated. Image credit NASA/Jet Propulsion Laboratory/University of Arizona. Part of HiRISE image ESP_026051_2160. (b) Surface observation made from the “Spirit” Mars Rover showing two dust devils. The foreground dust devil is probably a few 10s of meters in diameter at most. Image #2N177950967EFFADNDP0645R0M1 acquired on Sol 581 of the mission using the NavCam camera.

Image credit NASA/Jet Propulsion Laboratory/California Institute of Technology.

On Earth, dust devils are small and transient phenomena and are rarely greater than a few 10s of meters in diameter, persisting (with rare exceptions) for a few minutes or less. On Mars, though, they can be kilometers across, several kilometers high, and leave tracks many kilometers long that attest to longer lifetimes than on Earth. Dust devils represent loci of enhanced windspeeds compared to the ambient conditions: the peak horizontal windspeeds contained within the swirling body of terrestrial dust devils can be greater than 25 ms–1. On Mars, few direct measurements have been made, but windspeeds are thought to be even higher (e.g., Ringrose et al., 2007), offering an explanation as to how dust devils can lift such large amounts of loose material. Thus dust devils are important injectors of dust into the Martian atmosphere—and in fact might be responsible for maintaining the dust in Mars’s hazy atmosphere (e.g., Newman, Lewis, Read, & Forget, 2002a). On Earth they play a smaller role in terms of global contributions to aerosols but, in locations where they are very active, can impact local air quality.

Understanding dust devils requires a multidisciplinary approach: collecting “real” data requires in situ field studies, which during most of the 20th century was only possible on Earth. Hence, dust devils on Mars have been mainly observed using remote sensing studies——the opposite case of Earth, where the number of field studies far outweighs studies from orbital observations. In addition, laboratory and numerical studies have helped us to understand specific aspects of dust lifting processes and to explore how dust devils form. Here, only a brief summary of the history and state of dust devil research is summarized and is focused mainly on dust devils on Mars. For more information, the reader is directed to the review paper of Balme and Greeley (2006) or, for a complete, in-depth review, the edited reference volume Dust Devils (Reiss et al., 2017).

Evolution of Dust Devil Studies

Motivations for Studying Dust Devils

As described in the historical account of dust devil studies of Lorenz et al. (2016), there are records of dust devils (although often described by other terms) going back thousands of years, including references to “whirlwinds” by the Greek playwright Sophocles (ca. 450 bce; Bowker, 2011) and mentions of swirling dust columns in poems written in ancient Rome and 8th-century China. Dust devil are clearly interesting phenomena that are worth recording. Detailed “scientific” studies of dust devils, however, arguably began with 19th-century European records (e.g., Baddeley, 1860), which detail scientific adjuncts to exploration of desert environments by the representatives of expansionist European powers (Fig. 3). In the 20th century, the motivation for the study of dust devils evolved from descriptions of unfamiliar phenomena and the search for explanations for the unexplained to directed research aimed at answering specific questions. For example, Lorenz et al. (2016) suggest that a motivation for dust devil population research in the United States (specifically, the study of Snow & McClelland, 1990) was to understand whether dust devils in the arid southwest of the country would transport radioactive dust widely from nuclear test sites in these regions. The motivation for many dust devil studies during the 21st century is to understand their impact on air quality at both regional and global scales (see review by Klose et al., 2016) and to understand the role of dust devils on Mars.

Figure 3. A sketch of dust devils in India from Baddeley (1860).

Image credit: P. F. H. Baddeley.

Early 20th-Century Studies

Early- to mid-20th-century dust devil studies measured key physical parameters, such as size, shape, rotation direction, wind speed, and temporal and spatial distribution. A study by Flower (1936) of dust devils in Egypt, Palestine, and Iraq provided data on height, lifetime, and rotation sense and remains a relevant data set in the 21st century. Flower’s work demonstrated that dust devils occurred most often between late morning and mid-afternoon during the summer months, that they had no preference for rotations sense (disagreeing with some contemporaries [e.g., Brooks, 1960; Durward, 1931] but in agreement with all modern observations), and that their lifetimes were generally of the order of a few minutes. Ives’s (1947) study is notable for its observations of long-lived dust devils in the Utah desert: the author records that dust devils have “one hour of duration for each thousand feet of height” and also that exceptional cases of dust devils lasting seven hours have been observed. Ives also made an early measurement of the drop in atmospheric pressure at the center of a dust devil (known as Δ‎P in dust devil research), although his measurement of “½ to 2½ inches of Mercury” was about a factor of 10 higher than modern measurements (Table 1). Ives also reported an estimate of the vertical wind speeds within a dust devil, made by combining observations of small rodents (kangaroo rats) being entrained into a dust devil with measurements of the terminal velocity of the creatures—an experiment which in the modern era would probably be considered ethically dubious, given that it involved dropping animals from airfield control towers.

Focused Dust Devil Studies in the 1960s and 1970s

During the 1960s, the first integrated, purposeful studies of dust devils were performed, contrasting the earlier opportunistic or isolated research reports. Two research programs stand out: the first by P. Sinclair, then a research student at the University of Arizona, and the second by J. Ryan and J. Carroll of the Douglas Aircraft Corporation (later, McDonnell Douglas). Sinclair’s goal was to understand dust devils as phenomena relevant to convective circulations on Earth, and his work, consisting of four published papers and his PhD thesis (Sinclair, 1964, 1965, 1966, 1969, 1973) was pioneering in many ways. He combined surveys of dust devil location and time of initiation with in situ measurements of parameters such as wind speed, temperature, and Δ‎P. Noting that studies using fixed recording stations resulted in very few dust devil encounters (e.g., Lambeth, 1966), Sinclair created dedicated instrument suites and mounted them on a mobile platform. This platform was a Jeep vehicle, on which a metal framework was mounted to support a ~10 m high meteorology tower (Fig. 4). Rather than waiting for dust devils to come to the instruments, this system enabled Sinclair to “chase’” the dust devils, thereby increasing the number of dust devils that could be sampled during a field study. As a final part of his study, Sinclair also used an instrumented “sailplane” (or “glider” in U.K. English) to analyze the thermals above dust devils.

Figure 4. Sinclair’s mobile dust devil sampling platform. From Sinclair (1966) and Sinclair (1973).

Image credit: P. Sinclair. ©American Meteorological Society. Used with permission.

In contrast to the terrestrial focus of Sinclair’s study, Ryan and Carroll’s research sprang out of an interest in Mars: Ryan had studied “Martian yellow clouds” (Ryan, 1964)—now known to be Martian dust storms—and produced a ground-breaking report (Ryan, 1969) that made important predictions about dust devils on Mars: (a) that there would be dust devils on Mars—this was also predicted by others around this time (e.g., Neubauer, 1966), (b) that the thin atmosphere would require very large wind speeds to entrain material—and hence only the largest dust devils would be highly particle-loaded, (c) that dust devils would be most common at the equator at perihelion and there would be fewer dust devils in the northern hemisphere than in the south, and (d) that vertical wind speeds in Martian dust devils would be large enough to support particle transport, once mobilized from the surface.

In a field campaign, Ryan and Carroll (1970) measured the in situ wind speeds of about 80 dust devils, monitored the rotation direction, and estimated the diameter of many more. They used a 300 × 500 m study area of the “El Mirage” playa in the Mojave Desert in southern California, which they “scraped” each day to provide tracer materials. Unlike Sinclair, they used a fixed sensor suite, not a mobile platform, but they also measured ambient meteorological parameters, such as atmospheric vorticity and atmospheric temperature profiles. The abundant (although artificially provided) loose materials allowed them to sample many more dust devils than previous fixed studies (e.g., Lambeth, 1966), showing that many potential dust devils went undetected due to not being dust loaded. As well as providing a large and consistent data set of dust devil wind structure (Ryan & Carroll, 1970), they found no preference for rotation direction, but they did find agreement between local atmospheric vorticity and dust devil rotation sense (Carroll & Ryan, 1970).

These two studies demonstrated that the physical parameters of dust devil could be systematically measured in situ and were much enhanced by simultaneous monitoring of the local meteorology and complementary surveys of dust devil frequency, location, size, and lifetime. Furthermore, both studies showed that sufficient encounters between dust devils and sensors could be recorded within a single field study. In some ways, the differences between the Sinclair and Ryan and Carroll studies highlight a methodological difference that persists in modern studies: Should dust devil instrument suites be mobile, actively chasing their targets, or should they be fixed, thereby taking human decisions (i.e., which dust devil to chase) out of the reckoning?

The 1980s and 1990s: Dust Devils on Mars

The prediction (e.g., Neubauer, 1966; Ryan, 1969) that dust devils would be found on Mars took more than a decade to be proved, awaiting the results of the Viking Orbiter (Snyder, 1977) and Lander missions (Soffen, 1977), so the 1980s proved to be a fruitful era for the study of dust devils. The first dust devils on Mars were identified in Viking Orbiter images (Thomas & Gierasch, 1985) as small bright clouds with elongated shadows, and the detection of more than 100 low-pressure vortices passing over the Viking landers almost certainly included true (i.e. dust-loaded) dust devils (Ryan & Lucich, 1983). Furthermore, many observations of “tornado tracks” (Grant & Schultz, 1987), sinuous albedo features on the Martian surface, were made and were later shown to be caused by the passage of dust devils (Edgett & Malin, 2000).

Identification of dust devils from surface platforms on Mars (Ferri, Smith, Lemmon, & Renno, 2003; Metzger, Carr, Johnson, Lemmon, & Parker, 1998) was first achieved using the Imager for Mars Pathfinder (IMP) instrument on the 1997 Mars Pathfinder lander (Golombek et al., 1997). Coupled with the IMP discovery were the detections of the passage of convective vortices by the Pathfinder atmospheric pressure sensor (Ferri et al., 2003; Murphy & Nelli, 2002; Schofield et al., 1997). These data suggested that real, “dusty” dust devils on Mars could be as small as a few 10s of meters across—in contrast to the early prediction of Ryan (1969). While imaging and measurement of dust devils from the Martian surface was ongoing with Pathfinder instruments, orbital observations from the Mars Global Surveyor (MGS; Albee, Arvidson, Palluconi, & Thorpe, 2001) began to demonstrate that dust devils were globally important phenomena. For the first time, dust devils with a range of sizes were observed in high-resolution imaging and conclusively linked to the sinuous surface albedo features that became known as “dust devil tracks” (Malin & Edgett, 2001).

A Renaissance of Terrestrial Dust Devil Research Around the Turn of the Millennium

Encouraged by the IMP and MGS observations, new modeling, remote sensing, and laboratory studies began. The questions that drove this research included “how do dust devils form?”, “how do dust devils lift dust on Mars?” (essentially the same question asked by Ryan in his 1969 report), and “do dust devils on Mars trigger global dust storms?” To test this, Balme et al. (2003) examined dust devil tracks in the Martian southern hemisphere but found no significant difference in dust devil track numbers in the period before a global dust storm than during a year with no global dust storm. Cantor et al. (2006) found that both active dust devils and dust devil tracks were anticorrelated with dust storms. Modeling studies (Newman et al., 2002a; Newman, Lewis, Read, & Forget, 2002b) showed that dust devil activity reduces with increasing dust loading in the atmosphere (in contrast to winds and larger-scale lifting), again suggesting that dust devils do not cause dust storms. Thus a key conclusion of this era of dust devil research was that dust devils do not trigger global or major dust storms.

Inspired by Sinclair’s mobile dust devil sampling methodology. Metzger (1999) used a similar mobile chase methodology at test sites near Eloy, Arizona, and in El Dorado Valley, Nevada, to characterize dust devils. This included field experiments to understand how dust devils lifted dust (Metzger, 1999) and attempts measure the particle load, and hence dust flux, within dust devils (reported much later than the field studies themselves, in Metzger et al., 2011).

“How do dust devils form?” was tackled by Renno and colleagues in a series of papers treating dust devils and similar vortices as convective heat engines (Renno, Burkett, & Larkin, 1998; Renno, Nash, Lunine, & Murphy, 2000; Renno & Ingersoll, 1996). They found that the “intensity” of dust devil can be described as a function of the sensible surface heat flux and the depth of the convective boundary layer (Renno et al., 1998). To test this assertion, Renno et al. (2004) utilized modern data logging and sensor developments to develop new mobile platforms and sensor suites to collect field data including triaxial windspeeds, Δ‎P, and temperature, finding broad agreement of measurements with theory.

The question of “how dust devils lift dust” was approached using field, laboratory, and theoretical studies. A key suggestion, apparently first posed by Greeley and Iverson (1985) in response to early Martian dust devil observations, was that dust devil could “suck up” materials from the surface as their low-pressure cores passed over loose materials. This became known as “the Δ‎P effect” and was tested in the laboratory between 2000 and 2003 using the Arizona State University Vortex Generator (ASUVG) apparatus (Fig. 5A). Using the ASUVG under both terrestrial and Martian atmospheric pressure conditions, Greeley et al. (2003) showed that dust devil-like vortices are far better than unidirectional winds at lifting smaller particles (Fig. 5B), suggesting that the Δ‎P effect was responsible. At the same time, in the field, Balme et al. (2003) used profiling wind speed measurements to measure the friction wind velocity within dust devils and found that, at least for terrestrial dust devils, surface wind shear force alone was sufficient to lift fine materials without recourse to the Δ‎P effect. Later modeling studies (Balme & Hagermann, 2006) identified how the Δ‎P effect could work on Mars, showing that small, intense, rapidly translating dust devils would be most likely to lift dust with the Δ‎P effect. However, without in situ data, it remains unknown whether dust devils on Mars really do “suck.”

Figure 5. Arizona State University Vortex Generator. (a) Photography showing apparatus—dry ice used for flow visualization. (b) Results from Greeley et al. (2003) comparing vortex lifting threshold (points) against boundary layer theory threshold (line). F is the threshold parameter defined as MATH, where ΔP is particle density, g is gravity, Dp is particle diameter, and ρ‎ is air density. The circled points, falling far below the boundary threshold expression, are for the smallest particles and are therefore much more easily lifted by a vortex than by a boundary layer wind. After Greeley et al. (2003).

Copyright 2003 by the American Geophysical Union.

Also during this period the first high-resolution numerical atmosphere models were applied in which dust devils (or at least, convective vortices) could be resolved (Fig. 6). Early work by Kanak (2005, 2000) used a 2 m grid Large Eddie Simulation (LES). A key outcome of this work was the recognition that convective vortices can form in simulations without background wind and that the vertical vorticity source in dust devils could derive from the twisting upwards by convection updrafts of horizontal shear-rolls near the ground (Kanak, 2005). Kanak (2006) compared terrestrial and Martian LES simulations and found that, despite significant differences in the environments, convective vortices formed naturally in both. Rafkin et al. (2001), Toigo et al. (2003), and Michaels and Rafkin (2004) also found convective vortices in LES models of the Martian convective boundary layer, again emphasizing the importance of convective cells in controlling the location and source of vorticity of convective vortices. Such studies suggested that convective vortices and dust devils are a natural part of boundary layer convection. Other numerical modeling work sought to place Martian dust devils into a global context. For example, Newman et al. (2005, 2002a), Basu et al. (2004), and Kahre, Murphy, and Haberle(2006) used dust devil parametrization schemes to predict the contribution of dust devils to the Martian annual dust budget. These modeling studies found that the dust devil parameterization scheme contributed about half of the annual atmospheric dust loading during years without global-scale dust storms (Klose et al., 2016). New models and improvements to existing models—both global and local—have formed a vital part of dust devil research to this day.

Figure 6. Example LES dust devil simulation, after Raasch and Franke (2011). Flow direction is indicated by trajectories of particles moving passively with the resolved‐scale flow. Particles were released at 2 m height and at fixed time intervals of 7.5 s in those areas where the perturbation pressure was below −8 Pa. Color shows the magnitude of horizontal velocity (red, high; blue, low). The displayed vortex has a diameter of about 30 m and is plotted up to about 100 m height.

Copyright 2003 by the American Geophysical Union.

Post-Millennium Dust Devil Research

In a similar way to how Mars Pathfinder and MGS observations inspired new dust devil research, observations from the Mars Exploration Rover (MER) “Spirit” (Crisp et al., 2003; Squyres et al., 2004) provided a “step-change” in knowledge of Martian dust devils. The Spirit rover attained an elevated position by climbing a low hill within the ~170 km diameter Gusev crater in which it landed, allowing it to “look” down on the plains below. From this vantage point, it was able to image hundreds of Martian dust devils, picked out in good contrast against the surface, rather than being observed against a dusty sky. Dedicated dust devil imaging campaigns allowed 761 dust devils to be observed at the Spirit landing site (Greeley et al., 2006, 2010). This imaging data set allowed the seasonal and daily time of occurrences, sizes, durations, and translation speeds of the dust devils to be analyzed. What was particularly striking, though, was how Earth-like the images returned were (e.g., Fig. 2b) and that the observed dust devils were also Earth-like in scale and morphology. The volume of observations allowed Earth-like “fieldwork” to be performed for the first time, resulting in maps of dust devil motion across three different seasons (Greeley et al., 2010).

The Spirit data provided deep understanding of the dust devil population at a specific site on Mars, but only by deriving population statistics for dust devils in general can their global or regional impact on climate or air quality be calculated or modeled. Here, the contribution of R. Lorenz, of the Johns Hopkins Applied Physics Lab, should not be understated. Beginning in 2009 (Lorenz, 2009) Lorenz and his colleagues sought to analyze data from, and to understand the limitations of, human and robotic surveys of terrestrial and Martian dust devils and to bring them into a quantitative population framework. Like Ryan and Carroll, nearly 40 years previously, Lorenz challenged the assumption that mobile sampling is needed to measure statistically significant amounts of dust devils, arguing that a “chase” strategy places too much emphasis on human bias (chasing the most impressive or convenient examples only). Instead, he and his collaborators developed a series of automated sensors to collect data over long periods of time, amassing an impressive body of work describing the evolution of their techniques and the results obtained (Jackson & Lorenz, 2015; Lorenz, 2010, 2013b, 2013a, 2014; Lorenz & Jackson, 2016; Lorenz, Jackson, & Lanagan, 2018; Lorenz & Lanagan, 2014; Lorenz, Neakrase, & Anderson, 2015). Lorenz has also explored novel means to study dust devils, such as thermal imaging (Lorenz, 2004) and “seismic” signature analysis (Lorenz, Kedar, et al., 2015), as well as being the only researcher to have seriously explored the threat posed by dust devils to humans, documenting hazards to light aircraft and lightweight structures such as barns or temporary buildings and quantifying the “death rate by dust devil”: 0.001 per year per million population (Lorenz et al., 2016; Lorenz & Myers, 2005).

State of the Art and Future Directions


Dust devils are challenging phenomena to measure—they are transient; they are variable in size, shape, and intensity; and their occurrence is hard to predict. Nevertheless, over several decades of fieldwork the basic parameters that describe the terrestrial dust devil population have been well described, as well as their morphological types and diurnal and seasonal frequency of occurrence. These data are summarized briefly in Table 1, but much more information is available in detailed review papers (Balme & Greeley, 2006; Murphy et al., 2016).

In addition to providing understanding of dust devils for application to Mars, a key goal of terrestrial dust devil studies is to understand their local to global effects on aerosol budgets and air quality. Such studies have begun (e.g., Gillette & Sinclair, 1990; Jemmett‐Smith, Marsham, Knippertz, & Gilkeson, 2015; Klose & Shao, 2016; Koch & Renno, 2005) and take various approaches, as reviewed in Klose et al. (2016), but all need to integrate field studies. Hence, such synoptic studies require more information than is available as inputs into their calculations or to validate their models. The main aspects that have require more study include:


Measurements of dust loading and vertical dust flux. Strictly, dust flux within a dust devil is the integrated product of vertical wind speed and dust load per unit volume at all points in a horizontal plane through the dust devil. However, such a measurement is extremely difficult to make, given the complex three-dimensional nature of dust devils. Indeed, even assuming rotational symmetry and then estimating vertical dust flux as the mean suspended load multiplied by the integrated vertical wind velocity in a single profile across a dust devil is incredibly challenging, requiring simultaneous, high cadence, vertical wind speed and dust load measurements. Metzger et al. (2011) provide the largest data set to date; they found typical dust fluxes within dust devils of ~0.9 × 10–3 to 7.5 × 10–3 g m−2 s−1. They calculated these values by measuring vertical wind speed and dust load in dust devils as they passed over a mobile (but stationary, at least while the measurement was being made) sensor suite. They calculated the mean product of vertical wind speed and dust load for each measurement point in the profile across the observed extent of the dust devil. They noted that these “mean flux” values derived from profiling gave vastly smaller (1/10th to 1/50th) fluxes than the “peak flux” derived simply from multiplying peak vertical wind speed with peak dust load. Peak flux figures have been used in some extrapolations of field studies to regional or global estimates of dust devil particle lifting and hence could be in error. However, Metzger et al.’s measurements were derived from ~10 dust devils in only two locations, suggesting that there is much to be gained from a far larger study of dust loading and vertical wind speed in dust devils. Such a study should cover dust devils of different sizes and intensities and at different locations, to assess the broader range of dust fluxes and enable better parameterizations or ground truthing of regional and global studies.


A clear outcome of terrestrial studies is that there is no such things as a “typical” dust devil: size, rotational and vertical wind speed, dust loading, and lifetime of dust devils each range across one or more orders of magnitude, meaning that the amount of dust lifted by an individual dust devil is hugely variable. Yet there is a need to describe the amount of dust lifted by dust devils as a population. Early-21st-century studies (see review by Lorenz & Jackson, 2016) have attempted to pin down the shape of the dust devil size frequency population (e.g., Kurgansky, 2012; Lorenz, 2009, 2011; Pathare et al., 2010) and Δ‎P frequency population (e.g., Jackson & Lorenz, 2015; Lorenz & Lanagan, 2014) to provide statistical solutions that can be applied to different regions and times. However, population studies suffer from both a lack of data and poor data handling (e.g., Lorenz & Jackson, 2016, identify the binning of dust devil size data as a major impediment to their sound analysis). Also, human surveys are time consuming and challenging to perform, given the high labor cost and harsh environments. Automated detections of Δ‎P values are useful for determining the intensity of dust devils but suffer from an inability to detect dust loading—though new studies combining pressure detectors and solar flux measurements (Lorenz, Neakrase, et al., 2015) can detect size, intensity, and (if calibrated with in situ measurements) dust load simultaneously. The use of automated image analysis of time-lapse imaging data is also becoming practical (Lorenz et al., 2018). As sensor, data storage, and image analysis technology improves further, it is possible that long-term monitoring of dust devils could become practical. Linking this to simultaneous monitoring of the local boundary layer meteorology is a goal to aspire to and could provide the reliable data sets that describe dust devil populations in context with local conditions.

Table 1. Typical Terrestrial Dust Devil Parameters



Example References


Columnar, V-shaped, disordered. Column often surrounded by V-shaped “sand skirt.” Larger dust devils can contain subcortices

Metzger (1999)


Seconds to minute; longer-lived dust devils not uncommon; larger dust devil are longer-lived

Flower (1936); Ives (1947); Lorenz (2013b)

Diurnal occurrence

Late morning to mid-afternoon; earlier and later examples not uncommon

Jackson and Lorenz (2015); Ryan and Carrol (1970)

Rotation sense

No preference

Carroll and Ryan (1970)

Δ‎P (core pressure drop)

1–10 mb (mobile); <1 mb (fixed)

Metzger (1999); Jackson and Lorenz (2015)

Peak horizontal wind speeds within dust devils (@2 m height)

Usually 5–10 ms–1; rare measurements up to 25 ms–1

Ryan and Carrol (1970); Balme et al. (2003)

Temperature excursion

1–5° C; some larger values reported

Sinclair (1973); Tratt, Hecht,Catling, Samulon, & Smith(2003)

Horizontal translation speed

10%–20% greater than ambient winds at 10 m height; usually 1–10 ms–1

Balme et al. (2012)

Particle load (dust)

0.0008–0.042 g m-3(mean)

Metzger et al. (2011)

0.001–0.162 g m–3 (peak)

Particle load (total suspended load)

0.006–0.875 g m–3 (peak)

Metzger et al. (2011)


Much of our understanding of terrestrial dust devils—aside perhaps from studies of their contribution to local and regional aerosol budgets—comes from decades of research and is summarized in review papers such as Balme and Greeley (2006). On the other hand, the Martian dust devil literature is diverse and began only in the later part of the 20th century (though well described across the chapters of Reiss et al., 2017), so summary information about Martian dust devils is provided here.

Appearance and Size

Martian dust devils (Fig. 2) are similar in morphology to terrestrial examples, especially when observed from the surface (e.g., Greeley et al., 2010). However, most dust devil observations have been made from orbit, which disallows detailed morphological description. Orbital data do allow estimates of diameter and height (from shadow measurements) to be made, though, and measurements from 12 such studies are summarized by Fenton et al. (2016), who find median heights of 450 m and median diameters of 160 m. From these data they demonstrate that most Martian dust devils have height/diameter aspect ratio of 1:5, though aspect ratios of 5:20 are not uncommon. Orbital measurements are biased toward larger dust devils, of course, and such studies are not routinely performed on Earth. Lorenz and Jackson (2016) therefore consider whether surface observations of Martian dust devils show a different size population to those of terrestrial studies. They conclude that, despite problems caused by binning of terrestrial data, Martian dust devils are larger (at least in Gusev crater), having a modal size about three times greater than terrestrial examples. Lorenz and Jackson also note that if dust devil height is limited by the thickness of the planetary boundary layer (being convective systems), and if Martian dust devils have a similar aspect ratio to terrestrial ones, then the deeper Martian boundary layer (~10 km on Mars compared with 2–4 km on Earth) explains the factor of 3 size difference.

Global Distribution of Active Dust Devils

Active dust devils have been observed across all elevations and nearly all latitudes on Mars (Cantor et al., 2006; Fenton et al., 2016; Fisher et al., 2005; Stanzel et al., 2008; Towner, 2009). Their detection from orbit requires observation during times when they are active. Thus the combination of diurnal variability in dust devil formation times and variability (or fixed times) of the orbits of observing spacecraft can bias surveys. Certain areas of Mars have very high dust devil formation rates: northern Amazonis Planitia (~190–220° E, 30–50° N) is one such area (Cantor et al., 2006). A combination of low surface albedo (i.e., high summer surface temperatures), a deep convective boundary layer, and abundant dust might explain why this area is so productive, but it should be noted that apparently similar areas of Mars produce far fewer dust devils (Fenton et al., 2016). This spatial variability awaits a fuller explanation.

Dust Devil Tracks

Dust devil tracks (Fig. 7) are sinuous albedo features left on the Martian surface by the passage of dust devils, but not all dust devils form tracks—perhaps only 10% to 20% of dust devils, globally (Cantor et al., 2006). Dust devil tracks are generally darker than their surroundings, but bright tracks are also seen (Reiss et al., 2016). They have been only rarely observed on Earth (Reiss et al., 2016; Reiss, Raack, Rossi, Di Achille, & Hiesinger, 2010; Reiss, Zimmerman, & Lewellen, 2013; Reiss, Raack, & Hiesinger, 2011; Rossi & Marinangeli, 2004) but are found in huge numbers on the Martian surface, on virtually every terrain apart from the polar caps (Whelley & Greeley, 2008). Importantly, dust devil tracks form seasonally, persisting for only a few months (Balme, Whelley, et al., 2003; Verba, Geissler, Titus, & Waller, 2010) until erased by dust settling or seasonal frost. The length of dust devil tracks provides information about dust devil longevity, their width, dust devil size, and their sinuosity about local wind strength or variability. Terrestrial observations (e.g., Reiss, Raack, et al., 2011; Reiss et al., 2010, 2013) show that dust devil tracks are not necessarily formed by simple processes (i.e., removal of brighter dust over a dark substrate) but are instead formed due to complex interactions between wind and surface, such as destruction of aggregates or removal of finer components from mixed-sized sediments. Thus dust devil tracks record changes in surface photometric properties, rather than simple removal of material.

In some ways, dust devil tracks provide a better proxy for dust devil activity than active remote sensing measurements of active dust devils, in that they leave longer-lived signatures but are removed on a timescale shorter than a Martian season (conversely, variations in surface types provides a confounding factor that can complicate the link between dust devil tracks and active dust devils). This has been exploited to infer information about spatial and temporal variations in dust devil formation from dust devil track observations. For example, many fewer dust devil tracks are found in orbital images taken in autumn and winter, in agreement with terrestrial understandings. Also, more dust devil tracks are found at mid-latitudes during hemispherical spring and summer, with more found in the south than the north (Whelley & Greeley, 2008)—in agreement with Ryan’s (1969) prediction from nearly 50 years previously. Whelley and Greeley attribute this hemispherical difference to the position of Mars’ orbital perihelion, with the southern summer being warmer and shorter than the northern summer.

Figure 7. Dust devil tracks covering a part of the Martian surface. HiRISE image ESP_013538_1230, Image centered at 56.87° S, 198.10°E.

Image credit NASA/Jet Propulsion Laboratory/University of Arizona.
Wind Speeds

Aside from a very few measurements of wind speed in a dust devil made from rapid (< few seconds separation) repeat imaging from orbit (Cantor et al., 2006; Choi & Dundas, 2011) and the surface (Greeley et al., 2010), all Martian data for dust devil wind speeds come from in situ lander meteorology data. However, not all Mars rovers or landers have been equipped with meteorology instruments, and in many occasions wind sensors have failed, had serious problems, or lacked the sampling rate to adequately measure dust devil wind speeds. Peak wind speeds in dust devils have been measured directly using Viking (lander) Meteorology Instrument System data (Ryan & Lucich, 1983) or inferred for “near misses” by assuming a proscribed vortex structure (Ringrose et al., 2007; Ringrose, Towner, & Zarnecki, 2003). A few mechanical wind sensor measurements, which matched Δ‎P events in the pressure sensor, were made by the high-latitude Phoenix Lander (Ellehoj et al., 2010). Fenton et al. (2016) and Murphy et al. (2016) summarize the data available: dust devils on Mars have median vertical wind speeds of 1 to 2 ms–1 (maximum ~17 ms–1) and measured/inferred tangential velocities which are generally up to several 10s of ms–1, although some inferred measurements are closer to ~100 ms–1.

Thus while there are more than 100 direct measurements of wind speeds in Martian dust devils, generally these are not tied to other observations in such a way that their location within the dust devil is known, or if the vortex was actually dust loaded. Reliable measurements of wind speeds in dust devils probably await dedicated lander instruments. These could be long-lived, high cadence sampling; fixed surface meteorology experiments; or “triggered” meteorology experiments that activate their anemometry instruments when the approach of a dust column is detected, either through detection of insolation fluctuations or by automated imaging.

Δ‎P and Temperature Excursions

All measurements of pressure and temperature excursions within Martian dust devils (or, more strictly, convective vortices, as it is generally unknown whether such detections were of dust-loaded vortices or not) come from meteorology packages on landers or rovers. Pressure measurements were made by both Viking landers, but their sample rate and minimum detection limit were too poor to detect Δ‎P values of passing vortices and the Mars Exploration Rovers did not carry suitable sensors. Δ‎P data therefore come from Pathfinder (Ferri et al., 2003; Murphy & Nelli, 2002; Schofield et al., 1997), Phoenix (Ellehoj et al., 2010), and the Mars Science Laboratory (MSL) Curiosity rover (Kahanpää et al., 2016; Steakley & Murphy, 2016). In general, all data sets agree, finding Δ‎P values of <5 Pa (0.05 mb), with many more small than large Δ‎P values: Lorenz (2012) suggested a power law distribution with exponent of approximately –2 to describe the population. Interestingly, as pointed out by Murphy et al. (2016) and Lorenz and Jackson (2016), although Martian Δ‎P values are much lower in magnitude than terrestrial ones (reflecting the much lower surface pressure on Mars), they are higher as a percentage of ambient pressure, suggesting Martian dust devils are more “intense,” or perhaps that weaker ones have been more poorly sampled.

Peak temperature excursions of a few degrees Centigrade were measured within dust devils by Viking, Pathfinder, Phoenix, and MSL (Ellehoj et al., 2010; Ferri et al., 2003; Kahanpää et al., 2016; Murphy & Nelli, 2002; Schofield et al., 1997). These are broadly similar to terrestrial measurements, being generally of a few degrees Kelvin.

Daily and Seasonal Occurrence

Dust devil seasonal variations follow those seen on Earth. Surface and orbital observations of dust devils and their tracks show dust devils to be most common in spring and summer (e.g., Balme, Whelley, et al., 2003; Cantor et al., 2006; Greeley et al., 2006, 2010; Whelley & Greeley, 2008), though some examples are seen in autumn and winter. Martian dust devils also show similar diurnal patterns to Earth—forming mainly during late morning and afternoon, local time, although several studies of the passage of convective vortices (not necessarily dust-loaded) show modal occurrence times closer to midday (Murphy & Nelli, 2002; Steakley & Murphy, 2016), and the Phoenix lander data (Ellehoj et al., 2010) show a possible mid-morning “spike” in frequency. Most orbital measurements do not cover a range of local times due to their fixed orbits. The Mars Express spacecraft, however, carrying the High Resolution Stereo Camera (HRSC; Neukum & Jaumann, 2004), has an orbit that allows HRSC to image the surface over a range of local times. Using this instrument, Stanzel et al. (2008) found that dust devils occurred from local noon to late afternoon, peaking between ~14:00 to 15:00 local time. It is interesting to note that an often-used scheme for parameterization of dust devil activity in Martian Global Circulation Models (GCMs) can generate significant amounts of morning dust devil activity (Chapman, Lewis, Balme, & Steele, 2017), although whether this is an artifact of the parameterization or a true reflection of global dust devil behavior is unknown.


Martian dust devil lifetimes are hard to constrain, as measurement requires continual observation from start to finish—which is difficult on Mars from either surface or orbital platforms. Minimum duration time estimates can be made if multiple images of an individual dust devil are acquired, separated by known times. Lifetimes can also be inferred from observation of horizontal motion speeds of dust devils combined with measurement of the length of the track they leave behind them. Greeley et al. (2006, 2010) measured minimum dust devil lifetimes in Gusev crater, finding a mean of 2 to 3 minutes for dust devils with mean diameters of a few 10s of meters. Stanzel et al. (2008) and Reiss Zanetti, and Neukum (2011) used HRSC measurements of dust devil tracks to infer minimum lifetimes of 10s of minutes for dust devils with diameters of 100s of meters. It seems likely that, as on Earth, larger dust devils are longer lived and that lifetimes of an hour or more are possible.

Dust Lifting

Martian dust devils lift dust, or else they would not be visible in images. Determining the dust load per unit volume, and extrapolating that to a dust removal or vertical transport flux, is very difficult. On Earth, in situ methods have been used (Metzger et al., 2011), but the results for both particle load and flux span orders of magnitude (both due to variability within the dust devil population and to measurement error, caused partly by the difficulty in collecting such data in the first place). Also, the total suspended load is much higher near the bottom of the dust devil and includes dust to sand (perhaps up to granule or pebble) sized materials, but, higher in the dust devil, the material is almost entirely dust-grade and the particle load is far lower (Metzger et al., 2011; Oke, Dunkerley, & Tapper, 2007; Raack, Reiss, Balme, Taj-Eddine, & Ori, 2017). On Mars, imaging studies have been used to estimate particle load in dust devils. Estimates for dust load for Martian dust devils include ~0.07 gm–3 from Pathfinder data (Metzger, Carr, Johnson, Parker, & Lemmon, 1999), 2.1 × 10–6 gm–3 to 0.25 gm–3 from MER Spirit data (Greeley et al., 2010), 0.02 to 0.15 gm−3 from THEMIS orbital observations (Towner, 2009), and 0.004 to 0.122 gm−3 from HiRISE orbital observations (Reiss, Hoekzema, & Stenzel, 2014). Clearly, there is a huge variation in these data, but they show that dust loading in Martian dust devils can be as high as ~0.1 gm−3, surprisingly similar to terrestrial measurements of up to 0.04 gm−3 for mean dust load and three (or sometimes more) times this for peak load (Metzger et al., 2011).

Extrapolations of dust load to dust flux are fraught with uncertainty. As noted by Metzger et al. (2011), bulk dust load measurements cannot simply be multiplied by maximum vertical wind speed to give dust flux. This ignores variations in both vertical wind speed and dust load across the dust devil. Hence, many Martian flux calculations are likely to be overestimates. Mars estimates for dust flux range from ~10–6 gm–2s–1 to ~1 gm–2s–1, a range that encompasses the lower end of vortex dust removal flux measurements (0.02 to 500 gm–2s–1) made under Martian conditions using the ASUVG apparatus (Neakrase & Greeley, 2010). More data that report the dust loading and vertical wind speeds within Martian dust devils are needed if better estimates of dust flux are to be made. Again, this probably awaits dedicated instruments or instrument control systems.

The processes by which dust is lifted by dust devils are generally understood but poorly constrained (Neakrase et al., 2016). It is likely that wind shear is the main mechanism (Rafkin et al., 2016) for dust lifting, perhaps assisted by “saltation triggering” (Greeley, 2002) in which more easily lifted sand grade particles reimpact the surface, splashing up dust. Secondary mechanisms including the Δ‎P effect (Balme & Hagermann, 2006), and possibly even thermos-luminescent lifting (e.g., Küpper & Wurm, 2015) or electrification of dust particles in dust devils (see summary in Harrison et al., 2016), might also play a larger role on Mars than Earth but need to be investigated further.

Effects on Martian Climate

The effects of dust devils on the Martian dust cycle (a key component of the Martian climate) are mainly studied using GCMs, in which wind shear and dust devil lifting are “tuned” to fit observations, allowing inferences to be drawn about the relative importance of each. Dust devils contribute about half of the annual dust lifting (e.g., Kahre et al., 2006; Newman et al., 2005) but do not trigger global dust storms. Any albedo changes caused by track formation seem not to have long-term regional effects, such as stripping dust from given regions and depositing it in others (Klose et al., 2016) and thereby affecting global albedo patterns, although this has not yet been well studied. Future climate modeling studies could seek new ways to parameterize dust devils based on new observational spatial data (if dust flux and population studies can be refined), new studies of dust devil formation in LES models that include dust lifting (Spiga et al., 2016; Spiga, Forget, Lewis, & Hinson, 2010), or modifications to existing parameterizations.

Impact on Mars Exploration

Continued robotic, and perhaps human, exploration of Mars is likely to continue. Dust devils have been suggested as possible hazards to exploration due to their strong winds, electrification, and dust lifting. However, dust devils can also benefit exploration: the passage of dust devils has been shown to remove dust from the solar panels of Mars landers, improving their power generation and prolonging the lifetime of the missions (Lorenz & Reiss, 2015). Whether hazard, boon, or scientific target, the recognition that dust devils are common on Mars must be taken into account when future missions are planned.

Conclusion and Future Directions

Dust devils form from boundary layer convection and are efficient dust transport and lifting mechanisms. On Earth, they can enhance local aerosols and form a potential hazard to lightweight structures and even aircraft, but their contribution to the global dust cycle seems to be small. On Mars, however, they have a measurable effect on the dust cycle, which in turn plays a key role in the Martian climate. Dust devils are therefore important targets for future research. For both Martian and terrestrial studies, measurement of dust devil populations provide an ongoing challenge. Improved automation and availability of low-cost sensors mean that better terrestrial studies can be made to understand populations’ statistics and to link them with local meteorology. On Mars, where every kilogram and watt are important on a rover or lander, it is harder to see how rigorous populations studies can be made except at a very few sites. Nevertheless, the availability of high-resolution, global orbital imaging data, and new developments in automatic image recognition, might compensate for a lack of in situ measurements.

Measurement of dust load, and its extension to estimation of dust flux by measurement of vertical wind field in dust devils, is another area where more data are needed. Again, field studies on Earth will likely improve, with the combination of flux and population studies allowing more accurate ideas of the effects of dust devils on terrestrial aerosol budgets. The addition of more robust wind sensors to Mars landers would provide new data that could feed in to flux measurements.

As computer models improve, physical processes can be simulated in higher fidelity. High-resolution, large-domain LES models are beginning to be used to simulate convection on Mars and resolve vortices within the convective system. Future studies that include variable ambient winds, surface roughness and albedo, dust lifting processes, and radiatively active airborne dust will further improve understanding. Such models could provide important information about how dust devils form, what atmospheric parameters govern their size and intensity, and how much dust they can entrain. These results can in turn be fed into new parameterization schemes for GCM studies to improve understanding of the global impact of dust devils on Mars.

The future direction of travel in dust devil research will be to integrate the diverse elements (especially numerical models and field measurements) to make inferences about the regional to global impact of dust devils. It is likely that modeling studies will take big steps, but each new step will need to “go back” to field studies for validation and could require new ways of measuring dust devils in the field and hence new field studies.

Further Reading

  • Bagnold, R. A. (1941). The physics of windblown sand and desert dunes. London, U.K.: Methuen.
  • Balme, M. R., & Greeley, R. (2006). Dust devils on Earth and Mars. Reviews of Geophysics, 44(RG3003).
  • Greeley, R., & Iversen, J. (1985). Wind as a geologic process on Earth, Mars, Venus and Titan. Cambridge, U. K.: Cambridge University Press.
  • Read, P. L., & Lewis, S. R. (2004). The Martian climate revisited: Atmosphere and environment of a desert planet. Berlin, Germany: Springer.
  • Reiss, D., Ralph Lorenz, R. D., Balme, M. R., Neakrase, L.D.V. Rossi, A. P., Spiga, A., & Zarnecki, J. C. (2017). Dust devils. Space Sciences Series. New York, NY: Springer.
  • Shao, Y. (2000). Physics and modelling of wind erosion. Dordrecht, The Netherlands: Kluwer Academic.